Full text of Economic Report of the President : 2018

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Economic Report
of the President
Together with
The Annual Report
of the
Council of Economic Advisers
February 2018

Economic Report
of the President
Together with
The Annual Report
of the
Council of Economic Advisers
February 2018

x

Contents
Economic Report of the President.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
The Annual Report of the Council of Economic Advisers. . . . . . . . . . . . . . . . . . . . 13
Chapter 1

Taxes and Growth. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Chapter 2

Deregulation That Frees the Economy. . . . . . . . . . . . . . . . . . . . . . . 73

Chapter 3

Labor Market Policies to Sustain the Middle Class. . . . . . . . . . . . 103

Chapter 4

Infrastructure Investment to Boost Productivity.. . . . . . . . . . . . . 159

Chapter 5

Enhancing U.S. Trade in a Global Economy. . . . . . . . . . . . . . . . . . 219

Chapter 6

Innovative Policies to Improve All Americans’ Health. . . . . . . . . 279

Chapter 7

Fighting Cybersecurity Threats to the Growing Economy. . . . . . 323

Chapter 8

The Year in Review and the Years Ahead. . . . . . . . . . . . . . . . . . . . 387

References

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449

Appendix A	Report to the President on the Activities of the Council
of Economic Advisers During 2017.. . . . . . . . . . . . . . . . . . . . . . . . . 515
Appendix B S
 tatistical Tables Relating to Income, Employment,
and Production. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527

____________
*For a detailed table of contents of the Council’s Report, see page 21.

iii

Economic Report of the President

x

Economic Report of the President
To the Congress of the United States:
In the past year, I have taken several major steps to implement an economic
agenda that champions prosperity and success for the American people. This
began with two crucial pro-growth policies: massive tax cuts and sweeping
regulatory reform. These historic accomplishments, and many other achievements in our first year, have restored confidence within our families and
businesses that the United States is, and always will be, the greatest place
on Earth to pursue opportunity. I am proud to present the economic agenda
my Administration has designed and begun implementing to enable every
American to build a bright future—to achieve the American Dream.
The American people have long been awaiting an effective pro-growth
agenda that inspires robust economic activity, spurs innovation, creates jobs,
and improves families’ financial security. The Federal Government’s economic
policy in recent decades, however, has been a story of one broken promise
after another.
For too long, leaders have promised growth, but done little to change
policies that drove business away and pushed our factories and jobs offshore.
They promised prosperity, but layered on regulatory red tape that hurt workers and businesses in the struggle to keep up with the ever-increasing cost of
complying with new rules. They promised to empower American citizens with
opportunities and resources for their future. Instead, they enabled a bloated
bureaucracy, special interest groups, and unaccountable international institutions. Consequently, my Administration inherited an economy with relatively
slow growth. The median American’s real household income from working had
dropped lower than it was a decade ago. And the previous Administration’s
forecasts suggested that this would never get better.
My Administration’s pro-growth policy agenda is reversing these trends
and ending the disappointments of the past. No longer will we perpetuate the
illusion that policies that encourage economic growth, job creation, and wage
growth are out of our reach. No longer will we turn a blind eye to the regions of
our country that have suffered from politicians’ failed programs and misplaced
priorities. No longer will we tolerate unfair and nonreciprocal trade practices
that impoverish our workers. And no longer will we tolerate burdensome,
backward, or perverse economic policies that guarantee bleak outcomes for
our citizens.

Economic Report of the President | 3

Policies for a Pro-Growth Agenda
The primary components driving my Administration’s pro-growth policy
agenda—tax cuts, tax reform, and smart deregulation—have inspired enormous confidence in the economy and optimism that it will continue thriving.
We approached tax reform with the following principles: Our corporate
tax system was uncompetitive with the rest of the world; the American middle
class deserved a fairer, simpler tax code that lowered rates, exempted more
income from taxation altogether, and limited costly deductions favoring special interests; and a smarter tax system would encourage business expansion,
job creation, and wage growth.
On December 22, 2017, I signed into law the Tax Cuts and Jobs Act of
2017. My Administration worked closely with Congress to pass this historic
bill, which brings much-needed tax relief to the middle class and to small businesses. For too long, our tax policy squeezed families and small businesses
with unfairly high tax burdens. Our backward tax code drove companies, jobs,
and profits abroad. Our corporate tax rate was 60 percent higher than that of
our average economic competitor—it was the highest in the developed world.
This came at a huge cost to our citizens: the median American’s real household
income collapsed at the beginning of this century, and it took 15 years to
recover the losses.
The Report that follows shows that the corporate tax changes alone
are expected to increase annual income for families by an average of $4,000.
Americans at all income levels will receive a tax cut. The standard deduction
nearly doubles, and the child tax credit fully doubles. Bringing down the business tax rate from 35 to 21 percent, and allowing firms to fully expense their
investments in equipment and intellectual property, is encouraging robust
economic activity by making America a more competitive place to do business.
Other changes to the tax code are encouraging companies to bring back to
America the nearly $3 trillion in wealth that they have parked overseas.
The enactment of the Tax Cuts and Jobs Act has inspired businesses to
express their confidence in the economy even before the new law’s complete
implementation unleashes its full, tangible benefits, which include larger
paychecks, greater profits, and lower tax burdens. My Administration has kept
its commitment to making smarter, pro-growth economic policy, sending the
crucial signal to businesses that America is once again the most promising
place in the world to build, invest, create, expand, and hire. In response to our
historic changes to the tax code, over 350 companies have, to date, announced
billions in new investments in plant and equipment in the United States, along
with improvements to workers’ compensation or benefits, including raising the
minimum wage for their employees, increasing 401(k) contributions, and giving bonuses. These improvements will affect more than 4.1 million employees.

4 | Economic Report of the President

As we knew when we set out to reform taxes and rein in the regulatory
state, when America’s enterprises are optimistic about where the economy is
heading, they will expand their operations; invest more in plant and equipment, which raises workers’ productivity; and, as a result, raise wages. We
have recently seen capital spending again contribute to rising productivity,
after holding productivity growth back under the previous Administration, and
have also once again started to observe economic growth above the 3 percent
level—which many claimed was impossible. The stock market reached record
highs, creating trillions of dollars in wealth—reflecting consumers’ and investors’ confidence.
In addition to reforming tax policies that drove business away, my
Administration began a Government-wide effort to eliminate costly, obsolete
regulations soon after taking office. For too long, the regulatory state had
grown at the expense of our economy and wages. It ballooned far beyond what
is needed to protect citizens from harm. The effect of excessive regulation is
well known: it functions as a tax, stunting economic growth and generally discouraging productive activity. This undermined the operation of our markets,
increased costs for consumers, and empowered Washington bureaucrats over
American consumers. To date, we have stopped the growth of the regulatory
state in its tracks. In our first year, we adopted a miniscule fraction of the
“economically significant” rules churned out by other Administrations. We
exceeded our policy of striking two regulations for each one we create, eliminating 22 regulations for every 1 we have adopted. By creating fewer rules and
reliably committing to eliminating the unnecessary old ones, we have signaled
to firms that they can invest in growing their business. Business confidence has
skyrocketed in response.
But the best is yet to come. As we continue to implement the new tax bill
and cut regulatory red tape, the following priorities round out our pro-growth,
America-first agenda.

Infrastructure and Energy
A modernized, safer transportation infrastructure is imperative for connecting
our citizens to opportunities. Delayed projects, insufficient investment, traffic
congestion, and wear and tear are slowing America down. My Administration
is dedicated to generating $1.5 trillion in new infrastructure investment and
shortening the approval process so projects may be permitted in under two
years.
Renewing our infrastructure must prioritize accountability and enable
greater State and local control. We are committed to reversing the legacy
of prolonged deadlines and the wasted resources that befell infrastructure
projects in the past and that has deprived Americans of the infrastructure they
deserve. President Obama’s stimulus package was intended to significantly
increase investment in infrastructure, create jobs, and improve the economy;
Economic Report of the President | 5

however, only 3.5 percent of the over-$800 billion plan went to constructing
transportation infrastructure. Taxpayers felt severely misled. Bureaucracy created years-long delays to breaking ground on many projects. And millions of
promised “shovel-ready” jobs never appeared.
The American people are tired of empty promises. They know this Nation
is capable of doing so much better. They want accountability for promised
projects and jobs. They want decisionmaking to be done at the State and
local levels, where folks know what needs to be done and how to do it. My
Administration will work with Congress to develop an infrastructure plan that
enables sophisticated projects to be approved and finished promptly.
Advancing the capabilities of our infrastructure means more than just
rebuilding roads and bridges. Americans rely on the Internet to work, learn,
and connect with each other, and it must be modernized to suit their needs.
In his 1994 Economic Report of the President, President Clinton promised to
connect “every classroom, every library, and every hospital in America” to the
Internet by 2000. Decades later, 39 percent of rural Americans still lack highspeed broadband. And a quarter of America’s K-12 students lack adequate
Internet connectivity at school. It is intolerable to continue pretending that
this is the best America can offer to our students. My Administration is working
to expand accessibility and expedite the process of bringing the Internet to
hard-to-reach areas.
As we plan for a reinvigorated infrastructure, we must also plan for the
future of American energy by renewing our commitment to energy dominance.
This encompasses energy independence, which ensures energy security for
American families and businesses for decades to come, and global energy leadership, which consolidates our role as a major influence in the international
energy sphere.
The United States is the world leader in combined total production of oil
and natural gas. Technological ingenuity has unleashed the American energy
supply, and domestic energy production is increasing for both petroleum
and natural gas. Capitalizing on and expanding our energy supply solidifies
our position as a global energy leader, reduces our reliance on imported
energy, and provides opportunities for job creation and economic growth. Our
untapped oil and natural gas fields are crucial resources for solidifying our
energy security and independence. Increasing energy supply will also reduce
electricity costs, lowering utility bills for American households—particularly
low-income families, which are disproportionately affected by high utility bills.
Lower electricity costs can also reinvigorate American manufacturing and job
growth by decreasing manufacturing costs for American-made products.
To further unleash America’s energy potential, we must also reduce
the excessive regulatory burden that inhibits production. Since I took office,
my Administration has been working to remove these barriers, enabling the
private sector to create more jobs; increase wages for its employees; produce
6 |

Economic Report of the President

clean, affordable energy to power our economy; and increase American energy
exports. Steady and secure supplies of coal, liquefied natural gas, refined products, and crude oil enable the United States to maintain our energy leverage
in the international sphere. With our plentiful supply, the United States can
reduce other nations’ reliance on exporters that use energy as a geopolitical
weapon. Diversifying our energy sources and furthering research on next-generation energy technologies will enable our Nation’s firms to provide cleaner,
more efficient, and more affordable energy for American citizens.

Trade and the International Sphere
International trade offers an opportunity to grow the economy, but it must
be pursued with American interests in mind. We cannot continue to pursue
a global order that disregards America’s prosperity and the well-being of our
citizens. My Administration believes in the benefits of free trade, and it is committed to robust trade relationships with other nations that value fairness and
reciprocity. Some of our trading partners, however, do not share these values,
and this places unsustainable stress on global trade. As such, the United States
can no longer reward governments that distort the free market with illegal
subsidies or handcuff American exporters with high tariffs and nontariff barriers. We can no longer tolerate unfair trade practices, such as forced technology
transfer and industrial espionage. And we can no longer tolerate complacency
toward unfair and nonreciprocal trade that undermines America’s potential.
Immediately after taking office, I began revising our Nation’s trade policies to promote the interests of the American people. My Administration is in
the process of improving the North American Free Trade Agreement and the
Korea–United States Free Trade Agreement. In the coming year, we will also
seek to negotiate new, better trade deals, and we will hold accountable any
country that engages in unfair trade practices through tough and focused
action.
With a tougher, smarter, and more clear-eyed agenda, we can rejuvenate
our trade deals, boost our exports, and improve the economic prospects of
our citizens and communities. By following through on this agenda, we will
become a stronger and more secure Nation that can offer better opportunities to our people. For far too long, the United States put up with unfair trade
policies that left our workers behind. Previous Administrations pursued unenforceable multilateral trade deals and let other countries get away with unfair
practices. We will not perpetuate a trade agenda that exports our factories and
jobs, weakens our manufacturing and defense industrial bases, and diminishes
our economic and national security.

Innovation, Intellectual Property, and Cybersecurity
America has long been one of the most productive, innovative, and forwardthinking nations in the world. In recent decades, however, Government policy
Economic Report of the President | 7

has not properly recognized the importance of innovation for growth and the
value of America’s distinct entrepreneurial spirit. Renewed commitment to
enforcing our intellectual property rules encourages innovation by signaling
that inventions will be rewarded and protected, and it enables the United
States to maintain its long-standing advantages in science and technology.
Protecting our evolving technologies and ideas from theft and forced technology transfers, to which American enterprises are increasingly subject abroad,
is not only in America’s economic interest—it conveys our great pride in the
capabilities of America’s innovators and the value of their ideas.
Protecting the results of American ingenuity—such as new technology,
research findings, and other forms of innovation—is crucial to our future
economic and wage growth. The IP Commission estimates that stolen intellectual property reduces GDP by 1 to 3 percent a year—an annual loss in the
range of $185.7 billion to $557.1 billion. Given this threat to our economy,
my Administration is partnering with the private sector to protect American
technologies, intellectual property, and innovators. We are striving to ensure
an economic climate that values innovation, encourages the private sector to
invent, and enables the private sector to protect itself from all forms of industrial espionage, including cyber theft. And we will take action when necessary.
As the threats to our economy have become more complex and widespread, it
has become increasingly important to empower our companies and citizens
with tools that help protect against cyberattacks, cyber theft, and other malicious activities that endanger our citizens’ privacy, our national security, and
our economic success.

Investing in the Future Americans Deserve
My Administration is proud of the economic success that the United States
has achieved over the past year, and we are dedicated to sustaining this
progress. We are working hard to reverse policies that drove many businesses
to other countries, taking jobs with them. In 2017, our strengthened economy
created 2.2 million jobs, and the unemployment rate was at its lowest level in
seventeen years. Maximizing growth and opportunity, however, must include
two key priorities that did not receive enough attention for a very long time:
workforce enablement and support for distressed communities. We are
dedicated to ending the low labor force participation rate inherited from the
previous administration, and to uplifting our Nation’s communities that have
borne the highest costs of bad policies, in the forms of poverty, despair, and
drug addiction.
Within the communities that suffered the most when businesses left is a
silent Nation of Americans who have dropped out of the workforce. Although
our overall low rates of unemployment are encouraging, the labor force
participation rate for prime-age men is lower now than it was in every year
between 1948 and the start of the Great Recession. This means that as bad

8 | Economic Report of the President

policies ran their course, too many of our citizens lost hope, were discouraged
by Government programs from working, or struggled to find opportunities and
stopped looking. And while most of America is working and enjoying the benefits of our now-robust economy, we cannot forget that there are still millions
of people who have been left behind. We are committed to addressing this
deeply troubling situation by reversing the economic policies that disregarded
so many of our citizens for so long and to creating new, focused remedies.
My Administration’s Rural Prosperity Task Force is working to solve problems
such as disproportionately high rural poverty rates and inadequate broadband
access. The Investing in Opportunity Act of the Tax Cuts and Jobs Act provides
tax incentives to invest in low-income communities. Combined with our focus
on job creation and workforce development, we are driving a renewed commitment to the communities in our Nation that have been neglected for far
too long.
My Administration is not only concerned about the economic consequences of the dismal tax, trade, immigration, welfare, and regulatory policies
of the past, which drove Americans from the labor force. We are also concerned
that the low labor force participation rate excludes Americans from many other
benefits of civil society.
To address this, we are committed to making it possible for all working-age, able-bodied Americans to contribute to our economy and society.
Strengthening work requirements as a condition of welfare assistance will help
more individuals experience the many benefits of work and vastly improve the
financial situations of those receiving assistance. We are targeting for reform
all Government programs that provide incentives to avoid work. Workforce
development programs play a key role in helping individuals who are receiving
assistance to find jobs and earn incomes that will enable them to move from
dependency to self-sufficiency. For this reason, we are advancing programs
that teach or update job skills, promote lifelong learning, offer training in
trades, and better connect our industries with the future workforce. We are prioritizing greater availability of paid parental leave and affordable, high-quality
childcare for the workforce in order to strengthen our families and enable more
women to fully participate in the workforce. And we are committed to ensuring
that America’s elementary and secondary schools and institutions of higher
education equip students with the skills they need to achieve lifelong success.
To boost our workforce and heal our society, we must also contend with
crisis levels of drug abuse, addiction, and overdose. Drug overdose deaths
numbered 63,632 in 2016. Partly because of this crisis, life expectancy in the
U.S. dropped in 2015 and 2016—the first time it has fallen in consecutive years
since 1963. Not only do drug addiction and overdoses take lives too soon; they
leave traumatized family members and distraught communities in their wake.
When we speak of investing in the future, those affected by addiction are some
of our citizens who will need the most support.
Economic Report of the President | 9

To date, we have taken many measures to alleviate the opioid crisis. We
are focused on addressing the vast, illicit supply of these drugs, particularly
by investing in the infrastructure of interdiction. We are shoring up detection
capabilities in international mail-processing centers, using fentanyl-detecting
dogs and screening technologies, and attacking the light and dark webs where
transnational criminal marketplaces operate. And we have increased funding and resources to the States to strengthen their responses to the opioid
epidemic, including the creation of the State Targeted Response to the Opioid
Crisis grant program. We have improved policies to make it easier to find and
crack down on bad actors, to expand access to treatment for addicts, to invent
new solutions for treating addiction, and to further develop nonaddictive pain
management treatments through partnerships with innovative companies.
We continue to prioritize solving this terrible crisis, alleviating the suffering it
has caused, and preventing future anguish. As we discover the best ways to
diminish drug abuse and addiction in our country, we must have an economy
capable of giving opportunities to build a brighter future to the victims of the
opioid crisis and similar kinds of adversity.
Fulfilling the promise of a brighter future must also include developing
solutions to a concern shared by many Americans: the rising cost of healthcare.
In the past, government attempts to improve healthcare gave rise to mandates,
rules, and healthcare tax and spending programs that, perversely, increased
what the average American family paid for healthcare. My Administration fought
to repeal the individual mandate, no longer forcing the American people to buy
the expensive plans that bureaucrats designed for them. My Administration
is committed to providing Americans with more affordable health coverage
options by promoting choice and competition in healthcare markets, and by
addressing health behaviors that can promote a longer, healthier life—which
insurance alone cannot do. Troubled by the growing unsustainability of the
Federal-State Medicaid program, we are committed to improving the program
for both its beneficiaries and the taxpayers who finance it.
We are also committed to preventing Government expansion from stifling innovation in the healthcare sector, and to preventing international freeriding that takes advantage of our pharmaceutical innovation. We are focused
on correcting policies that hinder drug price competition, addressing a top
concern of the American people: that they cannot afford the medications they
need. We have made it a high priority to ensure that safe, generic alternatives
to important drugs are approved on a faster timeline, making last year a record
one for generic drug approvals.
Finally, my Administration has worked to counter the disappointing history of the services that have been provided to our veterans. We are pursuing
stringent accountability within the Department of Veterans Affairs and seeking
to provide much stronger support and opportunities to our veterans. The brave

10 |

Economic Report of the President

men and women who have served our Nation have made countless sacrifices.
We owe them nothing less as they build their futures and plan for retirement.

***

We have made great strides in our first year, and we remain committed to
policies that grow our economy and improve each citizen’s chance to succeed.
My vision for this term is to increase American families’ prosperity; to encourage job creation and wage growth by regaining a competitive business climate
with smarter tax and trade policy, and also deregulation; to capitalize on our
resources and technology to achieve energy dominance; to invest in infrastructure to make commerce more vibrant and connect citizens with opportunities;
to encourage innovation as one of our most powerful national security and
economic tools; and to enable our distressed communities to prosper by combating workforce development issues and the opioid crisis.
To achieve this vision, we will not rely on the belief that the Government
knows best and can solve every problem. Instead, we will continue crafting an
economic program that lays the groundwork for the conditions that will enable
our citizens to achieve success and prosperity. We will continue to take pride in
making policies that honor the dignity and ingenuity of the American people,
and in dedicating our work to the welfare of each of our citizens.
From the outset, my Administration has valued nothing more highly than
bringing freedom, prosperity, and opportunity to all American families. To that
end, we are dedicated to empowering them with a robust economy. The economic agenda outlined here is designed to make it possible for the American
people to dream of, and to achieve, the bright future they deserve.

The White House
February 2018

Economic Report of the President | 11

The Annual Report
of the

Council of Economic Advisers

Economic Report of the President | 13

x

Letter of Transmittal
Council of Economic Advisers
Washington, D.C., February 21, 2018
Mr. President:
The Council of Economic Advisers herewith submits its 2018 Annual
Report in accordance with the Employment Act of 1946, as amended by the Full
Employment and Balanced Growth Act of 1978.
Sincerely yours,

Kevin A. Hassett
Chairman

Richard V. Burkhauser
Member

Tomas J. Philipson
Member

Economic Report of the President | 15

Introduction
The purpose of the Annual Report of the Council of Economic Advisers is to
provide the public and the economic policy community with a detailed account
of the performance of the U.S. economy in the preceding year and with an
analysis of the Administration’s domestic and international economic policy
priorities for the years ahead. In this Report, we thus review the salient policy
developments of 2017 and preview policy aims for the coming years, in the
context of the Administration’s unified agenda to expand our economy and the
economic prosperity of all Americans.
The U.S. economy experienced strong and economically significant
acceleration in 2017, with growth in real GDP exceeding expectations and
increasing from 2.0 and 1.8 percent in 2015 and 2016 to 2.5 percent, including
two successive quarters above 3.0 percent. The unemployment rate fell 0.6
percentage point, to 4.1 percent, its lowest level since December 2000, while
the economy added 2.2 million jobs, an average of 181,000 per month. Notably,
manufacturing and mining—having lost 9,000 and 98,000 jobs, respectively, in
2016—added 189,000 and 53,000 jobs during 2017. Labor productivity grew 1.1
percent, compared with a decline of –0.1 percent in 2016, and average hourly
earnings of private employees rose 2.7 percent, compared with average growth
of 2.1 percent during the preceding 7 years. Reflecting the economy’s outperformance of expectations, the January 2017 Blue Chip consensus forecast of
2.3 percent GDP growth in 2018 was revised upward in February 2018 to 2.7
percent.
The four quarters of 2017 thus marked a nontrivial trend shift. From 2010
through 2016, real output in the United States grew at an average annual rate of
2.1 percent, while labor productivity grew, on average, by less than 1 percent.
The pace of economic recovery was slow by historical standards, particularly
because recent research has confirmed Milton Friedman’s original observation
that in the United States, deeper recessions are typically succeeded by steeper
expansions, and that this correlation is in fact stronger when the contraction
is accompanied by a financial crisis. Since the nineteenth century, the recent
recovery was one of only three exceptions to this pattern.
In the chapters that constitute this Report, we provide evidence that the
historically anemic recovery from the Great Recession was not independent of
policy choices, and accordingly we proceed to identify the exacerbating factors
in the weakness of the post-2009 recovery and the current Administration’s
strategies and menu of policy options to address them.
First and foremost, in chapter 1, on the historic Tax Cuts and Jobs Act
(TCJA), we find that investment and labor productivity have been inhibited
in recent years by the coincidence of high and rising global capital mobility
and an increasingly internationally uncompetitive U.S. corporate tax code and
Economic Report of the President | 17

worldwide system of taxation. This combination had the effect of deterring U.S.
domestic capital formation, thereby restraining capital deepening, productivity growth, and, ultimately, output and real wage growth, with the economic
costs of corporate taxation thereby increasingly and disproportionately borne
by the less mobile factor of production—namely, labor. Indeed, the five-year,
centered-moving-average contribution of capital services per hour worked to
labor productivity actually turned negative in 2012 and 2013 for the first time
since World War II. We estimate that by lowering the cost of capital and reducing incentives for corporate entities to shift production and profits overseas,
the corporate provisions of the TCJA will raise GDP by 2 to 4 percent over the
long run, and increase average annual household income by $4,000.
Similarly, in chapter 2 we discuss a large body of academic literature indicating that an excessive regulatory burden can negatively affect productivity
growth, and thus overall growth, by attenuating the flow of new firms’ entries
and established firms’ exits, and also by amplifying the spatial misallocation
of labor and creating employment barriers to entry. We furthermore highlight
actions the Administration has already taken to eliminate inefficient and
unnecessary regulations, with the effect of raising prospects for innovation,
productivity, and economic growth.
In chapter 3, on labor markets, we find considerable evidence suggesting, as with regulation, that postrecession efforts to strike a new optimum on
the frontier of social protection and economic growth may have sacrificed too
much of the latter in pursuit of the former. We also find that while demographic
shifts owing to the retirement of aging Baby Boom cohorts exerted strong
downward pressure on the labor force participation rate, factors other than
demography accounted for one-third of the overall decline in participation
during the recovery, and half the decline since the cyclical peak in the fourth
quarter of 2007 (also see chapter 8). For instance, we find that increases in fiscal transfers during the Great Recession intended to mitigate the demand-side
effects of rising unemployment generated persistent negative effects on the
prime-age labor supply. Meanwhile, structural unemployment coterminous
with imperfect geographic mobility—exacerbated by regulatory restrictions
(chapter 2), drug abuse (chapter 6), and inadequate investment in infrastructure (chapter 4)—have similarly intensified downward trends in labor force
participation among prime-age workers.
These challenges, however, particularly those of low labor productivity
growth and declining labor force participation, are not policy-invariant. As
we highlight in chapters 1 and 8, for example, policies that incentivize highly
skilled and experienced older workers to defer retirement, such as the marginal
income tax rate reductions enacted by the TCJA, can have important implications not only for labor force participation but also for productivity. Moreover,
by raising the target capital stock, we expect the TCJA to result in capital

18 |

Economic Report of the President

deepening, again contributing to productivity growth and rising household
earnings.
Relatedly, in chapter 4 we document the deficiencies of our current
public infrastructure, and investigate the adverse effects of these deficiencies
on economic growth and consumer welfare, as well as potential remedial
policy options. In particular, we examine how the fundamental mismatch
between the demand for and supply of public infrastructure capital could be
ameliorated by utilizing existing assets more efficiently and by adjusting longrun capacity to levels best matched with local needs, which would allow local
governments more flexibility in giving prices a larger role in guiding consumption and investment decisions, and in streamlining environmental review and
permitting processes. Moreover, addressing the current inadequacies of our
public infrastructure would help to attenuate the coincidence of structural
unemployment with imperfect geographic mobility—again, exacerbated by
the regulatory restrictions discussed in chapter 2—that has been a factor in the
decline of labor force participation identified in chapters 3 and 8.
Chapter 5 covers issues in international trade policy and actions the
Administration has taken and could take to generate positive-sum, reciprocal
trade agreements with our trading partners. Specifically, in addition to reviewing the benefits of economic specialization and consequent gains from trade,
we also demonstrate how instances of unfair trade practices by a subset of our
partners have had the effect of limiting the potential gains from trade to the
United States and the world, with particularly adverse consequences for the
U.S. manufacturing sector. Addressing these issues would raise productivity by
encouraging greater investment in sectors where the U.S. economy enjoys a
comparative advantage, especially but not exclusively energy and agricultural
products.
In chapter 6, we turn our attention to the health of the true catalyst of
U.S. economic growth: the American worker. Although the Affordable Care Act
(ACA) expanded insurance coverage to at most 6 percent of the U.S. population—through Medicaid, marketplaces, and the dependent coverage provision—we survey a large body of academic literature that estimates the effect of
insurance coverage on health to be substantially smaller than commonly presumed. Indeed, for the first time in over 50 years, U.S. life expectancy declined
in 2015 and 2016, suggesting that factors such as drug abuse, particularly of
opioids, and obesity may have a larger impact than insurance coverage alone
can redress. Instead, we find that increased choice and competition, along with
a recognition by policymakers that the determinants of health are multidimensional, may constitute more efficient avenues for improving health outcomes,
particularly among lower-income households. Fundamentally, it is the view of
this Council that healthy people not only live longer, more enjoyable lives but
are also an essential component of reversing recent trends in labor productivity and labor force participation, discussed in chapters 1 and 3.
Economic Report of the President | 19

Chapter 7 considers the emerging challenge of cybersecurity, particularly
in the context of our ongoing transition to an information economy. Malicious
cyber activity presents new threats to the protection of property rights, including rights to intangible assets and even information itself, and thus imposes
large and real costs on the U.S. economy. Given the existence of positive
externalities from investing in cybersecurity, we discuss policy options that
might shift this investment to its socially optimal level, including public-private
partnerships that promote basic research, protecting critical infrastructure
assets, disseminating new security standards, and expanding the cybersecurity
workforce.
Finally, in chapter 8 we examine the year in review and survey the
years ahead. Acknowledging underlying strengths and challenges, the
Administration’s November 2017 baseline forecast, which excludes the effects
of the TCJA, projects that output will grow by an overall average annual rate
of 2.2 percent through 2028. The policy-inclusive forecast, however, which
assumes full implementation of the Administration’s agenda, is for average
annual real GDP growth through 2028 of 3.0 percent. We expect growth to
moderate slightly after 2020, as the capital-output ratio approaches its new
steady state level and the pro-growth effect of the individual elements of the
TCJA dissipate, though the level effect will be permanent. However, expected
further deregulation and infrastructure investment will partly offset the declining contribution to growth of tax cuts and reforms toward the end of the
budget window. The policy-inclusive forecast is conservative relative to those
of previous Administrations, and in fact is slightly below the median of 3.1
percent. Moreover, the baseline forecast is precisely in line with the long-run
outlook given in the 2017 Economic Report of the President, reflecting our view
that nonimplementation of the current Administration’s policy objectives will
imply a reversion to the lower growth trend of recent years.
Preliminary indicators suggest that markets indeed detect a trend shift.
In the weeks immediately following the TCJA’s passage, over 300 companies
announced wage and salary increases, as well as bonuses and supplementary
401(k) contributions of $2.4 billion affecting 4.2 million workers, citing the new
law. In addition, by the end of January 2018, this Council tallied $190 billion in
newly announced corporate investment projects that were publicly attributed
to the TCJA, revealing that firms are responding to the TCJA as theory and
empirical evidence predicted.
As a society, we hold many values and aspirations, including but not
limited to expanding economic prosperity, that may not exist always and
everywhere in complete harmony. It is the view of this Council that in recent
years, the pursuit of alternative policy aspirations at the expense of growth has
imposed real economic costs on the American people, in the form of diminished opportunity, security, equity, and even health. We therefore endorse an
agenda for returning the American economy to its full growth potential.
20 |

Economic Report of the President

x

Contents
Chapter 1: Taxes and Growth. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
U.S. Taxation in Historical and International Perspective.. . . . . . . . . . . . . . . . .
Costs of Taxation in the United States. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Excess Burden of Taxation.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Compliance Costs of Taxation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

34
36
39
41

Tax Incidence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
The Incidence of the Corporate Income Tax.. . . . . . . . . . . . . . . . . . . . . . . . . 45
The Incidence of the Mortgage Interest and State and Local
Tax Deductions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Evidence on Taxes and Growth.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Effects on the Cost of Capital. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Effects on Net Capital Outflows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Effects on Profit Shifting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Effects of Crowding Out. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

54
57
59
63
64

Estimating the Growth Effects of Tax Reform. . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Chapter 2: Deregulation That Frees the Economy. . . . . . . . . . . . . . . . . . . . . . . . 73
Theories of Efficient and Inefficient Regulation and Deregulation.. . . . . . . . . . 75
Regulatory Benefits. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Regulatory Costs.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
The Impact of Regulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Measurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Aggregate Growth.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Business Dynamics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Productivity.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Employment .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Labor Mobility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

81
81
84
87
88
91
92

The Trump Administration’s Initiatives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

Chapter 3: Labor Market Policies to Sustain the Middle Class. . . . . . . . . . . . . 103
Trends in GDP and GDP per Capita across the Business Cycles. . . . . . . . . . . .
Median Income Trends across Business Cycles.. . . . . . . . . . . . . . . . . . . . . . . .
Six Measures of Median Income. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Decomposing Median Income Changes by Employment and Earnings. . .

110
114
115
123

Measures of Employment and Earnings Using Aggregate Data. . . . . . . . . . . .
Addressing America’s Employment Challenge.. . . . . . . . . . . . . . . . . . . . . . . . .
Labor Force Participation among the Older Population. . . . . . . . . . . . . . . . . .
Policy Options to Promote Participation among Older Americans. . . . . . . . .

125
127
131
136

Economic Report of the President | 21

Labor Force Participation among Teenagers and Young Adults.. . . . . . . . . . .
Policy Options to Promote Participation among Teenagers and
Young Adults. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Labor Force Participation among Prime-Age Adults. . . . . . . . . . . . . . . . . . . . .
Demand-Side Factors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Supply-Side Factors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Government Transfers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Physical and Mental Health. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Geographic Immobility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alternative Uses of Time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

137
143
144
145
147
150
151
152
153

Policy Options to Promote Participation among Prime-Age Workers. . . . . . . 153
Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Appendix.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

Chapter 4: Infrastructure Investment to Boost Productivity. . . . . . . . . . . . . . 159
U.S. Infrastructure’s Growing Problem of Excess Demand. . . . . . . . . . . . . . . .
How Increasing the Supply of Infrastructure Supports Economic Growth. . .
Recent Trends. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Evidence for the Growth Effects of Public Capital. . . . . . . . . . . . . . . . . . . .

164
166
167
170

Funding and Financing Needed Infrastructure. . . . . . . . . . . . . . . . . . . . . . . . .
Fiscal Roles for Federal, State, and Local Governments. . . . . . . . . . . . . . .
Funding Infrastructure Investment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Financing Infrastructure Investment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

176
177
177
192

How Core Infrastructure Ensures a Competitive Economy. . . . . . . . . . . . . . . . 205
The Energy Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Modernizing America’s Waterways. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

Chapter 5: Enhancing U.S. Trade in a Global Economy. . . . . . . . . . . . . . . . . . . 219
The Economics of Trade. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
U.S. Trade in the World Economy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Assessing the Gains from Trade. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
U.S. Trade Balances in Perspective. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Services Trade.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

222
222
224
228
231

Dividing the Pie: Defining Fair Trade between Countries. . . . . . . . . . . . . . . . .
Trade Barriers.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tariff Overhangs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Responses to Unfair—and Not-So-Unfair—Trade Practices.. . . . . . . . . . . .
Nontariff Measures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Nontariff Barriers to Automobile Trade. . . . . . . . . . . . . . . . . . . . . . . . . . . .
Nontariff Barriers to Agricultural Trade. . . . . . . . . . . . . . . . . . . . . . . . . . . .
Defining the Rules of the Game for Trade. . . . . . . . . . . . . . . . . . . . . . . . . .

233
236
238
240
242
246
248
250

Domestic Gains and Losses from Trade. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Domestic Gains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Domestic Losses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Manufacturing Job Losses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

254
254
255
255

Trade Opportunities for the United States.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 263

22 |

Economic Report of the President

U.S. Agricultural Trade.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
U.S. Energy Dominance Relies on Trade. . . . . . . . . . . . . . . . . . . . . . . . . . . 270
Updating American Trade Policy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

Chapter 6: Innovative Policies to Improve All Americans’ Health. . . . . . . . . . 279
Healthcare Insurance and Spending, and the ACA. . . . . . . . . . . . . . . . . . . . . .
The Impact of the ACA and Health Insurance on Health. . . . . . . . . . . . . . .
Why the ACA Expansions May Have Limited Health Effects. . . . . . . . . . . . .
The ACA’s Impact on Spending. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

282
283
285
287

Improving People’s Health by Limiting the Effects of Unhealthy Behavior. . .
Improving Health by Combating the Opioid Epidemic. . . . . . . . . . . . . . . .
Obesity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Smoking. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Policy to Address Health Behaviors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

291
292
297
298
300

Improving People’s Health through More Access to Medical Innovations. . . .
Encouraging Innovation, and Making It Affordable. . . . . . . . . . . . . . . . . . . . . .
Why Americans Pay High Prices for Biopharmaceutical Products,
and How to Lower Them.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Cutting High Prices Resulting from Price Manipulation . . . . . . . . . . . . . . .
Enhancing Price Competition in the Pharmacy Benefit Manager Market..
Raising Innovation Incentives and Decreasing Free-Riding to Reduce
the Price of Better Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Government’s Role in Reducing Prices and Stimulating Innovation..

302
306
308
312
313
314
320

Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320

Chapter 7: Fighting Cybersecurity Threats to the Growing Economy.. . . . . . 323
Malicious Cyber Activities and Cyber Threat Actors.. . . . . . . . . . . . . . . . . . . . .
The Costs of Adverse Cyber Events Incurred by Firms.. . . . . . . . . . . . . . . . . . .
Estimating the Costs of Adverse Cyber Events for Firms . . . . . . . . . . . . . . .
The Distribution of Adverse Cyber Events across Sectors. . . . . . . . . . . . . .
Externalities from Weak Cybersecurity and Underinvestment in Cyber
Protection.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Spillover Effects to Economically Linked Firms. . . . . . . . . . . . . . . . . . . . . .
Attacks through the Weakest Link in the Supply Chain. . . . . . . . . . . . . . . .
Using Cyber Vulnerabilities to Usurp Resources and Launch Attacks on
Other Firms.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Economy-wide Spillover Effects from Firms with Critical Infrastructure
Assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

327
331
332
344

Common Vulnerabilities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Software. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Firmware.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hardware. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Cloud Computing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

351
352
353
354
354

346
347
347
350
351

The Problem of Insufficient Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
Dark Cyber Debt. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
Cyber Insurance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359

Economic Report of the President | 23

The Costs of Malicious Cyber Activity for the U.S. Economy. . . . . . . . . . . . . . .
Devastating Cyberattack Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Financial Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Power Grid.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

362
365
366
367

The Rise of Quantum Computing and the Need for Better Encryption. . . . . .
Approaches to Reducing Cyber Risk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Public Sector Efforts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Private Sector Efforts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

368
369
369
377

Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384

Chapter 8: The Year in Review and the Years Ahead. . . . . . . . . . . . . . . . . . . . . 387
Output.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Consumer Spending. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Investment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Government Purchases.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Net Exports. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

392
393
397
404
405

The Labor Market. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Policy Developments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fiscal Policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Monetary Policy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

411
415
415
417

Productivity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Wage and Price Inflation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Financial and Oil Markets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Equity Markets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Interest Rates and Credit Spreads. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Oil Prices and Production. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

419
422
425
426
427
430

The Global Macroeconomic Situation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Developments in 2017.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Global Productivity Slowdown . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Advanced Economies’ Sluggish Increase in Wage Inflation.. . . . . . . . . . . .
Too-Low Price Inflation in Advanced Economies. . . . . . . . . . . . . . . . . . . . .
Financial Vulnerability in Emerging Market Economies. . . . . . . . . . . . . . .

431
431
434
435
436
437

The Outlook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
GDP Growth during the Next Three Years.. . . . . . . . . . . . . . . . . . . . . . . . . .
GDP Growth over the Longer Term.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Upside and Downside Forecast Risks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

439
443
444
447

Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449

24 |

Economic Report of the President

Appendixes
A.
B.

Report to the President on the Activities of the Council of
Economic Advisers During 2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  515
Statistical Tables Relating to Income, Employment, and
Production. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  527

Figures
1-1
1-2
1-3
1-4
1-5
1-6
1-7
1-8
1-9
1-10
1-11
1-12
2-1
2-2
2-3
2-4
2-5
2-6
3-i
3-ii
3-1
3-2
3-3
3-4
3-5
3-6
3-7
3-8
3-9
3-10

Federal Tax Revenue as a Percentage of GDP, 1792–2017. . . . . . . . . . . . . . . . . 35
Top Statutory Corporate Tax Rates in OECD Countries, 2000–2017. . . . . . . . 37
Effective Average Corporate Tax Rates of OECD Countries, 1999–2017 . . . . . 37
Corporate Tax Rates and Corporate Tax Revenue, 2000–2016. . . . . . . . . . . . . 41
The Incidence of a Tax on Sellers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Tax Incidence with Perfectly Inelastic Supply. . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Tax Incidence with Perfectly Elastic Supply. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Average Income Increases for U.S. Households from a Corporate Rate
Reduction, Using Literature Wage Elasticities. . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Distribution of Mortgage Interest Deduction, by Adjusted Gross Income. . . 53
Distribution of State and Local Tax Deduction, by Adjusted Gross Income. . 54
Contribution of Business Investment to Real GDP Growth, 2010–16. . . . . . . . 60
The Effect of an Exogenous Tax Cut on Nominal Interest Rates. . . . . . . . . . . . 66
Total Paperwork Burden Hours, FYs 1997–2015. . . . . . . . . . . . . . . . . . . . . . . . . . 81
Estimated Measures of Regulation, 1976–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Published Economically Significant Final Rules by First Year of a
Presidential Term. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Product Market Regulation, 2013 (35 OECD Countries). . . . . . . . . . . . . . . . . . . 85
Total Factor Productivity Growth, 1990–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Final and Economically Significant Rules, 2007–17. . . . . . . . . . . . . . . . . . . . . . . 96
U.S. Household Income Distribution in 2016. . . . . . . . . . . . . . . . . . . . . . . . . . .  107
Self-Reported Economic Class of Americans (percent), 2015 . . . . . . . . . . . .  108
Level of Real GDP and Peak-to-Peak Average Real GDP growth,
1948–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  111
Average Annual Real GDP Growth Rates during Business Cycles,
1948–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  112
Real GDP Growth Rates, 1948–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  112
Average Annual Real GDP per Capita Growth Rates over Business
Cycles, 1948–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  113
Alternative Measures of Trends in Real Median Income, 1979–2016. . . . . .  120
Alternative Measures of Real Median Income, 1979–2016 . . . . . . . . . . . . . . .  123
Age 16+ Ratio of Employment to Population Measured Relative to
Opening Business Cycle Peak, 1980–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  125
Prime-Age (25–54) Ratio of Employment to Population, Measured
Relative to Opening Business Cycle Peak, 1980–2017. . . . . . . . . . . . . . . . . . .  126
Real Average Hourly Earnings, Private Production and Nonsupervisory
Workers, Measured Relative to Opening Business Cycle Peak,
1980–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  127
Labor Force Participation Rates by Gender (Seasonally Adjusted),
1948–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  130

Economic Report of the President | 25

3-11
3-12
3-13
3-14
3-15
3-16
3-17
3-18
3-19
3-20
3-iii
3-21
3-22
4-1
4-2
4-3
4-4
4-5
4-i
4-6
4-7
4-8
4-ii
4-9
4-iii
4-iv
4-v
4-10
4-11
4-12
4-13
4-14
4-15
5-1
5-2
26 |

Annual Labor Force Participation Rate by Age Group, 1970–2017. . . . . . . .  130
Annual Labor Force Participation Rate for Older Workers by Age Group,
Relative to 1970, 1970–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Annual Female Labor Force Participation for Older Workers by Age
Group, Relative to 1970, 1970–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  132
Annual Male Labor Force Participation for Older Workers by Age Group,
Relative to 1970, 1970–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Annual Labor Force Participation Rate for Teens and Young Adults by
Sex and Age Group, 1970–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  136
Employment and Enrollment Status of 16- to 19-Year-Olds, 2001–16. . . . .  140
Employment and Enrollment Status of 20- to 24-Year-Olds, 2001–16. . . . .  140
Annual Labor Force Participation Rate for 16- to 19-Year-Olds, by
Enrollment Status, 2001–16. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  142
Change in Time Spent on Types of Activities for People Age 16 to 19
and 20 to 24 Years, 2003–5 to 2013–15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  142
Annual Labor Force Participation Rate for Prime-Age Workers by Age
and Sex, Relative to 1970, 1970–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  145
Annual Labor Force Participation Rates for Women, by Age of Youngest
Child, 1994–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  149
Share of U.S. Residents Who Moved during the Past Year, 1947–2016 . . . .  154
Alternative Measures of Trends in Real Median Income, Deflated Using
PCE Chain Price Index, 1979–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  157
Road Mileage and Vehicle Miles Traveled, 1980–2016. . . . . . . . . . . . . . . . . . .  165
Traffic Congestion Measures, 1982–2014. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  165
Inland Waterways System Lock Delays, 1993–2016. . . . . . . . . . . . . . . . . . . . .  167
Public Spending on Transportation and Water Infrastructure,
1956–2014 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  168
Contribution of Public Capital Stock to Productivity Growth, 1947–2016.  169
Educational Attainment of Infrastructure versus Noninfrastructure
Labor Force Participants, 2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  174
Public Spending on Mass Transit, Highways, and Water Infrastructure,
2014. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  178
Federal Fuel Tax Revenues, 1993–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  182
The Highway Trust Fund’s Highway Account, 1990–2016 . . . . . . . . . . . . . . .  182
Oregon Fuel Tax Revenues, 1993–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  184
New U.S. State and Local Government Debt Issues, 2004–16. . . . . . . . . . . .  195
P3 Enabling Legislation, by U.S. State and Territory. . . . . . . . . . . . . . . . . . . .  197
Pennsylvania Rapid Bridge Replacement Program Project Cost . . . . . . . . .  198
Access to Fixed Terrestrial Broadband Service, 2014 . . . . . . . . . . . . . . . . . . .  200
Renewable Energy’s Current and Projected Share of U.S. Generation
Capacity, 2005–50 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  208
Renewable Energy’s Share of Annual Generation Additions, 2006–17 . . . .  208
Total Coal Exports by Weight and Revenue, 2002–17. . . . . . . . . . . . . . . . . . .  210
U.S. Annual Natural Gas Exports, 2001–17. . . . . . . . . . . . . . . . . . . . . . . . . . . . .  210
Lost Transportation Time across the Inland Waterways System,
2000–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  213
Inland Waterways Trust Fund Balance, 2009–16 . . . . . . . . . . . . . . . . . . . . . . .  214
Goods versus Services Value Added and Employment. . . . . . . . . . . . . . . . . .  224
Trade Relative to GDP for the OECD and the United States, 1970–2015. . .  225
Economic Report of the President

5-3
5-4
5-5
5-6
5-7
5-8
5-9
5-10
5-11
5-i
5-12
5-13
5-14
5-15
5-16
5-17
5-18
5-19
5-20
5-21
5-22
5-23
5-24
5-25
5-26
6-i
6-1
6-2
6-3
6-4
6-5
6-6
7-1
7-2
7-3
7-i
7-ii
7-4
7-5

U.S. Goods-Alone Trade Balance and Share of GDP, 1976–2017. . . . . . . . . .  227
U.S. Monthly Goods Trade with the World. . . . . . . . . . . . . . . . . . . . . . . . . . . . .  229
U.S. Monthly Services Trade with the World . . . . . . . . . . . . . . . . . . . . . . . . . . .  229
Goods and Services Trade Balance, by Country, 2016. . . . . . . . . . . . . . . . . . . 230
Trade Balance as a Share of GDP, 1970–2016. . . . . . . . . . . . . . . . . . . . . . . . . .  231
U.S. Bilateral Trade Balances with Major Trading Partners, 2016 . . . . . . . .  232
U.S. Exports of Goods and Services, by Destination, 2016. . . . . . . . . . . . . . .  235
Trade Barriers around the World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  239
Antidumping Measures, 2001–15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  243
CSPV Module Prices, 2010–18. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  244
Estimated Annual Cost of IP Theft to the U.S. Economy. . . . . . . . . . . . . . . . .  253
U.S. Manufacturing Employment, 1965–2017. . . . . . . . . . . . . . . . . . . . . . . . . .  257
Displaced Workers by Industry, 2005–15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  260
Characteristics of Displaced Compared with Nondisplaced Workers,
2005–15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  260
Share of Displaced Workers Changing Industry or Location, 2005–15 . . . .  262
Probabilities of Employment and Labor Force Nonparticipation
Transitions for Displaced Workers, by Industry Relative to Finance,
Insurance, and Real Estate, 2005–15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  262
U.S. Agricultural Imports, Exports, and Trade Balance by Value, 1940–
2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  267
Government Support of Agricultural and Nonagricultural Goods in
High-Income Countries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  267
Comparing Agricultural Trade Restrictiveness . . . . . . . . . . . . . . . . . . . . . . . . .  269
Tariffs on Agricultural Imports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  269
U.S. Petroleum and Total Goods Trade Deficits, 1991–2016. . . . . . . . . . . . .  272
U.S. Monthly Coal Exports, 2003–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  274
U.S. Monthly Natural Gas Withdrawals and Net Imports, 1973–2017 . . . . .  275
U.S. Natural Gas Exports by Pipeline and LNG, 2000–17. . . . . . . . . . . . . . . . .  275
U.S. Quarterly Solar Photovoltaic Installations, 2011:Q1–2017:Q3 . . . . . . .  276
Federal Marketplace Premiums and Issuers, Plan Years 2014–18 . . . . . . . .  288
Opioid-Involved Overdose Deaths, 1999-2016 . . . . . . . . . . . . . . . . . . . . . . . . .  293
Overdose Deaths by Type of Opioid Involved, 1999-2016 . . . . . . . . . . . . . . .  294
Smoking Rates and the Price of Cigarettes, 1995-2015. . . . . . . . . . . . . . . . . .  301
U.S. Male Deaths Caused by HIV Disease, 1987-2015. . . . . . . . . . . . . . . . . . . .  301
Politico / Harvard Polling on Americans’ Top Priorities for Congress,
2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  308
OECD Patented Pharmaceutical Sales, 2016. . . . . . . . . . . . . . . . . . . . . . . . . . .  315
Cost Components of an Adverse Cyber Event. . . . . . . . . . . . . . . . . . . . . . . . . .  332
Cumulative Abnormal Return by Firm Size . . . . . . . . . . . . . . . . . . . . . . . . . . . .  336
Cumulative Abnormal Return by Type of Adverse Cyber Event . . . . . . . . . .  337
Equifax’s Cumulative Abnormal Returns After Its September 2017 Data
Breach Announcement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  341
SolarWorld’s Cumulative Abnormal Returns After Its Data Breach
Become Publicly Known. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  344
Distribution of Security Breaches by Industry. . . . . . . . . . . . . . . . . . . . . . . . . .  346
TransUnion’s and Experian’s Cumulative Abnormal Returns After
Equifax’s Data Breach Announcement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  348

Economic Report of the President | 27

7-6
7-7
7-8
8-1
8-2
8-3
8-4
8-5
8-6
8-7
8-8
8-9
8-10
8-11
8-12
8-13
8-14
8-15
8-16
8-17
8-18
8-19
8-i
8-ii
8-iii
8-iv
8-20
8-21
8-22
8-23
8-24
8-25
8-26
8-27
8-28
8-29

28 |

Portfolio of Finance Firms’ Cumulative Abnormal Returns After Equifax’s
Data Breach Announcement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  348
Investment Projections in Cybersecurity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  379
Frequency and Cost Savings, by Technology . . . . . . . . . . . . . . . . . . . . . . . . . .  380
Evolution of Blue Chip Consensus Forecasts for Real GDP Growth During
the Four Quarters of 2017 and 2018. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  390
Unemployment Rate, 1975–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  391
Real Growth in GDP, Private Domestic Final Purchases (PDFP), and Gross
Domestic Output (GDO), 2000–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  393
Personal Saving Rate, 1990–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  394
Real Consumer Spending and Real Compensation, 1975–2017. . . . . . . . . .  395
Consumer Sentiment, 1980–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  395
Household Debt Relative to Disposable Personal Income (DPI), 1995–
2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  396
Consumption and Wealth Relative to Disposable Personal Income (DPI),
1952–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  397
Composition of Growth in Real Business Fixed Investment, 2010–17. . . . .  398
Net Investment as a Share of the Capital Stock, Private Nonresidential,
1945–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  399
Total Payouts to Shareholders, Nonfinancial Corporate Businesses,
1952–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  399
Year-Round For-Sale Vacant Housing Units, 1996–2017. . . . . . . . . . . . . . . . .  401
National House Price Indices, 2000–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  401
Share of Mortgages with Negative Equity, Past Due Installments, and in
Foreclosure, 2009–17. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  402
Housing Affordability Index, 1990–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  403
Housing Starts, 2000–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  404
Contribution of Inventory Investment to Real GDP Growth, 2010–17. . . . .  405
Contribution of Government Purchases to Real GDP Growth, 2003–17 . . .  406
Government Purchases as a Share of Nominal GDP, 1948–2017 . . . . . . . . .  406
Real State and Local Government Purchases During Recoveries, 1960–
2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  407
Real State and Local Expenditures, 1955–2017. . . . . . . . . . . . . . . . . . . . . . . . .  408
State and Local Government Surplus as a Share of Nominal GDP, 1947–
2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  408
Unfunded Liabilities of State and Local Pension Systems Relative to
Revenues, 1951–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  409
Contribution of Exports to U.S. Real GDP Growth, 2000–2017 . . . . . . . . . . .  410
Contribution of Net Exports to U.S. Real GDP Growth, 2000–2017. . . . . . . .  410
Monthly Nonfarm Payroll Job Growth, 2007–17. . . . . . . . . . . . . . . . . . . . . . . .  412
Unemployment Rate by Duration, 1990–2017. . . . . . . . . . . . . . . . . . . . . . . . . .  413
Rates of Part-Time Work, 1960–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  413
Alternative Measures of Labor Force Underutilization, 2007–17 . . . . . . . . .  415
Federal Budget Deficit as a Share of Nominal GDP, FYs 1950–2017. . . . . . .  416
Federal Reserve Total Assets, 2000–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  418
Nonfarm Business Productivity, 1953–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . .  420
Sources of Labor Productivity Growth, 1953–2016 . . . . . . . . . . . . . . . . . . . . .  421

Economic Report of the President

8-30
8-31
8-32
8-33
8-34
8-35
8-36
8-37
8-38
8-39
8-40
8-41
8-42
8-43
8-44
8-45
8-46
8-47
8-48
8-49

Nominal Wage Growth for Private Sector Workers Over Past Year,
2007–17. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  423
Consumer Price Inflation, 2012–17 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  423
U.S. Prices versus Import Prices, 1955–2017. . . . . . . . . . . . . . . . . . . . . . . . . . .  424
Growth in Real and Nominal Average Hourly Wages of Private Sector
Workers, 2007–17. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  425
S&P 500 and VIX (Chicago Board Options Exchange) Indices, 2017–18. . . .  426
Nominal Long- and Short-Term Interest Rates, 2017–18. . . . . . . . . . . . . . . .  427
U.S. Treasuries Yield Curve, 2016–18. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  429
Brent Crude and West Texas Intermediate Prices, 2016–18. . . . . . . . . . . . . .  430
Crude Oil Production and Net Imports, 1990‒2017. . . . . . . . . . . . . . . . . . . . .  431
Foreign Real GDP and U.S. Real Export Growth, 2000–2017 . . . . . . . . . . . . .  432
European Union Consumer and Industrial Confidence Indicators,
2002–18. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  433
Average Annual Labor Productivity Growth in Selected Advanced
Economies, 1996–2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  434
Nominal Wage Inflation in Selected Economies, 2005–17. . . . . . . . . . . . . . .  436
Central Bank Inflation Measures for Selected Economies, 2001–17 . . . . . .  437
Emerging Market Credit to Nonfinancial Corporations, 2008–17. . . . . . . . .  438
Nonperforming Loans and Return on Assets for China’s Commercial
Banks, 2011–17. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  439
Projected Real GDP Growth Rate, 1975–2016. . . . . . . . . . . . . . . . . . . . . . . . . .  440
The Administration’s Forecasted Real GDP (Growth Rate), 2018–28. . . . . .  442
The Administration’s Forecasted Real GDP (Level), 2018–28. . . . . . . . . . . . .  443
Average 10-Year Forecast, by Administration, 1975–2018 . . . . . . . . . . . . . . .  446

Tables
1-1
1-2
1-3
4-1
4-2
4-3
5-1
6-1
6-2
7-1
7-2
8-1
8-2

Tax Brackets for Ordinary Income Under Previous Law and the Tax Cuts
and Jobs Act (2018 Tax Year). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Percentage of SALT and MID Deductions by Income Bracket. . . . . . . . . . . . . . 52
Summary Statistics for FDI Semielasticities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Annual Hours of Delay per Auto Commuter, 101 Urban Areas . . . . . . . . . . .  166
Average Public Nondefense Investment as a Percentage of Nominal
GDP, 1980–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  168
Sound Transit Fare Revenues and Boardings, 2016. . . . . . . . . . . . . . . . . . . . .  189
Tariff Policy for Selected Economies, 2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . .  240
Comparison of Estimated Costs of Opioid-involved Overdose Deaths,
CEA and Other Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  295
Summary of Annual Per-Patient Drug Expenditures, 2015. . . . . . . . . . . . . . .  307
Cumulative Abnormal Returns Following News of an Adverse Cyber
Event, 2000–2017. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  335
Number of Security Incidents and Breaches by Victim Industry and
Organization Size, 2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  345
Administration Economic Forecast, 2016–28 . . . . . . . . . . . . . . . . . . . . . . . . . .  445
Supply-Side Components of Actual and Potential Real Output Growth,
1953–2028 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  447

Economic Report of the President | 29

Boxes
1-1
2-1
3-1
3-2
3-3
3-4
3-5
4-1
4-2
4-3
4-4
4-5
4-6
5-1
5-2
6-1
6-2
6-3
6-4
6-5
6-6
6-7
6-8
6-9
6-10
7-1
7-2
7-3
7-4
7-5
7-6
7-7
7-8
7-9
7-10
8-1

30 |

Update: Post-TCJA Corporate Announcements. . . . . . . . . . . . . . . . . . . . . . . . . . 50
Determining the Future of the Fiduciary Rule. . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Who Is in the Middle Class? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  107
The Use of Survey Data versus Tax Record Data to Measure Income and
Its Distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Women’s Labor Force Participation after World War II. . . . . . . . . . . . . . . . . .  128
Military Spouses in the Labor Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  146
Paid Family Leave and the Challenges of Working Families . . . . . . . . . . . . .  148
Autonomous Vehicles: A 21st-Century Innovation. . . . . . . . . . . . . . . . . . . . . .  171
Labor Market Effects of Increased Infrastructure Investment. . . . . . . . . . . .  174
Oregon’s Tax on Vehicle Miles Traveled. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  184
Public-Private Partnerships. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  196
Bridging America’s Digital Divide. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  200
Transitioning to the 21st Century: The Case of 5-G Wireless. . . . . . . . . . . . .  200
Global Safeguards for Solar Products. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  244
Distressed Communities and the Tax Cuts and Jobs Act. . . . . . . . . . . . . . . .  264
The Trump Administration’s Actions to Restore Choice and Competition
to Healthcare. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  288
The Trump Administration’s Actions on Opioids . . . . . . . . . . . . . . . . . . . . . . .  296
The Administration’s Actions to Combat Smoking . . . . . . . . . . . . . . . . . . . . .  300
The Administration’s Efforts to Facilitate Innovation. . . . . . . . . . . . . . . . . . .  306
The Trump Administration’s Improvements to Medicaid. . . . . . . . . . . . . . . .  310
The Trump Administration’s Improvements to Medicare Part B . . . . . . . . .  311
The Trump Administration’s Improvements to Medicare Part D . . . . . . . . .  312
The Trump Administration’s Steps to End Global Free-Riding in
Biopharmaceuticals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  316
The Trump Administration’s Improvements to the 340B Program . . . . . . .  317
The Trump Administration Expedites Drug Approvals . . . . . . . . . . . . . . . . . .  319
Cyber Threat Actors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  329
PII Data Breach at Equifax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  341
Cyberattacks by a Nation-State: Sony Pictures Entertainment. . . . . . . . . . . 342
Cyber Theft of IP and Sensitive Corporate Information: SolarWorld. . . . . .  343
Supply Chain Attack: Home Depot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  349
Exploiting Cyber Vulnerabilities to Attack a Third Party: Mirai Botnet . . . .  350
Target’s Cyber Insurance Policy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  361
The Theft of U.S. Military Secrets through Cyber Means: The F-35. . . . . . . .  364
Law Enforcement’s Role in Mitigating Cyber Threats: The Kelihos Botnet.  377
Bitcoin. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  381
Challenges in the State and Local Sector. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  407

Economic Report of the President

x
Chapter 1

Taxes and Growth
In this chapter, we report evidence that the extensive use of itemized deductions in the U.S. income tax code can distort incentives, affect the distribution
of the tax burden, and reduce Federal income tax revenue. Lowering individual
income tax rates while simultaneously limiting use of distortionary deductions
can therefore facilitate tax relief to middle-income households—with corresponding supply-side benefits—while at the same time partially offsetting
short- to medium-term negative revenue effects.
In addition, because the magnitude of the corporate tax changes in the Tax
Cuts and Jobs Act—particularly the international aspect of corporate taxation—marked a more substantial break from its antecedent, we focus in more
depth on this part of the reform. In particular, we survey a large body of
academic literature on the effects of taxing corporate income and demonstrate
that the empirical evidence indicates that not only is capital highly responsive
to changes in corporate taxation but also has become more so over time. The
result is that not only have firms located less production and investment in
the United States, and correspondingly more abroad, but also that the cost
of this lower output has been increasingly and disproportionately borne by
the less mobile factor of production—namely, labor. Using estimates from this
literature, we then calculate that two salient corporate tax reforms—reducing
the top marginal Federal corporate tax rate from 35 to 21 percent, and allowing
firms to fully expense investments in nonstructure capital—would raise output
by 2 to 4 percent over the long run, and furthermore boost average annual
household wages by about $4,000.

31

The evidence presented in this chapter strongly suggests that the U.S. economy, and in particular U.S. workers, have been substantially harmed by the
convergence of two undisputed economic trends. The first is the high and
accelerating international mobility of capital, and the second is the increasingly
uncompetitive nature of U.S. corporate income taxation relative to the rest of
the world. The result has been throttled capital formation in the United States,
and consequently stagnant wage growth in the absence of capital deepening.
Under the Tax Cuts and Jobs Act, the shift away from worldwide taxation
toward a territorial system ends the penalty on companies headquartered in
the United States, because they will no longer pay additional taxes when they
bring overseas profits home. As a transition to the territorial system, income
that has already accrued offshore will be subject to a low, one-time tax, thereby
eliminating any tax incentive to keep funds offshore.

I

n December 2017, Congress passed and the President signed into law the
Tax Cuts and Jobs Act (TCJA)—the most significant combination of Federal
tax cuts and comprehensive tax reform the United States has experienced
in decades. The TCJA had four goals: tax relief for middle-income families,
simplification for individuals, economic growth through business tax relief,
and repatriation of overseas earnings. On the individual side, several reforms
were implemented in order to achieve these aims. The standard deduction
was approximately doubled, with the result of lowering taxes for millions of
families and simplifying tax filing because fewer households will itemize. In
addition, marginal tax rates were lowered (see table 1-1) and the Child Tax
Credit raised and expanded to apply to more families, among other changes.
Moreover, a number of popular deductions, such as the mortgage interest and
charitable contributions deductions were maintained, while certain deductions that primarily benefited higher income households were eliminated or
capped. Meanwhile, a 20 percent deduction was introduced for pass-through
business income, while on the corporate side, firms will now be able to fully
deduct investments in equipment and intangible assets, and will benefit from a
reduction in the top marginal Federal corporate tax rate from 35 to 21 percent.
The corporate tax cuts were implemented in tandem with a shift toward a territorial system of taxation.
In this chapter, we report evidence that the extensive use of itemized deductions in the U.S. income tax code can distort incentives, impact
32 |

Chapter 1

Table 1-1. Tax Brackets for Ordinary Income Under Previous Law and the
Tax Cuts and Jobs Act (2018 Tax Year)
Single filer
Previous law
10%
$0-$9,525
10%
15%
$9,52512%
$38,700
25%
$38,70022%
$93,700
28%
$93,70024%
$195,450
33%
$195,45032%
$424,950
35%
$424,95035%
$426,700
39.6% $426,700+ 37%

TCJA
$0-$9,525
$9,525$38,700
$38,700$82,500
$82,500$157,500
$157,500$200,000
$200,000$500,000
$500,000+

Married filing jointly
Previous law
TCJA
10%
$0-$19,050 10%
$0-$19,050
15%
$19,05012%
$19,050$77,400
$77,400
25%
$77,40022%
$77,400$156,150
$165,000
28%
$156,15024%
$165,000$237,950
$315,000
33%
$237,95032%
$315,000$424,950
$400,000
35%
$424,95035%
$400,000$480,050
$600,000
39.6% $480,050+
37%
$600,000+

Source: Tax Cuts and Jobs Act.

the distribution of the tax burden, and reduce Federal income tax revenue.
Lowering individual income tax rates while simultaneously limiting use of
distortionary deductions can therefore facilitate tax relief to middle-income
households—with corresponding supply-side benefits—while at the same time
partially offsetting short- to medium-term negative revenue effects.
In addition, because the magnitude of the changes on the corporate
side—particularly the international aspect of corporate taxation—marked a
more substantial break from its antecedent, we focus in more depth on this
part of the reform. In particular, we survey a large body of academic literature
on the effects of taxing corporate income and demonstrate that the empirical
evidence indicates that not only is capital highly responsive to changes in corporate taxation but also has become more so over time. The result is that not
only have firms located less production and investment in the United States,
and correspondingly more abroad, but also that the cost of this lower output
has been increasingly and disproportionately borne by the less mobile factor
of production—namely, labor. Using estimates from this literature, we then
calculate that two salient corporate tax reforms—reducing the top marginal
Federal corporate tax rate from 35 to 21 percent, and allowing firms to fully
expense investments in nonstructure capital—would raise output by 2 to 4
percent over the long run, and furthermore boost average annual household
wages by about $4,000.
The evidence presented in this chapter strongly suggests that the U.S.
economy, and in particular U.S. workers, have been substantially harmed by
the convergence of two undisputed economic trends. The first is the high and
accelerating international mobility of capital, and the second is the increasingly
uncompetitive nature of U.S. corporate income taxation relative to the rest of
the world. The result has been throttled capital formation in the United States,

Taxes and Growth

| 33

and consequently stagnant wage growth in the absence of capital deepening.
Under the TCJA, the shift away from worldwide taxation toward a territorial
system ends the penalty on companies headquartered in the United States
because they will no longer pay additional taxes when they bring overseas
profits home. As a transition to the territorial system, income that has already
accrued offshore will be subject to a low, one-time tax, thereby eliminating any
tax incentive to keep funds offshore.
We begin this chapter with a brief overview of U.S. taxation in historical and international perspective, focusing on the declining competitiveness
of U.S. taxation of corporate income relative to other advanced economies.
We turn next to the costs of taxation and a short review of the theory of tax
incidence with two examples that are relevant to the TCJA—the incidence
of corporate taxation and the mortgage interest tax deduction. Finally, we
provide an in-depth review of the growth literature in the context of corporate
taxation, focusing on the effects of the user cost of capital on optimal capital
accumulation and long-run growth.

U.S. Taxation in Historical and
International Perspective
Since before Independence, public taxation has been a contentious issue in
the politics of what is now the United States. In the wake of the Seven Years’
War, repeated attempts by successive British governments to raise revenue
from the North American colonies to cover the costs of colonial defense—the
Sugar Act, Stamp Act, and Townshend Acts—were met with increasing hostility
from colonial taxpayers, culminating with the Boston Tea Party after passage
of the so-called Tea Act in 1773. Within a decade of the Constitution’s commencement, efforts by the Federal government to assess taxes on whiskey
and property resulted in armed insurrections in the Whiskey Rebellion and
Fries’s Rebellion, the former of which required President Washington to lead a
13,000-strong army to confront the insurrectionists.
Perhaps not surprisingly, then, for the Republic’s first 150 years, taxation
at the Federal level remained a relatively small fraction of the nation’s total
economic output, as shown in figure 1-1. Before World War II, total Federal government revenue never exceeded 10 percent of gross domestic product (GDP)
and—with brief exceptions during wartime and the Great Depression—never
exceeded 5 percent of GDP. World War II thus marked a sharp discontinuity in
Federal taxation—by the end of the war, Federal tax revenues had grown from
6.7 to 19.9 percent of GDP, and thereafter never fell below 14 percent.
Moreover, the world wars also marked sharp discontinuities in the composition of Federal tax revenues. Before World War I, the Federal government
relied almost exclusively on excise taxes for revenue, and the single largest
source of revenue was tariffs on imported goods. While the Administration of

34 |

Chapter 1

Figure 1-1. Federal Tax Revenue as a Percentage of GDP, 1792–2017
Individual income
Excise

Corporate income
Miscellaneous other

Social insurance

Percent
25
20
15
10
5
0
1790

1830

1870

1910

1950

1990

Sources: Office of Management and Budget Historical Table 2.3 (2017); U.S. Census Historical
Statistics of the United States (1975).

President Lincoln implemented a tax of 3 percent on income over $800 in 1861
dollars (amended in 1862 to 3 percent on income between $600 and $10,000
and 5 percent on incomes exceeding $10,000), the tax was poorly enforced,
generated little revenue, and was ultimately repealed by the Revenue Act of
1870. A subsequent Federal income tax levied in 1894 was ruled unconstitutional by the U.S. Supreme Court a year later, in Pollock v. Farmers’ Loan & Trust
Co.
It was only after intense lobbying, particularly by the prohibitionist
movement, that the Sixteenth Amendment to the U.S. Constitution was ratified
in 1913, thereby granting the Federal government the authority to levy direct
taxes on income. The prohibitionists’ aim was to provide the Federal government an alternative source of funding, given that excise taxes on alcohol comprised 40 percent of revenue (Okrent 2010). Thereafter, the individual income
tax grew steadily as a share of Federal government revenue, reaching a peak of
50 percent in 2001.
Taxation of corporate income, meanwhile, began in 1909, when Congress
enacted an income-based tax on corporations. After ratification of the Sixteenth
Amendment, this constituted the corporate portion of the Federal income tax.
Though the statutory rate was initially a low 1 percent of corporate income,
it was repeatedly raised throughout the Great Depression and after, eventually reaching a peak of 52.8 percent in 1968, with intermediate rates as high
as 53 percent during World War II (IRS 2010). Thereafter, rates were gradually
reduced and thresholds raised, settling at a top statutory rate of 35 percent in

Taxes and Growth

| 35

1993. In the entire postwar period, the corporate tax contribution to Federal
revenue peaked in 1952, at 32.1 percent, before steadily declining to just over
9 percent in 2016 (OMB 2017).
While the U.S. held steady at a top Federal statutory tax rate of 35 percent
through 2017, the same was not true for other developed economies belonging
to the Organization for Economic Cooperation and Development (OECD) that
have experienced a steady downward trend in rates. Figure 1-2 shows the top
U.S. statutory corporate tax rate versus the OECD average (combined national
and subnational), excluding the U.S., since 2000. Though the latter declined
from 32.3 percent in 2000 to 23.8 in 2017, the combined U.S. rate only declined
from 39.3 to 38.9 percent, driven by reductions at the State level. During this
time, the U.S. went from being the developed economy with the seventhhighest corporate tax rate to that with the highest.
Relative to the rest of the world, the United States had the fourth-highest
combined statutory corporate income tax rate, after the United Arab Emirates,
Comoros, and the U.S. territory of Puerto Rico. It was one of only 13 national
jurisdictions levying corporate tax rates of 35 percent or more. Fourteen
jurisdictions levy no tax on corporate income at all. The global distribution of
corporate tax rates, moreover, has been consistently shifting toward a lower
average rate since 1980. Unsurprisingly, then, Spengel and others (2017) calculate that a reduction in the effective average U.S. corporate tax rate from 36.5
to 22.7 percent (Federal plus State, assuming a 20 percent corporate income
tax rate—the rate that was initially under consideration until it was amended
to 21 percent)—compared with a European Union average of 20.9 percent—
would substantially improve U.S. competitiveness, contributing to sharp shifts
in foreign direct investment toward capital projects in the U.S. The effective
average tax rate, a measure of the after-tax profit of an investment project
over its lifetime, is a crucial determinant of investment location. Reflecting
the declining competitiveness of U.S. statutory corporate tax rates, figure 1-3
reveals that before the TCJA’s enactment, the effective average tax rate of the
rest of the developed world was declining substantially relative to that of the
United States.

Costs of Taxation in the United States
If there is one principle on which economists agree—and, indeed, that in large
measure defines the profession—it is that people and firms respond to incentives. We may dispute the magnitudes of these responses, and sometimes even
their direction, particularly in the aggregate, but their existence is universally
acknowledged by economists of all stripes. The study of the economic effects
of taxation is, fundamentally, the study of responses to such incentives and
how these responses may offset—partly or, occasionally, totally—the benefits
of taxation.

36 |

Chapter 1

Figure 1-2. Top Statutory Corporate Tax Rates in OECD Countries,
2000–2017
Percent
60
50

Highest

40

United States
30

OECD ex-USA

20
10

Lowest
0
2000

2005

2010

2015

Source: OECD Statistics (2017).
Note: Series combine national and subnational rates.

Figure 1-3. Effective Average Corporate Tax Rates of OECD Countries,
1999–2017
Percent
40

35

United States

30

25

Other OECD

20

15
1999

2001

2003

2005

2007

2009

2011

2013

2015

2017

Source: Centre for Busines Taxation (2017).
Note: Series combine national and subnational rates. Because there are alternative ways to
compute the effective average tax rate, calculations may vary by source.

Taxes and Growth

| 37

The concept that taxation has benefits is not particularly controversial.
It allows for the provision of public goods and services—such as security,
justice, and official statistics—as well as of goods and services—such as health
and education—that may generate positive externalities and thus be prone
to underproduction by markets alone. At the same time, taxation may also
counteract the overproduction of goods and services that generates negative
externalities, such as pollution and poor health. It is especially important that
taxation further allows for the redistribution of resources in order to mitigate
instances of absolute inequality.
The benefits of taxation, however, are not without costs. These include
not only the direct costs of collecting taxes but also the opportunity costs
of tax compliance, along with the costs of reduced economic activity, owing
to the fact that taxation introduces a discrepancy—a “wedge”—between the
prices producers receive and the prices consumers pay for goods and services.
In markets where producers and/or consumers are highly responsive to small
changes in prices, the magnitude of this effect on economic activity can be substantial. Taxation may also generate additional adverse distortionary effects
by incentivizing the diversion of resources to less efficient economic activities.
The principle that efficient taxation minimizes the inducement to change
behavior is not new. In 1776, in the Wealth of Nations, Adam Smith articulated
four maxims regarding the administration of public taxation. In addition to the
maxims of equity, certainty, and convenience, his fourth maxim was that of
economy. “Every tax,” Smith writes, “ought to be so contrived as both to take
out and to keep out of the pockets of the people as little as possible over and
above what it brings into the public treasury of the state.” “A tax,” he elaborates, “may either take out or keep out of the pockets of the people a great deal
more than it brings into the public treasury.” That is, not only may the cost of
administering and levying the tax, along with the economic costs associated
with the “vexation” of tax compliance, consume a substantial share of the
revenue thus collected, but also the behavioral responses to tax incentives may
impose additional costs, as economic activity that would otherwise occur is
curtailed or moved underground or abroad for purposes of evasion.
To understand the scope of the issue, it is useful to think about the costs
of taxation in terms of direct and indirect costs. The direct costs constitute the
dollar amounts of taxes paid. The indirect costs, however, are twofold. First,
the excess burden of taxation refers to the reality that it typically costs the
economy more than $1 in the aggregate to raise $1 in tax revenue. Second,
compliance costs measure the amount of hours spent on filing tax returns and
provide an estimate of lost productivity. Inefficiently high costs of taxation can
have the adverse effects of translating into lower productivity and economic
growth. As noted above, before enactment of the TCJA, the U.S. tax code was
burdened with the highest corporate tax rate among advanced economies and

38 |

Chapter 1

an international tax system that encouraged capital outflows and the offshoring of profits.

The Excess Burden of Taxation
Tax rates affect incentives to work, save, and invest, and they can divert inputs
from more to less productive activities. These distortions in economic decisionmaking and the resulting inefficient allocation of resources lead to a reduction
in societal welfare beyond the amount of taxes collected. The more responsive
taxpayers are to higher taxes, the more they change their behavior, thereby
increasing the excess burden (or deadweight loss).
For instance, increases in marginal individual income tax rates induce
changes in labor supply through both income and substitution effects. The
income effect implies that hours worked would increase as tax rates increase as
workers seek to maintain the same level of income. Conversely, the substitution
effect implies that workers choose to work fewer hours as tax rates increase,
substituting other activities for labor as labor has become more expensive (and
leisure cheaper). Workers may also respond to changes in marginal income tax
rates along the extensive margin; some may simply choose to exit the labor
force altogether rather than adjust work hours. The labor economics literature
finds that the labor supply behavior of male workers is typically less responsive
to tax changes than that of females, especially if the former are married and
primary earners (Pencavel 1986; Keane 2011). Elasticities of labor supply along
the extensive margin are also typically larger than those at the intensive margin
(Heckman 1993; Blundell and McCurdy 1999). Although females are more likely
to respond at the extensive margin, they have also become less responsive to
tax changes in recent decades. This is due to such factors as greater labor force
participation rates and increased career orientation among married women
(Blau and Kahn 2007; McClelland, Mok, and Pierce 2014).
In addition to whether and how many hours to work, changes in marginal
tax rates can also affect the timing of retirement and the intensity and quality
of labor effort. Retirement could come earlier than otherwise planned, and
investments in human capital could decrease with higher tax rates. The decline
in the labor force participation of older workers could have adverse effects due
to an earlier loss of expertise and, along with diminished human capital investments, could contribute to lower aggregate productivity (Keane and Rogerson
2012, 2015). In addition to labor supply effects, individuals also increasingly
seek to avoid taxes as taxes increase. Saez, Slemrod, and Giertz (2012), for
example, find that tax avoidance increases with higher tax rates. Thus, by
lowering marginal tax rates, enactment of the TCJA effectively reduces the
incentive to avoid taxes by lowering the reward to doing so.
The U.S. tax code similarly affects corporate incentives. The deductibility
of interest payments on debt, for instance—reduced by the TCJA—incentivizes
debt over equity financing. Corporate income tax rates also affect choice of
Taxes and Growth

| 39

investment location. As noted above, the top marginal statutory corporate tax
rate in the U.S. before the TCJA was the highest among advanced economies,
and much higher than the OECD average. However, the U.S. collected less in
corporate taxes, relative to GDP, than the OECD average, due largely to high
capital mobility and corporate profit shifting, the latter in response to tax rate
differentials. In 2016, the average top statutory corporate tax rate (combined
subnational and national) in OECD countries excluding the U.S. was 24.2 percent, and corporate tax revenue totaled 3.0 percent of GDP. In comparison, the
combined (State and Federal) top statutory corporate tax rate in the U.S. was
38.9 percent, while corporate tax revenue was only 2.2 percent of GDP. Figure
1-4 plots top statutory corporate tax rates and corporate tax revenue (as a percentage of GDP) in all OECD countries since 2000. This figure shows that higher
rates do not necessarily lead to higher revenue collected as a fraction of GDP.
In addition, before the passage of the TCJA, and unlike any other developed country, the United States operated a system of worldwide taxation that
taxed U.S. corporations on their net income from any source, once repatriated.
This system encouraged deferral of overseas profit reporting by U.S. multinationals, as firms were incentivized to hold large volumes of cash at their foreign
subsidiaries in lower-tax jurisdictions. Other OECD countries instead follow
either a territorial tax system, whereby corporations are taxed only on income
generated domestically, or a hybrid tax system, whereby foreign income is
taxed only if the foreign country’s tax system is significantly different from that
at home. Due to the worldwide tax system, deferral, and high domestic corporate tax rates, U.S. companies were highly incentivized to shift their reported
income abroad, leading to lower domestic investment, to less physical and
intellectual capital formation within the United States, and to lost productivity
due to the offshoring of operations.

The Compliance Costs of Taxation
The compliance costs of the U.S. tax system are substantial. During the past
century, the length and complexity of the Internal Revenue Code have grown
considerably. While the tax code contained about 400,000 words in 1955, it
reached roughly 2.4 million words by 2016 (Tax Foundation 2016). In addition
to the tax code, an additional 7.7 million words of tax regulations are provided
by the Internal Revenue Service just to explain the tax statutes. Benzarti (2017)
estimates that the total cost of filing all schedules of the Federal income tax
increased from $150 billion in 1984 to $200 billion in 2006 (1.4 percent of 2006
GDP).
In 2016, the Office of Information and Regulatory Affairs estimated that it
took Americans almost 9 billion hours to file their tax returns (Tax Foundation
2016). The majority of these hours were spent complying with business returns
(2.8 million hours) and individual returns (2.6 million hours). The hours spent
on compliance could have instead been spent on other productive activities.
40 |

Chapter 1

Figure 1-4. Corporate Tax Rates and Corporate Tax Revenue,
2000–2016
Corporate tax revenue (% of GDP)
14
12
10
8
6
4
2
0
0

10

20
30
40
Top statutory corporate tax rate

50

60

Source: OECD Statistics (2017).

Rather than hiring tax professionals to file highly complex returns, businesses
could instead have invested in new plant and equipment. This opportunity cost
of compliance can reduce productivity in the long run. Based on the Bureau
of Labor Statistics’ hourly compensation estimates of $52.05 for professional
workers and $37.28 for all private sector workers in 2015, the hours spent on
complying with the tax code cost the U.S. economy an amount exceeding $400
billion, or 2.4 percent of 2016 GDP (Tax Foundation 2016).
Slemrod (2006) discusses the nature of compliance costs for corporations. Though compliance costs are higher for larger firms, the costs are regressive, in the sense that they constitute a relatively lower percentage of operating
costs for larger firms. In other words, the burden of complexity falls disproportionately on smaller firms. Costs include audits, planning, research, appeals,
litigation, and filing returns. Such costs also vary widely across industries, with
the highest costs incurred by firms in the communication, technology, and
media sectors, and the lowest costs by firms in the retail, food, and healthcare
industries. Multinational companies face higher compliance costs, especially
those with operations in multiple foreign countries. In addition, because larger
corporations often have the resources to navigate tax complexity to achieve
a lower effective rate, the recently enacted tax reform has the added benefit
of leveling the playing field between large multinational firms and smaller
domestic firms.

Taxes and Growth

| 41

Unfortunately, while tax complexity is costly, it is sometimes a necessary
consequence of the need to balance multiple objectives of tax policy, particularly efficiency and progressivity (economy and equity, according to Smith’s
maxims), that are not always in harmony. For example, the introduction of
targeted tax deductions and credits for reasons of economic efficiency may
also require the implementation of phase-outs at higher income levels so that
low- and middle-income households benefit relatively as well as absolutely.
While such provisions may add to complexity, they are nonetheless necessary
for the maintenance of a progressive tax code.

Tax Incidence
It is well known in economic theory that the individual (or corporation) who
makes a tax payment to the government is not necessarily the one who bears
the burden of this payment; this burden is the incidence of the tax. Legally, the
incidence of a tax is determined by who actually pays the tax to the government. For example, the statutory incidence of sales taxes is typically on the
seller of the goods sold. This incidence, however, does not take into account
the fact that markets adjust in response to the imposition of the tax—and
it is this market reaction that determines who actually bears the burden or
economic incidence of a particular tax. We illustrate this concept using a simple
graph of demand and supply (figure 1-5). Absent any intervention in the market, q* units would be sold at a price of p* per unit. We illustrate the imposition
of a per-unit tax by shifting the supply curve in by $t units.
An alternative to a per-unit tax is an ad valorem tax, such as a sales tax.
Rather than shift the supply curve in by $t units, the supply curve would instead
rotate by t percent. In addition, for simplicity, we focus on shifting the supply
curve. The demand curve could, alternatively, shift in by the same number of
units. Economic theory shows that it does not matter who bears the statutory
incidence of the tax, as the economic incidence is the same in either case.
As figure 1-6 shows, even though the supplier is responsible for remitting
the money to the government, the increase in price to the consumer means
that the consumer is responsible for at least part of the tax. Crucially, the
amount born by the consumer depends upon the relative elasticities of supply
and demand. After the tax is implemented, consumers pay pb = p + t (> p*) per
unit, producers keep ps = p (< p*), and the government collects $t per unit sold.
Consider figures 1-6 and 1-7, which are variants of figure 1-5.
In figure 1-6, suppliers exhibit perfectly inelastic supply. That is, for a
given change in price, suppliers continue to supply the exact same quantity
of the good. In this case, the entire tax will be pushed onto them in the form
of lower prices (ps = p – t < p*). This result is very intuitive; if suppliers are
completely unresponsive to price then it makes sense to push the tax onto
them. The after-tax price for consumers will stay the same as the pretax price.

42 |

Chapter 1

Figure 1-5. The Incidence of a Tax on Sellers
Price

S2
S1
pb = p + t
Tax

p*
ps = p

D1

qtax

q*

Quantity

Figure 1-6. Tax Incidence with Perfectly Inelastic Supply
Price

S1
p* = pb = p
Seller's
share of tax
burden

Tax

ps = p - t
D1

q* = qtax

Quantity

Taxes and Growth

| 43

A typical example of perfectly inelastic supply is land. The United States, for
example, has a fixed amount of land. Likewise, geography can sometimes
limit usable land, rendering its supply more inelastic. Local ordinances and
regulations can also impact the supply of land available for building, making
the supply of housing inelastic—an example we will turn to below. Under this
scenario, suppliers will bear the burden of the tax. The after-tax price will not
change from the pretax price for consumers. Because suppliers must remit the
tax, they collect $p from the consumer, send $t to the government and are left
with $(p – t). In figure 1-7, however, sellers exhibit perfectly elastic supply. That
is, if the price falls below p*, the supply falls to zero. Thus, in this case, the entire
tax will be paid by consumers (the price they pay increases to p* + t), and sellers
continue to receive p*.
One final detail in these figures is important in our subsequent discussion
on tax incidence in the TCJA. Until now, we have considered the impact of a tax
on prices. Now we consider the impact of a tax on the quantity of the good sold.
Returning to figure 1-5, before the imposition of the tax, q* units of the good
were sold at a price of p*. After the tax, only qtax units are sold. The triangular
area between the demand and supply curves between qtax and q* is known as
deadweight loss or the excess burden, as discussed above. Because the competitive equilibrium at p* and q* maximizes social efficiency, any movement
away from that caused by per unit or ad valorem taxation is an efficiency loss to
society. This is the price that we pay for a redistributive tax system. Deadweight
loss is higher when demand and/or supply are very elastic—that is, the imposition of a tax generates large behavioral responses. On the opposite side, when
either demand or supply are perfectly inelastic—that is, when the imposition
of the tax does not generate any movement away from the efficient quantity,
deadweight loss is equal to zero. It is true that suppliers may bear the entire
burden of the tax (in the case of perfectly inelastic supply) but, given that they
still provide q*, there is no loss to society overall—simply a transfer of income
from suppliers to the government.
The incidence of a tax has important implications for the distributional
aspects of tax reform. Much of the argument on whether particular tax reforms
help or hurt certain groups depends upon how elastic we believe demand or
supply to be. Here, we highlight two important elements of the TCJA whereby
tax incidence has played a key role in our understanding of how taxes ultimately affect households. We first consider the corporate tax. Though corporations pay the tax, they do not ultimately bear its burden. A main tenet of
public economics is that people—whether they are shareholders, the owners
of capital, or workers—bear the burden of a corporate tax. Therefore, which
people in particular pay the tax is an empirical question. A second example is
the incidence of the mortgage interest deduction (MID), which is a subsidy for
home ownership given to households that itemize their taxes. Though the MID
is not eliminated in the new tax law, fewer households will claim the deduction
44 |

Chapter 1

Figure 1-7. Tax Incidence with Perfectly Elastic Supply
Price

pb = p + t
Buyers'
share of tax
burden
p* = ps = p

Tax
S1

D1

qtax

q*

Quantity

because they will be better off claiming the higher standard deduction. This
has the potential to have an impact on home prices, though evidence indicates
that the impact will be quite modest.

The Incidence of the Corporate Income Tax
In analyzing where the burden of the corporate tax falls on capital relative to
labor, it is important to distinguish between short-run and long-run burdens.
In the short run, increases in the corporate tax are borne by current owners of
corporate capital, through a drop in asset values, and by investors, through
lower after-tax rates of return. In the long run, labor bears more of the burden
of the corporate tax. This is because an increase in the effective tax rate on
capital income from new saving and investment leads to a reduction in capital
accumulation. The resulting decline in the capital-to-labor ratio decreases
labor productivity and leads to a fall in wages.
Exactly how much of the burden labor bears is a subject of much academic research. Piketty and Saez (2007) assume that the burden of the corporate income tax falls on owners of capital income. Several think tanks and
public research services—including the Tax Policy Center and Congressional
Budget Office—assume in their current tax models that most of the corporate
tax burden (about 80 percent) is borne by capital, and the rest by labor. The
empirical literature places the corporate tax burden borne by workers at
between 21 and 75 percent, with higher figures generally representing more
recent studies that assume freer movement of capital across borders. These

Taxes and Growth

| 45

incidence estimates can be interpreted as the share of the total burden of
taxation borne by workers, where the total burden is the surplus eliminated
from the private market by corporate taxation. The burden includes not only
the government revenue raised but also the deadweight loss from imposition
of the tax. With no deadweight loss, applying these incidence estimates to the
static change in government revenue from corporate rate reductions would
provide the lower bound on the additional surplus (or dollars) accruing to
workers under the rate changes. The existence of a positive deadweight loss
implies that the total burden would be more than 100 percent.
In a paper for the Federal Reserve Bank of Kansas City, Felix (2009) estimates an elasticity of workers’ wages with respect to corporate income tax
rates based on the variation in the marginal tax rate across the 50 U.S. States.
In these estimates, a 1-percentage-point increase in the top marginal State corporate rate reduces gross wages by 0.14 to 0.36 percent over the entire period
(1977–2005), but the dampening effects of corporate tax rates on wages are
growing over time. For the most recent period in Felix’s data (2002–5), a 1-percentage-point State corporate tax increase reduces wages by 0.45 percent.
These estimates imply an elasticity of roughly –0.1 to –0.2 for the U.S. statutory
corporate tax rate. Carroll (2009) corroborates Felix’s results. Again, using data
on changes in the corporate tax rate across States, Carroll (2009) estimates
coefficients that are consistent with an elasticity of –0.1 to –0.2 for workers’
wages with respect to the U.S. statutory corporate tax rate. Fuest, Peichl, and
Siegloch (2018), studying nearly 6,800 tax changes in German municipalities
between 1993 and 2012, identify the wage effects of municipal corporate rate
changes. Their point estimates imply a wage elasticity of –0.14 with respect to
the local business tax. An additional contribution of the study by Fuest, Peichl,
and Siegloch (2018) is their analysis of the distributional consequences of the
corporate taxation burden, which shows that low- and medium-skilled workers are differentially disadvantaged by higher tax rates. They find that these
consequences, which are large enough to significantly affect tax progressivity,
would decrease the U.S. tax system’s overall progressivity by an estimated 25
to 40 percent.
Other country-based studies, like those assessing the effects of corporate
income tax rate changes in Canada, may be more applicable to the United
States. Still, labor union membership is higher in Canada, suggesting that there
may be some limits to applying these estimates to the U.S. Using corporate
rate changes across and within Canadian provinces between 1998 and 2013,
Ebrahimi and Vaillancourt (2016) estimate the effects on workers’ wages,
analogous to the analyses by Felix (2009) and Carroll (2009) for U.S. States.
The study finds that a 1 percent increase in the statutory corporate tax rate is
associated with a reduction in workers’ hourly wages of 0.15 to 0.24 percent.
These results—which control for observable worker characteristics, including
labor union membership—hold for both public and private workers. In new
46 |

Chapter 1

research, McKenzie and Ferede (2017) also use changes in corporate tax rates
within Canada to develop an estimate of the impact on workers’ wages. The
baseline elasticity estimate is –0.11, with alternative estimates giving absolute
values as large as –0.15.
Looking at other countries with similar market-based economies,
Dwenger, Rattenhuber, and Steiner (2013) estimate the effects on workers’ wages of corporate tax rate changes in Germany. Their results imply a
semielasticity of wages of about –1.24 with respect to the average tax rate,
without accounting for employment effects, and of –2.36 when employment
effects are included.
A cross-country study by Hassett and Mathur (2015), based on 65 countries and 25 years of data, finds that the elasticity of workers’ wages in manufacturing after 5 years with respect to the highest marginal tax rate in a country
is –0.5 in the baseline case, which includes the addition of spatial tax variables.
An expanded analysis by Felix (2007) follows the strategy used by Hassett and
Mathur (2006), but incorporates additional control variables, including workers’ education levels and countries’ degree of economic openness. Felix’s estimates imply a semielasticity of between –0.7 and –1.23. When she replicates
Hassett and Mathur’s specification, the semielasticity is –0.43.
A set of recent papers also seeks to measure the rent-sharing, or “bargaining,” channel directly, including papers by Liu and Altshuler (2013) and
by Arulampalam, Devereux, and Maffini (2012). Liu and Altshuler measure an
elasticity of between –0.03 and –0.04 for U.S. workers’ wages with respect to
effective marginal tax rates, which represent these workers’ profit sharing with
respect to their employers’ tax liabilities. Research by Desai, Foley, and Hines
(2004) also relies on wage data for U.S. multinationals to assess the relative
share of the corporate tax burden borne by labor, measuring the labor share
at between 45 and 75 percent, near the higher end of theoretical predictions.
However, because these papers do not assess the economy-wide effects of
corporate tax reform, they are excluded from figure 1-9, which summarized the
studies discussed here.
Results from Azémar and Hubbard (2015) also utilize cross-country
changes in the corporate tax rates of OECD countries (generally high-income,
developed countries, like the U.S.) to measure the effects of corporate tax rate
changes on workers’ wages. The paper measures changes in workers’ wages
with and without controls for changes in value added (labor productivity). The
results imply a semielasticity of –0.43 (–0.17 for the U.S.) for workers’ wages
with respect to the corporate tax rate, of which about three-fourths is related to
the indirect channel and one-fourth to the direct channel. Azémar and Hubbard
note that the estimates without value added (those corresponding to the combination of both direct and indirect channels) may be overestimates, given the
correlation between value added and corporate tax rates. We include them in
figure 1-8, but note this caution on interpretation.
Taxes and Growth

| 47

Applying the results in each paper to the rate reduction in the TCJA,
figure 1-8 summarizes the estimated changes in U.S. household wages implied
by each paper. For results where semielasticities are reported, we multiply
the semielasticity by the change in the statutory rate, a 14-percentage-point
change (from 39.6 to 25.6 percent). This is the percentage change in wages
implied by the point estimates. These changes are then applied to the mean
value of household income reported by the U.S. Census in 2016, $83,143,
and multiplied by the share of average household income that is wage-andsalary income (78 percent) (BLS 2016, 2017). For results where elasticities are
reported, we calculate the percent change in the tax rate in the U.S. as 0.14
divided by 0.396 (the U.S. statutory rate, including State taxes), or 35.4 percent.
What do these empirical results imply for the likely effects of corporate
tax reform in the United States? Within each of the four estimation strategies shown in figure 1-8, there is a range of estimated worker wage effects.
Overall, the estimated impact of the 14-point reduction in the U.S. corporate
tax rate varies from $2,400 (based on the cross-Canadian province results from
McKenzie and Ferede 2017) to just over $12,000 based on the longer-run effects
of corporate tax rate changes observed in the Hassett and Mathur data. The
average result is $5,500. Removing the two lowest and two highest estimates
gives a range of $3,400 to $9,900.
Cross-country estimates made by Felix (2007) and by Hassett and Mathur
(2006, 2015) imply far larger effects on wages from corporate reforms, ranging
from $6,000 to $12,000 for Hassett and Mathur (2015), who take into account
the spatial correlation of corporate rate changes. Azémar and Hubbard’s (2015)
estimates are closer to $4,000, although they caution that this number may
be too high because the data suffer from omitted variable bias. For Hassett
and Mathur, the larger estimates may partially reflect the intentional measure
of longer-term wage outcomes; both Azémar and Hubbard (2015) and Felix
(2007) measure cross-country differences in the cross section. This larger range
of estimates is also consistent with estimates made by Dwenger, Kübler, and
Weizsäcker (2013) of the wage effects of German corporate tax reforms, but the
differing nature of wage bargaining across countries is an important limit on
transferring these results.
In all cases, the corporate tax rate changes used for identification are
smaller than the 14-percentage-point reduction proposed under the TCJA. If
the effects of corporate taxes on wages are not linear, then the outcome for
U.S. workers may be different from the estimates given in figure 1-8. One final
consideration relates to changes in employment, which these wage estimates
do not incorporate. If the effect of corporate tax reform is to raise U.S. workers’ wages primarily through wage bargaining rather than through enhanced
productivity, employers could reduce their demand for labor as a result. This
seems far less likely to be the case in the U.S. than in countries with centralized

48 |

Chapter 1

Figure 1-8. Average Income Increases for U.S. Households from a
Corporate Rate Reduction, Using Literature Wage Elasticties
Azémar and Hubbard (2015)*
Global
estimates

Felix (2007)
Felix (2007)
Hassett and Mathur (2015)*

U.S. Statebased
estimates

Felix (2009)°
Carroll (2009)
Carroll (2009)

Canadian Ebrahimi and Vaillancourt (2016)*
province Ebrahimi and Vaillancourt (2016)*
estimates
McKenzie and Ferede (2017)
German
estimates

Dwenger et al. (2013)
Dwenger et al. (2013)
Fuest et al. (2018)*
$0

$5,000

$10,000 $15,000

Source: See reference list and discussion in text.
Note: *indicates paper was published in a peer reviewed journal. ° indicates paper was
published in the Economic Review of the Federal Reserve Board.

wage setting and stronger labor bargaining power. For a discussion of the
German case, see Dwenger, Kübler, and Weizsäcker (2013).
As a whole, these estimates suggest that a U.S. Federal corporate rate
reduction from 35 to 21 percent is likely to result in wage increases for U.S.
households of $4,000 or more. Though this is a long-run outcome, box 1-1
discusses how corporate investment and bonus announcements immediately
following passage of the TCJA are consistent with estimated effects.

The Incidence of the Mortgage Interest and State and Local
Tax Deductions
The U.S. Treasury estimates that in fiscal year 2017, the American taxpayers
who used the mortgage interest deduction reduced their Federal tax liabilities
by $65.6 billion (U.S. Department of the Treasury 2017). The TCJA does not
eliminate the MID. However, the law nearly doubles the standard deduction,
making it likely that more households will achieve greater tax savings by
claiming the standard deduction instead of itemizing. In addition, for new
mortgages, the law reduces the amount of mortgage indebtedness for which
interest can be deducted from $1 million to $750,000, so that less interest can
be deducted on more expensive homes.
To determine the impact of these changes, it is useful to consider the
theoretical exercise of eliminating the MID, which would provide an upper

Taxes and Growth

| 49

Box 1-1. Update: Post-TCJA Corporate Announcements
In the weeks immediately following passage of the TCJA, more than 300
companies announced wage and salary increases, as well as bonuses and/
or 401(k) match increases affecting 4.2 million workers, citing the TCJA. As
of February 5, 2018, the Council of Economic Advisers tallied a total of over
$2.4 billion in new bonuses that have been publicly announced and explicitly
attributed to the TCJA, as well as 46 employers announcing starting wage
increases. In addition, by that date, the CEA counted $190 billion in newly
announced corporate investment publicly attributed to the TCJA.
Two of the most prominent examples are Walmart Inc. and Apple Inc.
Walmart, the largest private employer in the United States, announced on
January 11, 2018, that it was raising its starting wage by 10 percent, from $10
to $11, expanding maternity and parental leave benefits, introducing new
financial assistance for employees looking to adopt, and providing a onetime cash bonus for eligible employees of up to $1,000. Meanwhile Apple, as
of December 2017 the largest publicly listed company in the world by stock
market capitalization, on January 17, 2018, announced employee bonuses of
$2,500 worth of restricted stock units in response to the TCJA. The company
also announced that it would be incurring a $38 billion tax bill in order to
repatriate offshore cash in order to invest $30 billion in the U.S. In addition,
JPMorgan Chase, the largest bank in the U.S. by assets, announced a $20 billion investment program that will open 400 new branches and add 4,000 jobs.
The bank also announced that it would be raising hourly wages for 22,000
full- and part-time U.S. employees.
Though subject to change and evolving circumstances, the immediate
corporate response to the TCJA therefore offers provisional confirmation of
the theoretical and empirical evidence on profit sharing, the link between
corporate taxation and labor earnings, and the effect of corporate taxation on
investment in the context of internationally mobile capital. Moreover, though
the primary channel through which we expect corporate tax reductions to
affect wages is that of long-run capital deepening raising labor productivity,
it is also rational in a tightening labor market for forward-looking employers
to raise wages and offer bonuses in the short run in order to retain similarly
forward-looking workers.

bound on any potential effects from the TCJA. When considering the tax incidence of the MID, it may be that buyers gain and current homeowners lose
(or are at least made no worse off). The opposite may also be true. The statutory incidence falls upon the potential homebuyer because he or she can no
longer use mortgage interest to reduce his or her Federal income tax burden.
In other words, the subsidy for housing is removed, and a subsidy is just a
negative tax—the economic theory discussed above applies in the same way,
except that the government, rather than collecting taxes, provides a financial

50 |

Chapter 1

payment. Thus, both home buyers and sellers benefit to some extent from the
subsidy—the subsidy allows sellers to receive a higher price and allows buyers
to pay a lower price (assuming standard demand and supply curves, but, unlike
figure 1-6, the demand curve shifts out with a subsidy for buyers.) Removing the
subsidy (and moving back toward the competitive equilibrium) has the effect
of potentially modestly lowering the price that sellers receive and raising the
price that buyers pay. As always, the precise economic incidence is determined
by the relative price-elasticities of demand for and supply of the housing stock.
There is a body of academic literature that has studied the effect of
the MID on housing prices. Early studies found substantial effects on housing prices. For example, Poterba (1984) estimates that there would be a very
large housing price response to the MID’s elimination—a 26 percent decline.
However, this result was estimated more than 30 years ago, in an environment
of 10 percent inflation and may not be relevant in today’s economic setting.
Capozza, Green, and Hendershott (1996) estimate that eliminating the MID
(along with ending the deduction for property taxes) would decrease house
prices by an estimated 13 percent. Harris (2013) estimates that eliminating the
MID would reduce home prices by 12 to 19 percent, depending upon the model.
Other contributions to the academic literature that consider the MID
within the context of the larger economy find significantly lower price effects
from its elimination. These studies examine a more flexible model of the housing market that allows demand and supply to respond to reductions in the
demand for housing, or incorporate spillover effects in the rental housing market. Most recently, Sommer and Sullivan (2017) find that eliminating the MID
would reduce home prices by 4.2 percent in the long run, although the effect
is only half this size in an environment with the low interest rates observed
today. In a similar model, Floetotto, Kirker, and Stroebel (2016) estimate that
eliminating the MID would decrease home prices by only 1 percent in the long
run. Given that the value of housing is equal to roughly 30 percent of total
household wealth, a 4 percent fall in housing prices translates into about a 1.2
percent decline in total household net wealth (Federal Reserve Board 2017).
Recent research (e.g., Hilber and Turner 2014) also indicates that the
impact of eliminating the MID would vary depending on the elasticity of supply
of housing in different areas. In markets where supply is constrained, eliminating the MID is more likely to reduce prices because supply does not adjust
downward in the long run and with little to no impact on homeownership
rates. Rappoport (2016) uses a structural model, which allows housing supply
elasticities to vary across areas, and he finds that eliminating the MID would
decrease house prices by 6.9 percent on average, but with considerable variation across markets, again, depending on the elasticity of supply.
In sum, the most recent academic literature suggests that the impact of
eliminating the MID on house prices is likely to be more modest than those in
the earlier literature, and its magnitude in different areas will depend on the
Taxes and Growth

| 51

Table 1-2. Percentage of SALT and MID Deductions by Income Bracket
Share of filers

Share of SALT
deductions

Share of MID
deductions

Under $1

1.43

0.00

0.00

$1 to $9,999

14.46

0.61

1.19

$10,000 to $24,999

21.99

2.89

3.67

$25,000 to $49,000

23.43

12.52

12.43
16.83

Income bracket

$50,000 to $74,999

13.32

16.90

$75,000 to $99,000

8.63

16.45

16.38

$100,000 to $199,000

12.25

34.73

34.76

$200,000 to $499,999

3.62

12.76

12.15

$500,000 to $999,999

0.58

2.07

1.80

$1,000,000+

0.29

1.07

0.78

Sources: Internal Revenue Service; CEA calculations.

extent to which housing supply can respond to reduced demand. Cities like
San Francisco, where the housing stock is relatively inelastic, may experience
greater price responses compared to cities with relatively few land-use restrictions, like Dallas.
As noted above, the TCJA does not eliminate the MID. As a result, any
potential impact on housing prices is expected to be more muted than that
100of households itemsuggested by this review of the literature. The proportion
izing is estimated to decrease from 29.2 percent to 13.4 percent—that is, 23.1
million more filers taking the standard deduction as a result of the TCJA. Many
of these households would have claimed the MID in the past, but after the
enactment of the TCJA no longer find it beneficial to do so. In the same vein,
other households with particularly large mortgage interest obligations may
still find it in their interest to claim the MID. As such, the studies that consider
the full elimination of the MID can be considered as upper bounds on the anticipated impact of the TCJA and, in all likelihood, ultimately will be far smaller.
Nonetheless, the TCJA is likely to offset any potential harm by nearly doubling
the standard deduction, lowering statutory tax rates and substantially increasing the Child Tax Credit.
Similarly, the extant academic literature suggests that capping the
Federal income tax deductibility of State and local taxes at $10,000 would have
only modest, though potentially progressive, economic impacts. Feldstein
and Metcalf (1987) find strong evidence that deductibility affects how State
and local governments finance spending, and limited evidence that it affects
overall levels of State and local government spending. Specifically, they find
that States where Federal deductibility implies a relatively low cost of financing via deductible personal taxes (e.g., income, sales, and property taxes), rely
more heavily on those taxes versus business taxes and other revenue sources.

52 |

Chapter 1

Figure 1-9. Distribution of Mortgage Interest Deduction, by Adjusted
Gross Income
$1 to $9,999

$10,000 to $24,999

$25,000 to $49,000

$50,000 to $74,999

$75,000 to $99,000

$100,000 to $199,000

$200,000 to $499,999

$500,000 to $999,999

$1,000,000+

Average MID deduction (dollars)
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
0

10

20

30

40
50
60
Percentage of filers

70

80

90

100

Sources: Internal Revenue Service; CEA calculations.
Note: Excludes those earning less than $1 in adjusted gross income.

Inman (1989) and Metcalf (2011) observe similar outcomes. Inman finds
that though eliminating State and local income tax (SALT) deductibility would
increase the progressivity of the Federal income tax code, it could also result in
higher local property taxation and lower fees in larger U.S. cities. Elimination
would only reduce total local government spending in large cities if property
taxes are constrained, in which case local government revenues and spending
would be expected to decline by roughly 3 to 7 percent. Inman further finds
that the Federal government would unambiguously collect more tax revenue
from taxpayers in large cities. Metcalf, meanwhile, shows that deductibility
leads to greater reliance on income and property taxes at the State and local
levels, while having no observable impact on nondeductible taxes and fees.
Though the deduction is significantly regressive at the Federal level, he finds
some evidence that the SALT deduction may support progressive taxes at the
subnational level.
The aim of the individual elements of the TCJA is, therefore, to simplify
the income tax code and attenuate the distortionary effects of itemized deductions such as the SALT and MID by lowering marginal income tax rates while
simultaneously limiting the applicability of such deductions. Table 1-2 shows
that households that have adjusted gross income of at least $100,000 constitute less than 17 percent of the population but claim about half of SALT and
MID deductions. Thus, it is possible to partially offset the Federal revenue loss
of marginal rate reductions, whilst still delivering tax relief targeting middleincome households. Moreover, figures 1-9 and 1-10 show how MID and SALT

Taxes and Growth

| 53

Figure 1-10. Distribution of State and Local Tax Deduction, by Adjusted
Gross Income
$1 to $9,999

$10,000 to $24,999

$25,000 to $49,000

$50,000 to $74,999

$75,000 to $99,000

$100,000 to $199,000

$200,000 to $499,999

$500,000 to $999,999

$1,000,000+

Average SALT deduction (dollars)
300,000
250,000
200,000
150,000
100,000
50,000
0
0

10

20

30

40

50

60

70

80

90

100

Percentage of filers
Sources: Internal Revenue Service; CEA calculations.
Note: Excludes those earning less than $1 in adjusted gross income.

deductions are concentrated at the upper end of the income distribution—
higher income households disproportionately take advantage of such deductions. Accordingly, the nonpartisan Joint Committee on Taxation (2017a)—the
official scorer of legislative bills—finds that by 2025, before the scheduled
expiration of the individual elements of the TCJA, households earning between
$40,000 and $100,000 will enjoy a net reduction in their total Federal tax liability of between 2 and 5 percent. Meanwhile, the share of Federal taxes paid by
households earning more than $1,000,000 is estimated to increase slightly,
from 19.1 to 19.5 percent. The persistence of this middle-class income tax relief
will then depend on whether Congressional representatives elect to extend the
reforms enacted under the TCJA or allow them to expire.

Evidence on Taxes and Growth
A fundamental challenge to estimating the effects of changes in income tax
rates on economic growth is that the timing of the tax changes are not random. Historically, legislators have tended to lower tax rates during periods of
economic contraction and raise taxes during periods of expansion. This high
correlation of tax changes with economic conditions can negatively bias estimates of the effect of tax rate reductions on investment and output. Estimated
effects of tax changes may also be biased by the correlation of those changes
with unobserved factors.

54 |

Chapter 1

Recent empirical studies have employed two techniques to address
these challenges. One is the approach called structural vector autoregression,
following Blanchard and Perotti (2002), in which the identification of causal
effects relies on institutional information about tax and transfer systems and
the timing of tax collections to construct automatic fiscal policy responses to
economic activity. In their original study, Blanchard and Perotti (2002) find an
initial tax multiplier of 0.7 on impact, with a peak impact of 1.33 after seven
quarters. In contrast, using sign restrictions to identify tax shocks, Mountford
and Uhlig (2009) find a peak-to-impact multiplier that is substantially larger.
A second technique, originating with Romer and Romer (2010), uses
narrative history from Presidential speeches and Congressional reports to
identify exogenous tax changes with political or philosophical, as opposed to
economic, motivations. These changes are unlikely to be correlated with other
factors affecting output. Tax changes unrelated to the business cycle can be
used as a quasi-natural experiment to estimate the effect on economic output;
this matters because if tax cuts are a response to deteriorating economic conditions, the data will show a spurious negative correlation between taxes and
growth. Romer and Romer estimate that a 1-percentage-point increase in the
total tax share of GDP decreases GDP by 1 percent in the first year and up to 3
percent by the third year. They further find that a 1-percentage-point increase
in the total tax share of GDP decreases investment by 1.5 percent in the first
year and up to 11.2 percent by the third year.
Using Romer and Romer’s (2010) series as an external instrument for
changes in average individual marginal tax rates, Barro and Redlick (2011) similarly find that a permanent 1-percentage-point reduction in the average marginal tax rate raises real GDP per capita by 0.5 percent in the subsequent year,
corresponding to a conventional tax multiplier of 1.1. Applying the narrative
approach to U.K. data, Cloyne (2013) finds that a 1-percentage-point reduction
in the total tax share of GDP increases GDP by 0.6 percent on impact and by 2.5
percent over three years, and raises investment by 1.2 percent immediately
and by 4.6 percent by the third year. Hayo and Uhl (2014), using German output
data, estimate a maximum response to a 1-percentage-point drop in total tax
liability (as a percentage of GDP) of 2.4 percent. Applying a similar approach to
fiscal consolidations (tax revenue increases) across the OECD countries, Leigh,
Pescatori, and Guajardo (2011) find that a tax-based fiscal consolidation of 1
percentage point of GDP reduces GDP by 1.29 percent.
Mertens and Ravn (2013) develop a hybrid approach that combines both
methods. Because narratively identified shocks may be prone to measurement
error, and identification in a structural vector autoregression framework can
require questionable parameter restrictions, Mertens and Ravn develop an
estimation strategy that utilizes Romer and Romer’s (2010) narrative tax shock
series as an external instrument to identify structural tax shocks, avoiding
the need to impose parameter restrictions. Utilizing this hybrid approach to
Taxes and Growth

| 55

analyze U.S. data, they estimate that a 1-percentage-point cut in the average
corporate income tax rate raises real GDP per capita by 0.4 percent in the first
quarter and by 0.6 percent after a full year, with the effect persisting through
20 quarters. Mertens and Ravn additionally estimate that a 1-percentage-point
cut in the average corporate income tax rate generates an increase in nonresidential investment of 0.5 percent on impact, with a peak increase of 2.3 percent
after six quarters. Also employing a hybrid approach, Mertens and Montiel-Olea
(2017) find that in the first two years following a tax decrease of 1 percentage
point, real GDP is expected to be higher by about 1 percentage point.
On the individual side, meanwhile, Mertens and Ravn estimate that a
1-percentage-point cut in the average personal income tax rate raises real GDP
per capita by 1.4 percent on impact and by a peak of 1.8 percent after three
quarters. Though they find that a 1-percentage-point reduction in the average personal income tax rate has a negligible impact on inflation, short-term
nominal interest rates, and government debt, they do find significant positive
effects on employment, hours worked, consumption, and durable goods purchases and nonresidential fixed investment. In particular, they observe that a
1-percentage-point decrease in the average personal income tax rate results in
a peak employment response of 0.8 percent after 5 quarters, and peak durable
goods and nonresidential investment effects of 5 and 4 percent, respectively,
beyond one year.
Though the estimated coefficients found in these studies are not directly
comparable, the signs, sizes, and statistical significance of the estimates—
combined with their replication across time and geography—provide strong
evidence of a positive effect of tax cuts on economic growth. Although some
of this literature relies on changes in overall tax liabilities, the most recent
research allows us to specifically simulate the impact of corporate tax changes.
Moreover, dynamic stochastic general equilibrium models of newer vintage—
for example, research by Lizarazo Ruiz, Peralta-Alva, and Puy (2017)—are now
generating growth effects from changes in income tax rates that are in the
range of the findings of Mertens and Ravn (2013) and Barro and Redlick (2011),
which suggests an increasing convergence of estimates derived under alternative modeling frameworks. This development is important, because some critics of the macroeconometric literature have asserted that the results are too
large to be theoretically plausible.
Moreover, recent academic research suggests that labor supply effects
among older workers may be contributing to observed growth effects of reductions in marginal individual income tax rates. Keane and Rogerson (2012, 2015)
observe that the effect of work experience and on-the-job training on the net
present value of lifetime earnings will vary with worker age. Because the net
present value of additional human capital acquisition on the job is quite large
for younger and relatively less experienced workers, labor supply responses
to marginal income tax rates are low among these workers. In contrast, labor
56 |

Chapter 1

supply responses to marginal income tax rates are much higher among older,
more experienced workers, which implies potentially significant effects on
productivity.

Effects on the Cost of Capital
A primary mechanism through which changes in corporate tax rates and
depreciation allowances affect business investment is their effect on the user
cost of a capital investment—which can be thought of as the rental price of
capital, and is the minimum return required to cover taxes, depreciation, and
the opportunity costs of investing in capital accumulation versus financial
alternatives. A decrease in the user cost increases the desired capital stock, and
thereby induces gross investment.
By increasing the after-tax return on capital assets, a decrease in the tax
rate on corporate profits decreases the before-tax rate of return required for
the marginal product of new physical assets to exceed the cost of producing
and using these assets, increasing firms’ desired capital stock. Conversely, by
decreasing the after-tax return on physical assets, a decrease in the net present
value of tax deductions for investment expenses increases the before-tax rate
of return required.
Several factors may tend to bias empirical estimates of the user-cost
elasticity of investment, and early studies (e.g., Eisner and Nadiri 1968) tended
to find estimates that were considerably smaller than the benchmark unit
elasticity of demand for capital of Jorgenson (1963) and Hall and Jorgenson
(1967). First, a reliance on aggregate data potentially biases elasticity estimates downward, due to simultaneity between the user cost of capital and
investment shocks. Second, aggregate data suffer from limited variation and
unobserved firm heterogeneity, as demonstrated by Goolsbee (1998, 2004).
Second, as Goolsbee (2000) and Cummins, Hassett, and Oliner (2006) demonstrate, Tobin’s q-based empirical evaluations of neoclassical models will
tend to suffer attenuation bias when the fundamentals that drive investment
are mismeasured. Third, as noted above, estimates of the effects of changes
in corporate taxes on economic output can be biased by the timing of tax
reform. Historically, legislators have tended to lower corporate tax rates and
raise investment tax incentives during periods of economic contraction, and to
raise corporate taxes (and withdraw investment credits and other incentives)
during periods of economic expansion. Insofar as investment is correlated with
general economic conditions—for instance, in standard accelerator models, in
which the change in the growth of output drives investment—estimates of the
user-cost elasticity of investment will therefore be biased toward zero. Studies
that fully address these biases therefore tend to exploit large tax events that
differentially affect various types of firms or asset classes; in these instances,
the change in tax “treatment” is plausibly uncorrelated with underlying economic conditions.
Taxes and Growth

| 57

Exploiting instances of major corporate tax reforms, Cummins and
Hassett (1992) estimate user-cost elasticities of investment of roughly –1.1 for
equipment and –1.2 for structures. Auerbach and Hassett (1992) and Cummins,
Hassett, and Hubbard (1994, 1996) exploit differences in the composition of
investment across industries to identify user-cost elasticities, and they find an
estimated long-run elasticity of the capital stock of –0.67. Djankov and others
(2010) find an elasticity of –0.835 at the mean, based on their own database of
corporate income tax rates for 85 countries in 2004.
Using cointegration and plant-level microeconomic data, Caballero,
Engel, and Haltiwanger (1995) report estimated long-run user-cost elasticities
of investment by Standard Industrial Classification, two-digit industry codes
ranging from –0.01 for transportation to –2.0 for textiles and –1.0 on average.
These results imply a generally high long-run responsiveness of investment
to changes in the user cost of capital. Schaller (2006) also uses cointegration
techniques to estimate long-run user-cost elasticity. Assuming that user costs
will largely be exogenous in a small, open economy, Schaller estimates a
user-cost elasticity of –1.6 from quarterly Canadian aggregate data spanning
1962 through 1999. Using Bundesbank data to specifically estimate user-cost
elasticities with respect to the German tax system, and employing generalizedmethod-of-moment techniques to instrument for potentially endogenous
investment decisions, Harhoff and Ramb (2001) find a smaller user-cost elasticity of –0.42.
More recently, Dwenger (2014) has used German panel data and a distributed lag model based on research by Chirinko, Fazzari, and Meyer (1999).
Dwenger’s baseline estimates are about twice as large as the elasticity of –0.25
estimated by Chirinko and colleagues. However, after properly accounting for
the equilibrium relationship in the error correlation model, Dwenger (2014)
finds point estimates of the user-cost elasticity of investment to be –0.9; and a
two-sided, chi-square test suggests that the elasticity is not statistically different from the neoclassical benchmark of –1.0.
Approaching the question from a somewhat different angle, Giroud
and Rauh (2017) employ Romer and Romer’s (2010) narrative approach to
estimate the impact of U.S. State-level corporate taxes on establishment
counts, employment, and capital. They find short-run statutory corporate tax
elasticities of both employment and establishment counts of about –0.5 (–1.2
over a 10-year horizon), and short-run statutory corporate tax elasticities of
capital of –0.24 to –0.25. Exploiting quasi-experimental variation created by
the Domestic Production Activities Deduction, which allowed firms to deduct a
percentage of their “qualified production activities income” from their taxable
income, Ohrn (2017) finds that a 1-percentage-point reduction in the corporate
tax rate increases investment by 4.7 percent of installed capital and decreases
debt by 5.3 percent of total assets.

58 |

Chapter 1

There is, therefore, a generally emerging consensus within the academic
literature, as summarized by Hassett and Hubbard (2002) and Dwenger (2014),
that places the estimated user-cost elasticity of investment at about –1.0,
consistent with the neoclassical benchmark. These estimates imply that a tax
change that lowers the user cost of capital by 10 percent would raise demand
for capital by up to 10 percent. The tax rate or deduction change required to
affect a 10 percent reduction in the user cost of capital varies with the values of
other relevant user cost parameters, which we discuss below.
As evidence of the need for increased capital investment in the United
States, figure 1-11 shows that business equipment investment has weakened
substantially since 2014. The figure shows the contribution to GDP growth from
each of three business investment categories: equipment, structures, and intellectual property. Investment in equipment and structures, and their resulting
contribution to real GDP growth, has slowed in recent years and was negative
in 2016, as gross investment was less than depreciation, with the result that net
investment turned negative. In contrast, growth in intellectual property investment remained positive in 2016. Reductions in the user cost of capital that spur
equipment investment could reverse these trends and boost GDP growth.

Effects on Net Capital Outflows
One component of investment is foreign direct investment (FDI), and numerous empirical studies, which are discussed below, have observed that FDI is
highly responsive to cross-border differences in tax rates. Furthermore, this
responsiveness may have increased in recent years. These predictions are relevant to GDP estimates because, for a given level of domestic savings (S), any
increase in inward FDI constitutes a decline in net capital outflows (NCO) and
a corresponding decline in net exports (NX), in accordance with the national
income accounting identity S = I + NX = I + NCO. Intuitively, a decline in the
user cost of capital attracts capital inflows (both foreign firms investing more
in U.S. capital stock formation and U.S. firms choosing domestic capital stock
formation over foreign), leading to an exchange rate appreciation that lowers
exports and raises imports, resulting in a decline in net exports. As a capital
inflow, however, FDI is an important funding source for increased investment,
because I = S – NCO.
De Mooij and Ederveen (2003, 2005, 2008) provide extensive literature
reviews of the impact of tax rates on FDI. As most papers utilize different data
and empirical specifications to isolate this impact, these literature reviews
transform the coefficients in each study into a uniform semielasticity of FDI with
respect to the corporate tax rate. In their 2003 paper, de Mooij and Ederveen
average across 351 elasticity estimates, finding a mean elasticity value of –0.7,
which corresponds to a mean semielasticity (with respect to a percentage
point on the tax rate) of –3.3. In their 2005 paper, they extend the 2003 result by
considering alternative classifications of literature and including new studies.

Taxes and Growth

| 59

Figure 1-11. Contribution of Business Investment to Real GDP
Growth, 2010–16
Intellectual property

Equipment

Structures

Business fixed investment (net)

Percentage points (annual rate)
1.20
0.80
0.40
0.00
-0.40
-0.80
2010

2011

2012

2013

2014

2015

2016

Source: Bureau of Economic Analysis (2017), private fixed investment by type.
Note: Business fixed investment contribution shows the net contribution of private
nonresidential structures, equipment, and intellectual property spending to real GDP
growth.

Instead of averaging across all studies, they estimate average semielasticities
by study type: time series, cross-sectional, discrete choice, and panel. They
find an average semielasticity of –2.61 across time series studies, –7.16 across
cross-sectional studies, –3.43 across discrete choice models, and –2.73 across
panel data studies. Across all 427 estimates, they find an average semielasticity
of –3.72. In their most recent paper (de Mooij and Ederveen 2008), they predict
semielasticities based on study characteristics. For studies that use financial
data such as FDI or property, plant, and equipment, they predict an effective
marginal tax rate semielasticity of –4.0, an effective average tax rate semielasticity of –5.9, and a country statutory tax rate semielasticity of –2.4. For count
data, such as the number of new plants and/or plant expansions, they find an
effective marginal tax rate semielasticity of –1.3, an effective average tax rate
semielasticity of –3.2, and a country statutory tax rate semielasticity of 0.3.
Summarizing their work, table 1-3 contains semielasticities based on the coefficients within the described studies.
In the first study on taxation and FDI, Hartman (1984) examined aggregate inflows into the U.S. between 1965 and 1979 as a ratio of gross national
product, leading to a mean elasticity of –2.6, as calculated by de Mooij and
Ederveen (2003). Several papers then extend Hartman’s analysis by using a
longer time series and slightly adapting Hartman’s model (Boskin and Gale
1987; Young 1988; Murthy 1989), suggesting mean semielasticities of –5.8, –1.1,
and –0.6 (de Mooij and Ederveen 2003). Newlon (1987) criticized the data on
60 |

Chapter 1

Table 1-3. Summary Statistics for FDI Semielasticities
Study and year
Number
Hartman, 1984
6
Bartik, 1985
3
Boskin and Gale, 1987
12
Newlon, 1987
2
Young, 1988
12
Murthy, 1989
4
Slemrod, 1990
58
Grubert and Mutti, 1991
6
Papke, 1991
2
Hines and Rice, 1994
4
Jun, 1994
10
Swenson, 1994
10
Devereux and Freeman,
4
1995
Hines, 1996
46
Pain and Young, 1996
6
Cassou, 1997
17
Shang-Jin, 1997
5
Devereux and Griffith,
10
1998
Billington, 1999
2
Broekman and van Vliet,
3
2000
Gorter and Parikh, 2000
15
Grubert and Mutti, 2000
15
Altshuler, Grubert, and
20
Newlon, 2001

Mean
−2.6
−6.9
−5.8
−0.4
−1.1
−0.6
−5.5
−1.7
−4.9
−10.7
−0.5
1.3

Median
−3.5
−6.6
−2.7
−0.4
−2.1
−0.7
−3.5
−1.6
−4.9
−5.0
−1.3
2.7

Max.
2
−5.7
0.3
3.5
5.3
0.5
17.8
−0.6
−0.9
−1.2
5.9
5.1

Min.
−4.0
−8.5
−21.2
−4.3
−9.2
−1.6
−84.5
−3.3
−8.8
−31.7
−5.4
−8.1

S.D.
2.3
1.4
7.6
5.5
4.2
1
14.4
1.2
5.6
14.1
3.2
4.3

−1.6

−1.6

−1.4

−1.7

0.1

−10.9
−1.5
−7.5
−5.2

−10.2
−1.4
−2.8
−5.0

−1.1
−0.4
3.1
−4.7

−36.7
−2.8
−44.7
−6.2

8.2
1.2
13.5
0.6

−0.8

−0.9

0

−1.2

0.4

−0.1

−0.1

−0.1

−0.1

0

−3.3

−3.5

−2.5

−4.0

0.8

−4.5
−4.0

−4.3
−4.2

4.2
−1.7

−14.3
−5.8

4.2
1.2

−2.7

−2.6

−1.4

−4.0

0.8

Benassy-Quere,
Fontagne, and LahrecheRevilthers, 2001

4

−5.0

−5.0

−2.2

−7.9

3

Swenson, 2001
Buettner, 2002*

95
23

−3.9
−1.52

−3.2
−1.59

8

−29.9

8.4
0.58

Benassy-Quere,
Fontagne, and LahrecheRevilthers, 2003*

19

−5.37

−4.22

3.21

Stöwhase, 2003*

5

−7.36

−6.82

1.12

Buettner and Ruf, 2004*

15

−0.42

−0.39

0.35

2

−0.64

−0.64

0.02

14

−5.26

−4.30

2.71

Desai, Foley, and Hines,
2004*
Stöwhase, 2005*

Sources: de Mooij and Ederveen (2003); de Mooij and Ederveen (2005).
Note: S.D. = standard deviation. * indicates an update from de Mooij and Ederveen (2005).

Taxes and Growth

| 61

the rate of return for FDI used by Hartman and similar studies and highlighted
spurious correlation in the data, but found a similar semielasticity of –0.4 (de
Mooij and Ederveen 2003). Slemrod (1990) also criticized Hartman’s use of FDI
flows, raising concerns about using aggregate FDI flows to analyze the relationship between FDI and tax rates. Auerbach and Hassett (1992) then showed
that different components of FDI respond differently to tax rates, with mergers
and acquisitions constituting a particularly responsive form of FDI, moving
researchers to use data on property, plant, and equipment (PPE).
Grubert and Mutti (1991) analyzed the distribution of plant and equipment in manufacturing affiliates in 33 countries, leading to a mean semielasticity of investment of –1.7 with respect to foreign effective tax rates (de Mooij
and Ederveen 2003). Hines and Rice (1994) consider the distribution of PPE in
all affiliates in 73 countries, and they estimate a much larger semielasticity of
PPE ownership with respect to tax rates of –10.7, though the mean significant
semielasticity is –5.0 (de Mooij and Ederveen 2003; they estimate the mean significant semielasticities in their earlier working paper, De Mooij and Ederveen
2001). Altshuler, Grubert, and Newlon (2001) similarly compared the tax sensitivity of PPE and inventories in 58 countries between 1984 and 1992, finding
that the elasticity of both to changes in after-tax returns increased between
1984 and 1992, leading to an estimated average semielasticity of –2.7 across
both years (de Mooij and Ederveen 2003).
Another set of studies analyzes the impact of a host country’s tax rates on
firms’ location choices using discrete choice models. For example, Bartik (1985)
estimates the probability that a multinational chooses a given U.S. State for the
location of new plants as a function of State statutory corporate income tax
rates. De Mooij and Ederveen (2003) estimate that Bartik’s mean semielasticity
is –6.9. Using the same concept, Papke (1991) also finds a negative relationship
between U.S. States’ corporate income tax rates and location decisions, with
a mean semielasticity of –4.9 (de Mooij and Ederveen 2003). Devereux and
Griffith (1998) expand the discrete choice model outside the United States, by
looking at U.S. firms’ decisions to locate in France, Germany, and the United
Kingdom. De Mooij and Ederveen (2003) show that Devereux and Griffith’s
logit model implies an average semielasticity of –0.8. Buettner; and Ruf (2004)
and Stöwhase (2003) look at the location choices of German multinationals in
the European Union’s member countries. Buettner and Ruf find mixed results,
while Stöwhase find that firms respond to effective tax rates but not statutory
tax rates. The average semielasticity from Buettner and Ruf is –0.42, compared
with –7.36 from Stöwhase (de Mooij and Ederveen 2005).
Since de Mooij and Ederveen’s (2008) meta-analysis, several notable
studies have been published. Using a novel data set on corporate tax rates
across 85 countries in 2004, Djankov and others (2010) find large effects of
corporate tax rates on FDI where raising the effective tax rate by 10 percentage
points reduces FDI by 2.3 percentage points after one year. Looking at affiliates
62 |

Chapter 1

in low-tax jurisdictions, Dischinger and Riedel (2011) find a significant inverse
relationship between the average tax differential to other group affiliates and a
subsidiary’s intangible (intellectual) property investment, with estimates suggesting a semielasticity of about –1.1. Karkinsky and Riedel (2012), meanwhile,
examine whether patent applications are more likely to be made by lower-tax
affiliates, and find a semielasticity evaluated at the sample mean of –2.3, meaning that a 1-percentage-point increase in the corporate tax rate differential
reduces the number of patent applications by 2.3 percent.
The estimated tax elasticities and semielasticities of cross-border investment and of profit shifting (the latter are discussed below) are relevant to an
analysis of the effects of changes in the user cost of capital on investment—not
only because they have a direct bearing on investment financing but also
because they help explain weak investment and the absence of capital deepening in the U.S. before the TCJA. Elasticities from this literature suggest that
reducing the U.S. corporate tax rate will have two effects. First, U.S. net capital
outflows (NCO) would decline. This is both because foreign firms would invest
more in U.S. capital and because U.S. firms would invest less in capital abroad.
Second, U.S. firms would be less incentivized to shift their profits abroad, as
discussed below. The former effect will tend to result in a dollar appreciation, which will reduce net exports (NX). The latter effect will tend to raise net
exports. Provisions of the Tax Cut and Jobs Act to reduce the abuses of a territorial tax system would be critical to the realization of these gains.

Effects on Profit Shifting
An additional margin along which changes in corporate tax rates are likely to
affect growth is through profit shifting by U.S. firms to their foreign subsidiaries
or by U.S. subsidiaries to their foreign parents, typically by mispricing sales of
goods, services, and intangible capital between affiliates in high- and low-tax
jurisdictions. (Profit shifting, otherwise referred to as base erosion and profit
shifting, refers to tax avoidance strategies that exploit gaps and mismatches in
tax rules to artificially shift profits to low- or no-tax locations. This is different
from the legitimate earning of profits abroad from investment.) Guvenen and
others (2017) focus their analysis on U.S. multinational enterprises, and argue
that profit shifting by these enterprises leads to some economic activity being
credited to their foreign affiliates, resulting in an understatement of U.S. GDP.
This profit shifting has increased substantially since the 1990s. The authors
correct for this mismeasurement by “reweighting” the amount of consolidated
firm profits that should be attributed to the United States by apportioning
profits according to the locations of labor compensation and sales to unaffiliated parties.
Applying these new weights to all U.S. multinational firms and aggregating to the national level, the authors calculate that in 2012, about $280 billion
in so-called foreign profits could be reattributed to the United States. Given

Taxes and Growth

| 63

that the trade deficit was equal to about $540 billion, this reattribution would
have reduced the trade deficit by over half in that year. Extrapolating the 2012
findings to subsequent years, the CEA estimates that transfer pricing continued
to account for at least half the trade deficit between 2013 and 2016.
Crucially, firms’ propensity to engage in profit shifting is highly responsive to tax rate differentials. Hines and Rice (1994), using aggregate countrylevel data from the Bureau of Economic Analysis, estimate a tax semielasticity
of profit shifting of –2.25, indicating that a 1-percentage-point decrease in a
country’s corporate tax rate is associated with an increase of 2.25 percent in
reported corporate income.
As discussed in a recent white paper on this topic, “Corporate Tax Reform
and Wages: Theory and Evidence” (CEA 2017), the tendency of U.S. firms to
hold corporate profits overseas reduces the wages of U.S. workers. Engaging
in profit shifting is highly responsive to tax rate differentials, as discussed
above, and corporate rate reductions are therefore likely to affect the share
of profits that is repatriated. Under the assumption that U.S. workers would
retain 30 percent of the 2016 profits of U.S. firms that were earned abroad and
not currently repatriated, U.S. households could earn a raise of up to 1 percent,
depending on the share of profits that was repatriated. (For an example of
workers capturing 29 percent of firm operating surplus, see Kline et al. 2017.)
The trajectory of foreign profits indicates that the value of these profit shifts
for U.S. workers would increase in the future. Household income boosts from
this channel may be additive to the estimated $4,000 in household income
discussed above, because the empirical literature is largely based on countries
and time periods with less foreign profit activity, and less existing capital
parked overseas as taxes changed.
In general, profits earned abroad show the willingness of U.S. firms to
invest in production and business operations overseas, at the expense of
domestic investment. Reductions in the corporate tax rate create an opportunity for U.S. firms to instead increase their domestic investment. Furthermore,
these multinationals are among the class of high-paying employers in the
United States. The rent-sharing literature discussed above implies that incentivizing these high-paying firms to locate more of their operations in the U.S. is
again constructive for U.S. wage growth.

Effects of Crowding Out
Decades of research suggest that the long-term benefits of tax reform may be
attenuated by the revenue changes in the Federal budget. Decreases in tax
revenues that enlarge the government deficit and thus raise public borrowing
may crowd out private sector investment, which reduces long-term economic
activity. In the literature, crowding out occurs in both real and financial ways
(Blinder and Solow 1973). There is real crowding out when increased public
investment displaces private capital formation. This direct crowding out occurs
64 |

Chapter 1

through public consumption and investment (Buiter 1990). Conversely, financial crowding out influences private behavior by altering budget constraints or
by affecting prices through interest rates. Increased interest rates, a result of
the Federal Reserve’s bond-financing of fiscal deficits, can lead to the partial
loss of private capital formation.
The effect of increasing public spending or lowering taxes on future
national income is dependent on the return from public investment. Gale and
Orszag (2004) examine three models of how fiscal policy affects the economy.
The Ricardian equivalence theorem holds that government deficits will be offset by private saving through the purchase of public bonds, so that there are no
resulting changes in interest rates or capital flows. The small, open economy
view suggests private savings is less than the deficit, but that capital inflows
make up the shortfall. This means that economic growth and interest rates
remain stable. The conventional view is that private savings and capital inflows
are less than the change in deficit, so that GDP falls and interest rates rise to
induce more savings. If this holds, then public debt would be expansionary in
the short term and contractionary in the long term. Using reduced-form models, they conclude that increases in the budget deficit by $1 reduces national
savings by 50 to 80 cents. This suggests that Ricardian equivalence does not
hold.
Gale and Orszag (2004) also compare the small, open economy and conventional views by testing whether budget deficits affect interest rates. They
find that larger projected budget deficits equivalent to 1 percent of GDP are
associated with increases in long-term interest rates of 25 to 35 basis points.
Calomiris and Hassett (2002), however, observe little evidence that modest
temporary increases in government deficits have a significant effect on interest
rates, and further provide evidence that the empirical counterfactual should
account for the probability that a substantial portion of additional Federal tax
revenue would be spent rather than saved.
Laubach (2009) controls for the business cycle and monetary policy
effects on deficits by examining the relationship between the long-term forward rate and projected deficits projected by the Congressional Budget Office.
He finds that a 1 percent increase in the projected debt-to-GDP ratio raises
future interest rates by about 4 to 5 basis points. When the deficit increased
by the same measure, interest rates increased by 25 basis points. Engen and
Hubbard (2005) also find that when government’s debt increased by the equivalent of 1 percent of GDP, interest rates rose by 2 to 3 basis points. They find
that these results are sensitive to model specification. When the dependent
variable is the change in the forward rate rather than the interest rate level,
the result is statistically insignificant. Results from several economic studies
suggest that a deficit increase over the long term would result in higher interest
rates. Some researchers consider these results evidence that the conventional
view is a more accurate model than a small, open economy view.

Taxes and Growth

| 65

Figure 1-12. The Effect of an Exogenous Tax Cut on Nominal
Interest Rates
95% Confidence interval

Impulse‐response function

Interest rate (%)
1
0.5
0
-0.5
-1
-1.5
0

5

10
Quarters

15

20

Sources: Romer and Romer (2010); BEA NIPA; Board of Governors of the Federal Reserve
System; CEA calculations.

More recent research by Traum and Yang (2015) finds that no systematic
relationship exists between debt, real interest rates, and investment when
using a New Keynesian model that accounts for the interaction between monetary and fiscal policies. Using data from the tax increases of the 1990s and
the deficit-financed tax cuts of the early 2000s, they conclude that short-term
private investment crowding in or out depends on the fiscal or monetary shock
that leads to a deficit. Crowding in can occur when deficits result from lower
capital tax rates or increased public investment, because both raise the net
return to capital. In the long term, deficits can crowd out investment.
Utilizing their hybrid approach, meanwhile, Mertens and Ravn (2013) find
that a 1-percentage-point reduction in the average corporate tax rate has no
statistically significant effect on short-term interest rates (neither the Federal
funds rate nor three-month Treasury bill), over any time horizon. Mertens and
Ravn do, however, observe a significant short-run disinflationary effect of
reductions in the average corporate tax rate, which they note “might instead
have been expected to trigger a stronger monetary policy accommodation.” A
simpler approach, following Ramey (2016), is to set Romer and Romer’s (2010)
exogenous tax shock series as the impulse in a standard fiscal policy vector
autoregression, which generates the impulse response function reported in
figure 1-12. These results indicate that there is no statistically significant effect

66 |

Chapter 1

on short-term nominal interest rates, and possibly even a slight negative effect
over the first two years, which is consistent with Romer and Romer (2010),
whose results are generally inconsistent with large interest rate effects of tax
changes. In contrast, to completely offset our estimated long-run increase
in GDP of 2 to 4 percent due to the corporate rate cut to 21 percent and the
introduction of full expensing, discussed below, real interest rates would need
to rise by between 300 and 400 basis points.

Estimating the Growth Effects of Tax Reform
The particular challenge in translating the Mertens and Ravn (2013) estimates
into growth projections is that their estimated coefficients are based on an
explanatory variable equal to Federal tax receipts on corporate income relative
to corporate profits. Applying these estimates to a reduction in the statutory
corporate income tax rate from 35 to 21 percent and the simultaneous introduction of full expensing for nonstructure investment requires calculating the
effect these changes would have on Federal tax liabilities. As such, it is not
valid to simply treat changes in average effective tax rates as changes in the
relevant independent variable of Mertens and Ravn’s model. Corporate profits
in 2017 and preliminary estimates of the combined 2018 revenue effect (Joint
Committee on Taxation 2017b) suggest a roughly 5.7-percentage-point decline
in the ratio of Federal corporate tax liabilities to corporate profits in the first
year of implementation. Mertens and Ravn’s estimates suggest this could raise
GDP per capita by as much as 3.4 percent. Their estimates further suggest that
the proposed reduction in the average corporate tax rate would generate an
increase in nonresidential investment of about 13 percent.
Alternatively, converting the estimated 2018 revenue effect (Joint
Committee on Taxation 2017b) of the 14-percentage-point statutory rate
reduction plus full expensing into a change in the total tax liability share of GDP
of 0.8 percent, Romer and Romer’s (2010) estimates suggest these two reforms
will raise GDP by 0.8 percent in year one, and by 2.4 percent over three years.
Romer and Romer’s estimates also imply an increase in investment of about 9
percent by year three.
Approaching the question from a more structural angle, a primary
mechanism by which changes in corporate tax rates affect business investment
is through their effect on the user cost of capital. As discussed previously, by
increasing the after-tax return on capital assets, a decrease in the rate of tax
on corporate profits decreases the before-tax rate of return required for the
marginal product of new capital assets to exceed the cost of producing and
using those assets, thereby increasing firms’ desired capital stock. By raising
firms’ desired capital stock relative to current stock, decreases in the user cost
of capital thereby require positive net investment to offset depreciation, implying an increase in gross investment.

Taxes and Growth

| 67

The user cost modeling is simple if expensing is permanent, and though
the Administration expects Congress to eventually make the provision permanent, the TCJA as passed calls for 100 percent bonus depreciation to expire
after five years. In a forward-looking, rational expectations model, firms would
look ahead to the expiration of the provision and respond less to the tax
change because their long-run target capital stock would be lower than for a
permanent change in policy.
Accordingly, estimating the impact of an expiring provision necessitates
estimating the impact of permanent expensing and the impact of the expiring expensing. The correct growth effect would then be between the two,
depending on the extent to which firms expect the provision to be renewed in
subsequent legislation.
We begin with the effect of permanent expensing. The emerging consensus in the academic literature places the user-cost elasticity of investment at
–1.0, which implies that a tax change that lowers the user cost of capital by 10
percent would raise demand for capital by 10 percent. Computing the effect on
the average user cost of capital of reducing the statutory corporate tax rate to
21 percent and introducing immediate full expensing of nonstructure capital
investment depends on the values of the relevant parameters. Calculations of
these parameters—following Devereux, Griffith, and Klemm (2002); and Bilicka
and Devereux (2012)—yield an estimated decline in the average user cost of
capital of about 9 percent, though the percentage change in the user cost varies across asset types. Assuming a consensus user-cost elasticity of investment
of -1.0, a 9 percent decline in the average user cost of capital would induce a
9 percent increase in the demand for capital. Following Jensen, Mathur, and
Kallen (2017), it is then possible to use the Multifactor Productivity Tables from
the Bureau of Labor Statistics in a growth accounting framework to increment
the Congressional Budget Office’s 10-year GDP growth projections by the additional contribution to output from a larger capital stock, assuming constant
capital income shares, until we attain a new steady state.
Based upon user-cost elasticity estimates, our calculations show that a
reduction in the statutory Federal corporate tax rate to 21 percent combined
with full expensing of capital investment would raise long-run GDP by between
2 and 4 percent. Our estimates indicate a 0.4 percent increase over the baseline forecast by the Congressional Budget Office in year one and a 3.8 percent
increase in the long-run steady state if full expensing is made permanent. If full
expensing of nonstructure assets expires in year five as legislated, however,

68 | Chapter 1

the long-run steady state increase in GDP will be 2.9 percent. The economy
will achieve the higher growth path if firms expect the policy to be continued.1
Previous estimates of similar tax reform proposals yielded similar results.
A report released by the President’s Advisory Panel on Federal Tax Reform
(2005) evaluated a “Growth and Investment Tax Plan” that implemented a
business cash flow tax, allowed for the immediate expensing of capital investment, and set a flat corporate tax rate of 30 percent. Analysis of this plan by
the U.S. Treasury Department using a variety of alternative models found that
these reforms would have generated an increase in the capital stock by 5.6 to
20.4 percent over the long run, raising output by 1.4 to 4.8 percent during the
same period.
Financing of the additional investment implied by the reduction in the
user cost of capital would depend, as noted above, partly on repatriation
of previous profits and decreased profits attributed to foreign subsidiaries,
and partly on changes in savings and capital flows. Our preliminary calculations suggest that funding the estimated increase in gross investment could
potentially be achieved almost entirely by increased capital inflows by both
U.S. and foreign multinational firms. Given that in 2016 private nonresidential
fixed investment totaled $2.3 trillion, a 9 percent change would constitute an
increase of about $207 billion in investment. The mean estimate of the tax
semielasticity of FDI of -3.3 reported in de Mooij and Ederveen (2003), corresponding to a user-cost elasticity of 2.8 (Δ log[I] = αlog[ϰ] = [α / (1 – EMTR)]
Δ EMTR), suggests that a 14-point reduction in the statutory Federal corporate
tax rate along with the introduction of full expensing would raise FDI in the
United States by $121 billion and reduce U.S. direct investment abroad by
$79 billion, for a combined reduction in net capital outflows of $200 billion. As
noted above, increased capital inflows would result in an appreciation of the
dollar, thereby reducing net export demand in tandem with the increase in net
capital inflows.
Moreover, our calculations also indicate a positive contribution to
growth from reduced profit shifting by U.S. firms to foreign affiliates. Applying
Hines and Rice’s (1994) estimate of the tax semielasticity of profit reporting
to a statutory corporate rate reduction of 14 percentage points suggests that
reduced profit shifting could add up to $142 billion to GDP (0.8 percent), based
on 2016 numbers. Analyzing the TCJA specifically, private subscriber reports
1 Devereux and Griffith (1998) have argued that marginal analysis like this may be less relevant
for high-value, discrete projects that should be more responsive to average tax rates over time.
They develop a measure of the effective average tax rate, EATR, and show that capital spending is
highly responsive to it. Jensen, Mathur, and Kallen (2017) calculate the impact of moving toward
full expensing and dropping the statutory corporate tax rate to 20 percent, finding that the EATR
declines by about 11 percentage points. They estimate that, under the assumption that the move
to full expensing remains permanent, corporate investment would increase as a result by up to
34 percent in certain asset classes, thereby raising GDP by 4.7 percent over 10 years and by 8.4
percent in the long run.

Taxes and Growth

| 69

from Goldman Sachs and Deutsche Bank estimated in late 2017 that the
reduced incentives to shift profits overseas could reduce the U.S. trade deficit
by as much as 50 percent on impact, with a permanent level shift. Along with
the lower corporate tax rate that would make profit shifting less attractive for
many firms, the TCJA has also put in place a number of significant anti-base
erosion measures such as the “global intangible low-taxed income” provision
that imposes a minimum tax on foreign earnings, and the “base erosion antiabuse tax.”
On the individual side, Barro (2018), using estimates from Barro and
Redlick (2011), finds that the tax bill’s cut to the weighted-average marginal
individual income tax rate will expand the economy by 1.6 percent through
2019, corresponding to additional growth of 0.8 percent a year. Barro notes
that though the growth effect is temporary, the extra contribution to the level
of GDP is permanent. Using the Joint Committee on Taxation’s (2017b) 2018
revenue score and applying Mertens and Montiel-Olea’s (2017), Romer and
Romer’s (2010), and Mertens and Ravn’s (2013) estimates to the TCJA, meanwhile, suggest that the reductions in individual tax liability will raise GDP by 0.4
to 1.3 percent within three years.

Conclusion
We started this chapter with a look back at the United States’ tax system in a
historical and international perspective and the major tax reform achievement
of 2017—the enactment of the Tax Cuts and Jobs Act. On the individual side, we
demonstrated that the extensive use of itemized deductions in the U.S. income
tax code not only generates regressivity at the Federal level but also can distort
incentives and reduce Federal income tax revenue. Lowering individual income
tax rates while simultaneously raising the standard deduction and limiting the
use of distortionary deductions—as the TCJA does—can therefore facilitate
tax relief for middle-income households. These changes also provide supplyside benefits to economic growth, while at the same time partially offsetting
short- to medium-term negative revenue effects. Specifically, we find that the
net individual tax cuts of the TCJA can be expected to raise GDP by up to 1.6
percent by 2020.
On the corporate side, the TCJA is meant to address, in large part, the
increasing uncompetitiveness of U.S. corporate income taxation relative to the
rest of the world, due to a relatively high statutory rate and worldwide taxation. This, along with the observed high and increasing international mobility
of capital in recent years, has deterred U.S. capital formation. An analysis of
tax incidence also reveals that labor—as the relatively less mobile factor of
production—has borne a disproportionate burden of high corporate taxation.
We have cited a wide range of academic studies demonstrating that
reductions in corporate tax liabilities have significant, positive short- and

70 |

Chapter 1

long-run effects on GDP growth and wages—in particular by lowering the user
cost of capital and thereby inducing higher investment in capital formation,
financed principally by increased capital inflows. On the basis of these studies,
we have calculated that a reduction in the statutory Federal corporate income
tax rate from 35 to 21 percent, simultaneously with the introduction of immediate full expensing of capital investment, would generate an increase in GDP
of between 2 and 4 percent over time. We have also studied the impact of this
growth on a typical household, and find that the average household would,
conservatively, realize an increase in wage and salary income of $4,000.
In addition, estimated user-cost elasticities of foreign direct investment
suggest that the corporate tax changes of the TCJA will not only mitigate the
migration of U.S. investment capital abroad but also will serve to attract inward
capital investment by foreign companies, including overseas affiliates of U.S.
companies that are currently holding an estimated $2.6 trillion in unrepatriated profits. Increased international tax competitiveness will also contribute to
the net export component of GDP—by as much as 50 percent—by attenuating
corporate profit shifting through transfer pricing, and thereby partly offset the
effects of increased capital inflows on exchange rates.

Taxes and Growth

| 71

x
Chapter 2

Deregulation That Frees
the Economy
By limiting the costs of unnecessary regulation, by reviewing and eliminating
ineffective rules whose costs exceed their benefits, the Administration’s agenda
of deregulation is unleashing the talents of the American people and the true
potential of American businesses. Although some regulation can be beneficial—for example, to protect the environment and health—when job creators
must abide by overly burdensome rules, Americans lose opportunities to transform their own ideas into new businesses and into even more opportunities.
Regulation’s dynamism-dampening effects are evident in empirical analyses
of its influence on the economy. Increases in regulation decrease rates of
new business entry, and newer firms tend to make greater contributions to
economy-wide productivity, which in turn means higher wages for employees.
Increased regulation may even explain a nontrivial portion of the productivity
slowdown observed in recent years, which has exacerbated the stagnation of
wages. Moreover, the effects of regulation extend beyond business dynamics.
For example, overregulation has a negative impact on people’s ability to relocate to where jobs exist. Geographic mobility in the United States has ebbed to
an all-time low, as regulatory barriers, especially at the State and local levels,
make living in high-priced cities unattainable for many Americans. According to
one estimate, for instance, the relaxation of restrictive land-use regulations in
just the three cities of New York, San Jose, and San Francisco between 1964 and
2009 would have increased the 2009 U.S. gross domestic product (GDP) by 8.9
percent, and would have given more Americans the freedom of movement that
has been such a tradition in the United States. Additional barriers to mobility

73

come from, among other things, State-level occupational licensing restrictions
that prevent Americans from pursuing opportunities. These regulatory distortions of the labor allocation across borders also cause economic distortions,
along with regulation’s overall negative impact on job growth. In addition to
preventing people from moving to new jobs, regulation may prevent jobs from
being created in the first place, and can reduce the number of jobs in the overall
economy.
To put the economic burden of regulation into context, consider a thought
experiment: Imagine that each of the 9.8 billion hours devoted to compliance
paperwork in fiscal year (FY) 2015, according to the Office of Management and
Budget, were instead used by employees to create output equal to average
hourly earnings. These earnings would total $245.1 billion, equal to 1.35 percent of that year’s GDP and 41.6 percent of that year’s Federal national defense
budget.
To prevent these unintended consequences, the Administration is dedicated
to eliminating excessive regulation. In the Administration’s first eight months,
Federal agencies issued 67 deregulatory actions and only 3 regulatory actions,
far outpacing the goal of 2 deregulatory actions for every regulatory action. This
effort has created more than $8.1 billion in present-value cost savings. Given
the evidence regarding the impact of poor regulation on the economy, continued deregulatory efforts in the coming years can lead to further cost savings for
both firms and consumers as the U.S. economy grows.

G

overnment regulation affects firms and individuals pursuing various
types of economic activity. Examples of firm-level activity influenced
by government regulation include how and when one firm may merge
with another, how public utilities set prices, the amount of pollutants a firm
may generate in the course of producing its goods, and how much risk a firm
in the financial sector can take on without endangering the wider financial
system. Examples of the impact of regulation in the lives of individuals come in

74 |

Chapter 2

the form of things like the seatbelts and airbags found in automobiles and in
the insurance policies that individuals buy.
Regulations have intended as well as unintended consequences. In
some cases, the intended and unintended benefits of a regulation outweigh its
intended and unintended costs. The benefits of regulations that outlaw child
labor, for example, outweigh the costs they impose. In other cases, however,
the intended and unintended costs of a regulation may instead outweigh its
benefits. And the quantity of regulations in the United States, regardless of
how they are measured, has rapidly increased in recent decades. In light of the
reality that regulations can impose costs that exceed their benefits, this proliferation of regulations underscores the importance of ensuring that existing
regulations do not impose excessive costs.
The Trump Administration has prioritized the identification—and
removal—of regulations that fail to generate benefits that outweigh their
costs. This agenda of deregulation stands poised to increase economic growth
and improve the economic opportunities available to American businesses and
employees. Economists and other academics have, over the years, produced a
body of literature on regulation that provides the economic rationale for the
Administration’s current agenda of deregulation. After all, as Gayer and Viscusi
(2016, 1) note, the intellectual basis of cost/benefit assessments of government
regulation date back to Jeremy Bentham’s 1776 adage that “it is the greatest
happiness of the greatest number that is the measure of right and wrong.”
To put this Administration’s regulatory priorities in context, this chapter
first explores the theoretical justifications for regulation. Then it synthesizes
the economics literature’s empirical documentation of the effects that regulations have on economic activity in places like the United States. These empirical analyses demonstrate the benefits that can be generated by deregulation
along the lines of the Trump Administration’s agenda. Finally, the chapter
describes the actions undertaken by the Administration so far in order to
deliver these benefits to the American people.

Theories of Efficient and Inefficient
Regulation and Deregulation
Classical economic theory argues that economic agents, whether firms or
individuals, acting in their own self-interest (which they are in the best position
to know) will, via voluntary trading, come up with the most efficient allocation
of goods and services. Such an allocation of resources will maximize social
welfare. In this theoretical world, the only role for the government is to protect
property rights—no regulation of a market is necessary. So if this is the case,
why do we have regulations?
Two broad and influential schools of thought lay out the economic basis
for regulation. For the first school, regulations improve welfare by correcting
Deregulation That Frees the Economy | 75

“market failures.” For the second one, in additional to correcting market failures, regulations can improve welfare by addressing “internalities,” which
entails correcting individuals’ or firms’ failures to behave in their own (i.e.,
internal) self-interest.

Regulatory Benefits
The scope of benefits one counts in justifying the creation or the removal of a
regulation varies between these two different views. The set of benefits generated by a regulation, after all, depends at least in part on the nature of the
economic activity affected by the regulation.
In the first, and traditional, view of the conditions whereby regulations
can improve economic welfare, they correct for the failures of markets to generate the socially optimal outcome (e.g., when they address market failures).
The economics profession has identified the circumstances in which market
failures can result. These include when firms have excessive market power,
preventing competition within a market; when there are externalities imposed
by one individual or firm on another; and when there are information asymmetries between different market participants. The textbook example of a market
failure due to the existence of market power and the absence of competition comes from the leverage of the Standard Oil Company, a firm that once
held a monopoly on the production of oil in the U.S. The textbook example
of a market failure due to an externality that leads to overproduction in the
absence of regulation may come from the example of pollution in the environment, because firms and the consumers of their products do not fully pay for
the costs that the pollution they generate impose on others. An example of
information asymmetry is the market for used cars, because a car dealer may
know the defects of a car better than a buyer. As a result of this asymmetry, in
the absence of regulation, the market for used cars may not produce efficient
outcomes.
In all these cases, the measure of a regulation’s benefits and costs must
be in accord with their effects on the entire economy, rather than vis-à-vis
the specific firm or individual that produces or consumes the product that
the regulation affects. The regulatory breakup of Standard Oil generated benefits for U.S. consumers and other U.S. producers that exceeded the costs to
Standard Oil. U.S. citizens writ large have benefited from reduced pollution of
the environment, despite the fact that reducing it has made both producers’
costs and consumers’ prices higher than they were in the absence of pollution
regulations. Likewise, U.S. consumers, rather than used car dealers, benefit
from steps to ensure that used car dealers do not exploit asymmetries of information by reducing the costs of information for buyers.
In contrast, the second set of economic rationales go much further than
market failure rationales in determining the benefits of regulation. Rather than
addressing the externalities discussed above, regulations can also improve
76 | Chapter 2

welfare by correcting for “internalities” that lead individuals to make decisions
that do not serve their own best interests. Those who suggest that internalities
can render regulation welfare-improving depend on evidence in the behavioral
economics literature purporting to document cognitive biases in support of
this proposition. (For a representative exposition of the internalities-based
school of thought on regulation, see Allcott and Sunstein 2015; and for a representative exposition of its contrasts with the traditional view of regulation, see
Mannix and Dudley 2015.)
The departure of the internalities-based approach to regulation from the
approach that centers on the redress of market failure is not merely an abstraction. Indeed, some regulations in recent years have been justified on the basis
of cost/benefit analyses that include internalities. An example comes from
a regulation on energy conservation standards for commercial refrigeration
equipment. As part of the Energy Policy and Conservation Act of 1975, this regulation set energy efficiency standards for commercial refrigerators, effectively
circumscribing the type of commercial refrigerators that would be available
on the market. From a decrease in the set of possible refrigerators available
to choose from in the first place, the cost/benefit analysis assumes that those
who operate commercial refrigerators would experience a decrease in the cost
of operating them over the lifetime of the product. To assert that commercial
refrigerator operators would benefit from a restriction on the set of available
commercial refrigerators, one would need to assume that some subset of
commercial refrigerator operators would choose some subset of commercial
refrigerators that makes them worse off relative to what they could have chosen in the absence of the regulation. This regulation, then, appears to justify
its benefits in part on the basis of its purported remedying of an “internality”
suffered by the operators of commercial refrigerators.
Nonetheless, whether internalities render regulation welfare-improving
as much as the proponents of this school of thought would suggest remains
controversial. Some question whether the government regulators themselves
suffer from behavioral biases that distort their decisionmaking (e.g., Viscusi
and Gayer 2015). In this view, regulation serves to increase institutional behavioral biases rather than overcome them. Others question the reliability of the
research that purports to document the existence of the cognitive biases that
give rise to alleged internalities (e.g., Shrout and Rodgers 2017).
The idea that a regulation may generate benefits that accrue directly to
the actors that are the focus of the particular regulation, rather than to other
participants in economic activity, is not as new as the internality-focused
school of thought and the rise of the behavioral economics that is its foundation. At least since the 1990s, some have argued that the regulated experience
net benefits from regulation, though the focus tended to be on firms rather
than on individuals. This corner of the economics literature has focused, in
particular, on the possibility that environmental regulations can incentivize
Deregulation That Frees the Economy | 77

firms to innovate, offsetting compliance costs through increased efficiency
and enhanced productivity (Porter 1991; Porter and Van der Linde 1995;
Ambec et al. 2013). This is formalized in the economics literature as the Porter
Hypothesis, which can be presented in several different forms.
The first, or “narrow,” form of the Porter Hypothesis distinguishes
between market-based instruments and prescriptive regulation. The use of
market-based instruments incentivizes firms to innovate by working within a
competitive market, while prescriptive regulation discourages innovation by
defining how activities should be undertaken. This form emphasizes that flexible regulatory policies are more efficient than prescriptive ones.
The “weak” form of the Porter Hypothesis claims that environmental
regulation results in increased innovation. The “strong” form of the Porter
Hypothesis not only claims that well-designed environmental regulation
can increase innovation, but also can increase a firms’ competitiveness and
productivity enough to offset compliance costs. However, empirical evidence
supporting Porter’s view is anecdotal, and more rigorous empirical studies have provided mitigating and even contradictory results (Lanoie, Patry,
and Lajeunesse 2008; Jaffe and Palmer 1997; de Vries and Withagen 2005;
Brunnermeier and Cohen 2003). This lack of clear evidence in support of the
Porter Hypothesis makes its use in the estimation of regulatory benefits and
costs difficult to justify.

Regulatory Costs
Regulations can impose costs through a number of different channels. First,
there is the cost of complying with a regulation that businesses pay, both in
demonstrating that they are complying with the regulation and in changing
their production processes to do so. Second, though these regulations are
often placed on businesses, those who buy their products will pay part of the
regulations’ costs. Increases in firms’ costs via regulation will increase the
prices of products for consumers in a competitive market (as with the Fiduciary
Rule; see box 2-1). Third, there are costs that accrue to would-be consumers
who do not engage in a transaction, due to the effect of a regulation, or to
would-be businesses that cannot enter a market or stay in operation due to
the existence of a regulation. Fourth, costs can accrue to would-be employees
if firms decrease hiring in response to a regulation. Of particular importance,
for an externality like pollution, if these costs are equal to or less than the
benefits of reduced pollution, then the externality is appropriately internalized
and social welfare increases. But if this is not the case, the regulation does not
improve social welfare.
Of all regulatory costs, an easily identifiable one is the cost of collecting
information used by the government to determine compliance with a regulation. The Paperwork Reduction Act of 1980 was designed to reduce the total
paperwork burden that the Federal government imposes on private businesses

78 | Chapter 2

Box 2-1. Determining the Future of the Fiduciary Rule
In 2015, the U.S. Department of Labor released an updated rule proposal
to amend the definition of a fiduciary under the 1975 Employee Retirement
Income Security Act, known as ERISA. This rule change would expand those
with a fiduciary duty to include those providing investment advice to a retirement plan, participant, or individual retirement account owner. Imposing a
fiduciary duty requirement has a clear benefit, in that financial advisers would
be required to act in the best interest of their clients. Also, there is a large
academic literature finding that conflicting investment advice imposes substantial costs on retirement savers (Chalmers and Reuter 2010; Christoffersen,
Evans, and Musto 2013; Del Guercio and Reuter 2014; Foerster et al. 2017).
However, the rule would also impose large costs.
The Fiduciary Rule would immediately make an entire class of retirement planning professionals comply with those responsibilities associated
with being a fiduciary. Given the rule’s ambiguous language, it also creates
regulatory uncertainty as to whether the fiduciary duty exists for certain
investment educators and investment advisers. This increases the costs to
provide retirement investment advice, as advisers are now forced to comply
with a whole host of new regulations. This cost will be passed onto consumers
in the form of reduced availability of investment education and advice, or
higher fees for said advice. The industry points out that the additional compliance costs may make it unprofitable to provide individual investment advice
for small retirement accounts.
In response to these concerns, President Trump ordered the Department
of Labor to study if the fiduciary rule harms investors by decreasing access
to retirement savings products, information, or related financial advice.
Additionally, the department is asked to determine if the Fiduciary Rule has
disrupted the retirement services industry. The deadline for compliance with
the prohibited transaction exemptions accompanying the Fiduciary Rule has
been postponed until July 2019. For advisers, this delay will allow time to
comply with the extensive requirements associated with being a fiduciary.
Consumers of retirement investment advice should not see dramatic changes
in the availability of retirement products or advice during this period. The
Administration is continuing to review this rule, and hopes to tailor it more
narrowly so it has a less dramatic impact on the retirement investment
market.

and citizens. The act imposes procedural requirements on agencies that wish
to collect information from the public, including an estimate of the hours
necessary to collect the required information and an estimate of the personnel cost that reflects the burden of the collection. As part of the Paperwork
Reduction Act, agencies must seek and consider public comment on proposed
collections of information with 10 or more respondents, and receive approval

Deregulation That Frees the Economy | 79

from the Office of Management and Budget before beginning to collect information from the public.
In spite of the Paperwork Reduction Act, the Office of Management
and Budget’s estimated paperwork burden for regulatory compliance has
increased steadily over the years. This can be seen in the blue trend line in
figure 2-1—but with one important caveat: the methods used to capture total
paperwork burden changed between FY 2009 and FY 2010. Hence, it is not
possible to compare values across these two periods. In fact, doing so would
suggest that the burden barely changed between FYs 2009 and 2015. But when
one looks at the hours before and after this break in the data, it is clear that
the total burden increased in both. It went from 7.0 billion hours in FY 1997
to 9.8 billion hours in FY 2009—an average annual increase of 2.8 percent. It
then went from 8.8 billion hours in FY 2010 to 9.8 billion in FY 2015—an average
annual increase of 2.2 percent. The red trend line in figure 2-1 looks at the total
paperwork burden coming from the Treasury Department alone. Paperwork
from the Treasury, which accounts for more than 70 percent of the total burden
every year between FYs 1997 and 2015, follows a similar trajectory during the
two periods.
To put the economic impact of the paperwork burden into context, consider a thought experiment: Imagine that each of the 9.8 billion hours devoted
to paperwork in 2015 instead were used by employees to create output equal
to their average hourly earnings. These earnings would total $245.1 billion, an
amount equal to 1.35 percent of that year’s GDP and 41.6 percent of that year’s
Federal national defense budget. One potential benefit of the Administration’s
deregulatory efforts will be to slow down the growth in costs related to the
paperwork burden.
In some cases, regulators underestimate costs or additional, unanticipated costs arise. Although the original regulatory impact assessment may
have estimated a net benefit from a regulation, rising costs over time could
reduce or eliminate this benefit. This often occurs in situations where technology brings unanticipated change to a market that is heavily burdened by
regulation. Such unanticipated or underestimated costs are often cited as a
justification for instituting periodic retrospective reviews of existing regulation.
In addition, it is worth noting the distinction between the direct and
indirect costs imposed by a regulation. A firm’s direct costs are those that are
attributed directly to complying with a regulation—for instance, its costs to hire
a compliance officer to handle the regulatory paperwork. Its indirect costs are
the opportunity costs of investing its funds in a regulatory compliance activity
that could have been used for another part of its business. For instance, if a firm
must invest $1 million in a compliance activity, its indirect cost of compliance is
the profit that the $1 million could have generated if it had been invested in a
revenue-producing activity. These costs are difficult to measure, given that it is
difficult to know what a firm would have done with funds allocated to financing
80 |

Chapter 2

Figure 2-1. Total Paperwork Burden Hours, FYs 1997–2015
Hours (billions)
12

Total

9

Treasury
6

3

0
1995

2000

2005

2010

2015

Source: Office of Management and Budget, Information Collection Budget.
Note: At the end of 2006, the Office of Management and Budget introduced an electronic system to
process agencies’ paperwork burden estimates. Because the system was not fully implemented
until the end of 2007, values from about this time may contain errors, and comparison between
prior values and values afterward may not be reliable. Between 2009 and 2010, values underwent a
one-time adjustment, affecting all future values. The FY 2015 estimate will be updated in the FY
2016 report.

compliance activity if those funds had been used elsewhere. Nonetheless,
indirect costs are imposed on the economy by regulation.

The Impact of Regulation
There is a sizable body of economics literature on regulation’s impact on
various measures of a country’s economic health. Deriving a causal estimate of
the effect of regulation is difficult, in part due to the difficulty of formulating a
reasonable measure of regulation. Despite this, the economics literature does
include efforts by researchers to overcome such difficulties and document
the economic effects of regulation. Estimates of the effects’ magnitude vary
substantially, but this literature does highlight cases where regulation lowers
the level and rate of economic activity and can harm firms, employees, and
consumers.

Measurement
Many of the methods used by researchers to quantify the stock of current
regulations or the flow of all regulations are imperfect. Nonetheless, in spite of
these imperfections, these measures do allow one to draw at least some inferences about regulation and its effects. Moreover, measurement error in regulation measures will, in many cases, bias estimates of the impact of regulation
toward zero. To the extent that the literature finds effects, then these effects
should be notable to analysts.

Deregulation That Frees the Economy

| 81

One common variable used to study the impact of regulation is the number of pages in the Federal Register or the Code of Federal Regulations—two
measures with known limitations. For example, one could object to measurements derived from either on the grounds that they fail to measure the semantic content of what the text says (e.g., a few paragraphs of text could prohibit
a vast amount of economic activity, or vice versa). Even then, these caveats
illustrate the specificity of the conditions in which measurement error invalidates attempts at statistical inference. For instance, whether measurement
error causes an attempt at statistical inference to generate a “false positive”
(e.g., the effect of regulation on growth) depends on how the measurement
error correlates with the other variables relevant to the statistical technique at
hand. Thus, the presence of measurement error requires careful consideration
of appropriate statistical analysis but does not necessarily eliminate the usefulness of these imperfect measures of regulation.
As an alternative solution to a lack of a good measure of total regulation,
researchers have introduced RegData, an index derived from a textual analysis
of the Code of Federal Regulations, to more accurately measure the quantity
and impact of regulation (McLaughlin and Sherouse 2017; RegData is maintained by the Mercatus Center at George Mason University). RegData breaks the
data down by paragraph and by title, allowing analysis by different aggregation
levels; and it then counts the number of keywords that are indicative of restrictions on the economy—such as “shall,” “must,” “may not,” “required,” and
“prohibited.” RegData also closely tracks the number of pages in the Code of
Federal Regulations, from which it is derived.
In spite of the imperfections of any one measure, measures of regulation in the United States as a whole seem to support the idea that regulation
has increased in the country. Figure 2-2 illustrates the growth measures of
total regulation—using the Federal Register, the Code of Federal Regulations,
and the sum of all industry-relevant restrictions from RegData. Each series is
represented as an index, such that 1976 is equal to 100. Series are set to begin
in 1976 to reduce the impact of changes in the underlying construction of the
Federal Register in earlier years. Despite their shortcomings, each measure
shows an increase in the quantity of regulations of almost 2 to 2.4 times over
the last 40 years.
Another alternative is to measure the subset of regulations that are
classified as “economically significant”—which the 1993 Executive Order (EO)
12866 defines as those estimated to have “an annual effect on the economy
of $100 million or more.” Figure 2-3 highlights the economically significant
final rules published by selected agencies and administrations. A total of 29
economically significant final rules were published in the first year of the
George W. Bush Administration, and 45 were published in the first year of the
Obama Administration. In the first year of the Trump Administration, agencies
published only 18 final rules—most at zero net cost.
82 |

Chapter 2

Figure 2-2. Estimated Measures of Regulation, 1976–2016
Index (1976 = 100)
300
250

CFR pages
index

200
150

FR pages index
100

RegData index
50
0
1975

1980

1985

1990

1995

2000

2005

2010

2015

Sources: GW Regulatory Studies Center; McLaughlin and Sherouse (2017).
Note: CFR = Code of Federal Regulations. FR = Federal Register. RegData index is the sum of all
industry-relevant restrictions.

Figure 2-3. Published Economically Significant Final Rules by First
Year of a Presidential Term
Agriculture

EPA

Transportation

Health and Human Services

Other

Count per year
45

30

15

0
G. W. Bush

Obama

Trump

Sources: RegInfo; Office of Management and Budget.
Note: A “Presidential term” is defined as starting on January 20 and ending on January 19.
President Trump’s term accounts for regulation through January 19, 2018. The George W. Bush
Administration’s counts began before an electronic system was implemented by the Office of
Management and Budget in 2004 and should be compared with later Administrations’ counts
with caution.

Deregulation That Frees the Economy | 83

Additional measures of the regulatory burden include survey-based indices that allow regulations to be compared across countries. For example, the
World Bank’s worldwide governance indicators measure six dimensions of governance, including regulatory quality. This measure captures “perceptions of
the ability of the government to formulate and implement sound policies and
regulations that permit and promote private sector development” (Kaufmann,
Kraay, and Mastruzzi 2011, 233). The World Bank also publishes Ease of Doing
Business rankings, calculating each country’s distance from benchmark economies (those that implement the best regulatory practices). Comparing these
rankings over time and across countries highlights the relative changes of each
economy’s regulatory scheme. In the most recent data, the United States was
6th out of the 190 rated countries in the Ease of Doing Business ranking, lagging
behind New Zealand, Singapore, Denmark, South Korea, and Hong Kong.
The Organization for Economic Cooperation and Development (OECD)
also publishes a series of regulation-related indices, including the Indicators
for Regulatory Policy Governance (known as iREG), Indicators of Regulatory
Management Systems, Product Market Regulation indicators, Competition
Law and Policy Indicators, Indicators for Employment Production, and the FDI
[Foreign Direct Investment] Regulatory Restrictiveness Index. Each measures
evaluates a different aspect of regulation. For example, the Product Market
Regulation indicators assess how regulation affects competition in the product
market, with the understanding that increased competition results in a more
robust economy and greater economic growth. Meanwhile, the FDI Regulatory
Restrictiveness Index measures statutory restrictions on FDI for more than 50
countries. This ranking relies on four measures: equity restrictions, screening
and approval requirements, restrictions on foreign personnel, and such other
restrictions as limits on land purchases and the repatriation on profits.
In one ranking in particular, the OECD Product Market Regulation indicator, the United States tends to be more-regulated than its OECD peers. The
OECD’s calculations place the United States 27th out of 35 countries, behind
France, Chile, and the Czech Republic (Koske et al. 2015; see figure 2-4). This
suggests that the United States has the opportunity to exploit the gains from
deregulation in product markets.
Researchers typically use these indices to compare regulatory regimes
across countries. As long as the measurement error of the index does not
systematically vary across the countries included in the analysis, cross-country
analyses that draw on these regulation indices to make inferences about
growth will be as reliable as any other inference from cross-country data.

Aggregate Growth
Estimates of the impact of regulation on economic growth vary, not least
because estimated effects will depend on the category of regulation considered,

84 |

Chapter 2

Figure 2-4. Product Market Regulation, 2013 (35 OECD Countries)
Index scale from 0 to 6 (least to most restrictive)
Netherlands
United Kingdom
Austria
Denmark
New Zealand
Australia
Germany
Slovak Republic
Italy
Estonia
Finland
Portugal
Hungary
Belgium
Czech Republic
Japan
Canada
Spain
Ireland
Luxembourg
Norway
France
Iceland
Switzerland
Chile
Sweden
United States
Latvia
Poland
Slovenia
Greece
Korea
Mexico
Israel 1
Turkey

0.00

0.50

1.00

1.50

2.00

2.50

Sources: OECD; Koske et al. (2015).

the nonrandomness of regulatory implementation or withdrawal, and possible
general equilibrium effects that complicate identification. Nonetheless, there is
evidence within the academic literature supporting the conclusion that higher
levels of regulation in the aggregate can result in lower economic growth.
Djankov, McLiesh, and Ramalho (2006) estimate that moving from
the 25th to the 75th percentiles on the World Bank’s Ease of Doing Business
Index increases average annual per capita GDP growth across 10 years by 2.3
percentage points. Loayza, Oviedo, and Servén (2010) and Jacobzone and
others (2010) similarly observe a negative relationship between regulation and
economic growth. An OECD study by Egert and Gal (2016), meanwhile, estimates that a 0.31-point reduction in a country’s score on the OECD’s 6-point
energy, transportation, and communications indicator of regulatory intensity
in product markets is associated with a 0.72 percent boost to GDP per capita
over 5 years. The estimated effect rises to 1.02 percent over 10 years, and 2.09
percent in the long run. Egert and Gal also estimate that 0.30-point reduction
in a country’s score on the OECD’s 6-point employment protection laws indicator of regulatory intensity in labor markets is associated with a 0.22 percent
increase in GDP per capita over 5 years, rising to 0.57 percent over 10 years and
1.83 percent in the long run.
Ciccone and Papaioannou (2007) measure the burden imposed by regulation across countries by constructing a proxy for regulation’s red tape, which

Deregulation That Frees the Economy | 85

reflects the time it takes to start a new business in each country. They find that
responsiveness to industry-specific global demand shocks, as measured by
new firm entrants, increases as the volume of red tape decreases. Alesina and
others (2005) also find that more stringent regulation of product markets has
large and negative effects on aggregate investment. Finally, Justesen (2008)
observes that, in contrast to other variables plausibly determined by governments, the level of regulation is correlated with subsequent economic growth
in a panel of countries spanning the period from 1970 through 1999.
One primary channel through which increased regulation appears to
affect growth is through its effect on productivity growth. Exploiting a new
time series measure of the extent of regulation by the U.S. Federal government,
Dawson and Seater (2013) find that regulation lowers total factor productivity
(TFP) by distorting the mix of inputs in production, thereby adversely affecting overall output growth. Bailey and Thomas (2017), meanwhile, find that
regulation may also affect productivity through its effect on firm entry and exit.
Using a fixed-effects model, they observe that industries with more Federal
regulation experienced fewer new firm births and slower employment growth
between 1998 and 2011, and that large firms are less likely to exit more heavily
regulated industries than small firms. More specifically, they estimate that a 10
percent increase in the intensity of regulation leads to a 0.47 percent reduction
in new firm births.
At the more local level, Hsieh and Moretti (2017) estimate that with
decreased zoning restrictions in three cities—New York, San Jose, and San
Francisco—the growth rate of aggregate output could have increased by 0.795
percent to 1.49 percent a year between 1964 and 2009, thereby increasing GDP
in 2009 by 8.9 percent. The authors find that zoning restrictions increased the
spatial misallocation of labor, with the result of lowering labor productivity
growth. Herkenhoff, Ohanian, and Prescott (2017) reach similar conclusions,
finding that U.S. labor productivity and consumption would be, respectively,
12.4 and 11.9 percent higher if all states moved just halfway from their current
land-use regulations to the current Texas level.
Although local zoning restrictions specifically may lie beyond the scope
of Federal policy, we can apply estimates from the academic literature on
regulation generally to create a back-of-the-envelope projection of the impact
of the current Administration’s deregulatory agenda. Because Égert and Gal
(2017) suggest that decreasing a country’s Product Market Regulation Index
by 0.31 (the typical decrease in an episode of deregulation in OECD countries)
would increase its GDP per capita by 1.02 percent within 10 years, we can apply
their estimate to moving the United States from 27th in the Product Market
Regulation Index to 1st. If the United States achieved the same level of product
market regulation as the Netherlands from structural reform, U.S. real GDP
would increase 2.2 percent over 10 years, assuming a constant population
growth rate and constant inflation. If the United States instead implemented
86 | Chapter 2

the typically observed reform—decreasing its index by 0.31—U.S. real GDP
would increase by 1.0 percent over 10 years. And if the U.S. moved up the ranking a few places to achieve Canada’s level of product market regulation, U.S.
real GDP would increase by 0.5 percent over the same time frame.

Business Dynamics
Evidence across sectors indeed appears consistent with what one would
expect if an increase in regulatory burdens were impeding the dynamism of
American business. Regulations that impose fixed costs on businesses double
as barriers that prevent new businesses from entering markets and competing
with established firms.
Trends across a number of indicators of business dynamism appear
consistent with what one would expect if regulatory burdens were increasing, in a trend that favored large, well-positioned businesses over newer and
smaller firms. First, the net rate of new establishment creation in the United
States has trended downward over time (e.g., Decker et al. 2014; Hathaway
and Litan 2014). Second, the degree of competition appears to have decreased
in most industries in the United States (Gutierrez and Philippon 2017). And
this decrease coincides with an increase in firm profits coupled with stagnating investment; profits are rising, yet firms seem to be investing less in capital
assets that produce their goods and services. In particular, for companies
to invest in the United States, the break-even rate of return on a U.S. capital
investment must be higher than in alternative, lower-regulation jurisdictions.
If firms’ regulatory costs increase, fewer companies are likely to invest in the
United States. The economics profession is only beginning to develop an
understanding of the causes of these trends.
However, some evidence points to increases in Federal regulation as a
causal mechanism that could explain the apparent decline in new firm creation
and decrease in new firm dynamism. Bailey and Thomas (2017) exploit the
variation in regulatory trends across industries at the level of the four-digit
North American Industry Classification System code offered by the RegData
database. The baseline specification of Bailey and Thomas (2017) includes
year and industry fixed effects, allowing the isolation of variation in Federal
regulations that are idiosyncratic to a given industry within a given year—and
addressing the concerns raised by estimates of Federal regulations’ effects,
based only on variation across time. And according to the results from this
approach developed by Bailey and Thomas, an increase in Federal regulations
tends to decrease rates of new business entry. Bailey and Thomas also note
that large, established firms tend to be less likely to exit when their industry
has more regulation. Though the complexity of trends in the dynamism of
America’s businesses belies the possibility that any one piece of evidence
could have the final word, Bailey and Thomas nonetheless points to a causal

Deregulation That Frees the Economy | 87

link between the decline in U.S. start-up rates and competition and increased
Federal regulation.
Even if evidence for the relevance of government regulation to contemporary business dynamics may be new, the possibility that government regulation would reflect the preferences of interest groups like those representing
established businesses rather than merely maximize aggregate social welfare
has a long history in the economics profession (Stigler 1971; Peltzman 1976;
Becker 1976). Meanwhile, measures of policy uncertainty have trended upward
over time since the 1960s, and additional research has documented that firms
appear to vary their lobbying expenditures and political donations in response
to fluctuations in political risk (Baker, Bloom, and Davis 2016; Hassan et al.
2017).
Also, according to new research, the existence of regulatory barriers
to entry that influence rates of business formation would have an impact on
more than the distribution of benefits between firms. Research suggests that
the contributions of firms to productivity tend to decrease rapidly as firm age
increases. Established businesses whose longevity may be prolonged by the
existence of regulations tend to make less of a contribution to productivity
than the new firms that could replace them, according to new research from
Alon and others (2017). They estimate that the aging of established firms
since 1980 had by 2014 lowered aggregate productivity to 3.1 percent below
the level where it would otherwise have been. To the extent that government
regulation has decreased start-up rates and prolonged the existence of established firms—as Bailey and Thomas’s (2017) results suggest—then regulation
may have generated a causal contribution to the decline in productivity in the
United States.

Productivity
The influence of regulation on business dynamics, however, is only one of the
possible channels through which regulation can exert an effect on productivity. A useful measure for exploring these channels is TFP, which is the portion
of output not explained by the quantity of inputs, measuring how efficiently
and intensely inputs are used. Annual TFP growth for the private business sector averaged 1.7 percent from 1995 to 2005 but slowed down after the Great
Recession, growing 0.05 percent on average annually from 2007 to 2016. This
has been the slowest TFP growth rate of any recent business cycle expansion.
In 2016, TFP decreased 0.1 percent for the private business sector, its first
decline since 2009. Figure 2-5 illustrates these trends in TFP growth.
Although changes in productivity—both increases and decreases—have
been explained by a wide variety of factors, regulation has been shown to be an
important determinant of productivity (Baily 1986; Maddison 1987). Because
TFP is measured as the output per combined inputs, an increase in regulatory

88 | Chapter 2

Figure 2-5. Total Factor Productivity Growth, 1990–2016
Percent change (annual rate)
4

3

2

Centered fiveyear moving
average

1

0

-1

-2
1990

1995

2000

2005

2010

2015

Source: Bureau of Labor Statistics, Multifactor Productivity Tables.

costs results in an increased level of total inputs for the same level of output,
decreasing the ratio of output to inputs.
For example, hiring a compliance officer increases labor input, but may
not increase a firm’s output, as hiring a worker in another role would do. Thus,
the compliance burden lowers measurements of productivity by inducing firms
to allocate funds to compliance that generate no output included in the TFP
measure, rather than to output-generating activity included in the measure.
And CEA analysis suggests that the allocation of funds toward compliance is
nontrivial in magnitude. Using Bureau of Labor Statistics data on compliance
officer wages (Ruggles et al. 2017), CEA estimates that businesses spent $19.8
billion in 2016 on compliance officers’ wages—which constitutes a real increase
of 202 percent since 2000, with compliance costs growing an average of 7.16
percent each year.
Regulation can also result in disincentives to invest and innovate, further
decreasing TFP. If regulation diverts funding from otherwise productive uses
like innovation, regulation then limits a firm’s ability to increase efficiency
and output, and thus TFP. Finally, regulation can create barriers to entry that
reduce competition between firms. And without competition, a profit-maximizing firm may not be incentivized to innovate and increase its TFP to maximize
profit (Bloom, Draca, and Van Reenen 2016; Syverson 2004; Schmitz 2005).
The available empirical evidence suggests that regulation’s effect on
productivity can help explain industry-level trends over fairly long time

Deregulation That Frees the Economy | 89

horizons. For example, the regulations issued by the Occupation Safety and
Health Administration and the Environmental Protection Administration (EPA)
have been found to reduce productivity growth in the typical manufacturing
industry by 0.44 percentage point per year, accounting for over 30 percent of
the productivity slowdown in the 1970s (Gray 1987). These estimates are higher
than others that look at industries beyond manufacturing (Denison 1979;
Portney 1981; Norsworthy, Harper, and Kunze 1979; Christainsen and Haveman
1981; Crandall 1981). But they are in line with or smaller than results found
studying pollution control expenditures (Siegel 1979), changes in productivity
in the electric utilities sector from emissions regulation (Gollop and Roberts
1983), and the impact of occupational safety and health and environmental
regulations on the rate of productivity growth in Quebec between 1985 and
1988 (Dufour, Lanoie, and Patry 1998).
Additional evidence from the implementation of the Clean Air Act suggests that regulation harms productivity. The act’s stricter air quality regulations are associated with an almost 2.6 percent decline in TFP for manufacturing plants, though the impact of regulations specifically governing ozone is
particularly large (Greenstone, List, and Syverson 2012). After controlling for
confounding price increases, output declines, and sample selection biases, TFP
decreases an estimated 4.8 percent due to the Clean Air Act, which is equivalent to roughly $21 billion (in 2010 dollars) annually, or about 8.8 percent of the
manufacturing sector’s profits during the relevant period.
Although most contributions to the literature assessing the relationship
between regulation and productivity have focused on environmental regulation, other types of regulation have also been shown to decrease productivity.
For example, the Sugar Acts of 1934, and their repeal in 1974, illustrate how the
rise and fall of regulation can influence productivity—productivity within the
sugar industry appears to decline upon the introduction of these regulations
and to rise upon their repeal (Bridgman, Qi, and Schmitz 2007, 2009).
Other research exploits variations across the OECD countries to examine
the effects of regulations on growth and productivity. In a panel of OECD countries, Bourlès and others (2013) find that anticompetitive, upstream regulation
in advanced economies causes a decrease in productivity in high-technology
sectors. An analysis of OECD countries by Barone and Cingano (2011) suggests
that less regulation leads to an increase in the value added to the economy by
private firms. Nicoletti and Scarpetta (2003, 26) find “empirical results [that]
seem to suggest sizable benefits from further progress in reforming the regulatory environment and in reducing the role of the state in business activities”—
at least in part because of the productivity channel.
In addition, regulation-caused delays in bringing products to market
can lead to decreases in investment in sectors with intensive research and
development that may be disproportionately likely to generate productivityenhancing innovations. For instance, policies limiting government uncertainty
90 |

Chapter 2

about regulatory approval in the pharmaceutical drug context could have
led to a more than doubling of medical research and development and could
increase the current share of healthcare spending by more than 3 percent of
GDP (Koijen, Philipson, and Uhlig 2016).

Employment
The impact of regulation on employment is, in theory, ambiguous. The burden
imposed on businesses may decrease the number of individuals employed.
But one can also imagine that compliance burdens may have an ambiguous
employment effect—if a firm is required, for instance, to hire new employees in
order to comply with new regulations. Indeed, some research has found little
effect of regulation on employment. However, other research suggests that
regulations can decrease employment, and some research even shows that
deregulation can specifically increase employment.
Some of the evidence indicates that, in certain circumstances, the
employment effects of regulation may be lesser in magnitude than one would
expect on the basis of the overall burden imposed by a new regulation on
business. For instance, Berman and Bui (2001a) find that though the particular regulations they study do impose large costs, the air quality regulations
enacted by the South Coast Air Quality Management District, which includes
and surrounds Los Angeles, have an effect on employment that is lesser in
magnitude than the overall burden imposed on businesses. They find similar
results regarding employment and larger effects of regulation on abatement
investment when looking at oil refineries (Berman and Bui 2001b). These
results are consistent with what one would expect if the regulations compelled
firms to reallocate resources away from their most productive use and toward
a less productive but labor-intensive use—the drop in employment is lesser in
magnitude than the drop in productivity.
Other research, however, finds an effect of regulation on employment
(List et al. 2003). For example, as a result of the Clean Air Act, pollutant emitters in counties above a certain standardized pollutant level are subjected to
stricter regulatory oversight. These highly regulated counties, relative to less
regulated ones, lost close to 590,000 manufacturing jobs (Greenstone 2002). As
a result of the same act, the strengthening of emissions standards led to a 15
percent decline in the size of the newly regulated, pollution-generating sector
within 10 years (Walker 2011).
A period of deregulation undertaken in Portugal directly addresses the
effects of deregulation rather than increases in regulation on employment.
Analyzing this period in Portugal, Branstetter and others (2014) document
evidence of gains in employment and firm formation. They estimate that gains
accrue disproportionately to small businesses and to businesses in bricksand-mortar, low-technology sectors, such as agriculture, construction, and
retail. These results are consistent with a standard model of regulation as a
Deregulation That Frees the Economy | 91

fixed cost—the type of cost that larger firms can shoulder, but that drive small
firms out of business or prevent them from entering in the first place. Small
businesses suffer more from the costs of regulation, Branstetter and others’
(2014) results show. Portugal’s experience also demonstrates the benefits of
deregulation for employees as well as business owners; Fernandes, Ferreira,
and Winters (2014) document that Portugal’s deregulation increased the
returns to skills as well as the returns to the possession of a university degree.
To deregulate, this evidence shows, is to unleash the economic potential of
employees and owners alike.

Labor Mobility
Regulations also are imposed at the State and local levels. When such regulations differ across localities or States, they can have a negative effect on labor
mobility, making it difficult for labor supply to respond to geographic differences in labor demand. Examples of such regulations include those pertaining
to land use and occupational licensing. Because regulatory barriers to labor
mobility undermine labor’s capacity to be allocated toward its most efficient
use, regulations of this type can have nontrivial macroeconomic effects.
Land-use regulations govern the private uses of land resources and
include housing codes, zoning ordinances, and building codes. In cities experiencing high growth and productivity, land-use regulations often restrict housing availability, increasing housing prices and limiting the number of potential
employees who can respond to the high labor demand. Exploiting variance in
construction costs across housing markets, researchers use the ratio of priceto-minimum profitable construction cost to identify the impact of regulatory
construction constraints. A higher ratio indicates that the price of the house
cannot be explained by its physical construction costs and may be accounted
for by the regulatory burden imposed. In 2013, 26.4 percent of a sampling of
single-family houses were priced above minimum profitable production costs
by more than 25 percent. When looking at production costs at a metropolitanarea level to account for unobserved variation, only three markets reported
median ratios of greater than 2, while 11 percent reported ratios between 1.25
and 2. In comparison, in 1985, over 90 percent of metropolitan areas reported
median ratios near or below 1, meaning that the share of the median price-tocost ratios by area that were above 1.25 increased from 6.4 percent in 1985 to
15.4 percent in 2013. These high ratios suggest that physical construction costs
cannot explain rising housing prices and instead point to the role of regulation
in limiting the supply (Gyourko and Molloy 2014; Glaeser and Gyourko 2017).
In an efficient allocation of labor, potential employees will move from
low-productivity regions to seek better opportunities in higher-paying, higherproductivity regions (Ganong and Shoag 2016). If people are unable to move
to higher-productivity cities, low-productivity cities will have too many workers, leading to overall lower aggregate employee output across all U.S. cities.
92 |

Chapter 2

Indeed, between 1965 and 2009, labor misallocation due to housing supply
constraints ended up lowering aggregate growth by almost 50 percent (Hsieh
and Moretti 2017). In addition to limiting aggregate employment inflows by
limiting the housing supply, housing prices that are rising in productive areas
due to regulation can then further deter low-skill migration by pricing houses
above low-skill employees’ budgets, leading to increased segregation based
on skills (Ganong and Shoag 2016). American workers’ proclivity to move is at
an all-time low (see chapter 3), implying that they are increasingly unlikely to
relocate in pursuit of labor market opportunities. Policy solutions to address
this weak mobility could include adjustments to land-use regulations that
would lower the price of housing and encourage a more robust alignment of
employees with jobs.
Occupational licensing is another geographically based regulation that
has an impact on labor mobility by varying licensing requirements by each
State. Occupational regulation generally requires individuals to file registration paperwork, acquire certification, or receive a license, often referred to as
“the right to practice” (Kleiner and Vorotnikov 2017). All forms of occupation
regulation can involve costs, but occupation licensing is typically the most
intense form of regulation, given that governments evaluate the legal qualifications of a given potential employee. Licensing laws make it illegal to practice a
given occupation without a license.
For example, California’s Board of Barbering and Cosmetology requires
1,600 hours of education and hands-on training to take a licensing test for cosmetology. An additional 3,200 hours of apprenticeship and 220 hours of related
training are required for licensing. The adjacent State of Oregon requires 1,450
hours of education and training for hair design licensing and 350 hours for nail
technology, along with 150 hours of safety or infection control training and 100
hours of career development at a State-licensed career school. All potential
licensees must then pass a practical examination at one of these schools. A
California-certified cosmetologist is not authorized to practice in Oregon without receiving Oregon’s certification, creating a barrier to mobility.
The share of the U.S. workforce in a licensed profession has steadily
increased. In the 1950s, less than 5 percent of the workforce was licensed,
compared with about 18 percent in the 1980s. By 2000, this had grown to at
least 20 percent; and in 2003, more than 800 occupations required licensing in
at least one State (Kleiner and Krueger 2013). In 2008, 35 percent of employees
across the United States were either licensed or certified by the government,
with 29 percent being licensed. A total of 85 percent of licensed employees
were required to take an exam, while almost 70 percent were required to take
continuing education courses (Kleiner and Krueger 2013).
Occupational licensing requirements impose both direct and indirect
costs, discouraging labor mobility by creating barriers to entry. For example, if
the costs of becoming licensed in a new State is less than the expected returns
Deregulation That Frees the Economy | 93

from moving, people will not relocate. A recent study found that greater regulatory harmonization affecting the accounting profession across the European
Union’s member countries led to increased international labor migration
in comparison with other professions (Bloomfield et al. 2017). In the United
States, occupations that experienced a decline in employment from 1990 to
2000, such as librarians and dietitians or nutritionists, faced a larger decline in
States where the occupations were licensed (Kleiner 2006). As with land-use
regulation, this lack of mobility creates inefficiencies by preventing workers
from moving to high-productivity areas.
These barriers to entry also create wage differences between licensed
and unlicensed employees. As both Adam Smith and Milton Friedman observe
in their descriptions of economic markets, occupational licensing creates
barriers to entry by imposing a quantity restriction on the labor supply that in
turn increases wages. An opposing view implies that wages increase because
licensing imposes a quality restriction on the labor supply, meaning that the
higher wage in this case reflects higher-quality employees. Although the evidence is not clear on the quality or quantity driver in increasing wages, there
is strong evidence that wages increase as a result of licensing (Kleiner and
Krueger 2013; Gittleman and Kleiner 2016). And though wage increases can
signal a strong economy, wages that are artificially raised for some can come
at the cost of other employees losing their jobs or being excluded from the local
labor market.

The Trump Administration’s Initiatives
The Trump Administration has committed to reducing the burden of regulation
on the U.S. economy through the elimination of inefficient, duplicative, and
obsolete regulations that prevent beneficial economic activity. Rather than
suppressing the innovation and entrepreneurship that are central to America’s
economic growth, regulatory policy should instead simply administer the law
with respect for due process and fair notice. Toward this end, in 2017 President
Trump issued four EOs directing agencies to review current regulations. The
first, EO 13771, instructed agencies to repeal two regulations for every new
regulation and to ensure that the total incremental cost of all new regulations
does not exceed zero. EO 13772 provided core principles for regulating the U.S.
financial system that emphasized the priority of empowering individuals to
make informed, independent financial decisions. EO 13777 required agencies
to review all existing regulations in order to highlight excessive regulation. And
finally, EO 13783 focused on energy regulations, requiring agencies to review
existing regulations that potentially burden the development of domestically
produced energy resources. These orders have led to the identification of
economically beneficial deregulatory opportunities as well as a more careful

94 |

Chapter 2

examination of future regulations, as is evident in the Fall 2017 Unified Agenda
of Federal Regulatory and Deregulatory Actions.
The Unified Agenda—which is published by the Office of Information and
Regulatory Affairs, a unit of the Office of Management and Budget—provides
transparency to the public regarding anticipated Federal regulatory and
deregulatory policy. In this agenda, more than 60 cabinet, executive, and independent agencies compile information on upcoming rules, long-term actions,
and completed actions. The Fall 2017 Unified Agenda reports that agencies
withdrew 635 proposed actions that had been included in the Fall 2016 Unified
Agenda. In the Fall 2017 Agenda, agencies also reclassified another 944 active
actions from the Fall 2016 Agenda as long term (700) or inactive (244). Inactive
regulations include those that are still being reviewed or considered. All these
actions reflect the Administration’s commitment to meaningful consideration
of regulations.
Of the new proposed rules and rules already under review, the
Administration published only 89 final rules (figure 2-6), about 42 percent of
the average number of final rules published annually during the past 10 years.1
Though these averages are inflated due to the fact that many administrations
substantially increase regulation in their last year in office, the number of
final rules published in 2017 is still about 46 percent of the average, when the
counts for the years 2008 and 2016 are removed. Many administrations also see
a decline in regulation in their first year. Still, the 2017 decline in the annual
number of economically significant rules published was the largest percentage
decrease since 2007, with 61 percent fewer than the previous year. The total
number of published final rules also fell at a faster rate than any other year
since 2007, signaling a dedication to eliminating excessive regulation.
The Fall 2017 Unified Agenda also outlines the regulatory goals for FY 2018
that reflect the regulatory outlook of the Trump Administration. For example,
the Department of the Interior intends to finalize 28 deregulatory actions, leading to a reduction in costs of more than $1 billion (in net present dollars). The
Bureau of Land Management, a part of the Interior Department, has proposed
repealing rules regulating hydraulic fracturing that duplicate State regulatory
efforts. The Department of Labor plans to streamline its approval process
for apprenticeship programs to help workers looking to participate in such
programs. The Department of Transportation plans to issue a rule that would
give passenger railroads increased flexibility in designing trains, including easing the regulatory burden for high-speed rail operation, which would increase
competition in the passenger train market. These deregulatory initiatives are
likely to reduce unnecessary burdens on individuals, businesses, and State and
local governments.
1 Because the Office of Information and Regulatory Affairs frequently updates its data on its
website, www.reginfo.gov, these counts are estimated as of February 9, 2018.

Deregulation That Frees the Economy | 95

Figure 2-6. Final and Economically Significant Rules, 2007–17
Count per year
300
250

Published final
rules

200
150
100

Published economically
significant final rules

50
0

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Sources: RegInfo; Office of Management and Budget.

In his first EO, 13771, addressing regulation, President Trump instructed
administrative agencies to consider whether earlier regulations are unnecessary before creating new ones by repealing two prior regulations for every new
regulation. For example, the Department of Housing and Urban Development
recently announced a top-to-bottom review of its manufactured housing rules
to evaluate whether the compliance costs of these rules are justified given the
shortage of affordable housing. By requiring the removal of two regulatory
actions to offset the implementation of each new regulatory action, the “twofor-one rule” limits future regulatory costs. In this way, agencies can ensure an
overall outcome of zero net costs, or even cost savings.
Both Canada and the United Kingdom have implemented similar processes for administrative rulemaking. In 2012, Canada enacted a “One-for-One”
for regulatory requirement, while the United Kingdom imposed a “One-In, OneOut” rule beginning in January 2011. Between 2012 and June 2014, Canada
removed 19 regulations, reducing the annual burden on businesses by over
C$22 million. Meanwhile, the U.K. government’s statistics suggest that its
initiative reduced business burdens by £963 million, and it has since changed
the rule to “One-In, Three-Out” through 2020 (Renda 2017). The academic
literature on the effectiveness of such efforts is limited, and both governments
have documented mixed results.
In complying with EO 13771, U.S. agencies outperformed the two-for-one
goal by issuing 67 deregulatory actions while only enacting 3 new regulatory
actions, a ratio of 22:1. These actions achieved $8.1 billion in cost savings

96 |

Chapter 2

in present value terms, or $570 million per year. The Administration aims to
continue its deregulatory agenda in 2018, with Federal agencies planning to
complete four deregulatory actions for every new regulatory action. Agencies
anticipate that this will save $9.8 billion, or $686.6 million per year in every year
that agencies adhere to the 2018 regulatory cost caps.
In response to EO 13772—which was signed on February 3, 2017—the
Department of the Treasury has issued three reports and plans on issuing a
fourth. The first report was released on June 12, 2017, and discussed regulation pertaining to the depository system, banks, savings associations, and
credit unions of all sizes, types, and regulatory charters (i.e., the Banking
Report). The second report was released on October 6, 2017, and discussed
regulation pertaining to capital markets, including debt, equity, commodities
and derivatives markets, central clearing, and other operational functions (the
Capital Markets Report). The third report was released on October 26, 2017,
and discussed regulation pertaining to the asset management and insurance
industries, and retail and institutional investment products and vehicles (the
Asset Management and Insurance Report). The final report will discuss the
regulation of nonbank financial institutions, financial technology, and financial
innovation.
The Banking Report outlined five reforms key to achieve a less burdensome regulatory system: improve regulatory efficiency and effectiveness by
evaluating duplicative regulations across numerous agencies; better align the
financial system to support the U.S. economy; reduce the regulatory burden
by decreasing unnecessary complexity; tailor the regulatory approach to firms’
size and complexity and better coordinate these efforts across regulations; and
align regulations to support market liquidity, investment, and lending.2 The
report also makes specific recommendations to improve legislation, regulations, and policy that run counter to President Trump’s core principles outlined
in EO 13772. In keeping with these principles, therefore, the report emphasizes
the need to refine, consolidate, and better define financial regulations across
agencies to reduce the outsized costs imposed on smaller banks and create a
more harmonized financial regulatory environment. As of December 2017, the
Senate is considering legislation to raise the threshold to which many of the
more onerous banking regulations apply.
The Capital Markets Report provides specific recommendations for
changes in legislation, regulation, and policy in order to support U.S. capital
markets. Recommended changes are intended to promote access to capital for
2 The depository system is affected be regulations issued by, among others, the Financial
Stability Oversight Council, the Board of Governors of the Federal Reserve System, the Office
of the Comptroller of the Currency, the Consumer Financial Protection Bureau, the Securities
and Exchange Commission, the Federal Deposit Insurance Corporation, the Commodity Futures
Trading Commission, the Federal Housing Finance Agency, and the National Credit Union
Administration.

Deregulation That Frees the Economy | 97

all types of companies, including small and growing businesses, by reducing
regulatory burdens and improving market access; fostering robust secondary
markets in equity and debt; appropriately tailoring regulations on securitized
products to encourage lending and risk transfer; recalibrating derivatives
regulation to promote market efficiency and effective risk mitigation; enabling
proper risk management for central counterparties and other financial market
utilities in recognition of the critical role they play in the financial system;
rationalizing and modernizing the U.S. capital markets regulatory structure
and processes; and advancing U.S. interests by promoting a level playing field
internationally.
The Asset Management and Insurance Report provides specific recommendations for changes in legislation, regulation, and policy in order to support
the U.S. asset management and insurance industries. Recommended changes
are intended to promote efficient regulation by adopting a principles-based
approach to liquidity risk management rulemaking for registered investment
companies; instituting a “plain vanilla” rule for exchange-traded funds that
allows new entrants to avoid the cost and delay of obtaining individual exemptive orders; modernizing shareholder reports to permit the use of implied consent for electronic disclosures; realigning the Federal Insurance Office around
five pillars, including the promotion of the U.S. State–based insurance regulatory system and the U.S. insurance sector; recommending that the Federal
Reserve Board leverage the information that is received by State insurance
regulators and the National Association of Insurance Commissioners from savings and loan holding companies, and recommending that the Federal Reserve
Board harmonize its financial reporting and recordkeeping requirements
with corresponding State regulatory requirements; encourage the States to
expeditiously pass uniform legislation regarding data security and breach notifications for insurers; and improving coordination and collaboration among
federal agencies and State insurance regulators on insurance issues.
On February 24, 2017, President Trump signed EO 13777, which requires
Federal agencies to review all existing regulations and identify and revise those
that meet criteria to isolate inefficient regulations. These include regulations
that eliminate jobs, that are outdated or ineffective, that impose costs in excess
of their benefits, that interfere with other regulatory reform initiatives, or that
are the result of since-rescinded EOs. Agencies are asked to make recommendations regarding these regulations and consider combining overlapping
regulations. For example, the Department of Defense identified approximately
500 regulations that are subject to review under EO 13777 and that apply to
everything from real estate to flood control. A notice of these regulations was
then published in the Federal Register, providing the public with the opportunity to comment on their effectiveness. These regulations are now being
reviewed by a Department of Defense task force, which will then offer recommendations to the Secretary of Defense in the coming year. With his approval,
98 | Chapter 2

these recommendations—including actions to repeal, replace, or modify these
identified regulations—will be implemented.
Similar actions are taking place in other U.S. Federal departments. Each
component within the Department of Homeland Security has designated a
senior official to oversee their component’s regulatory reform efforts and
report to a task force that oversees the department’s deregulation efforts. Like
the Department of Defense, the EPA issued a Federal Register notice evaluating
existing regulations and received over 460,000 public comments. The EPA also
created a Regulatory Reform Task Force to coordinate public input with regulation recommendations.
The Trump Administration has also applied its deregulatory philosophy
specifically to energy production. EO 13783 encourages energy independence
by both promoting the clean and safe development of U.S. energy resources
and avoiding unnecessary regulatory burdens. This EO requires agencies to
review all existing regulations and similar agency actions that could burden
the development or use of U.S. energy resources, including natural gas, coal,
and nuclear energy resources. After review, agency heads have been required
to submit recommendations that could alleviate or eliminate any unnecessary
regulation burden on domestic energy production. As with EO 13777, these
recommendations have included suspending, revising, or rescinding unnecessary regulations. For example, the “Waters of the United States” rule, which
would have greatly expanded the purview of the Clean Water Act and imposed
significant regulatory burdens on both America’s farmers and ranchers and its
energy producers, is also undergoing review and potential replacement.
In addition, the Administration has taken a number of steps to allow
American firms to harness the economic value of America’s coal reserves for
themselves and for their employees. Pursuing the energy dominance agenda
of EO 13783, the department also revoked a previous moratorium on new
leases for coal production on Federal land—estimated by the U.S. Department
of the Interior (2017) to produce about 40 percent of America’s coal. Although
there are many exacerbating factors besides changes in regulations, the coal
industry has responded as expected. According to 2016 and 2017 data from
the Energy Information Administration, coal mining employment increased 2.4
percent year-on-year from November 2016 to November 2017, and coal exports
in the first two quarters of 2017 rose by more than 55 percent above their 2016
level.
The Department of the Interior has also worked to streamline the
application and permitting process for oil and gas wells on Federal lands. In
December, the Administration repealed a regulation covering hydraulic fracturing (“fracking”) on federal lands, on the grounds that it was unnecessary,
burdensome, and duplicative of existing State and some tribal regulations.
Other efforts include an ongoing review of new regulations for venting and
flaring natural gas at well sites located on Federal lands. Other mechanisms
Deregulation That Frees the Economy | 99

have also addressed regulatory burdens—a Department of the Interior solicitor’s opinion reversed Obama-era guidance on criminal penalties under the
Migratory Bird Act, which removes a substantial risk for the development of
wind energy resources.
Finally, fulfilling the pro-growth agenda envisioned in EO 13783, the EPA
has taken multiple steps to evaluate and decrease its regulatory burden. In
October 2017, the EPA proposed the repeal of the Clean Power Plan in order to
alleviate the burden it would impose on America’s job creators and consumers
(EPA 2017). The EPA estimates that this plan’s repeal could lower compliance
costs in 2030 by as much as $33 billion, with these cost savings passed along to
businesses and consumers in the form of lower electric bills.
As part of the review process for EO 13783, and in coordination with
efforts to address EO 13777, the EPA has also identified four key initiatives
to reduce the unnecessary burden of these regulations: comprehensive New
Source Review reform, National Ambient Air Quality Standards reform, evaluation of the employment effects of EPA regulations, and a sector-based regulatory outreach program. Specifically addressing EO 13783, the EPA created its
Smart Sectors program to better coordinate its efforts with industry stakeholders on regulatory developments, with the understanding that smart regulation
requires improved relationships with the regulated community. For example,
as part of Smart Sectors, EPA sector liaisons are focusing on building relationships with sectors and improving customer service for them, improving expertise vis-à-vis industry’s specific factors, and using this information to better
inform future regulatory directions. The overall goal is to engage stakeholders
early in the development of policy through collaborative problem-solving—an
approach that will improve environmental outcomes.
Along with EO 13783, this Administration has encouraged energy development by facilitating the construction of the Keystone XL pipeline, which
would transport Canadian crude oil to U.S. refineries. In March 2017, the U.S.
Department of State issued a Presidential permit to TransCanada, which
enabled construction of the pipeline to proceed. This permit is necessary for
the construction, connection, operation, and maintenance of facilities exporting or importing petroleum products between the United States and foreign
countries. Though the Keystone XL Pipeline is far from complete, and regulatory hurdles at the State level remain, the final Presidential permit removed a
cloud of uncertainty that had surrounded the project.
Through this Administration’s efforts, including its EOs and Federal agencies’ resulting actions, it has taken steps to reduce economically inefficient
regulation. The Administration’s EOs discussed above require regulators to
critically examine both existing and potential regulations. With the adoption of
task forces and programs such as Smart Sectors, agencies are actively seeking
to remove costly regulations and create new, beneficial regulations with the
assistance and knowledge of field experts. The effects of this work are already
100 |

Chapter 2

evident, with the significant slowing down of proposed regulations and increasing deregulatory efforts. In the coming year, the effects of the Administration’s
actions will continued to be felt, given that many agencies have only begun
suspending, revising, or rescinding unnecessary regulations.

Conclusion
Government regulation pervades the lives of ordinary Americans, making an
impact on decisions made by both firms and individuals. Regulations have
intended benefits, along with expected and unexpected costs. Though individual regulations may be expected to generate net benefits when imposed,
the economics literature contains a multitude of studies providing evidence
that individual regulations can generate unexpected costs that are larger than
the realized benefits. These costs accrue in the form of dampened growth,
diminished capital formation, stunted business dynamics, hampered productivity, decreased employment, and lower labor mobility. As regulation in the
United States has marched upward in recent years, many of these maladies
have been manifested in the U.S. economy.
The Trump Administration, however, has prioritized the elimination of
unnecessary regulations in the U.S. economy. The Administration’s specific
and far-reaching actions will ensure that only those rules that provide benefit in excess of their costs will be imposed on Americans. The record of the
Administration’s first year reflects these efforts, because the number of deregulatory actions that eliminate unnecessary and harmful regulations exceeds
the number of new regulations. These actions will reduce the costs imposed
on America’s businesses and employees, significantly expanding economic
opportunities.

Deregulation That Frees the Economy | 101

x
Chapter 3

Labor Market Policies
to Sustain the Middle Class
The Great Recession of 2007 to 2009 sharply reduced what the middle class
earned from work. By 2016, the household labor earnings of the median
American were still below their prerecession high of nine years earlier, despite
the fact that the recession had officially ended in 2009. This unprecedentedly slow recovery, especially for middle-class labor incomes, is perhaps the
primary economic problem our Nation faces. While labor incomes stagnated,
a marked increase in net government transfers (government benefits received
minus taxes paid) to some degree offset the decline in the median American’s
household income from all sources between 2007 and 2016, finally even surpassing its prerecession high in 2016. But without substantial increases in economic growth, this level of redistribution is unlikely to be sustainable. Clearly,
the best possible outcome is for labor incomes to return to normal levels of
growth. In the interest of guiding policymakers in their efforts to stimulate such
a recovery, this chapter takes a deep dive into describing the recent failure of
American labor markets to deliver the prosperity to which Americans had previously grown accustomed. The patterns this chapter describes will clarify the
motivation of many of the Administration’s policy initiatives that this Report’s
other chapters discuss.
Why have American workers had the worst labor earnings experience in modern
history for the past nine years? As chapters 1 and 2 discuss, the Administration’s
actions on tax and regulatory reforms will stimulate lagging economic growth
and increase the demand for workers, addressing this key factor—as will
rebuilding our Nation’s infrastructure, as outlined in chapter 4.

103

However, other factors are important as well, and these will become clear
from the evidence this chapter presents. Reductions in the disincentives to
work—alongside increased enforcement of work provisions and eligibility
requirements in the country’s welfare programs—are also needed. Continued
government transfers mean that some prime working age Americans find
themselves facing a trade-off between staying on the sidelines, while continuing to receive government transfers, and coming back to work, which would
result in a forfeiture of those entitlements.
This Administration is also deploying other supply-side tools. Policies that
enable workers to get retraining to meet current market needs through apprenticeships and other programs will encourage work. Stemming the opioid crisis
and enacting policies to better connect workers with jobs, including broadband
access and improvements in geographic mobility, would also increase the
labor supply. Other pro-work policies proposed by this Administration, such as
paid family leave for new parents, should also raise the long-run probability of
parental employment.
The experience of older Americans who are now staying in the workforce longer
indicates that government policies can indeed affect decisions to work, and
that demography need not be destiny. Public policies enacted since the 1980s
with respect to retirement have played an important role in incentivizing work
at older ages—for instance, by raising the age at which full Social Security benefits can be claimed and ending the earnings test at that age for those who wish
to continue working while receiving benefits. Eliminating the earnings test for
workers between age 62 and normal retirement age would likely also increase
participation. Policy may also help nudge employers to fill the unmet demand
of older workers for jobs with flexible, reduced hours as a gradual entry into
retirement in lieu of full nonparticipation by these workers.
Finally, younger workers have become increasingly detached from the labor
force. Although more teens are enrolled in school, evidence suggests this
group, relative to 10 years earlier, spends more time on unproductive activities,

104 |

Chapter 3

neither at work nor in school. Early employment can be especially critical for
the life prospects of lower-income American teenagers, and policy efforts to
encourage the integration of practical labor force training into high school
programs may help stem this tide by providing teens with employment and
occupational direction early on.
Because past public policies are responsible for some of the drag on employment growth, changes in these policies can be important mechanisms for
getting workers off the sidelines and again fully participating in our economy
and enjoying its benefits. A coordinated effort encouraging people to do this
could—as the economics literature we describe below suggests—significantly
reinforce the positive effects of labor demand policies, such as the major tax
reform just passed.

T

wo measures of the United States’ economic success are its rate of real
economic growth, and how this growth translates into resources for the
average American.1 This chapter begins by documenting trends in real
GDP growth from quarterly U.S. National Accounts data. We measure the peakto-trough-to-peak average rate of growth within each business cycle, starting
with the first post–World War II cycle and continuing through the current one,
which began in December 2007. This method disentangles cyclical changes in
output from long-term, secular trends in economic growth.
To determine how these secular GDP growth patterns have translated
into resources for the American middle class since 1948, we first follow the
literature and use GDP divided by the total U.S. population, or per capita GDP,
as a measure of individual well-being. GDP is the most common measure of
aggregate economic growth, and GDP per capita has historically been the most
common way to estimate how this growth is distributed to the average person
(Jorgenson, Landefeld, and Schreyer 2014). But a growing body of literature
(e.g., Atkinson, Marlier, and Guio 2016; Nolan, Roser, and Thewisen 2016)
argues that the income of the median person (i.e., the person whose income is
exactly at the 50th percentile—above half the population, and below the other
1 We use the Bureau of Economic Analysis (BEA) GDP price deflator in all this chapter’s graphs
that are related to aggregate GDP values. We use the Consumer Price Index Research Series
(CPI-U-RS) (Stewart and Reed 1999) in all the graphs related to ASEC-CPS data. We do so because
these are the standard deflators used in the income inequality literature. However, we test the
sensitivity of our results using the PCE Price Index, which is a chain-type price index. We use this
price index rather than the Chain CPI, which only begins in 1990, because it has been available
since 1947. Our main findings are not sensitive to this deflator choice.

Labor Market Policies to Sustain the Middle Class

| 105

half) is a better way to estimate how yearly resources produced via GDP are distributed to “the middle class” (box 3-1).2 This is especially true when, over time,
systematic changes in income inequality across the entire income distribution
could cloud interpretations of mean GDP.3
It is easy to envision where this middle (median) person is in the income
distribution, but it is far more difficult to identify this person’s income using
aggregate data. The share of GDP going to specific individuals and the households in which they live cannot be directly determined from aggregate data. To
do so, the economics literature has turned to survey data.4
The longest continuous series of cross-sectional data providing information on the economic resources that, over a given year, flow to individuals and
the households in which they live comes from the Annual Social and Economic
(ASEC) Supplement to the Current Population Survey (CPS). It is a nationally representative annual survey of 60,000 households and approximately
200,000 individuals who live in them conducted by the U.S. Census Bureau.
In this chapter, for each year since 1979, we show the sensitivity of trends in
median income to alternative measures of income (e.g., wage earnings, market
income, and disposable income), of the sharing unit over which that income
is shared (e.g., only the individual who receives it, the tax unit, family, and
household), and the unit of analysis (e.g., individuals, tax units) one uses to
determine the median.
2 “Middle class” is a term of art that potentially has many definitions. Here we focus only on two:
the mean and median of the total population of the United States. Other measures are possible.
The main results given in this chapter, in figures based on data from the Current Population
Survey, are not sensitive to using the mean value of the middle quintile.
3 We use median income because it is not as sensitive as mean (per capita) income to both
changes in the income distribution over time—e.g. substantial increases in the share of income
held by the very top of the distribution—and to the under coverage of income at the two tails
of the distribution in survey data. This latter point is important because the U.S. top income
literature finds that insufficient income from top income groups is captured by the ASEC-CPS (see
Atkinson, Piketty, and Saez 2011) and the U.S. poverty literature finds that this is also the case at
the bottom of the distribution due to under coverage of government transfers (see Meyer, Mok,
and Sullivan 2015). In addition, because a major task of this chapter is to provide better ways to
measure how resources are distributed to the average American than by simply using GDP per
capita, we focus on the median American rather than some subset of Americans by age in the first
part of our analysis. We do, however, look at different labor force participation outcomes by age,
sex, and education later in the chapter.
4 Efforts are now ongoing to do so by assigning aggregate data from the National Accounts to
administrative tax record data; see Piketty, Saez, and Zucman (2016). However, as detailed
later in the chapter, such efforts still rely on the tax unit as the sharing unit. They also do not
follow the Canberra Group’s (2011) criteria with respect to capturing all persons in the tax unit,
including dependents, so that a country’s entire population is accounted for. See Larrimore et al.
(2017), as well as Auten and Splinter (2017), for recent critiques of the methods used by Piketty
and Saez (2003) and Piketty, Saez, and Zucman (2016). In their critiques, the authors begin with
the personal income tax record–based data used by Piketty and Saez (2003). But they then
substantively add to these data. This allows them to first make like-to-like comparisons and then
to show the sensitivity of Piketty and Saez–style results to alternative measures of income.

106 |

Chapter 3

Box 3-1. Who Is in the Middle Class?
Despite its frequent use in policy debates, there is no universally agreed-upon
definition of the term “middle class.” According to the U.S. Census Bureau,
in 2016 the median household had $59,039 in income (Semega, Fontenot,
and Kollar 2017). That is, half of all U.S. households had incomes below this
amount, while the other half had incomes above this amount. Meanwhile, in
2016 the average (mean) income of American households was $83,143. This
average is higher than the median income because the distribution is not
symmetric, as is evident in figure 3-i.
Although the median identifies the household exactly in the middle of
the income distribution, it does not identify which other households should
belong to the middle class. One approach is to include the middle quintile—
the 10 percent of households directly below the median and the 10 percent
of households directly above it. In 2016, the middle quintile in the U.S. had
roughly $45,000 to $75,000 in annual income; 40 percent of U.S. households
had income less than the lower end of the range ($45,000), and 40 percent had
income more than the top end of the range ($75,000).
It is notable that this statistical definition of the middle class is much
narrower than Americans’ self-perceptions of their own economic situations.
According to a study by the Pew Research Center (2015), almost 90 percent
$"0- 
тҊ$ѵ ѵѵ
 *0. #*'
 )*( 
$./-$0/$*)
$)
спрх
Number of households (in thousands)
рпѶппп

 $)ѷ ‫ڦ‬фшѶптш

 )ѷ
‫ڦ‬чтѶрут

чѶппп

хѶппп

уѶппп

сѶппп

п

п

сф

фп
цф
рпп
рсф
рфп
Income range (thousands of dollars)

рцф

спп‫۔‬

*0- ѷ ѵѵ
 ).0.
0- 0ѵ
*/ ѷ
$' 
,0$)/$' 
-)" җ.# 
4 ''*2Ҙ
$.
)
 ./$(/ ѵ
/
- 
-$"#/Ҋ ).*- 
/
‫ڦ‬сппѶпппѵ

Labor Market Policies to Sustain the Middle Class

| 107

of Americans consider themselves to be either lower-middle, middle, or
upper-middle class, with almost half (47 percent) considering themselves to
be exactly middle class (figure 3-ii). Only 1 percent consider themselves upper
class, according to Pew, while 10 percent consider themselves lower class.
$"0- 
тҊ$$ѵ
 '!Ҋ +*-/ 
*)*($
'..
*!
( -$).
җ+ - )/ҘѶ

спрф
*2 -
'..
++ -
($' 
'..

*2 -
($' 
'..
++ -
'..

$' 
'..

р‫ڿ‬

рр‫ڿ‬
рр‫ڿ‬

уц‫ڿ‬
уц‫ڿ‬

сш‫ڿ‬
сш‫ڿ‬
рп‫ڿ‬
рп‫ڿ‬
рп‫ڿ‬

*0- ѷ  2
 . -#
 )/ -
җспрфҘѵ
Note: Responses of “do not know” are not shown (figure percentages add up to 98%).

We find that average growth in real GDP and real GDP per capita are
lower during the current business cycle than during any other post–World War
II cycle. The story for real median income is more nuanced. The fall in median
labor earnings during the Great Recession (2007–9) was deeper than during all
other recessions after World War II, but government tax and transfer policies
greatly offset this decline. Using our preferred measure of median income—the
household size–adjusted, after-tax and after-transfer income of persons (real
disposable income)—we find that it took nine years (2007–16) before median
income returned to its 2007 level, just before the start of the Great Recession.
This feat, which has been achieved in each business cycle since 1979, was
accomplished most slowly in the current cycle.
In contrast, our median income measure that only includes the income
derived from work by all members of the median tax unit—the labor earnings
of the median tax unit—is still well below its level at the start of the business
cycle. This was also the case for this measure of median income during the

108 |

Chapter 3

2000–2007 business cycle. Hence, in 2016, the labor earnings of the median tax
unit remained substantially below its 2000 high point. This income measure,
which uses the tax unit rather than the household as both its sharing unit and
its unit of analysis, is commonly used in the tax record–based literature.
The failure of this tax record-based measure of median income to
return to 2000 levels is somewhat muted when we use our preferred measure
of median labor income—the household size–adjusted labor earnings of the
median person. Using this survey-based measure of income which uses the
household as its sharing unit and the person as its unit of analysis, median
labor income was only slightly lower at the end of the 2000–2007 business cycle
than at its beginning. And, thanks to substantial growth in 2015 and 2016, the
household size–adjusted labor earnings of the median person in 2016 are now
well above those for the Great Recession trough year of 2009 but still noticeably
below the level at the start of the 2007 business cycle.
Using shift-share analysis, Larrimore, Burkhauser, and Armour (2015)
account for changes in the household size–adjusted after-tax and after-transfer
income of persons (i.e., real disposable income). In doing so, they show that
unlike previous business cycles since 1979, the declines in the household
size–adjusted labor earnings of persons during the recession periods of the
first two business cycles of the 21st century predominantly came from declines
in employment rather than declines in earnings conditional on employment.
We come to a similar conclusion. Using aggregate statistics on employment
from the monthly CPS and earnings from the Establishment Survey, both series
maintained by the Bureau of Labor Statistics (BLS), we show that the Great
Recession was a downturn with more substantial employment effects than
previous downturns.
The lack of a return to the employment-to-population rates (where the
population is age 16 years and over) that prevailed at the start of the Great
Recession, even eight years after the end of the recession in 2009, is partially
demography-driven. We provide perspective on the important changes in
labor force participation rates by gender and age that account for declines
in the overall employment-to-population rates of those age 16 and over that
we are now observing. As the members of the Baby Boom generation, those
Americans born in the post–World War II era through 1964, have moved into
“retirement ages,” the overall participation rate has drifted downward. This
movement of the Baby Boom generation into older ages has greatly increased
the share of older persons in the overall population, and the employment rates
of older persons remain below those of younger persons. The aging of this
group first began to contribute to an overall decline of the participation rate in
2008; after the remainder of this cohort moves into retirement, the participation rate will stabilize. Partially offsetting the aging effect since 2008 are the
increased employment rates of older persons, as discussed later in the chapter.
Further, the incomplete recovery of overall employment rates to pre-2000
Labor Market Policies to Sustain the Middle Class

| 109

levels is not simply the story of the aging Baby Boom generation, because,
unlike those over the age of 55, the labor force participation and employment
rates of prime-age workers remain below those in 2007, immediately before
the Great Recession.
Because of these factors, we expect that the negative contribution of
the Baby Boom generation to GDP growth will wane over the next 10 years.
Further, policies that increase the labor force participation rate of all workers
would have a material impact on long-run economic growth. For example, a
combination of policies and economic conditions that return the prime-age
participation rate to the rate apparent in 2007 (still well below the rate apparent in 2000) would return about 1.7 million U.S. workers to the labor force over
10 years and raise the overall participation rate by 0.065 percentage point per
year, resulting in a 0.1-percentage-point increase per year in the rate of GDP
growth over the next 10 years.
In the rest of this chapter, we examine these employment dynamics more
closely, separately analyzing the labor force behavior of the population 55 and
over, of young persons (16–24), and of prime-age workers (25–54). We argue
that demography is not destiny, and that the labor force participation rates of
these groups are not only sensitive to economic growth and the accompanying
cyclical changes in the demand for their services, but also to changes in tax and
transfer policies that encourage them to increase or decrease their supply of
labor to the economy.

Trends in GDP and GDP per Capita
across the Business Cycles
Figure 3-1 reports real GDP by quarter and the average annual real GDP
growth rate for each of the nine post–World War II business cycles, beginning in the fourth quarter of 1948 (hereafter, 1948:Q4, etc.). We collapse the
double-dip recessions of 1953:Q2–1957:Q3 and 1957:Q3–1960:Q2, as well as
1980:Q1–1980:Q3 and 1981:Q3–1982:Q4, into single peak-trough-peak periods
of the late 1950s and 1980s respectively. The average yearly GDP growth
rate for each business cycle—as defined by the National Bureau of Economic
Research (NBER)—is measured from peak to peak, and the shaded areas of
figure 3-1 identify the beginning and ending peak of each business cycle.
Although a rule of thumb for the start of a recession is often described as two
or more consecutive quarters of negative GDP growth, the NBER committee’s
procedure for identifying turning points differs from this rule in a number of
ways. As the NBER (2010) committee states:
First, we do not identify economic activity solely with real GDP and real GDI
[gross domestic income], but use a range of other indicators as well. Second,
we place considerable emphasis on monthly indicators in arriving at a

110 |

Chapter 3

$"0- 
тҊрѵ
 1 '
*!
 '

)
 &Ҋ/*Ҋ &
1 -" 
 '


-*2/#Ѷ
ршуч–спрц
Real GDP (billions, 2009$)
рчѶппп

Real GDP (left axis)
Average annual real GDP growth (right axis)

Average annual real GDP growth rate (%)

спрцѷу

6.0

рфѶппп

5.0

рсѶппп

4.0

шѶппп

3.0

хѶппп

2.0

тѶппп

1.0

п
ршуф

ршфф

ршхф

ршцф

ршчф

ршшф

сппф

спрф

0.0

*0- .ѷ
/$*)' 0- 0
*!
*)*($
 . -#
җҘѸ
0- 0 *!
*)*($
)'4.$.
җҘѸ


'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .

0.$) ..
4' ѵ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!

ршфтѷс–ршфцѷт
)
ршфцѷт–ршхпѷсѶ
.
2 ''
.
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷуѶ
$)/*

.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*.
*!
/# 
'/ 
ршфп.
)
ршчп.Ѷ
- .+ /$1 '4ѵ
# 


+-$ 

 !'/*-
$.
0. 
/*
*)1 -/
0-- )/
/*
- '
*''-.ѵ

monthly chronology. Third, we consider the depth of the decline in economic
activity. Recall that our definition includes the phrase, “a significant decline
in activity.” Fourth, in examining the behavior of domestic production, we
consider not only the conventional product-side GDP estimates, but also the
conceptually equivalent income-side GDI estimates. The differences between
these two sets of estimates were particularly evident in the recessions of
2001 and 2007–2009 (for more details, see NBER 2010).

Although GDP growth has been positive over all nine business cycles,
the average yearly growth from peak to peak has greatly varied, as reported in
figure 3-2. The greatest average growth in GDP occurred during the first post–
World War II business cycle (1948:Q4–1953:Q2)—5.5 percent a year. But average
growth fell to 2.6 percent during the 1950s business cycle (1953–1960:Q2).
Average yearly GDP growth then rose to 4.5 percent during the 1960s business
cycle (1960–69:Q4), far surpassing average growth in the 1950s. Growth fell
to 3.7 percent during the 1969–73:Q4 business cycle of the early 1970s, and
further still, to 2.9 percent, from 1973 to 1980:Q1. But then came two decades
of economic growth greater than 3.0 percent. Average GDP growth increased
to 3.1 percent during the 1980s business cycle (1980–90:Q3), and to 3.3 percent
during the 1990s business cycle (1990–2001:Q1). Average GDP growth then fell

Labor Market Policies to Sustain the Middle Class

| 111

$"0- 
тҊсѵ
1 -" 
))0'
 '

-*2/#
/ .
0-$)"
0.$) ..

4' .Ѷ
ршуч–спрц
Peak-to-peak NBER business cycles
фѵф

ршучѷу–ршфтѷс
ршфтѷс–ршхпѷс

сѵх

ршхпѷс–ршхшѷу

уѵф

ршхшѷу–ршцтѷу

тѵц

ршцтѷу–ршчпѷр

сѵш

ршчпѷр–ршшпѷт

тѵр

ршшпѷт–сппрѷр

тѵт

сппрѷр–сппцѷу

сѵх

сппцѷу–- . )/

рѵу
пѵп

рѵп
сѵп
тѵп
уѵп
фѵп
Average annual growth rate during
peak-to-peak NBER business cycles (%)

хѵп

*0- .ѷ
/$*)'
0- 0
*!
*)*($  . -#
җҘѸ 0- 0
*!
*)*($

)'4.$.
җҘѸ


'0'/$*).ѵ
*/ ѷ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!
ршфтѷс–ршфцѷт
)
ршфцѷт–ршхпѷсѶ
.
2 ''

.
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷуѶ
$)/*
.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*.
*!
/# 
'/ 

ршфп.
)
ршчп.Ѷ
- .+ /$1 '4ѵ
# 


+-$ 
 !'/*-
$.
0. 
/* *)1 -/
0-- )/
/*
- '

*''-.ѵ

$"0- 
тҊтѵ  '

-*2/#
/ .Ѷ
ршуч–спрц
Quarterly real GDP growth (seasonally adjusted at an annual rate)
сп

спрцѷу

рх
рс
ч
у
п
Ҋу
Ҋч
Ҋрс

ршуф

ршфф

ршхф

ршцф

ршчф

ршшф

сппф

спрф

*0- .ѷ
/$*)' 0- 0
*!
*)*($
 . -#
җҘѸ
0- 0
*!
*)*($
)'4.$.
җҘѸ


'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .

0.$) ..
4' ѵ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!

ршфтѷс–ршфцѷт
)
ршфцѷт–ршхпѷсѶ
.
2 ''
.
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷуѶ
$)/*

.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*.
*!
/# 
'/ 
ршфп.
)
ршчп.Ѷ
- .+ /$1 '4ѵ
# 


+-$ 

 !'/*-
$.
0. 
/*
*)1 -/
0-- )/
/*
- '
*''-.ѵ

112 |

Chapter 3

$"0- 
тҊуѵ
1 -" ))0'
 '

+ -
+$/
-*2/#
/ .
*1 -

0.$) ..
4' .Ѷ
ршуч–спрц
Peak-to-peak NBER business cycles
ршучѷу–ршфтѷс

тѵч

ршфтѷс–ршхпѷс

пѵч

ршхпѷс–ршхшѷу

тѵр

ршхшѷу–ршцтѷу

сѵх

ршцтѷу–ршчпѷр

рѵч

ршчпѷр–ршшпѷт

сѵр

ршшпѷт–сппрѷр

сѵр

сппрѷр–сппцѷу

рѵх

сппцѷу–- . )/

пѵц
пѵп

пѵф

рѵп

рѵф

сѵп

сѵф

тѵп

тѵф

уѵп

Average annual growth rate during
peak-to-peak NBER business cycles (%)
*0- .ѷ
/$*)'
0- 0
*! *)*($
 . -#
җҘѸ
0- 0
*!
*)*($ )'4.$.
җҘѸ

'0'/$*).ѵ
*/ ѷ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!
ршфтѷс–ршфцѷт
)
ршфцѷт–ршхпѷсѶ
.
2 ''

.
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷуѶ
$)/*
.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*.
*!
/# 
'/ 

ршфп.
)
ршчп.Ѷ
- .+ /$1 '4ѵ
# 


+-$ 
 !'/*-
$.
0. 
/*
*)1 -/
0-- )/
/* - '

*''-.ѵ

to 2.6 percent during the 2001–7:Q4 business cycle, before collapsing to 1.4
percent over the current, ongoing business cycle (2007–present) that includes
the Great Recession years (2007–9:Q2) and the long, but relatively slow, quarterly growth in GDP through 2017:Q4, the latest quarter of data at the time of
this writing.5 As can be seen in figure 3-3, though there were several quarters
of very small positive growth (under 1.0 percent), and even two quarters of
negative growth, during the period from 2009:Q3 to 2014, quarterly economic
growth has picked up since then, and especially in 2015 and 2017.
Figure 3-4 reports average annual growth rates in GDP per capita for each
of the nine NBER-defined post–World War II business cycles described above.
Annual growth rates are lower for GDP per capita because the United States’
population has increased each quarter since 1948. But the general trend in real
growth in GDP per capita during business cycles over the nine business cycles
is the same as that of real GDP. The highest post–World War II average business
cycle growth rate in GDP per capita occurred during the first postwar business
5 The current business cycle, which began in 2007:Q4 and is ongoing as of the publication of this
report (2017:Q4), is 9 years long, approaching the longest on record, the 9.5 year business cycle of
the 1960s. The business cycle is not yet complete, and average growth rates for the cycle will rise
provided growth in the upcoming years is higher than the 1.4 percent per year apparent in Figure
3-2. But the long cycle implies that each additional year of growth will have a more muted impact
on the average than would be the case under a shorter cycle.

Labor Market Policies to Sustain the Middle Class

| 113

cycle (3.8 percent). It faltered in the 1950s, to 0.8 percent, before rising to its
second-highest point (3.1 percent) during the 1960s business cycle. Average
growth in GDP per capita again faltered in each of the two business cycles
of the 1970s, and especially during the 1973–80 business cycle (2.6 and 1.8
percent, respectively). But it then rose from this late 1970s average low during
both the 1980s (2.1 percent) and 1990s (2.1 percent) business cycles, before
once again declining during the 2001–7 business cycle to 1.6 percent—and
dramatically falling to 0.7 percent during the current, ongoing business cycle.
This is the lowest average business cycle growth rate in GDP per capita during
the postwar era.

Median Income Trends across Business Cycles
Economic growth, measured in aggregate or per capita terms, indicates progress in delivering more resources to a nation’s economy. How this progress
translates into resources for the middle class, however, depends on how gains
are distributed across the population. Household survey data are the usual
source for monitoring income and its distribution—at the household, family,
and individual levels. Each year the U.S. Census Bureau uses household survey data to derive its official statistics on income and poverty. (See Semega,
Fontenot, and Kollar 2017 for the most recent available year in this series.)
Household survey data are also the basis for cross-national comparative studies and are the source for most other distributional analyses, such as those
done by the Organisation for Economic Cooperation and Development (OECD
2008, 2011, 2015). The definitions that underlie the way that income questions
are asked in these household surveys provide best-practice measures of personal living standards.
The “income-sharing” unit that researchers choose when using these
data is virtually always the household (all persons living in the dwelling), and
the “income definition” is disposable (posttax, posttransfer) income, adjusted
for differences in household size and composition using an equivalence scale.6
The “unit of analysis” is the individual (regardless of age). Hence, median
income is based on the equivalized income assigned to each person in the

6 Size-adjusted household income accounts for economies of scale in household consumption
by dividing income by the square root of household size. This income measure is commonly
used in U.S. and cross-national studies of inequality (see, e.g., Gottschalk and Smeeding 1997;
Atkinson and Brandolini 2001; Burkhauser et al. 2011), as well as by the OECD in its official
measures of income inequality and poverty (d’Ercole and Förster 2012). It also closely matches
the adjustments for household size implied by the Census Bureau’s poverty thresholds (Ruggles
1990). This measure assumes that income is shared equally among all household members, so
each member receives the same amount for personal consumption.

114 | Chapter 3

population. Gottschalk and Smeeding (1997), d’Ercole and Förster (2012), and
the Canberra Group (2011) make the case for this standard methodology.7
A long-standing challenge to survey-based estimates is that they do not
provide a complete picture of the income distribution and its trends because
they fail to fully capture the highest incomes. In contrast, the tax-based data
used in the top income shares literature do a much better job of capturing the
highest incomes. (For the seminal article on U.S. top incomes, see Piketty and
Saez 2003; for a review of this literature, see Atkinson, Piketty, and Saez 2011.)
However, this tax data benefit is gained at the cost of being constrained to use
the definitions of income and income-sharing unit mandated by each country’s
tax administration (definitions that differ from the survey-based ones), and
being restricted to summary inequality measures that do not incorporate differences across the full income range (i.e., top income shares).
However, because we are focusing on the middle class and are using a
median rather than a mean income measure to track changes in middle-class
income, under coverage of income at the very top of the distribution is unlikely
to affect the results we report. In addition, survey data allow us to consider
various definitions of income and sharing units, and to consider different units
of analysis (box 3-2). This is not possible when using tax record–based data.
Likewise, those using tax record–based data in the standard labor economics
literature are forced to focus on the median wage earnings of workers. As a
result, they do not account for the fact that many workers live in households
that share earnings and other resources, which can lead to a misrepresentation
of the distribution of income available to all Americans.
The set of six measures of median income from the survey- and tax-based
literatures we discuss in the next subsection are all derived, as noted above,
from data contained in the unrestricted, public-use ASEC Supplement to the
CPS. This is the most common cross-sectional survey-based source of data
for those interested in measuring the incomes and income distributions of
Americans. The supplement contains a detailed questionnaire on the sources
of income for household members and is commonly used to evaluate levels
and trends of income and income inequality (see, e.g., Gottschalk and Danziger
2005; Daly and Valletta 2006; Blank 2011; and Burkhauser et al. 2011).

Six Measures of Median Income
We document median income trends on the basis of six measures which we
describe below. But in all cases we will compare trends in real median income
7 The International Expert Group on Household Economic Statistics (Canberra Group) was
convened as an initiative of the Australian Bureau of Statistics under the auspices of the United
Nations Statistical Commission. Its report was largely adopted as the standard for measuring
household income by the International Conference of Labour Statisticians. In 2011, the United
Nations Economic Commission for Europe provided an updated reference, outlining its latest
standards and recommendations.

Labor Market Policies to Sustain the Middle Class

| 115

Box 3-2. The Use of Survey Data versus Tax Record
Data to Measure Income and Its Distribution
This chapter argues that the household size–adjusted posttax, posttransfer income of the median person measure, based on U.S. household survey
data, provides a better estimate of the levels and trends in the resources going
to the American middle class than do studies that rely solely on tax record–
based data. It nonetheless recognizes that household surveys do a poorer job
of capturing the resources controlled by those at the very top of the income
distribution than do tax-based studies. However, the most recent income
distribution studies in other countries not only recognize the limitations of
both tax record–based and survey-based data, but also overcome them.
They do so by directly linking tax record data to survey data (see Burkhauser
et al., forthcoming, a, b), or they use survey data to capture the bottom part
of the distribution and tax record data to capture the top part, and then add
these separate estimations together (e.g., see Atkinson 2007; Alvaredo 2011).
Jenkins (2017) reviews this literature and provides an alternative method
from that of Alvaredo (2011).
The major advantage of linking tax record data to data from a random
sample household survey is that a tax unit is a subset of a household. So
one can, for instance, assign the income values of the top 1 percent of tax
units from the tax-based data to the top 1 percent of tax units that are
subcomponents of the households in the survey. It is much more difficult to
do this starting with the tax record data, unless one can link these tax units
to the households in which they reside. Matching efforts of this sort are still
at the experimental stage in the United States. See Larrimore, Mortenson,
and Splinter (2017) for an attempt to use the mailing addresses on tax forms
contained in Internal Revenue Service records to do so.
Another major way in which the top income literature based on tax
record data differs from the survey-based inequality literature that is not
resolved, even by linking data sets, is in the treatment of capital gains.
Contrary to the Canberra Group’s (2011) conventions, the top income literature not only includes capital gains as a source of income, but also does
so by including taxable realized capital gains rather than Haig-Simons
recommended accrued capital gains. Most household surveys follow the
Canberra Group’s conventions and do not provide information on capital
gains. Armour, Burkhauser, and Larrimore (2013, 2014) and Larrimore and
others (2017) discuss the merits of including taxable realized capital gains.
They argue that if one chooses to use a measure of capital gains as a source
of income, then it is preferable to include all accrued capital gains based on
Haig-Simons principles.

116 |

Chapter 3

using the Consumer Price Index Research Series (CPI-U-RS).8 The first is the
labor earnings of the median tax unit. This income measure only looks at one
source of market income, labor earnings (i.e., wages and salaries, self-employment income, and farm income), and uses the tax unit as both its sharing unit
and its unit of analysis. Such a measure is in the style of the tax record–based
literature because labor earnings are a component of market income and the
sharing unit is the tax unit.
The second measure is the household size–adjusted labor earnings of
the median person. Although this measure also only includes labor earnings, it expands the sharing unit to the household, estimates the equivalized
value of those labor earnings, and assigns it to each person in the household.
Burkhauser, Larrimore, and Simon (2012) first showed that because the number of tax units within households has grown over time, while the number of
people in those households has fallen, these demographic characteristics will
tend to increase this measure of median income over time relative to a tax
unit–based measure of labor earnings.
The third measure is the household size–adjusted market income of the
median person. This uses Piketty and Saez’s measure of pretax, pretransfer
market income. Its sources of income include labor earnings, interest, dividends, rents, trusts, and pension income received in retirement. But it excludes
public transfers. It does so using the household as the sharing unit and the person as the unit of analysis. Although the level of median income of this measure
will be greater than one that looks at labor earnings alone, its trend will also
depend on the relative growth of other sources of market income.
The fourth measure is the household size–adjusted pretax, posttransfer
income of the median person. This is the measure of income that the U.S.
Census Bureau uses in its household income series (Semega, Fontenot, and
Kollar 2017); it adds government cash transfers to the income measure used
in the previous series. This includes income from cash transfer programs such
as Aid to Families with Dependent Children and its successor, Temporary
Assistance for Needy Families, as well as from social insurance programs such
as Social Security and Workers’ Compensation. It excludes, however, transfers
directly tied to the tax system, such as the Earned Income Tax Credit (EITC).
It also excludes any in-kind government transfers, such as food and housing
assistance, and the value of Medicare or Medicaid insurance. The U.S. Census
Bureau reports the “pretax, posttransfer income of households” in the first
8We do so since the CPI-U-RS is the standard deflator used in the survey-based income inequality
literature. However, we test the sensitivity of our results using the PCE price index which is a
chain-type (or Tornqvist) price index, so it does not systematically overstate inflation like the
CPI-U and its variations, which are Laspeyres indices. Again, we use this price index rather than
the Chain CPI, which only begins in 1990, because it has been available since 1947. See figure 3-22
in the appendix to this chapter, where we show that though this deflator slightly increases real
growth in median income over each of the four business cycles we explore in this section, our
main findings are not sensitive to this deflator choice.

Labor Market Policies to Sustain the Middle Class

| 117

figure of its annual report (Semega, Fontenot, and Kollar 2017); it is also the
measure reported in box 3-1. But the U.S. Census Bureau, in its more sophisticated discussions of income trends, uses this fourth measure we define here,
which takes into consideration the number of persons in the household and
assigns them an equivalized income value. Because this measure adds government transfers but does not subtract government taxes, its level of median
income will be greater than one that looks at market income alone, but its
trend will depend on the relative growth of other sources of government transfers to market income.
The fifth measure is the household size–adjusted posttax, posttransfer
income (including some in-kind transfers) of the median person. This disposable
income measure more fully captures the importance of government tax and
transfer policies for the resources of the median person. It uses NBER’s TaxSim
9.3 (Feenberg and Coutts 1993) to estimate Federal and State taxes and liabilities, including Social Security and Medicare payroll taxes. In addition, it captures the market value of some in-kind transfers. The Census Bureau imputes
the value of SNAP (the Supplemental Nutrition Assistance Program—food
stamps), housing subsidies, and school lunches on an annual basis. We use
these values in our estimates. All are now generally recognized as important
resources that are primarily available to low-income households, and the
Census Bureau now includes them as resources in its Supplemental Poverty
Measure (Garner and Short 2012). Larrimore, Burkhauser, and Armour (2015)
use this measure in their analysis. Because it both adds government in-kind
transfers and tax credits (e.g., the EITC) but subtracts taxes, the level of median
income of this measure could be higher or lower than the Census Bureau’s
median (pretax, postcash transfer) income values as well as median market
income alone. Its trends will depend on the relative growth of net government
transfers to market income.
The sixth measure is the household size–adjusted posttax, posttransfer
income (including in-kind transfers as well as the insurance value of employer
provided health insurance and Medicare and Medicaid) of the median person.
Burkhauser, Larrimore, and Simon (2012) were the first to use the market
value of health insurance in a disposable income measure, in order to show the
growing importance of access to health insurance for explaining differences
between survey- and tax record–based analyses of income and its distribution.
The U.S. Congressional Budget Office (CBO), in 2012, was the first government
agency to include the market value of both government- and employerprovided health insurance in their measure of income (CBO 2013). Larrimore,
Burkhauser, and Armour (2015) include this measure in an appendix table.
Lyons (2015)—as well as Burkhauser, Larrimore, and Lyons (2017)—show its
importance for estimating the income of working age people with disabilities;
and Elwell and Burkhauser (2017) show its growing importance for measures
of median income that they estimate back to 1959. Due to the rapid growth
118 |

Chapter 3

in employer- and government-provided health insurance, its median income
values will be greater than all other measures in levels and trends.
Figure 3-5 reports the trends for these six measures of median income,
normalized to 100 percent in 1979. The values given in this figure come from
extensions of the findings of Larrimore, Burkhauser, and Armour (2015) and
Elwell and Burkhauser (2017) to income year 2016 for all measures except
the sixth which extends only to 2014. NBER’s business cycles are denoted by
alternated shading between business cycle peaks, as in figure 3-1.9 Note that,
though the total population included in each of our six trend lines is the same,
the median person in that population will not be the same person because the
income sources and sharing unit used to capture income differ.10
There are a number of similarities in the trends of five of these six
measures of income. The median value of all five income measures that use
the household as their sharing unit and the person as their unit of analysis is
greater at the ending peak of the 1980s and the 1990s NBER business cycles
than at their starting peaks. During both cycles, median income falls from its
prerecession high to a trough (with the year varying by measure). But in both
cycles, strong postrecession growth increased median income well above its
initial prerecession business cycle high.
This is not the case for the growth in the labor earnings of the median tax
unit. The median value of this measure is noticeably lower at the end of the
1980s business cycle than at the beginning. Although it recovers somewhat
from its 1984 trough, it is only at 94 percent of its 1979 high in 1989. During
the 1990s cycle, postrecession growth is strong enough to lift median labor
earnings above its prerecession high, but it only manages to return it to just
above its 1979 prerecession high in 2000, well below the other five measures
9 CPS income values are annual. We convert to quarterly values by assigning annual values to Q4
and linearly interpolating. There are two important breaks in the CPS data during this period. The
first is the well-known break in the data that occurs between income years 1992 and 1993. We
follow Larrimore, Burkhauser, and Armour (2015) and adjust for this by assuming that the entire
decrease in median income between 1992 and 1993 is caused by the improvement of CPS data
collection efforts, and therefore we decrease median income in 1992 and in all preceding years by
the same percentage. The second occurs for income year 2013. In that year CPS used past years
methods for one part of the survey population and a new method for the other part, to test the
impact of the new method on outcomes. We follow Elwell and Burkhauser (2017) and use the
median value based on these new methods for 2013 and thereafter and raise median income in
all preceding years by the ratio of median values in 2013 based on the new and the old methods.
10 The median individual for each measure will also change year-to-year. Substantial shifts in
the composition of the population, such as through the immigration of low skilled workers or
the aging of the population into retirement, may increase the share of the population living in
households with low labor earnings, reducing the household size–adjusted labor earnings of the
median person even when, over the same period, the median earnings of employed individuals is
rising. Alternatively, the increase in the share of persons living in two or three earner households
may reduce the share of the population living in households with low labor earnings, increasing
the household size–adjusted labor earnings of the median person, even when over the same period
the earnings of employed individuals is falling.

Labor Market Policies to Sustain the Middle Class

| 119

$"0- 
тҊфѵ
'/ -)/$1 
 .0- .
*! - ).
$)
 '
 $)
 )*( Ѷ

ршцш–спрх

Index (1979:Q4 = 100)

1: Tax unit labor earnings
2: Household size–adjusted labor earnings of individuals
3: (2) + Other market income
4: (3) + Cash transfers
5: (4) + Taxes and in-kind transfers
6: (5) + Health insurance

спрхѷу

руп
рсп
рпп
чп
хп
ршцф

ршчп

ршчф

ршшп

ршшф

сппп

сппф

спрп

спрф

*0- .ѷ
/$*)' 0- 0
*!
*)*($
 . -#
җҘѸ
'2 ''
)
0-&#0. -
җспрцҘѸ
0-- )/

*+0'/$*)
0-1 4Ѹ
0- 0
*!
 *-
//$./$.Ѹ

'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .
) 2


0.$) ..
4' ѵ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!

ршфтѷс–ршфцѷт
)
ршфцѷт–ршхпѷсѶ
.
2 ''
.
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷуѶ
$)/*

.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*.
*!
/# 
'/ 
ршфп.
)
ршчп.Ѷ
- .+ /$1 '4ѵ
))0'
/
- 

'$) -'4
$)/ -+*'/ ѵ
*).0( -
-$ 
 ) 3
 . -#
 -$ .
$.
0. 
/*
*)1 -/
0-- )/
/*
- '

*''-.ѵ

of income in 2000. Those focusing on the growth in the labor earnings of the
median tax units will greatly understate the actual increase in labor earnings
available to the median American during this period because that median
American lives in a household that may contain more than one tax unit. The
same is true for using growth in the labor earnings of the median worker to
make inferences about the labor earnings available to the median American.
The other five measures all take into consideration the fact that workers
live in households, not in tax units or by themselves, and that these household
members share their individual labor earnings. Some of the measures also
include other sources of market income, as well as the net returns of government taxes and transfers. All show substantially higher growth in the resources
available to the median American over these first two business cycles than
does the measure of the labor earnings of the median tax unit.
This income measure’s inconsistency vis-à-vis the other five measures
continues during the 2001–7 business cycle. The labor earnings of the median
tax unit dramatically falls from 2001 to 2004, and though it increases thereafter
until 2006, it is substantially below its 2001 value by the end of the business
cycle in 2007. It then falls precipitously during the Great Recession and does
not turn upward until 2013. Since then, it has slowly recovered, but by 2016
it was well below its value at the start the current business cycle in 2007, and
even further below its 2001 peak.

120 |

Chapter 3

However, this measure fails to recognize the social insurance value of
living in a household—which is the pooling of labor earnings over all household
members. Thus, sharp reductions in earnings from one tax unit in a household
are softened by the continued earnings of its other tax units. At the same
time, the number of people living in each household falls during this period,
so household resources are shared among fewer people. These are important
distinctions. Although our preferred measure of median labor earnings, the
household size–adjusted labor earnings of the median person, also falls at the
start of the 2001 business cycle, during the recovery years, it increases and
almost reaches its 2001 level by 2007. This measure then falls precipitously during the Great Recession and does not turn upward until 2012; it then increases
and is closer to it 2007 peak level by 2016 than is the flawed measure the labor
earnings of the median tax unit. More important, the pooling of labor earnings
in households reduces the depth of the drop in median income in the years
between the business cycle’s prerecession and postrecession peaks.
The household size–adjusted market income of the median person follows
a very similar path within business cycles. Growth by the end of the 2001–7
cycle was not enough to raise median market income above its level in 2000,
but because nonwage market income has grown faster at the median during
the 2007 cycle, this measure of median income almost reaches its 2007 pre–
Great Recession peak by 2016 and experiences a less severe drop in the years
between the two cycle peaks.
The household size–adjusted pretax, posttransfer income of the median
person, as used by the U.S. Census Bureau—which adds cash transfers to market income—closely follows the market income trends. Growth by the end of
the 2001–7 business cycle was not quite enough to raise it to its level in 2000,
but government transfers offset market income declines during the cycle, so
its interim-year declines were smaller. Because government transfers have
grown faster than market income during the current cycle, this measure of
median income finally exceeded its pre–Great Recession high by 2016, greatly
offsetting market income declines in the interim years. What is less clear is
the degree to which this observed growth in net government transfers for the
median American had negative effects on their employment, and hence on
measures of labor and market income in the previous series.
The measure household size–adjusted posttax, posttransfer income
(including some in-kind transfers) of the median person—which is recommended by the OECD, and is used by most European Union members—is also
our preferred measure of median total income, because it more fully takes
into consideration both government taxes and transfers. Doing so shows how
effective government tax and transfer policy has been in increasing the median
income of Americans and in offsetting the decline in their market income during both the 2001–7 business cycle and the present cycle. Although the growth
of median income during the first two cycles is much greater than during the
Labor Market Policies to Sustain the Middle Class

| 121

last two, this fuller measure of income is the only one that shows growth over
all four. More important, it shows that government tax and transfer policies
since 2001 have largely offset the interim-year declines in median market
income during this period.
The measure household size–adjusted posttax, posttransfer income
(including in-kind transfers as well as the insurance value of employer provided
health insurance and Medicare and Medicaid) of the median person is somewhat
controversial, because it adds the market value of health insurance provided
by employers and the government (i.e., Medicare and Medicaid) to the previous measure. However, this measure is now used by the CBO in its measures
of income and in the economics literature (see Lyons 2015; Burkhauser,
Larrimore, and Lyons 2017; and Elwell and Burkhauser 2017). Because of the
rapid growth in health insurance provided by employers and the government
since the mid-1980s, its median income trends are considerably greater than
all the other such trends shown in figure 3-5 through 2006. Median values fall
somewhat until 2008 and are flat through 2014, which is the last year of our
data.11
Figure 3-6 reports real median levels from 1979 to 2016:Q4 for our preferred measures of labor earnings and disposable income, as well as for the
U.S. Census Bureau’s pretax, posttransfer measure of income. All three use
the same household sharing unit, with household income equivalized and the
individual as the unit of analysis. All values are in 2016 dollars.12 The only difference in these measures is in the sources of income considered.
When we only consider the labor earnings of all household members,
median income was $28,203 in 1979, and rose to a peak of $33,663 in 2000, but
was only $32,320 in 2016:Q4—still below its 2000 peak. When we include other
forms of market income as well as government cash income, median income
levels increase, respectively, to $32,696, $40,148, and $40,600. When we use
our preferred disposable income measure, its 1979 median value of $27,255
is less than the value of median labor earnings. This shows both the greater
importance of labor earnings at the median and the fact that the median person lived in a household whose taxes paid to government were greater than
11 The Census Bureau discontinued its series on the market value of Medicare and Medicaid in
2015. It addition, Burkhauser, Larrimore, and Lyons (2017) argue that the Affordable Care Act’s
rules regarding community ratings of health insurance, which came into effect in 2014, by law
reduced the cost of private market health insurance to persons with above-average expected
healthcare costs. This, in turn, reduced the market value of Medicare and Medicaid to their
beneficiaries because they are now eligible for this less expensive community-rated private
market health insurance.
12 All these values are reported for the median person, and show this person’s household
size–adjusted income. Because we are using an equivalence scale of 0.5, if this person lived in a
household of four persons, his or her household income would be the square root of 4, or two
times the values reported in this figure (equivalence value = household income / square root of
the number of persons in the household).

122 |

Chapter 3

$"0- 
тҊхѵ
'/ -)/$1 
 .0- . *!
 '
 $)
 )*( Ѷ

ршцш–спрх
1: Household size–adjusted labor earnings of individuals
2: (1) + Other market income + cash transfers
3: (2) + Taxes and in-kind transfers

Real dollars (2016)
уфѶппп

спрх

упѶппп
тфѶппп
тпѶппп
сфѶппп
спѶппп
ршцф

ршчп

ршчф

ршшп

ршшф

сппп

сппф

спрп

спрф

*0- .ѷ
/$*)'
0- 0
*!
*)*($
 . -#
җҘѸ
'2 ''
)
0-&#0. -
җспрцҘѸ
0-- )/

*+0'/$*)
0-1 4Ѹ

'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .

0.$) ..
4' ѵ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!

ршфтѷс–ршфцѷт
)
ршфцѷт–ршхпѷсѶ
.
2 ''
.
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷуѶ
$)/*

.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*.
*!
/# 
'/ 
ршфп.
)
ршчп.Ѷ
- .+ /$1 '4ѵ
))0'
/
- 

'$) -'4
$)/ -+*'/ ѵ
*).0( -
-$ 
 ) 3
 . -#
 -$ .
$.
0. 
/*
*)1 -/
0-- )/
/*
- '

*''-.ѵ

transfers received in 1979. This no longer becomes the case as early as 1983.
Although median disposable income continues to be at about the same level as
median labor earnings for the rest of the 1980s, and somewhat above median
labor earnings during most of the 1990s, by 2000, at $33,950, it was only about
$300 above median labor earnings. However, thereafter, median disposable
income grows much more than median labor earnings, especially after 2007,
as taxes and transfers play a much more important role in sheltering the
median person’s household equivalized income from losses in market income
during and after the Great Recession. By 2016:Q4 median disposable income is
$35,152, exceeding its previous 2007 peak.

Decomposing Median Income Changes by Employment and
Earnings
The slow rebound of the household size–adjusted labor earnings of the median
person after the Great Recession, as reflected in figure 3-6, is remarkable—and
it indicates a departure from the experience of individuals after previous,
albeit weaker, recessions. What is the underlying source of this departure
from previous experience? There are several possible answers, but one is the
unprecedented increase in government social safety net programs. Although
this increased use of tax and transfer policies to redistribute market income
successfully cushioned the sharp decline in GDP during the Great Recession of

Labor Market Policies to Sustain the Middle Class

| 123

2007–9, as reported in figure 3-1, the subsequent negative employment effect
of these transfers could to some degree be responsible for the remarkably slow
pace of GDP growth in the recovery period since then.
For the last two business cycles pictured in figure 3-6, previous research
using shift-share analysis highlights the primacy of employment declines over
earnings declines in accounting for the drop in the household size–adjusted
posttax, posttransfer income (including some in-kind transfers) of the median
person. Larrimore, Burkhauser, and Armour (2015) calculate that declines in
the employment of heads of households and their spouses alone account for
79 percent (–3.25 percentage points, out of –4.10 total percentage points) of
the median recession-related income decline from 2007 to 2010, controlling
for all other factors. Likewise, between the recession-related years 2000 and
2003, the employment declines of household heads and their spouses on their
own would have caused median income to fall by 1.88 percentage points; it
was only because of policies that changed tax liabilities that median income
actually rose by 0.35 percentage point.
These declines in median income that are accounted for by the employment declines of household heads and their spouses in the 2001–7 and
2007–present business cycles are in contrast to earlier recessions, when such
declines were completely accounted for by declines in the earnings of male
household heads and their spouses conditional on employment.13 For example, 2.6 percentage points of the total recession-related 1989–92 fall in median
income of 3.5 percentage points were accounted for by the earnings declines of
male household heads and their spouses, while only 1.1 percentage point was
accounted for by the decline in their employment. The respective numbers are
3.9 percentage points for earnings and 2.0 percentage points for employment,
out of a total decline of 6.6 percentage points for the recession-related years
1979–82. Between both 1979 and 1982 and 1989 and 1992, median incomes
fell, while the employment and earnings of both female household heads and
their spouses rose. In the next section, we explore in more detail the possible
explanations for the long-term decline in the employment and labor force
participation of working age Americans.

13 In the CPS data, the head of a household (“the householder”) is defined as the household’s
primary earner. The household can contain both its head’s family members—i.e., those related
to the head by blood or marriage—and other unrelated individuals. One of the reasons to use the
household as the sharing unit rather than the tax unit is that an increasing share of households
contain two adults who share income and perhaps the parentage of their children but are neither
blood relatives nor married and who file separate personal income tax forms.

124 |

Chapter 3

$"0- 
тҊцѵ
" рх‫
۔‬/$*
*!
(+'*4( )/
/*
*+0'/$*)Ѷ


 .0- 
 '/$1 
/*
+ )$)"
0.$) ..
4' 
 &Ѷ

ршчп–спрц
ршчпѷр–ршшпѷт
сппрѷр–сппцѷу

ршшпѷт–сппрѷр
сппцѷу–- . )/

Ratio to business cycle peak
рѵр

рѵпф
р

пѵшф
пѵш

пѵчф
п

ф

рп

рф

сп

сф

тп

тф

уп

Quarters after business cycle peak
*0- .ѷ
/$*)' 0- 0
*!
*)*($
 . -#Ѹ
0- 0
*!
 *-
//$./$.Ѹ
 '0'/$*).ѵ
*/ ѷ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷу
$)/*


.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*
*!
/# 
ршчп.ѵ


Measures of Employment and
Earnings Using Aggregate Data
The attribution of income reductions in the most recent recessionary periods to
declines in the employment of working age Americans—detailed in Larrimore,
Burkhauser, and Armour (2015)—are consistent with the aggregate employment statistics produced by the Bureau of Labor Statistics and derived from
the monthly CPS. For each business cycle beginning with the 1980:Q1 peak,
figure 3-7 gives the quarterly ratio of employment to population for those age
16 and over as a multiple of the ratio observed at the opening cycle peak. Thus,
the value in quarter 0 is always 1, and as each recessionary period unfolds,
the value falls below 1. Viewing the data in this way facilitates comparisons of
employment and wage declines across business cycles.
For the business cycles beginning in 1980 and 1990, the ratio eventually returns to parity before the closing cycle peak, within 20 and 17 quarters,
respectively. But for the 2001–7 cycle, the ratio of employment to population
never recovered from its initial decline, and the employment declines during
the Great Recession simply layered on to these still-depressed rates. As a result,
not only is the ratio of employment to population in 2017:Q4 below the value in
2007:Q4, it also remains below the ratio at the beginning of the 2001 recession.

Labor Market Policies to Sustain the Middle Class

| 125

$"0- 
тҊчѵ
-$( Ҋ" җсф–фуҘ
/$*
*!
(+'*4( )/
/*

*+0'/$*)Ѷ
 .0- 
 '/$1 
/*
+ )$)"
0.$) ..
4' 

 &Ѷ
ршчп–спрц


ршчпѷр–ршшпѷт
сппрѷр–сппцѷу

ршшпѷт–сппрѷр
сппцѷу–- . )/

Ratio to business cycle peak
рѵр
рѵпф
р
пѵшф
пѵш
пѵчф

п

ф

рп

рф

сп

сф

тп

тф

уп

Quarters after business cycle peak
*0- .ѷ
/$*)'
0- 0
*!
*)*($
 . -#Ѹ
0- 0
*!
 *-
//$./$.Ѹ

'0'/$*).ѵ
*/ ѷ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷу
$)/*


.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*
*!
/# 
ршчп.ѵ


In addition to the sharp GDP declines during 2008 and 2009 that are
shown in figure 3-1, the Great Recession coincided with the movement of the
Baby Boom generation (those Americans born between 1946 and 1963) into
older age. Although declines in the ratio of employment to population may
have been spurred by GDP declines, the slow recovery over the ensuing years
may also be partly accounted for by the front end of the Baby Boom generation
moving past age 55 beginning in 2001 and past age 62 in 2008, ages at which
average employment rates are below those for younger workers. Yet Baby
Boomers’ aging into retirement cannot be the full story; the employment patterns for the prime-age population (25–54), shown in figure 3-8, parallel those
given in figure 3-7. The members of this group saw slightly shallower employment declines in the first eight quarters of the 2008 recession, and though their
employment rebound has been stronger, prime-age workers still have not
reached parity after 40 quarters of the current business cycle. Just as with the
full population age 16 and over, prime-age workers began the 2007 cycle with
employment rates below those at the start of the 2001–7 cycle.
In contrast, conditional on employment, real average hourly earnings
for private production and nonsupervisory workers (the longest continuous
series available) in the BLS Establishment Survey weakened only slightly after
2007, and not at all during the 2001 recession, as indicated in figure 3-9. It is
particularly important to note that these earnings values are only for employed

126 |

Chapter 3

$"0- 
тҊшѵ
 '
1 -" 
 *0-'4
-)$)".Ѷ
-$1/ 
-*0/$*)
)

*).0+ -1$.*-4
*-& -.Ѷ
 .0- 
 '/$1 
/*
+ )$)"
0.$) ..

4' 
 &Ѷ
ршчп–спрц
ршчпѷр–ршшпѷт
сппрѷр–сппцѷу

ршшпѷт–сппрѷр
сппцѷу–- . )/

Ratio to peak business cycle
рѵр
рѵпф
р
пѵшф
пѵш

п

ф

рп

рф

сп

сф

тп

тф

уп

Quarters after business cycle peak
*0- .ѷ /$*)'
0- 0
*!
*)*($
 . -#Ѹ
0- 0
*!
 *-
//$./$.Ѹ

'0'/$*).ѵ
*/ ѷ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷу
$)/*


.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*
*!
/# 
ршчп.ѵ
-)$)". - 
 !'/ 
0.$)"
/# 
*).0( -
-$ 

) 3
!*-
-)
" 
-) -.
)
' -$'
*-& -.ѵ

workers, and the disproportionate job loss at the lower end of the income
distribution (Elsby, Hobijn, and Sahin 2010; Farber, Hall, and Pencavel 1993)
implies a degree of selection into those reporting hourly earnings in the relevant surveys. Still, the return of average hourly earnings to their pre-recession
level after only four quarters of the current recovery implies that the sluggish
pace of median income growth since 2007 is—as Larrimore, Burkhauser, and
Armour (2015) found—the outcome of sluggish employment patterns rather
than earnings conditional on employment. This is in contrast to the failure of
real average hourly earnings to ever return to their average level at the start
of the 1980s business cycle and for the first 26 quarters of the 1990s business
cycle.

Addressing America’s Employment Challenge
The employment patterns highlighted in figures 3-7 and 3-8, and the resulting weakness in the labor earnings of the middle class, are of critical policy
importance. In the remaining sections of this chapter, we take a deeper look
at the labor market attachment of the noninstitutionalized civilian population
age 16 and over. Our analysis focuses on labor force participation rather than
employment per se, in order to more closely match the academic literature.
Participation rates also have the benefit of capturing job seeking as well as

Labor Market Policies to Sustain the Middle Class

| 127

Box 3-3. Women’s Labor Force Participation after World War II
The emergence of women from World War II with increased professional
ambitions has been well documented (Mosisa and Hipple 2006; Shank
1988; Acemoglu, Autor, and Lyle 2004). The postwar rise in the female labor
force participation rate—from 33 percent in 1948 to a high of 60 percent in
1999—thus represented a steady increase in women’s ability to manage both
career and family. Indeed, this change was largely attributable to married
women’s engagement with the labor force (Smith and Ward 1985; Costa
2000); previously, these women had left the labor force at the time of marriage
or childbearing, rarely to return. In 1950, only 10 percent of married women
age 15–64 with a child younger than five years participated in the labor force,
compared with 28 percent without a young child (Costa 2000). However, by
1998 the share of married women age 15–64 with a child under the age of five
who were in the labor force had risen to 64 percent; and for those without a
young child, the labor force participation rate had risen to 76 percent.
In the postwar period, until the 1960s, the rise of women’s labor force
participation was driven by their return to the workforce at the end of their
childbearing years. However, starting in the 1960s, their overall rise in labor
force participation is attributable not to their returning to work after their
childbearing years but by their increasing participation during these years
(Bailey 2006, 2013). In the 1960s, women’s increasing use of contraceptive
technology, which gave them more control over the timing and frequency
of births, was a critical part of these changes, as were changing social norms
related to working mothers (Bailey 2006; Goldin and Katz 2002). Their access
to family planning enabled women to pursue higher levels of education,
invest in other forms of human capital (e.g., on-the-job training), and participate in traditionally male-dominated professions (Bailey 2013).
Although, through the end of the 20th century, women increasingly
found their way into the labor force, and stayed there through their childbearing years, the female labor force participation rate peaked in 1999, at 60
percent. Since then, this rate has edged downward, to 57.0 percent in 2017,
paralleling the trend for men. However, 2016 marked the first year since 2008
(during the Great Recession) that the labor force participation rate for women
increased (by 0.1 percentage point from the previous year). The general
decline over the past 20 years is partly attributed to Baby Boomers aging into
age brackets with traditionally lower participation rates, along with increases
in teenagers’ school enrollment (Toossi and Morisi 2017). But it is also due to
declines in the probability of employment among married mothers (Cohany
and Sok 2007), many of whom saw marginal tax rate hikes in recent years.
Despite these declines, American women are far more likely than women in
the other OECD countries to work full time and to be in higher-level positions,
such as managers and professionals (Blau and Kahn 2013).

128 |

Chapter 3

employment, providing a measure of labor force attachment beyond binary
employment status. Labor force participation rates will be a bigger overstatement of employment early in business cycles, when GDP first begins to decline
and unemployment grows, relative to later points in the same cycle, when
unemployment rates have declined from their recessionary peaks.
The labor force participation rate of the U.S. noninstitutional civilian
population in 2017 was 62.8 percent, the culmination of a multiyear decline
beginning in roughly 1990, although the decline for men began much earlier
(see figure 3-10); the overall rate has stabilized over the past three years. In
earlier years, from 1948 through roughly 1998, the steady decline in the labor
force participation rate of men was offset by a rapid increase in the rate for
women (box 3-3). But after the late 1990s, the participation rate of women stabilized and then declined, serving to reinforce the decline for men and inducing
a reduction in aggregate participation rate.
The aging of the U.S. population is an important factor in the labor force
participation patterns shown in figure 3-10. The Baby Boom generation has
aged into retirement, though increased life expectancies for men and women
alike have served to increase the size of the postretirement population. A
clearer vision of the employment crisis, therefore, requires examining participation within age groups. Beginning with 1970 (the opening year of the 1970–73
business cycle), figure 3-11 shows the annual participation rate by age group.
One clear pattern in the post-1969 data is the rise and then post-1988 fall
in labor force participation for workers age 16–24. (This decline is discussed
more fully below.) In contrast, prime-age workers, those between 25 and 54
years of age, steadily increased their participation rate through 1997, experiencing a rise after 1970 of 10 percentage points to a series high of 84.1 percent,
matched again in 1998 and 1999. During this period, women continued to
join and persist in the labor market, but after 1999, prime-age participation
faltered, and the rate in December 2017 was 81.9 percent, reflecting a rate
of 89.0 percent among prime-age males and 75.0 percent among prime-age
females. Participation patterns for prime-age workers are discussed in more
detail below.
Finally, the population age 55 and older, the fastest-growing group
shown in figure 3-11 thanks to the effects of the Baby Boomers, has been the
bright spot in the otherwise disappointing two decades of labor force participation since 1997. For this group, the 1990s and 2000s brought returns to
participation rates not seen since the 1960s; in 2017, the rate for those 55 and
older was 1.1 percentage points higher than in 1970. Some of the increase can
be attributed to the aging of a cohort of women who experienced high rates of
participation while they were of prime age during the 1990s. But these changes
for older workers are also attributable to changes in policy toward older workers, and we discuss these policy choices and their implications for the labor
force participation rate of older workers in the next section.

Labor Market Policies to Sustain the Middle Class

| 129

$"0- 
тҊрпѵ
 *-
*- 
-/$$+/$*) / .
4
 ) җ .*)''4
%0./ ҘѶ
ршуч–спрц
Percent
рпп

 Ҋрц

шп

 )

чп
цп

*/'

хп

*( )

фп
уп
тп
сп
ршуф

ршфф

ршхф

ршцф

ршчф

ршшф

сппф

спрф

*0 ѷ 0- 0
*!
 *-
//$./$.ѵ

$"0- 
тҊррѵ
))0'
 *-
*- 
-/$$+/$*) / 
4
" 
-*0+Ѷ

ршцп–спрц
Percent
рпп

спрц

чп

сф
/*
фу
''

хп

рх
/*
су

фф‫۔‬

уп

сп

п

ршцп

ршчп

ршшп

*0- .ѷ
0- 0 *!
 *-
//$./$.Ѹ

'0'/$*).ѵ

130 |

Chapter 3

сппп

спрп

Labor Force Participation among
the Older Population
Figure 3-11 presents a striking picture of changes in labor force participation
over time for older workers: Since the mid-1980s, older workers have increasingly remained in the labor force, while younger workers have been leaving. The
overall participation rates of these older population groups are substantially
lower than those of prime-age workers, imposing a demographic drag on the
aggregate participation rate as the U.S. population ages. Thus, the U-shaped
pattern of decline, and then return, of participation among older workers that
is observable in figure 3-11 serves to partially offset this demographic drag.
To further examine the change in labor force participation among older
Americans, we use the CPS to subdivide participation rates for five-year age
brackets, separately examining the participation of those age 55–59, 60–64,
and 65–69. Our results extend through 2016, the latest year with the full set of
monthly CPS samples available for public use. Indexing the value in each year
to the observed participation rate in 1970 gives a measure of relative change
for each group. Figure 3-12 indicates that the decline and recovery in participation has been dramatic for the 65–69 age group and, to a lesser extent, for the
60–64 group. By 1985, the participation rate for those age 65–69 was only 62.1
percent of its 1970 level. It then rose to 114.2 percent of its 1970 value by 2016.
For those age 60–64, the decline was not as steep—a decline to 77 percent of
its 1970 value in 1986, followed by a rebound to 102.9 percent of its 1970 value
in 2016. Those age 55–59 experienced a milder decline through 1987, before
seeing an upward trend through 2008 and a plateau thereafter. For all older
age groups, the turning point for participation rates occurs sometime in the
middle to late 1980s. And for all groups, the changes are driven by women,
whose increased participation serves to mask weak participation for older men
relative to 1970 (figures 3-13 and 3-14).
The changes in labor force participation among older workers are likely
due, at least partly, to changes in policies to reduce explicit or implicit age
discrimination against older workers and to changes in private pensions. For
instance, the Age Discrimination in Employment Act of 1967 first prohibited
firms from forcing their employees to retire before age 65; and in 1986, this
act was further amended to prohibit mandatory retirement before age 70. But
it was argued at the time that the actual influence of these mandates would
have only a limited effect on the age when an employee would retire from a
firm. More important were the rules governing employees’ private pension
plans (Burkhauser and Quinn 1983; Quinn, Burkhauser, and Myers 1990). In the
1970s and 1980s, the vast majority of workers participating in a firm-sponsored
pension program were enrolled in a defined-benefit (DB) plan. These plans
frequently discouraged continued work with the firm by providing less than
actuarially fair yearly increases in benefits for those who postponed taking
Labor Market Policies to Sustain the Middle Class

| 131

$"0- 
тҊрсѵ
))0'
 *-
*- 
-/$$+/$*)
/ 
!*-
' -
*-& -.

4 " 
-*0+Ѷ
 '/$1 
/*
ршцпѶ
ршцп –спрх
Ratio relative to 1970
рѵс

рѵп

спрх

фф
/*
фш

хп
/*
ху
пѵч

пѵх

хф
/*
хш

ршцп

ршчп

ршшп

сппп

спрп

*0- .ѷ
0- 0
*!
 *-
//$./$.Ѹ 
'0'/$*).ѵ

$"0- 
тҊртѵ
))0'
 (' 
 *-
*- 
-/$$+/$*)
!*-
' -

*-& -.
4
" 
-*0+Ѷ
 '/$1 
/*
ршцпѶ
ршцп –спрх

Ratio relative to 1970

рѵч

спрх

 (' Ѷ
хф
/*
хш

рѵу

 (' Ѷ
хп
/*
ху

 (' Ѷ
фф
/*
фш
рѵп

пѵх

ршцп

ршчп

ршшп

*0- .ѷ
0- 0
*!
 *-
//$./$.Ѹ

'0'/$*).ѵ

132 |

Chapter 3

сппп

спрп

$"0- 
тҊруѵ
))0'
' *-
*- 
-/$$+/$*)
!*-
' -
*-& -.

4
" 
-*0+Ѷ
 '/$1 
/*
ршцпѶ
ршцп–спрх
Ratio relative to 1970
рѵс

спрх

рѵп

' Ѷ
фф
/*
фш
пѵч

' Ѷ
хп
/*
ху
' Ѷ
хф
/*
хш

пѵх

пѵу

ршцп

ршчп

ршшп

сппп

спрп

*0- .ѷ
0- 0 *!
 *-
//$./$.Ѹ

'0'/$*).ѵ

their benefit beyond the earliest age it was offered (Burkhauser 1979; Kotlikoff
and Wise 1987).
The introduction of the Employee Retirement Income Security Act in
1974, with new standards designed to better protect employees and ensure
that they would retire with a pension, mandated that employers more “fully
fund” their DB plans. Firms responded to this and other workforce changes
by shifting to defined-contribution (DC) plans. These new DC plans were “fully
funded” in the sense that employees received an employer-provided payment
to their own personal retirement fund each year, but they shifted the investment risk of these funds to the employee. It is particularly important that
unlike DB plans these DC plans had no built-in incentive for early retirement.
In addition, in 1978 Congress allowed employee contributions to 401(k) DC
plans to be paid with pretax dollars, which further encouraged the switch to
DC plans. In 1977, 65.8 percent of active participants in employer-sponsored
retirement plans were in DB plans and 34.2 percent were in DC plans. By 2014,
the rates had more than reversed, to, respectively, 16.1 and 83.9 percent (U.S.
Treasury 2010; U.S. Department of Labor 2017). This growth in DC plans may
have significantly increased labor force participation rates over what they

Labor Market Policies to Sustain the Middle Class

| 133

would have been under a DB system (Leonesio et al. 2012; Purcell 2005; Juhn
and Potter 2006).14
In like manner, it has been argued that the rules governing a country’s
government run pension system, for example the Social Security program
(Old-Age and Survivors Insurance) in the United States, offer incentives to leave
the labor force early even while providing protection against income loss in
retirement (National Research Council 2001). As with private DB pension plans,
public plans can have an important effect on work incentives. The age at which
benefits are initially available, called the early retirement age, coupled with the
patterns of benefit accrual, help determine the continuation of workers in the
labor force. The key element of benefit accrual is the adjustment to benefits if a
person works for another year; if the “actuarial” adjustment effectively offsets
the fact that they are received for fewer years this will not have an added effect
on labor force participation past early retirement age. However, if there is no
adjustment, or if the adjustment is not large enough to offset the fewer years
of receipt of benefits, the result is an incentive to leave the labor force. In many
countries, disability and unemployment insurance programs effectively provide for early retirement before the explicit early retirement age. Gruber and
Wise (1999) provide the first systematic evaluation of the incentives effects of
major industrialized countries publicly provided pension systems.
In the United States, the 1983 Social Security Amendments initiated a
plan to gradually increase the full retirement age from 65 to 67 years of age.
Although early retirement at age 62 was still permitted, this increase in the
normal retirement age effectively lowered the yearly benefits from doing
so. Beginning with those born in 1938 or later, the full retirement age will
gradually increase over 22 years until it reaches 67 for individuals born after
1959.15 The 1983 Social Security Amendments also phased in an increased
reward (i.e., provided a larger “actuarial” adjustment) for delaying entitlement
receipt (Delayed Retirement Credit) beyond the age of full retirement between
1987 and 2005, which further encouraged retirement delays (Gustman and
Steinmeier 2006). Blau and Goodstein (2010) estimate that the combination
of increases in the full retirement age and rewards for delayed entitlement14 Because the upturn begins around 1988, the upward trend in the labor force participation of
older persons predates the welfare reform of the late 1990s, implying little ability of this policy
change to explain the participation rates of the older population. Participation rates also appear
to be continuous over the 2008–9 period, implying a limited role for increased participation as a
result of retirement asset value destruction during the Great Recession. For further discussion,
see Gustman, Steinmeier, and Tabatabai 2010; and Daly, Hobijn, and Kwok 2010.
15 The amendments implemented the increase in the full retirement age to 66 years in two-month
increments annually for age cohorts born between 1938 and 1942, and the full retirement age
of 66 years holds for the 1943–54 cohorts as well. For individuals born between 1955 and 1959,
the age continues to incrementally increase by two months a year so that, for those born in 1960
and thereafter, normal retirement age will be 67. This later six year transition from a normal
retirement age from 66 to 67 began in 2017 for those who reached age 62 in that year.

134 |

Chapter 3

taking may have accounted for one quarter to one half of the increase in labor
force participation for men age 55–69 since the 1980s, or an increase in their
labor force participation rate of 1.2 to 2.4 percentage points. In addition, policy
changes to the Social Security retirement earnings test enacted in 2000 may
have further removed financial disincentives to work past normal retirement
age. The Senior Citizens Freedom to Work Act eliminated the Social Security
earnings test for individuals between the full retirement age and 70 years of
age. This amounted to the removal of a tax on earnings in the form of lost
Social Security benefits in each quarter worked. Its removal in 2000 would
have further increased labor supply of older employees in the years before the
actuarially adjustment over these ages fully offset the delay in taking benefits.
Gustman and Steinmeier (2009) estimate that these various changes increased
full-time employment for those age 65–67 by about 9 percent between 1992
and 2004, and that the changes in the Social Security rules may account for
one-sixth of the observed increase in labor force participation for men age
65–67 between 1998 and 2004.16
Because the participation rates of older age groups were paralleled
by those age 55–59, a group that is largely unaffected by changes in Social
Security rules, nonpolicy causes are also likely to explain part of the overall
participation increase for this older age group.17 For example, improved health
among this population leads to both a greater physical and mental capacity
to participate in the labor market, as well as a longer life expectancy, over
which older Americans will need to fund living expenses during retirement.
At the same time, the changing nature of work makes the physical health of
America’s population less relevant for participation; even with no change in
average health conditional on age, successive cohorts of Americans will find
it more physically possible to continue working at older ages because their
occupations are decreasingly dependent on physical work capacity (Maestas
and Zissimopoulos 2010). An increasing rate of self-employment and part-time
bridge jobs among older workers also gives them the flexibility they desire to
gradually consume more leisure as they age while retaining their participation

16 Song and Manchester (2007) note that the impact of the test’s removal is uneven across the
distribution of employees’ earnings. For example, the effect on incomes in lower percentiles is
statistically insignificant, while for incomes in the 50th to 80th percentiles, the effect is both large
and significant. This indicates that removing the earnings test only has an impact on earnings
levels above a certain threshold.
17 However, two important caveats are in order in this regard. First, to the extent that partners
make retirement decisions jointly, Americans age 55–59 may be affected by changes to Social
Security through the effects on their partners (Coile 2004). Second, the economic return to
working in any given period includes the option value over working in future years, given the
State-dependence of labor force status; those who work today are more likely to be employed
in the subsequent year. As a result, choices about whether to work at younger ages are not fully
independent of changes in policy affecting the return to work at higher ages.

Labor Market Policies to Sustain the Middle Class

| 135

(Karoly and Zissimopoulos 2004; Maestas and Zissimopoulos 2010; Cahill,
Giandrea, and Quinn 2013).

Policy Options to Promote Participation
among Older Americans
Unlike younger age groups, the labor force participation rates of older age
groups have increased substantially in recent decades, with policy changes
playing an important role. Nonetheless, continued progress might be made
with additional reforms. A combination of increasing the actuarial adjustment
in benefits for those who delay taking Social Security benefits past normal
retirement age and eliminating the earnings test for this age group effectively
raised labor force participation beyond normal Social Security retirement age.
While those who postpone retirement between age 62 and normal retirement
age do receive an actuarial adjustment that is approximately actuarially fair,
they still face an earnings test. The interaction of the earning test with the
actuarial adjustment process over this age range is extremely complicated
to understand. The test applies to individuals who (1) claim benefits before
they reach the normal retirement age, (2) continue to work, and (3) earn
above a certain limit. Individuals who meet these criteria face a temporary
reduction in monthly benefits. The benefits withheld under the earnings test
are deferred, not forfeited, and future payments are adjusted to be actuarially
neutral. Additionally, each year after the initial claim for benefits, new income
earned by the employee is reflected in a recomputation of the average indexed
monthly earnings and a potential increase in the primary insurance amount.
A 2016 report by the Government Accountability Office (GAO 2016) finds
widespread confusion about how the earnings test works and what purpose
it serves. Survey results show that more than 70 percent of individuals aged
52–70 who were eligible for Social Security benefits incorrectly believe that the
reductions in monthly benefits are permanent. These misconceptions likely
lead individuals to incorrectly view the earnings test as an incentive to retire or
reduce their earnings so they will stay below the test’s threshold. Indeed, the
earnings of many claimants cluster around this threshold, suggesting that they
believe any additional earnings will be permanently lost. The GAO suggests
that more and better information should be provided to individuals to help
overcome these misunderstandings. But entirely eliminating the test could be
more beneficial to society overall since it would increase labor supply even if it
meant more people would take Social Security benefits while doing so. Since
the deferred benefits are approximately actuarially fair there should not be
much of an effect of eliminating the earnings test on the Social Security Trust
Fund.
Additional policy changes beyond Social Security rules might serve to
further encourage work at older ages. Clark and Morrill (2017) provide one

136 |

Chapter 3

policy option: employees who reach the normal retirement age and have sufficient work history could have the option of opting out of continued benefit
accrual in the Social Security system. Individuals would receive Social Security
(computed through the normal retirement age) and Medicare, but they would
not pay further Social Security taxes into the system or receive enhanced benefits from additional work. These employees already paid their “fair share” into
the system, and they would continue to receive benefits based on their work
history before the normal retirement age. Still, any changes should be weighed
against opportunities for tax expenditures to affect employment probabilities
of other demographics, as well as the impact of the changes on the solvency of
Social Security trust funds.
Other suggested changes could be more difficult to enact. For example,
employers could be interested in retaining older employees in more limited
work hours by changing their job and reducing their pay, including changes to
the hourly rate, given productivity adjustments in part-time work and declining productivity at older ages. But these actions put employers in jeopardy of
violating age discrimination laws, and employers could instead opt to encourage full retirement (Clark and Morrill 2017). But new research indicates that
older employees who can find suitable work arrangements increasingly want
to combine work with leisure at older ages (Ameriks et al. 2017). This suggests
that policy nudges that induce employer-based changes to work arrangements
could help encourage additional labor force participation among America’s
oldest employees.

Labor Force Participation among
Teenagers and Young Adults
Examining the labor force participation rates of teenagers and young adults
is more nuanced than for older adults. Teens and young adults are at prime
human capital-building ages. For some youth and young adults, investments in
human capital via formal schooling will provide the greatest long-term return.
But for others, and in particular lower-income youth, work experience may be
a more valuable form of human capital development. Employment at younger
ages has been shown to affect the aspirations of youth and their academic
achievement, in addition to reducing their participation in violent or delinquent behavior and providing them with noncognitive skills (Duckworth et al.
2007; Heckman 2008; Lillydahl 1990; Mortimer 2010; Heller 2014; Modestino
2018; Kautz et al. 2014). For these and other reasons, early labor force attachment may serve to heighten longer-term labor market outcomes (Carr, Wright,
and Brody 1996; Painter 2010; Ruhm 1997).
Compared with older workers, teens (age 16–19) and young adults
(20–24) have exhibited much weaker patterns of participation relative to their
participation in 1970, particularly during the past 20 years. As figure 3-15 makes

Labor Market Policies to Sustain the Middle Class

| 137

$"0- 
тҊрфѵ
))0'
 *-
*- 
-/$$+/$*)
/ 
!*-

0'/.
4
 3
)
" 
-*0+Ѷ
ршцп–спрц


).
)
*0)"

Percent
рпп

спрц
' Ѷ
сп
/*
су

чп

 (' Ѷ сп
/*
су
''Ѷ
рх
/*
су

хп

' Ѷ
рх
/*
рш
 (' Ѷ
рх
/*
рш

уп

сп

ршцп

ршчп

ршшп

сппп

спрп

*0- .ѷ 0- 0
*!
 *-
//$./$.Ѹ

'0'/$*).ѵ

clear, the participation rates of teens and young adults (those age 16–24) rose
steadily during the 1970s and remained steady through the 1980s, before experiencing a steady, 20-year decline that abated in 2010. Within this broader age
group, the most dramatic declines have occurred for those age 16–19 and are
particularly sharp for males; as a result, the rates for males and females in this
youngest age bracket have fully converged.
Over the full post-1970 period, the labor force participation rate for
young women age 20–24 has risen, although exhibiting a leveling off and a
small decline during the past several years. Rates for males in this age range
hovered between 84.5 and 86.4 percent from 1973 through 1989, before declining steadily to 74.1 percent in 2017.
The period from 2001 through 2017 represents a particularly sharp
change in the labor force participation rates of young adults and, especially,
teenagers. This period contains two substantial macroeconomic downturns
which likely reduced the labor market attachment of teenagers. For young
Americans without access to college, or without plans to attend college, work
experience is a vital tool for human capital development, and the decline
in labor force participation shown in figure 3-15 also suggests a weakening
of those opportunities. But research has also indicated that the decline in
participation among young Americans has been driven by noncyclical factors.
The decline may reflect a crowding-out from an increased time investment in
developing human capital in other ways, including extracurricular activities
138 |

Chapter 3

(perhaps as a means to boost college application competitiveness) while in
high school and increased college-going. In the analysis below, we find that
the increased school attendance of teens has offset some, but not all, of the
decline in working hours, while young adults appear to have seen no net
change in time devoted to work over this period.
A substantial body of literature points to the “idle rate” of teens, in particular, as a way of controlling for school enrollment when examining employment rates (Neumark 1995; Modestino 2013). If members of this age group have
reduced their attachment to the formal labor market, but they have simultaneously increased their investment in secondary or postsecondary education
(the “crowding-out” effect), then no change in the idle rate would be apparent.
And reduced labor force participation that comes with human capital development through schooling might be an optimal outcome from the perspective of
lifetime welfare and, ultimately, for economic growth. The literature finds little
evidence of an increase in the idle rate of teens; reductions in their employment have for the most part coincided with increases in their school attendance. Figures 3-16 and 3-17 show these changes for teens and young adults,
separately, for the period from 2001 to 2016. The figures subdivide the teen and
young adult population in each year into those employed, unemployed, not in
the labor force but attending school, and a category for “idle” individuals who
are neither in school nor participating in the labor force. (Individuals who are
employed or unemployed while in school are categorized by their employment
status.) As the figures make clear, the idle group has been a relatively constant
share of young adults during the past 15 years; the idle rate rose minimally
from 11.2 to 11.7 percent between 2001 and 2016. Teens exhibit more evidence
of increased idleness; their idle rate in 2016 was 10.2 percent, compared with
8.8 percent 15 years earlier.
However, other data indicate that a relatively constant idle rate for young
adults and an uptick in the rate for teens may not provide a complete picture of
the time use of these Americans and may give the false impression that teens
and young adults have fully replaced time spent working with education activities. One way to see this is to chart the labor force participation rate for this age
group, conditional on school enrollment. Figure 3-18 provides the participation
rates for those enrolled in high school, in college full time, in college part time,
and unenrolled. Participation probabilities conditional on enrollment have
declined substantially since 2001. The decline in the participation rate among
those not enrolled in school, 10.1 percentage points between 2001 and 2016,
was smaller than the drop among full-time college students (10.8 percentage
points) and smaller than the drop among high school students (16.4 percentage points). Even for individuals in college part time, the fall in participation
was 9.0 percentage points.
One possible way of explaining the contradiction between relatively
stable “idleness,” as measured in figures 3-16 and 3-17, and a decline in labor
Labor Market Policies to Sustain the Middle Class

| 139

$"0- 
тҊрхѵ
(+'*4( )/
)
)-*''( )/
//0.

*!
рх- to ршҊ -Ҋ'.Ѷ
сппр–рх
Percent

Unemployed

Employed

Idle

School, NILF

рпп
чп

хп
уп
сп
п
сппр

сппт

сппф

сппц

сппш

спрр

спрт

спрф

*0- .ѷ
 )/ "-/ 
0'$
. 
$-*/
 -$ .Ѷ
0-- )/
*+0'/$*)
0-1 4Ѹ


'0'/$*).ѵ
*/ ѷ
 
( ). “not in labor force.” Enrolled ./0 )/.
2#*
- 
'*-
!*- 
+-/$$+)/.
- 

'..$!$ 
4
/# $-
 (+'*4( )/
.//0.ѵ

$"0- 
тҊрцѵ
(+'*4( )/
)
)-*''( )/
//0.

*!
сп- to суҊ -Ҋ'.Ѷ
сппр–рх
Percent

Unemployed

Employed

Idle

School, NILF

рпп

шп
чп
цп
хп
фп

уп
тп
сп
рп
п

сппр

сппт

сппф

сппц

сппш

спрр

спрт

спрф

*0- .ѷ
 )/ "-/ 
0'$
. 
$-*/
 -$ .Ѷ
0-- )/
*+0'/$*)
0-1 4Ѹ


'0'/$*).ѵ
Note: NILF means “not in labor force.” Enrolled students who are labor force participants are
'..$!$ 
4
/# $-
 (+'*4( )/
.//0.ѵ

140 |

Chapter 3

force participation conditional on school enrollment, as shown in figure 3-18,
is a rising intensity of education-related activities conditional on enrollment.
Although a declining participation rate conditional on enrollment would,
ceteris paribus, indicate an increase in idle time for teens and young adults,
students may be replacing the time they would have otherwise spent doing
market work with additional time spent on educational activities conditional
on enrollment. If so, teens and young adults have traded one set of human
capital-enhancing activities for another.
These possibilities can be explored more fully using the BLS American
Time Use Survey (ATUS) to estimate the weekly time spent on specific activities
in order to directly capture changes in time allocated to different purposes. The
ATUS data are a valuable source for understanding how Americans spend their
time, but the data set has some important limitations; the ATUS data were first
collected in 2003, and the sample sizes are small. The resulting estimates are
therefore somewhat volatile year-to-year, and we combine years of data to
achieve higher stability (BLS 2016a, 2016b).
Figure 3-19 details the changes in minutes per day spent on broad activity categories between 2003–5 and 2013–15, separately for teens age 16–19
and those age 20–24. Consistent with the differences by age group apparent in
figure 3-13, the reduction in time spent working is far larger for teens than for
young adults, who exhibit a slight increase in time spent on work during this
period. For teens, the reduction is 16.4 minutes per day over these 10 years.
For those age 20–24, the greater reductions in time use (11.8 minutes
per day) came in the “caring for others” category, which includes taking care
of family members such as children. This decline is consistent with declining
birthrates for females in this age range (Martin et al. 2017).
Both groups experienced declines in time spent on household and leisure
activities, and teens, in particular, reduced their time spent on organizational,
social, and religious activities. These minutes were replaced primarily with
personal care activities, which include sleeping and grooming, and, to a lesser
extent, with “other activities” (the set of activities that BLS does not specify
elsewhere), eating and drinking, and educational activities. Teens and young
adults dedicated an additional 6.6 minutes and 3.7 minutes to educational
activities, respectively, over these 10 years.
Thus, between 2003–5 and 2013–15, teens substantially reduced the time
they spent working and reallocated about 40 percent of that time to education, while the remainder has been dedicated to activities perhaps less critical
for longer-term human capital development. Some caution is warranted here
as increased time spent in personal care and other activities may well be
productive in ways difficult to discern from the broad categorizations in these
data and without knowing more about individual circumstances. During the
same period, the population age 20–24 does appear to have made a net time
investment in educational activities, even without a reduction in time spent
Labor Market Policies to Sustain the Middle Class

| 141

$"0- 
тҊрчѵ
))0'
 *-
*- 
-/$$+/$*)
/ 
!*-

рх- to ршҊ -Ҋ'., 4
)-*''( )/
//0.Ѷ
сппр–рх
Percent
рпп

спрх

шп

чп

*'' " Ѷ
+-/
/$(

цп

*
 )-*''( )/

хп

фп

*'' " Ѷ
!0'' /$(

уп

тп

$"#
.#**'

сп

рп
п
сппр

сппт

сппф

сппц

сппш

спрр

спрт

спрф

*0- ѷ
0-- )/
*+0'/$*)
0-1 4Ѹ

'0'/$*).ѵ

$"0- 
тҊршѵ
#)" 
$)
$( 
+ )/ *)
4+ .
*!
/$1$/$ .
!*-
 *+' 

" 
рх
/*
рш
)
сп
/*
су
 -.Ѷ
сппт–ф
/*
спрт–рф
" 
рх–рш

" 
сп–су

Change in time spent (minutes per day)
 -.*)'
-

цѵф
–фѵу
–уѵц

*0. #*'

хѵх
тѵц
фѵс
тѵч

0/$*)'
/$)"
)
-$)&$)"

-$)"
!*-
*/# -.

–рѵш

–ррѵч

–сѵш
–фѵф
–тѵш
–рѵп

$.0-")$5/$*)'
*-&
)
2*-&Ҋ- '/ 

–рхѵу

рѵт

/# -

фѵф
Ҋсп

Ҋрф

Ҋрп

Ҋф

п

*0- .ѷ
( -$)
$( . 
0-1 4Ѹ

'0'/$*).ѵ
*/ ѷ
-")$5/$*)'
/$1$/$ .
$)'0 
/#*. 
/#/
- 
.*$'
*-
- '$"$*0.ѵ


142 |

Chapter 3

рпѵп

ф

цѵч
рп

on work or work-related activities. This gain has been largely at the expense of
time spent on leisure and caring for others in the household. The results for all
young adults age 20–24, shown in figure 3-19, stand in contrast with those for
males age 21–30 who are not enrolled full time in school, as documented by
Aguiar and others (2017), who find that leisure time has grown for this subset of
the population, and especially for video game playing, a subcategory of leisure
time. The ATUS data do not allow for an easy identification of time spent on
social media or surfing the Internet.

Policy Options to Promote Participation
among Teenagers and Young Adults
Understanding the time use of teenagers and young adults is critical for advancing policy proposals that will encourage this population to become attached to
the labor market. Declining work hours that have been replaced with other
activities promoting human capital development signify a short-term loss
in labor force participation that will lead to longer-term gains in employee
productivity and likely reflect net positive trade-offs, both for individuals and
the U.S. economy. Such investments should be encouraged. The ATUS data
detailed above indicate that young adults are making a net investment in
education, even without reducing time spent working. But for teens, where
sharper drops in participation since 2001 are apparent, the evidence indicates
a rising level of time spent in activities outside of education and work. These
patterns are apparent across the distribution of parental income, indicating
that labor force attachment is declining among populations where the benefits
are likely highest.
An additional policy question is whether to encourage teens to increase
their labor force participation or to nudge them to reallocate the recent decline
in work hours towards formal schooling instead. But some policy levers may
bridge the gap between these two competing options. In particular, vocational
and technical education can smooth young people’s transition into the labor
force by providing work experiences while they pursue secondary education.
Since 1984, the Carl D. Perkins Career and Technical Education Act has been
the main Federal program to encourage practical work training among secondary and postsecondary students; 60 percent of the total funding provided
under the act goes to high school programs.
Research indicates that hybrid approaches combining formal learning
and work experience may be even more valuable for students, although the
evidence base is still thin. In particular, apprenticeships and work-based learning among secondary students may provide a strong boost to future labor
market outcomes (Neumark and Rothstein 2005; Lerman 2014). In contrast
to systems in other countries, the United States has few formal partnerships
between secondary institutions and employers to promote apprenticeships

Labor Market Policies to Sustain the Middle Class

| 143

or other work-based learning, although interest in these programs has been
growing. Perkins Act funds could be reallocated to develop these relationships, promoting employment for teens both while in high school and upon
graduation.

Labor Force Participation among
Prime-Age Adults
Finally, we turn to the decline in the labor force participation of prime-age
workers, a topic which has already received much empirical attention. The
decline in the participation rate for men is apparent in figure 3-20, where the
participation rate for each year is again indexed to the rate in 1970. Between
1970 and 2017, participation rates fell almost continuously and were at 91.6
percent of their 1970 level for men age 45–54, 93.6 percent for men age 35–44
and 92.1 percent for men age 25–34.
In 1970, women had not yet completed their full integration into the labor
force, and the labor force participation rate for all three of their age groups
rose steadily, peaking in 1997 for women age 35–44, in 1999 for women age
25–34, and in 2000 for women 45–54, before commencing a multiyear fall. The
declines in participation for women age 35–44 and 45–54 since 2000 have been
steeper than for men in these age groups.
The weighted sum of the six age/sex trends shown in figure 3-20 produces
the aggregate trend for all prime-aged workers over this period. The increase
in the labor force participation of women more than offset the decline in the
labor force participation of men from 1970 to approximately 1990 increasing
overall labor force participation rates. For the next decade the increased in
women’s participation approximately offset the decline for men, resulting in
constant labor force participation rates. But since then all age/sex labor force
participation groups are below their 2000 levels, producing a fall in the primeage participation rate over the past 17 years.
Because prime-age workers are historically the most productive age
group, and the most likely to be employed, much attention has been paid
to these patterns, especially regarding the causes of so-called missing men.
But despite a large body of literature on this topic, considerable uncertainty
remains about the underlying causes for these trends or the extent to which
numerous factors have interacted. More important, there is little consensus
regarding the policy remedies that could be pursued to abate this participation
decline.
In this section, we present a brief review of the literature on the causes
of the decline in labor force participation of prime-age workers, and refer readers to two previous CEA white papers related to this topic (CEA 2014, 2016).
Our review indicates that the reduction in participation is likely multifactorial,
including changes in government policy that have served to undermine labor
144 |

Chapter 3

$"0- 
тҊспѵ
))0'
 *-
*- -/$$+/$*)
/ 
!*-
-$( Ҋ" 

*-& -.
4
" 
)
 3Ѷ
 '/$1 
/*
ршцпѶ
ршцп–спрц


' Ѷ
сф
/*
ту
 (' Ѷ
сф
/*
ту
''Ѷ
сф
/*
фу

' Ѷ
тф
/*
уу
 (' Ѷ
тф
/*
уу

' Ѷ
уф
/*
фу
 (' Ѷ
уф
/*
фу

Ratio relative to 1970

рѵч

спрц
рѵх

рѵу

рѵс

рѵп

пѵч
ршцп

ршчп

ршшп

сппп

спрп

*0- .ѷ
0- 0
*!
 *-
//$./$.Ѹ

'0'/$*).ѵ

supply, along with shifts in labor demand that have reduced labor market
returns for low-skilled workers. The particular employment challenges of military spouses are explained in box 3-4.

Demand-Side Factors
Much attention has been paid to the hypothesis that there has been a fundamental change in the U.S. labor market via a reduction in the demand for
workers with fewer years of schooling, resulting in a decline in their labor force
participation (Autor and Duggan 2003; Juhn and Potter 2006; Daly, Hobijn,
and Kwok 2010). Trade and technology may be partially responsible for this
decreased demand in certain occupations, as discussed more fully in chapter 5.
The relative contribution of trade and automation or technology to the decline
in manufacturing employment is the subject of continued debate, but the net
effect has been reduced demand for lower-skilled workers and a widening of
the skill and education earnings gaps. The declining demand for less-skilled
workers is apparent in lower wages, and several studies cite the decline or
stagnation of wages for less-skilled workers as a significant factor in the decline
in the labor force participation rate of men (CEA 2016; Juhn, Murphy, and Topel
2002; Blank and Shierholz 2006; Krause and Sawhill 2017; Daly, Hobijn, and
Kwok 2010). Fifty years ago, participation rates for prime-age workers with

Labor Market Policies to Sustain the Middle Class

| 145

Box 3-4. Military Spouses in the Labor Market
Spouses of Americans who serve in the military face particular challenges
in the labor market. In 2012, more than half of active duty service members
were married, representing approximately 725,000 military spouses (GAO
2012). Although some of these spouses live abroad, the vast majority are
participants or potential participants in the U.S. labor market; 92 percent are
female (U.S. Chamber of Commerce Foundation 2017).
With military service comes frequent geographic relocation, and military families typically move every two to three years. With each move, military
spouses become “trailing” spouses, those whose labor market engagement
is constrained by the geographic mandates of their military spouse. Trailing
spouses must reset their careers in each new location, or find employment
which facilitates remote work, and employers may view military spouses
as suboptimal hires due to turnover risk, even when these spouses are
otherwise attractive job candidates. Alternatively, employers may agree to
employ military spouses, but at a lower wage, in reflection of higher turnover
expectations.
A survey conducted by the U.S. Chamber of Commerce Foundation indicates that military spouses experience both higher rates of unemployment
and lower wages than their nonmilitary peers. In 2017, 16 percent of military
spouses in the labor market were unemployed, more than three times the
rate for all American labor force participants. Unemployment costs appear to
be even higher for younger spouses, and an Institute for Veterans and Military
Families report found that 18- to 24-year-old military spouses experienced
unemployment rates of 30.3 percent in 2012. Adjustments for demographic
characteristics reduce, but do not eliminate, the employment gaps between
military spouses and other labor force participants (Lim and Schulker 2010).
Evidence from Meadows and others (2016) indicates that military spouses
earned substantially less than other labor market participants with similar
characteristics.
The Trump Administration has made a priority of supporting those who
serve both in and out of uniform. The Department of Defense houses a number of programs charged with facilitating the gainful, rewarding employment
of military spouses. Military Spouse Preference provides employment preference for spouses of active duty military members of the U.S. Armed Forces
(including the U.S. Coast Guard and full-time Reserve or National Guard)
who are relocating to accompany their military sponsor on a Permanent
Change of Station move. Relocating military spouses are also eligible for noncompetitive appointment to positions within the Department of Defense or
other Federal agencies. Military spouses also have access to individual career
coaches at no charge; to scholarships that provide eligible junior spouses with
up to $4,000 to be utilized for the pursuit of licenses; to certificates or an associate degree in portable career fields; and the Military Spouse Employment

Partnership, which includes more than 360 companies and businesses that
have committed to recruiting, hiring, and retaining military spouses.
Military spouses are more likely than other Americans to find themselves ensnared in this country’s patchwork of State-level occupational
licensing laws. Occupational licensing refers to mechanisms to impose
minimum standards (often educational standards) for entry and the ability
to continue working in an occupation. In 2016, 22.3 percent of all employed
Americans held a job-related government license, according to the most
recent survey from the BLS (2017b). About 35 percent of military spouses in
the labor force work in professions that require State licenses or certification,
and over 26 percent of military spouses work in healthcare or education, the
most licensed industries. Variation in requirements between States raises the
cost of cross-State mobility for workers in these occupations, and military
spouses are roughly 7 times more likely to move across State lines than civilian spouses are. State-level licensing requirements imply military spouses
face relicensing at every interstate move.
Progress is being made on these issues. The State Liaison Office in the
Department of Defense works closely with many organizations, including the
Council of State Governments, to promote license portability to facilitate the
ability of military spouses to gain employment in their new State of residence
for those professions that require licensing to practice a profession. The
Administration is currently pursuing an Executive Order to strengthen this
appointment authority and hold Federal departments and agencies accountable to exercise it to the maximum extent possible.

college degrees and high school degrees or less were similar; but there is now
an 11-percentage-point gap in their participation rates.
Although the narrative of declining demand has more frequently been
applied to “missing men,” it is also consistent with the labor force participation
of women starting in the late 1990s. According to the BLS, the female labor
force participation rate is higher among those with more education; between
2000 and 2015, the participation rates of women who did not attend college
fell by 7.9 percentage points, to 49.1 percent, compared with a decline of 0.6
percentage point, to 82.3 percent, for those with at least a bachelor’s degree.
Box 3-5 gives more detail on the participation of America’s mothers.

Supply-Side Factors
The decline in labor force participation resulting from low-skilled employees
dropping out of the workforce in response to falling labor demand is important, but this is an insufficient explanation for the full participation rate decline.
Coglianese (2016) summarizes the contribution of reduced labor demand for
men, in particular, finding that about 50 percent of the decline in the male

Labor Market Policies to Sustain the Middle Class

| 147

Box 3-5. Paid Family Leave and the Challenges of Working Families
An increasing number of children in the United States live in households
where both parents are employed, according to the BLS. American families
often struggle to balance the demands of employment schedules against the
time requirements of children, with women, in particular, experiencing dips
in labor force participation during prime childbearing years. Between 2014
and 2016, labor force participation rates for young women peaked at age 28,
and 74.9 percent of women of that age were in the labor force in those three
years. But for women just four years older at the time, participation rates were
2.1 percentage points lower, likely reflecting childbearing; the average age of
a first-time American mother is 26 (Martin et al. 2017). Declining rates of participation in those prime childbearing years eventually reverse, and by age 40,
when children are older, 75.0 percent of women were labor force participants.
The 30s decade employment dip is apparent in data from at least 1994, and
the magnitude has remained relatively unchanged over this period. (There is
no corresponding dip in male participation.)
Falling labor force participation rates for women in their 30s partially
reflect the intense time and care needs of children. Data on child time coverage from the 2012 National Survey of Early Care and Education for 2012 indicate that between 63 and 70 percent of childcare for newborns to 4-year-olds
in the United States was provided by parents in that year, with two-parent
families and low-income families displaying higher rates of parental time
coverage than other groups.
Parental time pressures are particularly acute in the first few weeks and
years following a child’s birth or adoption, as shown in figure 3-iii. In 2016, the
labor force participation rate for women with a child under 6 was more than
10 percentage points lower than the rate for women whose youngest child
was between 6 and 17. This latter group of mothers, which includes mothers
of all ages, had participation rates remarkably similar to the rate for all primeage women. The participation rate of women with young children has been
remarkably steady since roughly 1997, varying between 62.8 and 64.9 percent
for the entire period with the exception of a dip to 61.8 percent in 2004. The
labor supply effects of young children for women are not apparent for men.
Men with young children (under 6) have higher participation rates than those
with a youngest child between 6 and 17, and higher rates than all prime-aged
men.
In recognition of the high time costs of young children, the Family
Medical Leave Act of 1993 gives American workers the right to 12 weeks of
unpaid parental leave, with exceptions for worker tenure, employer size,
and highly compensated individuals. The BLS estimates that 88 percent of
Americans had access to unpaid family leave in 2016. Paid leave is rarer. Only
15 percent of U.S. workers had access in 2016, and workers in the highest
income quartile were nearly four times more likely to have access to paid
parental leave than workers in the lowest income quartile (BLS 2017a). For the

148 |

Chapter 3

$"0- 
тҊ$$$ѵ
))0'
 *- *- 
-/$$+/$*)
/ .
!*-
*( )Ѷ

4
" 
*!
*0)" ./
#$'Ѷ
ршшу–спрх
Percent
чф

спрх
*( )
2$/#
4*0)" ./
#$'
" 
х–рц

чп
цф

-$( Ҋ" 2*( )

цп

*( )
2$/#
#$'
0) -
" 
х

хф

хп
фф
фп
ршшу

ршшч

сппс

сппх

спрп

спру

*0- ѷ
0- 0
*! *-
//$./$.ѵ
*/ ѷ
-$( Ҋ" 
2*( )
җсф /*
фуҘ
- 
0. 
.

*(+-$.*)
"-*0+ѵ
*( )
2$/#
#$'- )

0) -
рчѶ
#*2 1 -Ѷ
(4
)*/
 
+-$( Ҋ" ѵ

workers themselves and their children, paid family leave has several benefits,
including higher labor force attachment after the first year of a child’s life
and higher rates of breastfeeding, which has been shown to be beneficial
for child development (Rossin-Slater, Ruhm, and Waldfogel 2013; Huang and
Yang 2015). These policies also assist parents in smoothing income over the
postnatal weeks; for low-income families, this smoothing may carry benefits
for all children in the household (Isaacs, Healy, and Peters 2017). Employers,
for their part, must cover the costs of worker pay and benefits over the leave
period, as well as perhaps replacing those absent workers, costs which may
be easier for larger business to absorb.
The Trump Administration is committed to working with Congress to
enact a working family agenda, including policies that would extend paid
family leave to more American workers. Such an effort would have multiple
benefits, including encouraging the labor force attachment of parents in the
years following the birth of a child.

labor force participation rates is accounted for by low-skilled men, but only 20
to 30 percent of this fall is attributed to fewer market opportunities for lowskilled labor. Rather, the same study highlights a rising return to leisure time
for nonparticipating men. The next subsections examine possible contributors

Labor Market Policies to Sustain the Middle Class

| 149

to higher returns to leisure, along with other supply-side factors depressing the
labor supply among prime-age workers.

Government Transfers
As discussed earlier in this chapter, taxes and government transfers played
an important role in sheltering the median person’s household equivalized
income from losses in market income during and after the Great Recession.
The increased generosity of government transfers during economic downturns
is an important source of household income smoothing over business cycles—
but in the face of diminishing utility from income, government transfers also
serve to reduce the incentive to work, and one possible explanation for the fall
in labor force participation of prime-age workers is a rise in the value of government transfers available for those who are not in the labor force.
Mulligan (2015) finds evidence of a negative relationship between government tax and transfer programs and aggregate labor supply. Mulligan
documents an increase in both earnings and employment taxes since 2007,
including the effects of expanded unemployment benefits in 2008 and 2009
and the Affordable Care Act’s employer penalty. For example, the level of benefits offered through food stamps (SNAP) were temporarily increased as eligibility requirements were relaxed and legislative changes to the Unemployment
Insurance program allowed for an increased duration of benefits and an
increase in the overall level of benefits while States were encouraged to adopt
broader eligibility requirements. Other social safety net programs, such as
Medicaid, also adopted more relaxed eligibility rules and became more generous beginning in 2007. Mulligan (2012) finds that the expanded safety net
substantially increased the marginal tax on work, not only because some of
these programs required new employer-provided benefits (like the Affordable
Care Act) but also because they increased the tradeoff workers faced when
transitioning from the social safety net into employment. Mulligan concludes
that these expansions caused at least half the drop in hours worked between
2007 and 2009, effects concentrated at the lower end of the skill distribution.
Moffitt (2015) presents a counterpoint to Mulligan’s analysis, arguing that
the marginal tax rates faced by workers moving from the social safety net to
employment in Mulligan’s analysis are overestimated and that the actual rates
imply much smaller reductions in labor supply as a result. Similarly, Moffitt’s
review of the literature on the labor supply effects of individual programs suggest much weaker labor supply responses to these changes, although Moffitt
acknowledges that these effects are generally estimated outside recession
periods.
Notably, other authors have highlighted the success of some transfer
programs, including the Earned Income Tax Credit (EITC), in encouraging work
among single parents by subsidizing employment (Meyer and Rosenbaum
2001; Eissa and Liebman 1996; Grogger 2003). Still, the program appears

150 |

Chapter 3

to have served as a disincentive to work for married couples (Eissa and
Williamson 2004). In addition, while a study by Barnichon and Figura (2015)
on the impact of “welfare-to-work” reforms on low-income women during the
1990s—for example, the EITC’s expansion in 1993 and the introduction of the
Temporary Assistance for Needy Families program in 1996 to replace Aid to
Families with Dependent Children (AFDC)—finds evidence of increased labor
force attachment for some workers following the reforms, they show that these
reforms lowered the desire to work for some nonparticipants. This pattern may
reflect a bifurcation among previous AFDC recipients: Those who were closer to
employment responded to the enhanced work requirements and the EITC by
moving into employment; but those previous AFDC recipients who faced higher
employment hurdles moved into disability receipt—although the authors do
not test for this directly.
There is somewhat more of a consensus about the depressive effects of
Social Security Disability Insurance (SSDI) on labor force participation among
prime-age workers (Maestas, Mullen, and Strand 2013; Autor and Duggan 2003;
Autor et al. 2016; Gokhale 2014; Juhn, Murphy, and Topel 2002; French and
Song 2014). The SSDI incidence rate for all age groups, including prime-age
men, has increased during the last 50 years, and this increase cannot be fully
explained by an aging workforce or other demographic factors. Autor and others (2015) identify two major channels through which SSDI reduces labor force
participation rates: benefit receipt and long application times. SSDI applicants
are not allowed to work while receiving SSDI benefits or waiting for their
application to process. Applicants who are denied benefits endure wait times
for the applications and appeals processes averaging 26 months. Upon denial,
their return to the labor force is hampered by atrophied human capital and
the negative signal to employers of 6 months of unemployment. At the peak
of the Great Recession, approximately 2.3 percent of the prime-age population
applied for SSDI. Autor and others (2015) estimate that the depressive effects of
SSDI on employment range from 3 percent in the short run to 6.5 percent in the
long run—conditional on application. The depressive effects of SSDI on labor
force participation are particularly acute for prime-age workers, and Liebman
(2015) estimates that reforms to reduce application waiting time for SSDI could
have a benefit-to-cost ratio of 10 to 1.

Physical and Mental Health
Another explanation for the decline in prime-age participation is deteriorating
health. A recent survey finds that taking pain medication and, presumably,
physical pain, is highly prevalent among prime-age individuals not in the labor
force (Krueger 2017). Prescription pain medication is taken regularly by 44
percent of the prime-age men not in the labor force, compared with 20 percent
of those employed, and Krueger estimates that 40 percent of those men not
participating in the labor force report that the pain they experience prevents
Labor Market Policies to Sustain the Middle Class

| 151

them from obtaining a job for which they would be qualified. For prime-age
women, 35 percent of those not in the labor force report taking prescription
pain medication, as opposed to 26 percent of those who are employed. These
data imply that pain and physical limitations may be a serious barrier to work
for both genders.
In addition to physical pain, mental health issues, such as depression and
stress, have become increasingly common. Case and Deaton (2015, 2017) find
that midlife mortality rates for white, prime-age males are rising, in conjunction with alcoholism, drug addiction, depression, and suicide. The authors do
not indicate a direct link between decreased mental health and labor market
inactivity; however, it may be beneficial to further explore these trends and
possible relationships.
Beyond painkillers, several studies have concluded that the opioid epidemic, in particular, is related to declining prime-age labor force participation,
particularly for men (Mericle 2017). Krueger (2017) explores the relationship
between prescription opioids and participation rates for prime-age men and
women using 2015 county-level opioid prescription rates and county-level
labor force data from 1999 to 2001 and 2014 to 2016. He finds that prescription
opioid consumption can account for 20 percent of the decline in participation
for men and 25 percent of the decline for women. Anecdotally, employers
report difficulty finding job candidates who can successfully pass a drug test,
but data from employers on this topic are not readily available. According to
information from the drug-testing company Quest Diagnostics (2016), between
2011 and 2015, the positivity testing rate for heroin increased 146 percent for
the general U.S. workforce. For safety-sensitive employees facing mandatory
Federal drug testing, positivity testing for heroin increased 84 percent over this
period.
Ultimately, it is unclear to what extent Americans’ opioid abuse, or prescription pain medication habits, causes them to exit the labor force (Mericle
2017). It is possible that other characteristics of these drug users drive their
lower labor force participation. Moreover, drug abuse leads to other barriers to
work, such as atrophied skills and other social issues. As a result, any estimate
of the number of Americans who are absent from the labor force due to drug
addiction is not the same as an estimate of the number of Americans who
might find work if the illicit drug epidemic were to be fully curtailed. The followon effects of abuse would remain as barriers to labor force entry. That said, a
curtailment of the illicit drug problem would reduce the flow of individuals into
addiction, alleviating the long-term drag on labor force participation even if
short-run effects persist.

Geographic Immobility
High levels of unemployment concentrated in particular locations may amplify
exit from the labor force, especially when workers are unwilling or unable to
152 |

Chapter 3

move to a location with a stronger job market. People move to improve their
life circumstances, but the share of Americans moving has been declining, and
is currently at its lowest value on record (figure 3-21).
In addition to family and social connections, deterrents to moving
include (1) search time, (2) State-specific occupational licensing requirements,
(3) local land use regulations that raise housing costs in places with the greatest potential growth, and (4) homeowners’ limited ability to sell their homes.
Such obstacles to finding better employment opportunities, coupled with high
local unemployment rates, may lead employees to ultimately exit the workforce. More research is needed to determine the direct relationship between
geographic immobility and labor force participation, specifically exploring
whether obstacles to moving have increased over time, and how prime-age
workers in particular have been affected.
Emerging technologies and investments that allow workers to access
job opportunities without geographic relocation—including expanded access
to broadband technology, telecommuting, Internet-based employment, and
falling commuting times—may make geographic immobility less relevant for
labor force participation.

Alternative Uses of Time
Another consideration for the decline in labor force participation, particularly
for men, is that people may be choosing to spend more of their time outside
the workforce, perhaps substituting their spouse’s income for their own. Juhn
and Potter (2006) find that this is an unlikely scenario to explain the labor force
participation changes for prime-age men. Data on time spent doing housework
show that these increases for men have been relatively small, while labor force
losses are greater for single men or men whose wives are not employed. For
example, participation declined nearly 5 percentage points from 1969 to 2004
for single men and 7.7 percentage points for men with nonworking wives.
Conversely, men with working wives only experienced a drop of 3.4 percentage
points during the same period.

Policy Options to Promote Participation
among Prime-Age Workers
Given the myriad explanations for the decline in labor force participation for
prime-age workers, there is no ready consensus on the policies most likely to
affect change. But five main options are likely to garner more consensus than
others.
First, the decline in demand for workers with fewer years of schooling
should be met with a concerted effort to aid workers in their retraining efforts
and to increase demand for these workers. Boosts to infrastructure spending
would shore up demand for workers without college degrees and would likely

Labor Market Policies to Sustain the Middle Class

| 153

$"0- 
тҊсрѵ
#- 
*!
ѵѵ
 .$ )/.
#*
*1 
0-$)"
/# 
./
 -Ѷ
ршуц–спрх



Percentage of U.S. residents
сф

спрх

сп

рф

рп
ршуф

ршфф

ршхф

ршцф

ршчф

ршшф

сппф

спрф

*0- .ѷ
ѵѵ
 ).0.
0- 0Ѹ
0-- )/
*+0'/$*)
0-1 4ѵ
*/ ѷ
# 
*) Ҋ4 -
" *"-+#$
(*$'$/4 ,0 ./$*)
2.
)*/
.& 
 /2 )
ршцс
/#-*0"#

ршцф
)
ршцц
/*
ршчпѶ
.*
/# 
1'0 
$.
$)/ -+*'/ ѵ

result in higher wages for workers in the skilled trades, encouraging more entry
into these occupations. The Administration’s focus on apprenticeships is one
method to help bridge the gap between the skills needed by expanded infrastructure activity and the skills of today’s unemployed and labor force nonparticipants. Ensuring government financial support can be used for short-term
retraining programs (programs that do not lead to a two- or four-year degree)
may be another way to facilitate the re-entry of workers into employment.
Given weak performance evaluations of Workforce Investment Act programs in
recent years, however, demonstrated effectiveness of particular programs and
institutions should be a precursor to receive further Federal support.
Second, the geographic immobility of workers is a conundrum with no
easy solution. Encouraging internal migration is one possible tactic; but there
has been little research on how the Federal government might do so. In addition, encouraging internal migration is likely to exacerbate the struggles of
distressed communities. Indeed, the low proclivity of Americans to migrate
points to the need for local economic development and labor market connectedness without relying on workers to change their residence. Place-specific
private investment incentives, such as those proposed by Bernstein and
Hassett (2015), would resolve issues of geographically concentrated labor force
nonparticipation without requiring migration. The newly passed Tax Cuts and
Jobs Act (TCJA) is a positive step in this direction. The TCJA allows would-be
154 |

Chapter 3

for-profit investors to defer capital gains recognition, and the associated tax,
from the sale of an appreciated asset if they invest the gains in Opportunity
Zones located in “low-income community” census tracts. State governors have
90 days after the enactment of TCJA to designate their State’s Opportunity
Zones. Similarly, well-designed infrastructure investments that enable workers
to connect with jobs without moving would likely facilitate the reentry of some
workers to the labor force.
Third, curbing the opioid crisis is of critical importance for ensuring a stable or growing employment rate among prime-age workers, and curtailing the
supply of these substances would reduce addiction rates among Americans.
Even with a curtailment of the illicit drug flow into the United States, individuals who are currently out of the labor force because of opioid addiction may
struggle to reenter without additional investments in skill upgrading. Still,
progress on the opioid addiction front might stem the tide of workers into
nonparticipation and, over time, prop up the participation rate.
Fourth, there is likely a substantial opportunity to modify government
transfers to more directly encourage, rather than discourage, work. One
example of this is SSDI. Maestas, Mullen, and Strand (2013) show that SSDI
applicants with similar profiles but whose applications are rejected by a
“tough” evaluator are significantly more like to return to the labor force than
are those whose applications are accepted by an “easy” evaluator, and who
thus go onto the SSDI rolls. European countries have long since recognized that
movement onto their long-term disability rolls should be a last resort, and thus
have substantially reduced their disability beneficiary rates using “work first”
reforms. Such reforms have raised eligibility standards and have increased the
liability of employers for the disability insurance uptake of their former workers. These reforms have been successfully implemented, especially in Germany
and the Netherlands, where the combination of tighter eligibility requirements
and cost sharing with private employers has resulted in lower disability insurance uptake (Burkhauser and Daly 2011; Burkhauser, Daly, and Ziebarth 2016).
Fifth and finally, policies that encourage business formation and capital
spending could, if successful, drive up the demand for labor and wages and
increase labor force participation (CEA 2017). These policies are discussed in
chapter 1.

Conclusion
Headline annual economic growth in real GDP and per capita real GDP during
the current business cycle has been the slowest of all post–World War II business cycles. This unprecedentedly slow recovery, especially for middle-class
labor incomes, is perhaps the primary economic problem facing our Nation. A
marked increase in net government transfers offset to some degree the decline

Labor Market Policies to Sustain the Middle Class

| 155

in the median American’s household income. But without substantial increases
in economic growth, this level of redistribution is unlikely to be sustainable.
Clearly, the best possible outcome is for labor incomes to return to
normal levels of growth. Public policies, as discussed in other chapters, have
focused on increasing economic growth—and with it, increasing labor demand.
We have argued in this chapter that policies to increase aggregate supply are
also necessary to set the United States on a path of higher employment growth.
Although demographic factors can account for part of the decline in overall
labor force participation rates, they cannot account for it all, and we have
shown that demography need not be destiny. Accordingly, a combination of
policies and economic conditions that raise labor force participation rates can
materially affect overall economic growth. As a simple benchmark, a return of
the prime-age participation rate to the rate apparent in 2007 (still well below
the rate apparent in 2000) would return about 1.7 million U.S. workers to the
labor force over 10 years and raise the overall participation rate by 0.065 percentage point a year, resulting in an increase of 0.1 percentage point a year in
the rate of GDP growth.
Because past public policies are responsible for some of the drag on
employment growth, changes in these policies can be important ways to get
workers off the sidelines and again fully participating in our economy and
enjoying its benefits. A coordinated effort encouraging people to do this could
significantly reinforce the positive effects of labor demand policies.

Appendix
As discussed in footnote 8 above, our results for income are not sensitive to the
choice of price deflator. Here, the PCE chain price index is used in figure 3-22
in lieu of the CPI-U-RS in figure 3-5. The main conclusions from figure 3-5 are
not affected under this alternative measure, despite the fact that the PCE index
is a chain-type index while the CPI is Laspeyres index, which systematically
overstates inflation and therefore would understate real income growth.

156 | Chapter 3

$"0- 
тҊссѵ
'/ -)/$1 
 .0- .
*!
- ).
$)
 '
 $) )*( Ѷ

 !'/ 
.$)"

#$)
-$ 
 ) 3Ѷ
ршцш–спрх
рѷ
3
0)$/
'*-
 -)$)".
сѷ
 *0. #*'
.$5 –%0./ 
'*-
 -)$)".
*!
$)$1$0'.
тѷ
җсҘ
‫
۔‬/# -
(-& /
$)*(
уѷ
җтҘ
‫
۔‬.#
/-).! -.
фѷ
җуҘ
‫
۔‬3 .
)
$)Ҋ&$)
/-).! -.
хѷ
җфҘ
‫' 
۔‬/#
/).0-)

Index (1979:Q4 = 100)
рхп

спрхѷу

руп
рсп
рпп
чп
хп
ршцф

ршчп

ршчф

ршшп

ршшф

сппп

сппф

спрп

спрф

*0- .ѷ
/$*)'
0- 0
*!
*)*($
 . -#
җҘѸ
'2 ''
)
0-&#0. -
җспрцҘѸ
0- 0
*!

*-
//$./$. җ ҘѶ
0-- )/
*+0'/$*)
0-1 4Ѹ
 Ѷ
*).0( -
-$ 
 ) 3Ѹ
 0- 0
*!

*)*($ )'4.$.Ѷ
/$*)'
 )*( 
)
-*0/
*0)/.Ѹ 
'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .

0.$) ..
4' ѵ

*''+. 
/# 
*0' Ҋ$+
-  ..$*).
*!

ршфтѷс–ршфцѷт
)
ршфцѷт–ршхпѷсѶ
.
2 ''
.
ршчпѷр–ршчпѷт
)
ршчрѷт–ршчсѷуѶ
$)/*

.$)"' 
+ &Ҋ/-*0"#Ҋ+ &
+ -$*.
*!
/# 
'/ 
ршфп.
)
ршчп.Ѷ
- .+ /$1 '4ѵ
))0'
/
- 

'$) -'4
$)/ -+*'/ ѵ
# 
.$3
( .0- .
*!
)*($)'
( $)
$)*( 
- 
 !'/ 
0.$)"
/# 


#$)
-$ ) 3ѵ

Labor Market Policies to Sustain the Middle Class

| 157

x
Chapter 4

Infrastructure Investment
to Boost Productivity
Concerns about the state of our Nation’s infrastructure have become commonplace. We systematically face excess demand, quality degradation, and congestion when using our public assets—as, for example, on many of our urban roads
and highways. Without price signals to guide the users and suppliers of our
Nation’s infrastructure, we use our existing assets inefficiently, fail to properly
maintain them, and do not invest to add needed capacity. Furthermore, complex, overlapping, and sometimes contradictory rules and regulations deter
and delay investors from adding to or improving existing capacity.
The central infrastructure problem facing policymakers is how to resolve this
mismatch between the demand for and supply of public sector capital, both by
using our existing assets more efficiently and by adjusting long-run capacity to
efficient levels. Allowing prices to have a larger role in guiding consumption and
investment decisions will be key to achieving the positive growth and productivity effects that infrastructure assets can provide. We estimate that a 10-year,
$1.5 trillion infrastructure investment initiative could add between 0.1 and
0.2 percentage point to average annual real growth in gross domestic product
under a range of assumptions regarding productivity, timing, and other factors.
To achieve growth at the higher end of this range, we suggest four key actions
for policymakers to consider. First, the Federal regulatory structure must be
streamlined and improved—while ensuring the achievement of health, safety,
and environmental outcomes. Conflicting, unduly complex, and uncoordinated
rules and regulations can impede investments in—and significantly delay—the
delivery of needed infrastructure, an especially salient issue in the energy and

159

telecommunications sectors, as discussed in this chapter. Addressing these
issues will take time but will generate significant public benefits, and several
recent Federal actions have begun this process, including President Trump’s
August 15, 2017, Executive Order to reduce unnecessary delays and barriers to
infrastructure investment.
Second, additional resources can be secured for infrastructure investment,
turning to some combination of user charges, specific taxes, or general tax
revenues. Although public resources are important, this chapter emphasizes
the role of user fees based on marginal costs. Such user charges—which
typically are set by States and local governments and are collected from those
who directly benefit from publicly provided roads, water facilities, and other
types of infrastructure—will encourage efficiency in use, provide signals from
consumers and to suppliers about the value of future investments, and generate revenues. In the case of roads and highways, for example, fuel taxes have
historically acted as imperfect user fees, but conventional funding models are
now under pressure from rising fuel efficiency and the use of electric vehicles,
and congestion costs are high and rising in many urban areas. Innovations
such as user fees for vehicle miles traveled—as are being piloted in Oregon, for
example—and highway tolls that vary with congestion can increase efficiency
and raise needed revenues to pay for infrastructure improvements and additions to capacity.
Third, the Federal government can support the use of innovative financing
options such as public-private partnerships that will more efficiently utilize the
total capital available from the public and private sectors and lower its cost.
Well-designed financial contracts, compared with conventional procurement
methods, can result in lower project costs, shorter deadlines, higher-quality
services, and decreased life-cycle costs of provision.
Fourth and finally, policymakers at all levels of government can improve project
selection and investment allocations to ensure that the highest-value projects
are chosen and funded. Expanding the role of competitive grant programs, such

160 |

Chapter 4

as the Department of Transportation’s Infrastructure for Rebuilding America
grant program, can increase the productivity impact of any given infrastructure
investment. Further, giving State and/or local governments more flexibility in
project choice can help ensure that local projects are aligned with local needs
and preferences, and encouraging the use of cost/benefit analysis to inform
project selection can also increase the efficiency of infrastructure investments.
On balance—with appropriate regulatory policies and infrastructure funding,
along with financing provisions, in place—the United States can look forward
to a productive and prosperous 21st century.

O

ur Nation has been rightfully proud of its infrastructure—the roads,
bridges, waterways, energy facilities, telecommunications networks,
and other physical and technological underpinnings that make
possible our economic activity, trade, and commerce, both domestically and
abroad. However, recent decades have seen sustained growth in the demand
for infrastructure services that has not been met with corresponding growth in
and maintenance of their supply—so concerns about overuse, congestion, and
poor service have become common. The supply of infrastructure has failed to
keep up with increases in demand in part because much access to infrastructure is underpriced or, in many instances, provided free of charge to users,
which systematically has led to excess demand, overuse, and congestion—as,
for example, on many of our urban roads and highways.
In the private sector, congestion and excess demand for goods and
services typically cause prices to rise, signaling to consumers that they should
curtail their consumption, while these same high prices signal producers about
the value of investing and expanding production. However, in the public sector,
which funds and often directly provides much of the Nation’s infrastructure,
investment and allocation decisions are made by tens of thousands of distinct
governmental entities based on little or no price information; hence, they have
inadequate information about the expected benefits and costs of proposed
investments and allocations. Without price signals to guide the users and suppliers of our Nation’s roads, highways, waterways, and other infrastructure,
we rely on inefficient, nonprice rationing of our existing assets; do not properly maintain existing assets or invest to add needed capacity; and instead
often experience rising levels of congestion, delay, and quality degradation.
Furthermore, complex, overlapping, and sometimes contradictory rules and
regulations deter and delay investors from making capacity additions or
improvements, exacerbating the imbalance between the demand and supply
of infrastructure.
Infrastructure Investment to Boost Productivity | 161

The central infrastructure problem facing policymakers is how to resolve
this mismatch between the demand for and supply of public sector capital,
both by using our existing assets more efficiently and by adjusting long-run
capacity to efficient levels, a challenge made even more complicated by the
fragmented roles of the Federal, State, and local levels of government, and private sectors. In many cases, this will mean expanding or relocating capacity to
meet demand. However, in some cases, the opposite will be true: Infrastructure
supply can exceed demand, either overall or regionally, and the challenge will
be to reduce capacity to efficient levels while ensuring that all Americans have
access to the 21st-century infrastructure services they deserve.
In this chapter, we propose features of a more efficiently financed capacity expansion of the infrastructure for the U.S. economy. We consider not only
“core” assets—such as roads, bridges, railways, transit systems, and water and
wastewater facilities—but also telecommunications and power sector assets.
Allowing prices to have a larger role in guiding consumption and investment
decisions will be key to achieving the positive growth effects that additional
infrastructure assets can provide. We estimate that a 10-year, $1.5 trillion infrastructure investment initiative could add between 0.1 and 0.2 percentage point
to average annual real growth in gross domestic product (GDP), under a range
of assumptions regarding productivity, timing, and other factors.
To achieve growth at the high end of this range, we suggest four key
actions for policymakers to consider. First, the Federal government can take
the lead in streamlining, developing, and updating the regulatory environment
to pursue appropriate health, safety, and environmental goals without hindering innovation, especially in forward-looking technologies. As explored further
below, regulatory impediments and barriers have figured prominently in the
energy and communications sectors, and addressing these constraints will
have a positive impact on productivity and growth.
Second, additional resources can be secured for infrastructure investment, turning to some combination of user charges, specific taxes, or general
tax revenues. Although Federal resources are important, States and localities
actually fund most of the Nation’s core infrastructure. Thus, increased funding
support throughout our governments will be essential, in addition to attracting
private sector capital in sectors where most assets are privately owned, such
as telecommunications and energy. Additional general tax revenues at the
Federal, State, and/or local government levels may be appropriate, especially
for infrastructure facilities that provide benefits beyond the borders of the
investing jurisdiction, but this chapter emphasizes the role of marginal costbased user fees. Such user charges—which typically are set by States and local
governments and are collected from those who directly benefit from publicly
provided roads, water facilities, and other infrastructure—will encourage efficiency in use, provide signals to consumers and suppliers about the value of
future investments, and generate revenues. Developing and incentivizing the
162 |

Chapter 4

use of value capture programs would also increase available funding resources,
as parties experiencing capital gains (e.g., increased property values) would be
taxed to help pay for the costs of the infrastructure investment responsible for
these gains.
Third, the Federal government can support the use of innovative financing options such as public-private partnerships that will more efficiently utilize
the total capital available from the public and private sectors and promote
more efficient infrastructure delivery. Well-designed partnerships can improve
incentives to lower project costs, meet deadlines, provide high-quality services, and minimize life-cycle costs of provision compared with conventional
procurement methods.
Fourth and finally, policymakers at all levels of government can improve
project selection and investment allocations to ensure that the highest valued
projects are chosen and funded. Using tools such as cost/benefit analysis can
increase overall efficiency, because directing limited investment funds to their
most valued uses will make any given infrastructure investment that much
more productive. Further, maintaining project selection at the State and/or
local government levels can help ensure that projects with limited spillover
effects are aligned with local needs and preferences.
We also note that enhanced infrastructure spending may have implications for America’s workers to the extent that labor demand rises in infrastructure construction and design occupations and related fields. Although it
is difficult to predict the net employment impact of increased infrastructure
investment, a demand shift toward these occupations may benefit workers
in those fields. The current stock of infrastructure workers in the labor force
is disproportionately drawn from the population with a high school degree
or less, indicating that enhanced labor demand would disproportionately
benefit those with fewer years of formal education, precisely the segment of
the population where there is the most excess supply. The Federal government
can minimize any remaining labor constraints by easing occupational licensing requirements for infrastructure workers on federally funded projects and
by enhancing the retraining options for workers interested in transitions into
these occupations.
The chapter proceeds as follows. The first section documents the status
quo and the demand and supply imbalances in America’s infrastructure, and
the second section discusses the economic evidence for the value of increasing public sector capital. The third section considers the roles of Federal,
State, and local governments in undertaking the needed capacity expansions
or enhancements, with an emphasis on funding resources and financing
arrangements. The fourth section examines particular aspects of the value of
additional or enhanced capacity in the energy and telecommunications sectors
and the inland waterways system. The fifth section concludes.

Infrastructure Investment to Boost Productivity | 163

U.S. Infrastructure’s Growing
Problem of Excess Demand
Although the Nation’s transportation network, water facilities, communications sector, and energy infrastructure are the envy of many, studies and media
reports increasingly point to problems with congestion, service quality degradation, insufficient funding, fairness and affordability, and the lack of coordinated, forward-looking infrastructure management in the public sector (e.g.,
Rosenthal, Fitzsimmons, and LaForgia 2017; Gregory et al. 2017; Blakemore
2016). The American Society of Civil Engineers (2017) gave the Nation a grade
of D+ in its most recent infrastructure report card, little changed from previous
years, putting a $4.6 trillion price tag on the needed upgrading of public assets
across many sectors, including surface transportation, aviation, water utilities
and water resource management, and energy. Though specific conditions vary
across sectors and regions of the country, recent overall assessments have
identified key infrastructure deficits with real consequences for U.S. consumers
and businesses. For example, between 1980 and 2016, vehicle miles traveled in
the United States more than doubled, while public road mileage and lane miles
rose by only 7 and 10 percent, respectively (figure 4-1).
Unsurprisingly, queuing caused by traffic congestion has risen, imposing both direct and indirect costs on business and leisure travelers alike. The
national average annual congestion delay per auto commuter reached 42
hours in 2014, according to the Texas A&M Transportation Institute (TTI 2015).
TTI’s travel time index reached an all-time high value of 1.22 in 2014, meaning
that a trip that would take 30 minutes without congestion (“free flow” conditions) takes 22 percent longer—between 36 and 37 minutes—when roads are
congested. Once the value of extra travel time and wasted fuel costs are taken
into account, TTI estimates that total congestion costs were $160 billion in
2014, equivalent to 0.9 percent of GDP that year (figure 4-2). Left unaddressed,
these estimated congestion costs would total over $1.4 trillion over 10 years’
time.
Average highway congestion increased across the country, and congestion has worsened far more in some cities than it has in others. Table 4-1
indicates not only that the auto-commuter-weighted average hours of delay
per auto commuter in the Nation’s 101 largest cities rose from 33 hours in 1990
to 52 hours in 2014, but also that the range across cities widened considerably
during this period, from 61 to 76 hours.
Aside from roads and highways, congestion and service quality problems
on our waterways are also evident. Average delays at locks along the inland
waterways system have crept up, from under 1 hour per tow in 2009 to nearly
2.5 hours in 2016 (figure 4-3), despite a 9.2 percent decline in the number of
vessels served during this period. Similarly, the share of vessels experiencing a
delay has risen from a low of 34 percent during the Great Recession to a 2016

164 |

Chapter 4

$"0- 
уҊрѵ *
$' " 
)
 #$' 
$' .
-1 ' Ѷ ршчп–спрх
Index (1980 = 100)
ссп

спрх

ршф

 #$' 
($' .
/-1 ' 
рцп

руф

рсп

) 
($' .
шф
ршчп

0'$
-* ($' "
ршчф

ршшп

ршшф

сппп

сппф

спрп

спрф

*0- .ѷ
ѵѵ
 +-/( )/
*!
-).+*-//$*)Ѷ
 $"#24
//$./$.Ѷ
спрхѸ 
'0'/$*).ѵ

$"0- 
уҊсѵ
-!!$ *)" ./$*)
 .0- .Ѷ ршчс–спру
Hours

Dollars (billions)

уф

спру

уп

рчп
рхп

тф

*/'
*./
җ-$"#/ 3$.Ҙ

 '4 + -
0/*

*((0/ -

җ' !/
3$.Ҙ

тп
сф

руп
рсп
рпп

сп

чп

рф

хп

рп

уп

ф

сп

п

ршчп

ршчф

ршшп

ршшф

сппп

сппф

спрп

п

*0- ѷ
 3.
о
-).+*-//$*)
 )./$/0/ ѵ

Infrastructure Investment to Boost Productivity | 165

' 
уҊрѵ
))0'
 *0-.
*!
 '4
+ -
0/*
*((0/ -Ѷ
рпр
-)
- .
1 -"

ршшп

0/*Ҋ*((0/ -Ҋ
2 $"#/ 

1 -"
тт

ст

рпѵш

т

ху

спру

фс

ур

ртѵр

х

чс

 -

-*..
рпр
0-)
- .
/)-
 1$/$*)

$)$(0(

3$(0(

*0- .ѷ
 3.
оѶ
 3.
-).+*-//$*)
 )./$/0/ Ѹ

'0'/$*).ѵ

*/ ѷ

4 -'4
.0(
*!
''
/# 
+ -Ҋ/-$+
 '4.
!*-
/#*. 
+ -.*).
2#*
/-1 '
$)
/# 
+ &
+ -$*
җх
/*
рп

ѵ(ѵ
)
т
/*
ц
+ѵ(ѵҘѵ

# 
 1 '*+ 
- 
җi.e., with a +*+0'/$*)
 ).$/4
of (*- 
/#)
рѶппп
+ -.*).

+ -
.,0- 
($' Ҙ
2$/#$)

( /-*+*'$/)
- "$*)ѵ
# 
0-)
- 
*0)-$ .
#)" 
!- ,0 )/'4
җ 1 -4

4 -
!*-
(*./
"-*2$)"
- .ҘѶ
.*
$)- . .
$)'0 
*/#
) 2
"-*2/#
)
 1 '*+( )/
/#/
2.

+- 1$*0.'4
$)
- .
 .$")/ 
.
-0-'ѵ

high of 48 percent (USACE 2017a). Such delays can be costly; the American
Society of Civil Engineers estimated annual delay costs of $33 billion along the
system in 2010; even if delays had not increased since then, that annual cost
corresponds to a nearly $300 billion cost over 10 years’ time.
Infrastructure needs in the water and wastewater sector are also considerable. The U.S. Environmental Protection Agency (EPA) estimates that
$655 billion will be needed over the next twenty years to upgrade and replace
infrastructure in the water and wastewater sectors, comprised of $271 billion
for wastewater collection and treatment facilities and $384 billion for drinking water facilities. Concerns include water loss from water main breaks, raw
sewage discharges into local water supplies, and overall water quality. For
example, the EPA estimates the annual cost of water main breaks to be $2.6 billion, implying over $20 billion in costs over 10 years’ time. More detailed needs
assessments at the regional or local level confirm similar needs but also reflect
significant heterogeneity, because some water and wastewater utilities face far
greater challenges than that of others, especially in larger cities with declining
populations (GAO 2016).

How Increasing the Supply of Infrastructure
Supports Economic Growth
The value of adequate public infrastructure in terms of both quantity and
quality comes from its role in strengthening the economy’s growth prospects.
Increases in public capital intensity (public capital stock per worker) can affect
productivity and growth through multiple channels. More generally, without
sufficient, high-quality infrastructure allocated efficiently across sectors—and
indeed, across the country—economic growth will be constrained. The simple,
back-of-the-envelope estimates of 10-year costs from delays and quality
problems discussed briefly above--$1.4 trillion congestion costs on our roads,
nearly $300 billion from delays on our inland waterways systems, over $20
billion lost from water main breaks—point to the value to users of improved

166 |

Chapter 4

$"0- 
уҊтѵ
 )')
/ -24. 4./ (
 *&
 '4.Ѷ
ршшт–спрх
Hours per tow
сѵфп

спрх

сѵпп

1 -" 
 '4

җ/*2.Ҙ
рѵфп

рѵпп

пѵфп
ршшп

ршшф

сппп

сппф

спрп

спрф

*0- ѷ
ѵѵ
-(4
*-+.
*!
)"$) -.Ѷ 0'$
 *&
." 
 +*-/ѵ

infrastructure in terms of its quantity, quality, and allocation. In general, the
gross benefits of these improvements include any revenues users are willing
to pay for the improvements as well as any consumer surplus they experience,
recognizing that some of the benefits also accrue to nonusers. Assessing these
ex ante benefits is relatively straightforward for a specific asset or project,
but for the economy as whole, economists often lack direct welfare measures
and instead consider the relationship between infrastructure and productivity or output. This section reviews recent trends in infrastructure investment
spending and capital accumulation and summarizes the evidence for the links
between infrastructure, economic growth, and productivity.

Recent Trends
Two key ideas emerge from a review of recent data. The first is that infrastructure investment spending, as a share of the economy, has remained fairly
steady in recent decades; and the second is that States and local governments
are more important than the Federal government with respect to the funding,
ownership, and management of core infrastructure assets. The Congressional
Budget Office (CBO 2015) reports that public spending on transportation
and water infrastructure has averaged about 2.4 percent of GDP since the
1980s, with a temporary increase in 2009 and 2010 due to additional spending under the American Recovery and Reinvestment Act (figure 4-4). In 2016,
nominal government fixed, nondefense investment spending was 2.5 percent

Infrastructure Investment to Boost Productivity | 167

$"0- 
уҊуѵ
0'$ + )$)"
*)
-).+*-//$*)
)
/ )!-./-0/0- Ѷ ршфх–спру
Percentage of GDP
тѵс

спру

тѵп

сѵч

сѵх

сѵу

сѵс
ршфп

ршхп

ршцп

ршчп

ршшп

сппп

спрп

*0- .ѷ
*)"- ..$*)'
0" /
!!$ Ѹ
0- 0
*!
*)*($
)'4.$.Ѹ
 ).0. 0- 0Ѹ

!!$ 
*!
)" ( )/
)
0" /ѵ

' 
уҊсѵ
1 -" 
0'$
*) ! ). 
 )1 ./( )/
.
a  - )/" 

*!
Nominal GDP, 1980–2016
$( 
+ -$*

  -'
)*) ! ). 


// 
)
'*'

*/'

1980–90

пѵчш

сѵпх

сѵшф

1990–2000

пѵчс

сѵрт

сѵшх

2000–2010

пѵцс

сѵту

тѵпх

2010–16

пѵцс

сѵпс

сѵцу

*0- .ѷ
0- 0
*!
*)*($
)'4.$.Ѹ

'0'/$*).ѵ


of nominal GDP, with the structures component accounting for 1.5 percent of
nominal GDP.
Table 4-2 shows that average nominal nondefense public investment
as a share of nominal GDP has averaged 2.74 percent since 2010, with States
пѵчш‫ڿ‬
сѵпх‫ڿ‬
сѵшф‫ڿ‬spending
and local governments accounting
for nearly
three times as much
пѵчс‫ڿ‬
сѵрт‫ڿ‬
сѵшх‫ڿ‬
as the Federal government. In fact, most of the Nation’s nondefense public
пѵцс‫ڿ‬
сѵту‫ڿ‬
тѵпх‫ڿ‬
infrastructure is owned by States and local governments; for every $1 in nonпѵцс‫ڿ‬
сѵпс‫ڿ‬
сѵцу‫ڿ‬
defense capital stock owned by the Federal government, States and localities
own more than $6 worth of public infrastructure.
Economists typically model the role of public sector capital in the economy by treating it as one factor of production, alongside labor, private capital,
and natural resources. Increased stocks of public sector capital mean increased
flows of capital services available to the economy’s workers, fueling growth
168 | Chapter 4

$"0- 
уҊфѵ
*)/-$0/$*) *!
0'$
+$/'
/*&
/*
-*0/$1$/4

-*2/#Ѷ
ршуц–спрх
*-
+-*0/$1$/4
"-*2/#

Percentage points
тѵф
тѵп

сѵчп

*)/-$0/$*)
*!
+0'$
+$/'

сѵхц

сѵф
сѵп

рѵуу

рѵф
рѵп

рѵрш
пѵчп

пѵф

пѵрп
пѵп

ршуц–цт

ршцт–шф

пѵпх
ршшф–сппц

пѵрф
сппц–рх

*0- .ѷ
0- 0
*!
*)*($
)'4.$.Ѹ

'0'/$*).ѵ

through at least two channels. First, by raising the productivity of other factors
of production—labor, private capital, and land—increased public capital services encourage firms to increase their own investments and expand economic
activity. This indirect, or “crowding in,” effect has been identified in numerous
studies (e.g., Aschauer 1989; Abiad, Furceri, and Topalova 2016). A second,
direct effect works through increases in public capital services per employee
hour, or public capital deepening, which typically accounts for between 0.05
and 0.20 percentage point of growth in labor productivity—not nearly as large
as the impact of private sector capital accumulation, but nonetheless important.1 Since 2007, public capital deepening has accounted for 0.15 percentage
point of the 1.2 percent growth in labor productivity (figure 4-5).

1 Recall that labor productivity growth comes from growth in capital deepening, or the amount
of private capital services per labor input; growth in the skills of workers—often called a labor
composition effect—and increased overall efficiency, calculated as a residual and called total
factor productivity. Historically, in the United States, capital deepening has driven a significant
share of labor productivity growth, though with a marked slowdown in the post–Great Recession
period. From 1953 to 2010, capital deepening accounted for more than 0.9 percentage point of
that era’s 2.2 percent labor productivity growth, but actually detracted from productivity growth
from 2010 to 2015.

Infrastructure Investment to Boost Productivity | 169

Evidence for the Growth Effects of Public Capital
The likely returns to prospective increases in public investment and capital
stocks depend on many factors—including the responsiveness of output to
increases in public capital, the economy’s initial level of capital intensity,
depreciation rates, how quickly assets can be installed and brought into productive service, and even how the investments are financed.
Although the evidence discussed here is based on traditional types of
infrastructure assets, it is important to note that technological innovation
and change will also affect the value of specific infrastructure investments.
Transformative and potentially disruptive technologies, such as those used for
autonomous vehicles and unmanned aviation systems (or drones), may alter
the future use of existing infrastructure, the organization of business activity,
and even residential density and location patterns. Adapting the regulatory
environment to remove barriers to investment and innovation in these technologies will be key to generating the greatest possible future benefits from
their use, and recent regulatory actions move in that direction. In 2016, the
U.S. Federal Aviation Administration issued operational rules (Part 107 of 14
the Code of Federal Regulations), providing a basic regulatory structure for
drones. In addition, a Presidential memorandum issued on October 25, 2017,
establishes a three-year pilot program to facilitate the integration of drones
into the national airspace and permit more advanced operations of unmanned
aviation systems that go beyond the limits set by Part 107, including flying
beyond the visual line of sight of the operator and flying over people. The program is intended to facilitate coordination of and collaboration between regulatory authorities, a key step in adjusting regulation to limit barriers to private
investment in this sector. Another example comes from autonomous vehicles
and related technologies, which may affect future use of roads, highways, and
public transit assets and have the potential to improve safety, decrease traffic
congestion, and raise productivity (box 4-1). In this sector, too, regulators face
challenges in adjusting to the new technology without discouraging innovation. To that end, the U.S. Department of Transportation issued guidelines in
September 2017 regarding automated driving systems, establishing principles
regarding safety, technological change, and technical issues of deployment.
The guidelines are intended to assist Federal, State, and local regulatory
authorities as well as industry and consumer stakeholders in maximizing the
future benefits of the new technologies.
Turning now to conventional approaches to exploring the relationship
between public sector capital, productivity, and output, we note that the CBO
(2016b) estimated that a $1 increase in public capital generated an output
increase of about 8 percent, somewhat lower than other recent estimates (CEA
2016; Bom and Ligthart 2014). Our current preferred estimate puts the corresponding return at just under 13 percent, which is further explained below.

170 |

Chapter 4

Box 4-1. Autonomous Vehicles: A 21st-Century Innovation
Autonomous vehicles provide a flexible and hands-free commute during
which people can engage in activities apart from driving such as office work or
entertainment. A key attraction of these vehicles is their ability to potentially
reduce congestion in highways. This is because driverless vehicles would be
able to drive much closer to other vehicles in a safe manner, and be able to
accelerate and decelerate more quickly. And these vehicles would have the
potential to prevent collisions and reduce regular and incident delays by
creating a smoother traffic flow.
The widespread adoption of driverless cars in the U.S. can increase
economic growth. Winston and Karpilov (2017) estimate that autonomous
vehicles would spur growth in the U.S. by reducing congestion. They focus
their analysis on California, which is home to 11 of the top 16 highway bottlenecks in the Nation, and then extrapolate their results to other areas of the
Nation. They find that highway congestion had adverse effects on the GDP
growth rate, wages, and commodity freight flows in California. Their findings
corroborate similar results that congestion in the Nation’s West Coast ports
from 2014 to 2015 led to a 0.2-percentage-point decline in GDP (Amiti et al.
2015), and that highway congestion is associated with slower job growth in
U.S. metropolitan areas (Sweet 2014; Angel and Blei 2015). Automobile commuting in congested conditions may also damage physical and emotional
health (Fottrell 2015; Knittel, Miller, and Sanders 2016). The benefits of
autonomous vehicles depend on market penetration. In a given year, a 50
percent penetration rate (i.e., half the vehicles in the U.S. would be driverless),
could add more than $200 billion to GDP, 2.4 million jobs, and $90 billion in
wages to the U.S. labor force.
These potential sizable macroeconomic effects of advances in transportation technology are not surprising in light of the historical evidence on
the positive benefits to the U.S. of improvements in mobility. Krugman (2009)
elaborates on how railroads, by reducing transportation costs, facilitated
large-scale production and radically transformed the U.S. economy into
differentiated agriculture and manufacturing hubs. Similarly, given their
potential to reduce congestion and increase safety, autonomous vehicles are
an exciting area of ongoing scientific research.

Calculating the marginal return to public capital requires an estimate
of the elasticity of output with respect to public capital, which has been the
subject of hundreds of studies since the late 1980s. Aschauer (1989) estimated
a U.S. elasticity of about 0.4, suggesting that public sector capital accumulation
was historically a key factor driving economic growth. More recent studies have
confirmed the finding of a robust qualitative and positive relationship between
infrastructure, output, and growth, though with considerable variation across
geographies, time periods, and specific infrastructure assets studied. However,

Infrastructure Investment to Boost Productivity | 171

most recent studies conclude that this elasticity is well below Aschauer’s earlier estimates.
Bom and Ligthart (2014)’s meta-analysis of 68 studies covering the
1983–2008 period yields a short-run elasticity estimate of 0.083 and longrun estimate of 0.122. When restricting their analysis to studies focused on
core infrastructure (transportation, water, and sewer facilities), the authors
report slightly higher elasticities of 0.131 and 0.170 in the short and long run,
respectively, highlighting the point that not all infrastructure is created equal.
The authors also report evidence that output elasticities have declined over
time, because studies using more recent data find smaller output elasticities.
Another recent meta-analysis by Nunez-Serrano and Velazquez (2017) finds
0.13 and 0.16 for short- and long-run elasticities, respectively, somewhat
larger than Bom and Ligthart’s baseline results. However, Nunez-Serrano and
Velazquez do not include more recent studies in their analysis, so their estimates may not reflect recent declines in the elasticity estimates found by Bom
and Ligthart.
The CBO (2016b) assumes an elasticity of output with respect to public
capital of 0.06, but this is likely to be too low in the present context, in which
we consider increased investment in core infrastructure, exactly the asset
types associated with higher elasticities (Bom and Ligthart 2014). Given a ratio
of public capital to output of about 0.75, the CBO (2016b) estimates that the
marginal return to public capital will be about 8 percent (0.06/0.75). However,
using Bom and Ligthart’s average elasticity estimate of 0.106 and an adjusted
capital-output ratio that excludes Federal defense capital assets (0.645 in
2016), we estimate the return to be more than 16 percent. In fact, even Bom
and Ligthart’s lower short-run elasticity estimate for centrally provided public
capital (0.083) still yields a return on public sector capital of 12.9 percent, well
above the CBO’s estimate of 8 percent. Below, we use 12.9 percent as our
preferred estimate.
With these data in mind, we can assess the output consequences for
a given increase in public sector capital. A marginal return of 12.9 percent
suggests that $100 billion in new public capital stock, when fully installed
and productive, would raise output by $12.9 billion, or just under 0.1 percent,
each year it was in use; note that this $100 billion in new infrastructure stock
would generate decreasing annual returns each year as it depreciates. This
supply-side channel for infrastructure investment can be used to estimate the
impact of a longer-term, debt-financed program of $1.5 trillion in infrastructure
investment spending over 10 years’ time. The CEA’s analysis of several different
models indicates that these supply-side effects alone would cumulatively add
0.2 to 0.4 percent to the level of GDP over 10 years, depending on the marginal
return to public capital.
However, as the CBO (2016b) notes, several factors may cause actual
output effects to be smaller than predicted. For example, delays in spending
172 |

Chapter 4

additional funds, constructing infrastructure assets, or bringing those assets
into productive service will decrease expected returns. Permitting and regulatory delays can also affect returns from infrastructure investments. To address
such concerns, on August 15, 2017, President Trump issued Executive Order
13807, “Establishing Discipline and Accountability in the Environmental Review
and Permitting Process for Infrastructure,” which pertains to projects in the
transportation, water and wastewater, energy, and telecommunications sectors. This Executive Order aims to reduce unnecessary delays and barriers to
infrastructure investment; and it outlines a number of steps to streamline regulatory and environmental review processes, establish meaningful deadlines
for reviews and related permitting decisions, and clarify the roles of different
governmental bodies.
Another potentially important factor affecting the output impact of an
infrastructure investment program is the response of States and local governments to an infusion of additional Federal funds for infrastructure investment.
Such an increase could lead to reductions in resources provided by States and
local governments if Federal money serves to crowd out nonfederal support.
The CBO estimates this crowding-out effect at about one-third; applying this
value would lower the CEA’s predicted impact of a federally funded increase
in infrastructure accordingly. Empirical evidence for the sign and size of this
crowding-out effect has been mixed. For example, Knight’s (2002) study of
the Federal Highway Aid program found nearly complete crowding out; under
Knight’s preferred estimates, States and localities cut back by $0.93 for every
additional $1 provided in Federal highway grants during the 1983–97 period.
At a marginal return of 12.9 percent, this implies that a $10 billion increase in
Federal highway funding would ultimately yield only a $0.09 billion impact on
GDP. Although the exact magnitude of this crowding-out effect is uncertain,
Federal policymakers may wish to set maintenance-of-effort provisions as a
condition for receipt of certain Federal funds, to limit States’ ability to curtail
nonfederal support in response to an infusion of Federal funds.
Other effects of increased infrastructure investment. Increased infrastructure investment can also have other important economic effects. Embarking
on an ambitious infrastructure program may create improved employment
opportunities for some U.S. workers (box 4-2). In addition, such a program
could generate meaningful short-run effects that may vary cyclically. In the
short run, deficit-financed additional infrastructure spending affects GDP in
the year in which the spending occurs, generating direct and possibly indirect
(“multiplied”) effects on GDP. Depending on the timing, the extent of possible
crowding out—or, conversely, multiplier effects—and the marginal product of
public capital, the CEA estimates that the 10-year, $1.5 trillion infrastructure
investment program discussed above would add an average of 0.1 to 0.2 percentage point to annual growth in real GDP. If investment is front-loaded, there
is no crowding out, the fiscal multiplier is consistent with Zandi (2012), and the
Infrastructure Investment to Boost Productivity | 173

Box 4-2. Labor Market Effects of Increased
Infrastructure Investment
In addition to raising U.S. productivity and competitiveness, a boost to infrastructure spending may increase demand for the workers needed to build and
construct these new public assets. Although it is difficult to predict the net
employment impact of increased infrastructure investment, a demand shift
to selected occupations may benefit workers in those fields. We term the set
of 31 occupations that are most likely to experience an increase in demand
“infrastructure occupations”; these occupations account for more than 1
percent of employment in at least one infrastructure-related industry’s total
private wage and salary employment (as defined in the note to figure 4-i).
These occupations include workers who design and carry out infrastructure
projects, including engineers, pipefitters, construction laborers, and the like.
But it also includes transportation and warehousing occupations, along with
workers in installation, maintenance, and repair occupations.
Workers in these occupations are far more likely to have a high school
degree or less than the overall U.S. labor force as shown in figure 4-i. The
unemployment rate for workers is strongly correlated with educational
attainment, and even in the current economic expansion, workers with
fewer years of education are disproportionately likely to find themselves
unemployed. As of December 2017, workers with a high school degree or less

$"0- уҊ$ѵ
0/$*)'
//$)( )/
o!
 )!-./-0/0- 
1 rs0.

*)$)!-./-0/0- 
 *-
*- 
-/$$+)/.Ѷ
спрц
..
/#)
#$"#
.#**'
*( 
*'' "

Percentage of all participants
рпп

$"#
.#**'Ѷ
)*
*'' "
# '*-Ѿ.
 "- 
*-
(*-

чп
хп
уп
сп
п

)!-./-0/0- 
*0+/$*).

*)$)!-./-0/0- 
*0+/$*).

*0- .ѷ
0-- )/ *+0'/$*)
0-1 4Ѹ

'0'/$*).ѵ
*/ ѷ
The уҊ$"$/
North American Industry Classification System ( )

* .Ѷ
 !$) 
.

$)!-./-0/0- 
$)0./-$ .Ѷ
$)'0 
$1$'
 )"$) -$)"
*)./-0/$*)
җ''
. /*-.
!''$)"
0) -

 
тҊ$"$/
* 
стцпппҘѸ
*/# -
.+ $'/4 /- 
*)/-/*-.
җ 
* 
стчшппҘѸ
)

- ( $/$*)
)
*/# -
2./ 
()" ( )/
. -1$ .
җ 
* 
фхсшппҘѵ
 )!-./-0/0- 

*0+/$*).
- 
/#*. 
/#/
(& 
0+
/
' ./
р + - )/
*!
 (+'*4( )/
$)
*) 
*!
/# . 
уҊ$"$/

. /*-.ѵ

174 |

Chapter 4

had an unemployment rate 2.6 percentage points higher than those with a
bachelor’s degree—4.7 and 2.1 percent, respectively. Our estimates of the
unemployment rate for workers who report an infrastructure occupation
indicates an even greater disparity; in the Current Population Survey, 6.1
percent of labor force participants who report an infrastructure occupation
reported being unemployed in 2017, reflecting an excess supply of nearly
350,000 infrastructure workers relative to the unemployment rate for workers
in noninfrastructure occupations.
Despite this excess supply, the geographic footprint and skill needs of
expanded infrastructure investments are unlikely to perfectly match those of
currently unemployed infrastructure workers, and the Federal government
could take an active role in easing the transition of workers into infrastructure
employment. One impediment to the free movement of skilled workers into
new infrastructure jobs is the country’s patchwork set of occupational licensing requirements, which depress the movement of licensed workers across
State lines (Johnson and Kleiner 2017). In 2016, 22.2 percent of all labor force
participants reporting an infrastructure occupation in the Current Population
Survey said they had an active professional license or certification; this is
slightly fewer than the average for all participants (24.4 percent) but substantially more than would be expected—given the education distribution across
infrastructure occupations, because the probability of occupational licensing
increases with educational attainment. Tying infrastructure funds to the loosening of occupational licensing (or to reciprocal agreements between States)
could help alleviate the depressive effects of these licenses on geographic
mobility. This topic is discussed in greater detail in chapter 2.
Furthermore, the Federal government has additional tools to ensure a
skilled workforce for expanded infrastructure activity. One clear disconnect
between the needs of the labor market and the supply of America’s workforce
is the current subsidization of higher education through Pell Grants. These
grants, which are generally only available to students without a bachelor’s
degree and who are enrolled in programs with more than 600 clock hours of
instruction over 15 weeks, do not provide support to workers who require
shorter-term investments. Workforce Innovation and Opportunity Act funds
could be used for these short-term programs, but funds from this program
are not dedicated to this purpose and are therefore subject to competing
priorities. Although it would require Congressional approval, expanding Pell
Grant eligibility to include investments in short-term training (or retraining)
programs would help ensure that financial constraints do not prevent workers
from pursuing infrastructure occupations.

marginal product of capital is as reported in the 2016 Economic Report of the
President, we expect the average annual contribution to be at the upper end of
this range. With crowding out, no multiplier, and assuming the CBO’s estimate

Infrastructure Investment to Boost Productivity | 175

of the marginal product of capital, we expect the contribution to growth to
instead be closer to 0.1.
In general, the sign and magnitude of these “fiscal multipliers” remains
a topic of active research, and recent evidence suggests that spending multipliers exceed zero, meaning that the net impact of additional government
spending is positive (Auerbach and Gorodnichenko 2012; Ramey and Zubairy
2017). Auerbach and Gorodnichenko find that these multipliers are larger
during recessions, while Ramey and Zubairy find no evidence that multipliers
are higher during periods of slack. Abiad, Furceri, and Topalova (2016) and the
International Monetary Fund (2014) find that increased infrastructure spending
in particular during recessions can raise GDP through demand-side multiplier
relationships. In fact, even a study of the Great Depression found that an
additional $1 in public works and relief spending per capita between 1933 and
1939 was associated with a 44 cent increase in retail sales in 1939 (Fishback,
Horrace, and Kantor 2005)! These short-run demand-side effects of increased
infrastructure investment are not unimportant, but infrastructure’s long-run
effects on productivity and growth may be better guides to policymakers about
the effects of future investment programs and policies.

Funding and Financing Needed Infrastructure
Both around the world and in the United States, governmental resources
provide and support infrastructure investment to promote both efficiency
and equity goals. On the efficiency front, the public goods nature of some
infrastructure assets will lead the private sector to underproduce such assets
relative to socially desirable levels. These goods are generally characterized by
some degree of nonexcludability, meaning that it is difficult or very costly to
exclude nonpayers from consuming the good; of nonrivalry, meaning that one
person’s consumption does not hinder the ability of others to also consume
it; or both. For example, flood control services provided by a system of dams,
levees, and reservoirs may provide benefits to a wide geographic area. In
this instance, excluding nonpayers from experiencing the benefits would be
difficult, and the benefits experienced by one local resident do not impair the
ability of other residents to experience benefits as well.
Infrastructure assets may also have other characteristics that lead to
inefficient resource allocations under market provision. For example, many
assets pertaining to transit, water and sewer utilities, water resource management, energy, and communications are characterized by increasing returns to
scale, with high fixed costs and sometimes quite low marginal costs. In these
situations, efficiency concerns suggest that the best industry configuration
will include one or only a few suppliers. In some cases, these providers may
have market power and can price at well above marginal cost, in opposition
to efficiency goals; in other cases, these providers may price below marginal

176 | Chapter 4

cost, making cost recovery and efficiency goals hard to reach. Historically, in
these types of situations, government officials have turned to government-run
monopolies or regulated utilities to meet efficiency, equity, and revenue goals.
Other sources of market failure may also be present. Some infrastructure
assets provide services that generate agglomeration effects, whereby efficiency
gains arise from the spatial concentrations of firms and workers—because
more efficient labor markets, better matching between firms and workers, and
a quicker dissemination of ideas and best practices all increase productivity.
Evidence suggests that such economies are present in the transportation,
communications, and power sectors. Network effects also characterize infrastructure in transportation and communications, because the value of the
network rises as other users join and more nodes and segments are added. A
robust transportation network also makes it easier for workers and firms to
locate near each other; thicker markets mean better matches between firms
and workers, increasing efficiency. Again, these effects can mean that private
actors lack the incentives to invest to the desired fully efficient level, motivating the public sector to offer support and/or invest. Given these considerations,
the rest of this section describes the fiscal roles currently played by Federal,
State, and local governments and explores issues in funding and financing
infrastructure investment.

Fiscal Roles for Federal, State, and Local Governments
In the United States, infrastructure investment, operations, and maintenance
responsibilities are shared across the Federal, State, and local public sectors,
and in some cases, by private sector entities. The CBO (2015) reported that
combined Federal, State, and local public spending on transportation and
water and wastewater infrastructure was $416 billion in 2014, with the Federal
government accounting for 23 percent of the total, and State and local governments for the remaining 77 percent. The allocation of responsibility varied
sharply, depending on the category of infrastructure assets. For example, the
Federal government funded 28 percent of total highways spending, 23 percent
of mass transit and rail spending, and only 4 percent of water utilities spending. Within sectors, Federal support also varies and typically focuses more on
capital spending, not spending for operations and maintenance (figure 4-6).

Funding Infrastructure Investment
Given the desire to maintain, upgrade, and expand infrastructure investments
in various sectors of the economy, policymakers must consider the best ways
of funding these investments. Resources generally come from one of two
principal sources: tax revenues or user charges (user fees). In this subsection,
we discuss and analyze funding options available to policymakers at different
levels of government, with a special focus on the role of user fees for use of

Infrastructure Investment to Boost Productivity | 177

$"0- 
уҊхѵ
0'$ + )$)"
*)
..
-).$/Ѷ
 $"#24.Ѷ
)
/ -
 )!-./-0/0- Ѷ
спру
  -'
..

-).$/

// 
)
'*'Ѷ
) /
*!
  -'

о
+$/'
*/'

$"#Ҋ
24.

о
+$/'
*/'
о

/ -

+$/'
*/'
п

сп

уп

хп
чп рпп
Dollars (billions)

рсп

руп

рхп

рчп

*0- .ѷ *)"- ..$*)'
0" /
!!$ Ѹ

'0'/$*).ѵ
*/ ѷ
о
- +- . )/.
*+ -/$*).
)
($)/ )) ѵ

selected infrastructure services, including roads and highways, transit, and
water and wastewater services.
General tax revenues are often used to support projects and investments
that provide benefits widely or are somewhat nonrivalrous or nonexcludable.
Specific or dedicated tax revenues are also commonly used by Federal, State,
and local governments—sometimes reflecting a goal of linking those who use
the services to the funds collected to pay for them. Governments also rely on
direct fees and charges paid by users and beneficiaries of a particular service.
The economic incidence of these fees—that is, who actually pays them in the
form of higher prices paid by consumers or lower net prices received by suppliers—varies by service, and the revenues collected can be considerable.
A distinct but related revenue source may sometimes also be appropriate. For example, a new transit project (e.g., a new or rehabilitated station
along a rapid transit line) may increase economic activity and/or raise property
values in the areas near the project. Using “value capture” techniques, such as
tax increment financing (TIF), can enable the public sector to access some of
the value generated by the public investment, making more revenues available
to support the project (Chapman 2017).
User fees in theory. Setting and collecting user fees to fund infrastructure
investments helps the government achieve two key goals: ensuring efficiency
in the use of public assets, and collecting revenues to defray the costs of
providing these assets. If users of the service experience a significant private

178 | Chapter 4

benefit from doing so, efficiency gains can be significant when user fees are
set correctly. As discussed earlier in the chapter, underpriced access to public
infrastructure will generate excess demand for use, leading to congestion and
inefficient allocations. Without price signals to guide supply and consumption
decisions, the public sector struggles to determine how much infrastructure to
build and how it should be allocated.
The rationale for imposing user fees is especially strong when the services in question provide significant private benefits relative to the overall
public benefits generated by use of the asset. For example, a shipper that
sends barges full of grain through locks along the Mississippi River obtains
private benefits from using the Nation’s inland waterway system. Similarly, an
airline that uses gate facilities at a particular airport and accesses the Nation’s
air-traffic control system is also receiving a private benefit. In these instances,
user fees should be a significant part of the funding structure, though not
necessarily the only revenue source. Note that though some user fees paid by
businesses will eventually be paid by consumers in the form of higher prices,
firms using public sector assets will recognize these charges as costs of doing
business, thus encouraging efficient choices of production and allocation.
Setting specific fee structures to achieve multiple policy goals can be difficult, and trade-offs between goals are likely. Attaining efficiency goals usually
means setting unit prices at the marginal cost of provision, but in sectors with
high fixed costs (e.g., water and wastewater, and transit), the revenues generated may not be enough to cover fixed costs. Setting unit prices at average
cost can improve revenue generation, but comes at the expense of decreased
efficiency, as some users cut back consumption at the margin. Turning to “twopart” tariffs can help achieve efficiency and revenue goals though may raise
affordability concerns. Under a two-part tariff, the customer is charged a fixed
fee that does not vary with use and a unit price per unit consumed. Essentially,
the fixed fee allows service providers to collect the revenues they need to
defray their fixed costs, and the unit price acts as a signal to consumers, who
will consume up to the point where their benefits and costs are balanced at the
margin, contributing to efficiency.
There are many examples of this two-part tariff approach. Water utility
customers pay a monthly connection charge, in addition to charges based on
monthly water use; and in the power sector, electricity users pay monthly fees
along with charges that vary with electricity use. Service providers may use
increasing block tariffs, charging low unit prices for low (“lifeline”) levels of
consumption and higher unit prices for higher consumption levels, a structure
that can preserve access for consumers with a low ability to pay. Alternatively,
decreasing block tariffs offer a reverse approach, with unit pricing that falls
as consumption levels rise; this structure allows offering quantity discounts,
which are common in industrial settings. Simple unit pricing includes a constant per-unit charge for all levels of consumption.
Infrastructure Investment to Boost Productivity | 179

In funding for roads and highways, State and local governments already
rely on an informal two-part tariff system of quasi-user fees to raise funds to
partially cover capital and operating expenses. For example, annual vehicle
registration fees and driver’s license fees can be viewed as fixed components
that do not vary with road use, while gasoline taxes, tolls, and other charges
are somewhat connected to usage levels, acting at the margin to affect drivers’
choices about consumption.
User fees for roads and highways. With the increasing prevalence of electric vehicles and high fuel economy vehicles, and with some fuel-based revenue sources not being indexed to inflation, the existing financing mechanism is
becoming increasingly unsustainable, with funding needs growing faster than
dedicated revenues. Here, we explore current funding practices and alternatives, considering the efficiency, equity, and revenue effects of these choices.
At present, the Federal, State, and local governments rely heavily on
dedicated fuel taxes and general taxes to pay for roads and highways, with
a much smaller role played by direct user fees, such as tolls. Toll revenue
collected by State and local governments in fiscal year (FY) 2015 was $14.0
billion, accounting for 6.0 percent of total spending on roads and highways,
a share that has crept up only slightly since 1993 (DOT 1993, IV-6; 2015, table
HF-10), when the Federal gasoline tax, currently 18.4 cents per gallon, was last
increased. Although the administrative costs of toll systems are significant, at
between 8 and 13 percent of receipts (Kirk 2017, 7), the economic arguments in
favor of using toll revenues to pay for roads and highways are solid. By collecting fees from the direct users of the assets (motorists, commercial carriers, et
al.), governments acquire revenues needed to maintain, operate, rehabilitate,
and expand the roads, and drivers use the roads up to the point at which their
marginal benefits equal the marginal costs they impose when driving.
Federal gasoline and diesel taxes have some characteristics of user fees
because the individuals and businesses that buy fuel for vehicles and drive
on public roads and highways pay them. However, these taxes are imperfect
because they fail to encourage efficient use of existing roadways and to signal
the value of any potential additional capacity. Highly fuel-efficient vehicles
(including electric vehicles) pay less than the marginal costs generated by
their use of roads in terms of wear and tear, congestion, and other external
costs. More generally, these taxes do not reflect the crowding or congestion
costs generated by drivers. That is, driving 100 miles on low-use rural roads
generates the same fuel tax revenues as driving that same distance on high-use
urban roads—during rush hour. Furthermore, evidence suggests that heavy
trucks in particular do not currently face taxes and charges that are aligned
with the negative externalities they generate, which include pavement damage, traffic congestion, accident risk, and emissions. Even excluding emissions,
these external costs are significant, with estimates ranging from 2.01 to 4.14

180 |

Chapter 4

cents per ton-mile, which is equivalent to between 10 and 20 percent of the
average price per ton-mile to ship by truck (Austin 2015).
More generally, because fuel taxes do not reflect congestion costs
imposed by drivers, scarce road access is not allocated efficiently. Implementing
congestion pricing would encourage only consumers with high valuations to
use highly congested roads during peak demand times, improving efficiency
but also potentially making some drivers worse off, particularly low-income
drivers who may be priced out of the tolled lanes (CBO 2009). Using toll revenues to improve other travel options, particularly transit, can counteract this
distributional effect, but most discussions of congestion pricing acknowledge
its potential to create both winners and losers from the policy. Even so, the
lack of appropriate congestion pricing mechanisms creates winners and losers
as well, and some evidence suggests that at least some low-income drivers in
practice find tolled lanes worth paying for (Federal Highway Administration,
n.d.). Furthermore, Hall (2015) shows that congestion pricing can be Paretoimproving, not just potentially Pareto-improving, especially under conditions
of bottleneck congestion, which occurs when the number of vehicles that
can use the road per unit of time (its “throughput”) decreases. An example of
bottleneck congestion is when traffic backs up at an exit ramp, slowing down
through traffic on the roadway. Tolling a portion of the highway’s lanes (value
pricing) serves to internalize both motorist travel time externalities as well
as these bottleneck effects, raising speeds on both the tolled and nontolled
highway segments. When drivers differ in terms of income and valuations of
their time, partial time-varying tolls will raise welfare for drivers along both the
tolled and nontolled segments as long as high-income drivers use the highway
during rush hour. Under the policy, drivers “sort” into the road segments and
are better off, even before accounting for how toll revenues are spent.
The recent introduction of dynamic tolling along Interstate 66 in Northern
Virginia offers an example of how congestion pricing can improve travel times
and raise revenues for transportation projects. Preliminary figures from the
Virginia Department of Transportation indicate that morning rush-hour tolls
averaged between $8.20 and $12.87 for the 10-mile segment but that peak tolls
reached $40.00 for a short time. Further, travel speeds in the tolled lanes were
far higher than during a comparable period a year earlier, and travel speed in
the nontolled lanes as well as parallel roadways were similar or improved.
In addition to falling short on efficiency grounds, fuel taxes have seen
diminished revenue productivity in recent years, as the twin factors of inflation
and increased fuel efficiency have sharply curtailed the growth of the Highway
Trust Fund’s real fuel tax receipts. The Federal gasoline tax of 18.4 cents per
gallon has not been raised since 1993, while construction prices have risen at
a 3.9 percent annualized rate. Figure 4-7 shows that in 2016 real Federal fuel
tax receipts were only 93 percent of their 1993 levels, even as nominal receipts
more than doubled over that period.
Infrastructure Investment to Boost Productivity | 181

$"0- 
уҊцѵ
  -'
0 ' 3
 1 )0 .Ѷ
ршшт–спрх
Index (1993 = 100)
сфп

спрх
*($)'

спп

рфп

 '
рпп

фп
ршшп

ршшф

сппп

сппф

спрп

спрф

*0- .ѷ
ѵѵ  +-/( )/
*!
-).+*-//$*)Ѹ

'0'/$*).ѵ

$"0- 
уҊчѵ
# 
 $"#24
-0./
0)Ѿ.
 $"#24
*0)/Ѷ

ршшп–спрх
  $+/.
) /
*!
/-).! -.
!-*(
" ) -'
!0).

Dollars (billions)
фп

3+ )$/0- .
спрх

уп

тп

сп

рп

п

ршшп

ршшф

сппп

сппф

спрп

спрф

*0- sѷ ѵѵ
 +-/( )/
*!
-).+*-//$*)Ѷ
  -'
 $"#24
($)$./-/$*); CEA calculationsѵ

182 |

Chapter 4

These fuel tax revenues have failed to grow as quickly as appropriations
for highway spending, putting pressure on the Federal Highway Trust Fund
(HTF) used to finance highway and transit projects. As figure 4-8 shows, outlays
from the HTF’s Highway Account have regularly exceeded revenues since 2008,
and the CBO (2017) projects that, absent any changes, the Highway Account’s
balance will fall below zero by 2021. Because, by law, the HTF cannot incur
negative balances, Congress has authorized multiple transfers from general
funds to shore up the HTF; the most recent one was in 2016, when $70 billion
was transferred—$52 billion to the Highway Account and $18 billion to the Mass
Transit Account.
States, too, rely heavily on excise and sales taxes on fuels, with similar
revenue pressures arising from inflation and increased fuel efficiency of
vehicles. According to Quinton (2017), 26 States have increased their fuel taxes
in the last four years to raise more transportation revenues for their roads and
highways.
The declining revenue productivity of existing gasoline taxes has led
policymakers to consider other options for funding highways. One innovative
approach is to consider supplementing or replacing fuel taxes altogether with
a user fee more closely related to a consumer’s use of the system—such as, in
the present context, a tax on vehicle miles traveled. Assessing a charge based
on mileage instead of gasoline consumed would link consumers’ choices more
closely to the costs they impose, including congestion, emission, pavement
damage, and so on. Such charges could also be structured to vary with the
time of day, region of use, and other factors, including vehicle weight, which
has a large impact on pavement wear-and-tear (Sorensen, Ecola, and Wachs
2012; TRB 2012; Kirk and Levinson 2016). Although the design and implementation of such taxes has many challenges, VMT taxes can raise needed revenues
in a sustainable way while providing the right signals regarding the value of
consumption and supply, helping public officials to understand the value of
current uses of roads and highways and to plan for the future.
In the context of freight and commercial shipping, Austin (2015) estimated that a VMT tax on commercial trucks would decrease external costs by
$2.1 billion and raise $43.0 billion in tax revenues; including vehicle weight as
a factor in the tax and raising diesel taxes in tandem would achieve similar
efficiencies but generate revenues of nearly $70 billion annually (in 2014 dollars). Another recent study (Langer, Maheshri, and Winston 2017) finds greater
efficiency benefits from a gasoline-tax equivalent VMT tax when the VMT tax is
higher in urban areas than in rural areas, reflecting differences in external costs
across regions. The intuition here is twofold. First, because the evidence suggests that congestion, accidents, and environmental externalities are higher in
urban areas than in rural areas, the differentiated VMT tax gives urban drivers
a stronger incentive to cut back on miles driven, improving efficiency. Second,
as vehicles’ fuel efficiency rises, the VMT tax does a better job than the gasoline
Infrastructure Investment to Boost Productivity | 183

Box 4-3. Oregon’s Tax on Vehicle Miles Traveled
Oregon has long been a pioneer when it comes to transportation funding.
Oregon was the first State to levy an excise tax on gasoline, setting a tax of
1 cent per gallon in 1919. More recently, Oregon has devoted considerable
time and effort to exploring options to replace its excise taxes on fuel to fund
its roads and highways. Its OReGO program, which started on July 1, 2015,
charges volunteer participants a mileage fee of 1.7 cents per mile for travel
on public roads inside the State and provides rebates or credits for State fuel
taxes paid. Though small, the program offers tangible evidence that a tax on
vehicle miles traveled (VMT) is a promising alternative to relying on fuel taxes
(Oregon Department of Transportation 2017).
Motivation and recent history. Like other States throughout the country, Oregon has seen the revenue productivity of its motor fuel taxes diminish
as the fuel efficiency of vehicles has improved; also, its State excise tax on
gasoline, like that in most States, is not indexed to inflation. Figure 4-ii shows
that since 1993, nominal motor fuels sales tax revenues have risen by 65.8
percent, but in inflation-adjusted terms, revenues have fallen by more than
30 percent during this period.
In recent years, the State has moved more aggressively than some
others to increase its tax rates to make up for revenue shortfalls. Its excise tax
on gasoline of 24 cents per gallon in 1993 was raised to 30 cents per gallon in
2011; and legislation passed in 2017 will increase the excise tax by 4 cents per
$"0- 
уҊ$$ѵ
- "*)
0 '
3
 1 )0 .Ѷ
ршшт–спрх
Index (1993 = 100)
рцф

спрх

рфп

*($)'
рсф

рпп

 '

цф

фп
ршшп

ршшф

сппп

сппф

спрп

*0- .ѷ
ѵѵ
 ).0.
0- 0Ѹ 0- 0
*!
*)*($
)'4.$.Ѹ

'0'/$*).ѵ

184 |

Chapter 4

спрф

gallon in 2018, with additional increases planned through 2024. Furthermore,
the State has continued its exploration of using taxes on VMT to supplement
and perhaps in the future replace the excise taxes now in place. In fact, the
Oregon Department of Transportation (2017) estimated that continued reliance on motor fuel tax revenues over the next 10 years would lead to a $340
million revenue shortfall relative to what could be raised by a “road usage
charge” or a tax on VMT.
Pilot programs. In 2001, Oregon established the Road User Fee Task
Force (RUFTF) to examine alternative revenue sources to fund construction, repair, operations, and maintenance of Oregon’s roads. The task force
established criteria that a new revenue source or structure should meet,
including the “user pays” principle discussed above. Revenue adequacy,
system transparency, and enforceability were also important. Ultimately, the
RUFTF recommended to the State legislature that Oregon develop and test
mileage-based fees (i.e., road usage charges, RUCs) for this purpose, and the
State created and ran its first pilot project in 2006. For 12 months beginning in
2006, 285 volunteers used on-board equipment to measure mileage traveled
inside identified zones and to transmit data to fuel pumping systems where
participants bought fuel. No specific location data were collected or transmitted, so only the general zone and accumulated mileage were recorded and
used to determine fees. The fee was collected at the point of sale, as the current gasoline tax was collected, and participants received immediate credit
for fuel taxes paid.
After concluding the program and reviewing its performance, the
RUFTF began to develop a second pilot program, which ran from November
2012 to March 2013. The goals of the second pilot included using an open
architecture, ensuring better and more flexible use of technologies then and
in the future, giving motorists choices about how mileage was reported, and
including private sector vendors as part of the administrative structure. Most
important, however, the RUFTF also wished to provide motorists with the
option of avoiding the usage of global positioning system–enabled devices if
they desired, allowing users more control over their private information and
data.
After concluding these two pilot programs, officials developed the
small, voluntary OReGO program, which currently operates in the State.
Initially, volunteer drivers were charged 1.5 cents per mile traveled on the
State’s public roads, receiving credits/refunds for fuel taxes paid and for miles
driven on nonpublic roads or out-of-State roads. For a car with fuel efficiency
of 20 miles to the gallon, the charge amounted to 30 cents per gallon, the
then-current State excise tax. As of January 1, 2018, the road user charge
rose to 1.7 cents per mile, aligned with an increase in the State’s gasoline tax
from 30 to 34 cents per gallon. The program enrolled 1,307 vehicles between
June 1, 2015, and December 31, 2016, though only 669 vehicles remained
active as of December 31, 2016. Note that the program restricts the number

Infrastructure Investment to Boost Productivity | 185

of participating vehicles with low fuel efficiency (below 17 miles per gallon),
whose drivers would be likely to pay less under a VMT than under a regular
gasoline tax.
Lessons learned and future plans. Oregon officials have a program that
allows consumer choice, is based on an open technological platform, and is
administratively feasible. The program is small, however, and it is unclear
how a scaled-up program would affect revenue generation, efficiency, and
equity. McMullen and others’ (2016) prospective analysis of a close-to-revenue
neutral RUC found that the RUC was less regressive than the gasoline tax—and
that restricting the RUC payment option to owners of new cars or high-fuelefficiency vehicles would make the RUC even less regressive. McMullen and
others also found that the effects of moving from the gasoline tax to the RUC
varied across regions of the State; areas with drivers who drove more miles
on average tended to fare worse under the RUC than under the gasoline tax.
Although OReGo is not yet ready to bear the full burden of funding Oregon’s
road expenditures, it has given policymakers some real experiences on which
to base future policy and program decisions.

tax in giving all drivers the right incentives about their use of the roads. In terms
of distribution, Langer, Maheshri, and Winston (2017) find that the differentiated VMT tax imposes the largest welfare losses on high-income drivers compared with low-income drivers, because high-income drivers are more likely to
live in urban areas (and are more likely to drive highly fuel efficient vehicles),
mitigating concerns about the equity effects of VMT taxes.
Thus far, the actual U.S. experience with VMT taxes and other alternatives
to gasoline and diesel fuel taxes has been limited. Concerns about privacy risks
and administrative and implementation costs have hindered program development, despite technological advances that have made it easier to record,
report, share, and manage the information that is needed to administer such
taxes. Oregon has been a pioneer in this space, having conducted two pilot
programs for VMT taxes on motorists and established an ongoing, small-scale
program called OReGO (box 4-3). California, too, facing significant funding
shortfalls for its roads and highways, has experimented with VMT taxes, testing
a program with simulated, though not actual, road use charges of 1.8 cents per
mile for volunteer drivers (California Department of Transportation 2017).
A few States—including Kentucky, New Mexico, New York, and Oregon—
do impose alternative taxes on heavy vehicles, via weight-distance or ton-mile
taxes. These taxes depend on distance traveled as well as vehicle weight.
For example, Kentucky’s “Weight Distance License” system imposes a tax
of $0.0285 per mile traveled on the State’s roadways for all carriers with a
combined license weight 60,000 pounds or more (TRB 2012); this tax generated $79.1 million for the State in FY 2015, about 5.2 percent of all Road Fund
186 | Chapter 4

revenues that year. Thus, a truck weighing 30 tons would face a tax of 2.85
cents per mile, far less than Austin’s (2015) estimates of unpriced external
damages per ton-mile—at 3 cents per ton-mile, about the midpoint of Austin’s
estimates excluding emissions damages, a 30-ton vehicle would face a charge
of 90 cents per mile.
Distance-based road user charges are more common in other countries,
most of which levy such taxes only on freight traffic, not individual drivers. For
instance, in 2001 Switzerland—motivated by concerns regarding traffic, wear
and tear on roadways, and emissions—established a distance-based charge
system for heavy commercial vehicles. Under this system, heavy vehicles pay
fees for travel on all Swiss roads based on distance traveled, permissible total
weight, and emission category, and the charges are substantial; in 2001, a
34-metric-ton truck (almost 75,000 pounds) faced charges ranging from $0.90
to $1.27 per mile, depending on the emission category. Luechinger and Roth
(2016) estimate that the introduction of the tax decreased truck traffic in
Switzerland by between 4 and 6 percent, with some evidence suggesting a corresponding mode shift to rail. Direct estimates on external effects were mixed,
with evidence suggesting significant declines in nitrous oxide emissions but no
impact on accidents. Kirk and Levinson (2016) report that the administration
costs of this fee system are between 5 and 6 percent of total receipts, which
compares favorably with the costs of toll collections.
Germany also taxes heavy commercial trucks using its main highways.
The charges, called LKW-Maut, vary only with distance, not weight, but are
assessed and collected in real time using a complex system of on-board units,
global positioning system technology, web payment portals, and payment
kiosks at gas stations and highway rest stops (Kirk and Levinson 2016). Some
empirical evidence suggests that the introduction of the charge in 2005 was
followed by improved efficiency, as shippers adjusted by routing fewer empty
trucks and by loading trucks up to their maximum allowable weight. Doll and
others (2017, 33) report that the costs of running the charging system were 12.4
percent of its revenues in 2015.
User fees for transit. Public transit sector ridership and fare revenues have
come under increasing pressure from the entry and expansion of ride-sharing
services, low gasoline prices, and other factors. Transit services are primarily
provided by local governments and agencies, but funding comes from all levels
of government, and public subsidies are significant. Direct user fees, primarily
in the form of farebox revenues, do not cover total operating expenses, let
alone contribute to covering capital costs. Passengers are typically charged
fares far below the true marginal operating cost of providing service, leading
to inefficiency in the form of congestion, overuse, and queueing as well as
revenue shortfalls. In 2016, passenger fares covered 32.0 percent of operating
expenses, with the next largest shares coming from localities (31.6 percent)
and States (24.4 percent). Federal support was modest, at 7.2 percent of
Infrastructure Investment to Boost Productivity | 187

operating expenses. Capital expenditures, however, receive significant Federal
funding, which covers 40.7 percent of capital expenditures; passenger fares
and other revenues directly generated by transit agencies themselves cover
only 11.7 percent of all capital expenditures. In a few cases, new transit projects
have been funded with value capture (e.g., TIF) funds; for example, the Chicago
Transit Authority plans to combine Federal grant funds with TIF revenues as
the primary funding sources for its Red–Purple Line Modernization project,
with the TIF revenues directed toward repaying debt issued to finance the
project. Other value capture examples are described in the EPA’s (2013b) study
of several recent large-scale, transit-oriented development projects across the
country.
Overall, the sector faces significant challenges, facing long-deferred
maintenance needs, changing transit use patterns, and continued reliance on
public subsidies. Raising passenger fares significantly, especially for expensive
rail service, would improve both efficiency and cost-recovery but, in principle, present affordability problems for some low-income users. In practice,
many transit riders are not low-income, so equity concerns regarding fare
increases may be overstated. For example, the American Public Transportation
Association (APTA 2016) reports that high-income households (those with
incomes of $100,000 or more) make up 12 percent of all bus users but 29 percent of all rail users.
At least one transit agency has implemented ORCA LIFT, an income-based
transit fare system. The program was introduced in March 2015 in Seattle, and
it now operates in both the city (via King County Metro Transit) and the wider
metropolitan area (via Sound Transit), charging reduced fares to adults with
household incomes below 200 percent of the relevant Federal poverty threshold. Because previous evidence suggested that low-income riders were more
likely to ride in off-peak hours, officials had few concerns about increasing
peak hour congestion. In effect, these reduced fares offered a way to engage
in peak-load pricing, which can increase revenues and improve allocative efficiency. Overall, Sound Transit (2016) reports that passengers paying ORCA LIFT
fares accounted for 1.4 percent of system fare revenues and 2.8 percent of all
boardings, with an average fare paid of $1.00 (table 4-3). In contrast, reduced
fare passengers, who qualify based on categorical measures (age, disability
status, etc.), accounted for 2.3 percent of revenues but 6.4 percent of boardings, paying an average fare of $0.70 per trip. Pursuit of equity objectives costs
revenues, as the average fares paid indicate.
The biggest risks and opportunities facing the transit sector, however,
likely come from the rapid technological change and disruptive entry of new
transportation services and providers in cities across the country. The introduction of autonomous vehicles and “smart” road and highway infrastructure
will surely influence transit use and patterns in the years ahead, and the entry
and expansion of ride-sharing services presents another challenge. Transit use
188 | Chapter 4

' 
уҊтѵ
*0)
-).$/
- 
 1 )0 .
)
*-$)".Ѷ
спрх
$& /
!- 
/ "*-y

*/'

.+ /

0'/

*0/#

 0 
!-

*2
$)*(

 1 )0 .
җ‫ڦ‬Ҙ

цфѶсфрѶфуш

сѶсхчѶхуш

рѶчццѶшшш

рѶрхсѶрпц

чпѶфхпѶтпу

*-$)".

тхѶстпѶпцу

рѶчсфѶфшу

сѶхчсѶцтх

рѶрхфѶцсц

урѶшпуѶртр

 1 )0 
+ -
*-$)"

җ1 -" Ѷ
‫ڦ‬Ҙ

сѵпч

рѵсу

пѵцп

рѵпп

рѵшс

 - )/" 
*!
- 1 )0 .

җ‫ڿ‬Ҙ

штѵу

сѵч

сѵт

рѵу

рппѵп

 - )/" 
*!

*-$)".

җ‫ڿ‬Ҙ

чхѵф

уѵу

хѵу

сѵч

рппѵп

*0- .ѷ
*0)
-).$/Ѷ
- 
 1 )0 
 +*-/
спрхѸ

'0'/$*).ѵ
*/ ѷ
$"0- .
- +- . )/
.0(.
*1 -
/#- 
+-$)$+'
/-).$/
+-*"-(.ѷ

3+- ..Ѷ
*0) -Ѷ
)
 $)&ѵ

and farebox revenues are under pressure in many cities. The Federal Transit
Administration (FTA 2017) reports that 2016 transit ridership was 10.2 billion
штѵуп‫ڿ‬
сѵчп‫ڿ‬
сѵтп‫ڿ‬
рѵуп‫ڿ‬
рппѵпп‫ڿ‬
unlinked passenger trips,чхѵфп‫ڿ‬
down 3.7 percent
from
its peak inсѵчп‫ڿ‬
2014, andрппѵпп‫ڿ‬
ridership
уѵуп‫ڿ‬
хѵуп‫ڿ‬
in the first 10 months of 2017 was down 2.5 percent from that same period in
2016. At the same time, some local agencies report far larger declines in ridership and revenues. Some observers have argued that the entry and expansion
of ride-sharing services by firms such as Uber, Lyft, and others are to blame.
More generally, the entry of these firms has had wide-ranging welfare
and transit effects across the country. Some evidence suggests that the services provided by Uber, Lyft, and other firms have made consumers better off
with the introduction of more affordable and reliable transportation options,
especially in traditionally underserved areas of cities (Hall, Palsson, and Price
2017), and at least one city, Boston, has piloted a paratransit program with
Uber and Lyft. In principle, ride-share services could complement transit’s
fixed-route, fixed-schedule service by extending its reach and flexibility, making transit more attractive and increasing ridership. On the other hand, these
services could directly substitute for transit trips, as consumers can enjoy
taxi-like service at reduced prices. Systematic evidence to date is limited, but
Hall, Palsson, and Price 2017 find that Uber’s entry into metropolitan statistical areas (MSAs) across the country does not have a statistically significant
impact on transit ridership. However, over time, as Uber’s presence grows,
transit ridership slowly increases, suggesting that Uber acts as a complement,
not substitute, for transit service. However, these effects differ by size of the
MSA and transit agency: Uber reduces transit ridership in smaller MSAs, where
transit’s inflexibility makes Uber an attractive substitute, but Uber increases
ridership in larger MSAs, where its ability to extend the transit system’s reach
makes it a good complement to transit. The researchers also find that smaller

Infrastructure Investment to Boost Productivity | 189

transit agencies, especially those in large cities, saw increased ridership after
Uber’s entry. For larger transit systems, Uber’s impact was to decrease transit
use by an estimated 2 percent.
Ultimately, Uber’s overall effects on welfare will include multiple effects
on consumer surplus, transit use and farebox revenues, congestion and safety,
and public officials will need to monitor and respond to these technologically
driven forces. Further, States and localities may need to adjust their tax and
regulatory regimes to insure that all users of public roads and transportation
infrastructure pay for congestion costs generated, including ride-sharing companies (Povich 2017). Chicago, Portland, and Seattle are among the cities who
have already begun regulating and taxing ride-sharing services, which should
aid in internalizing congestion effects as well as providing revenues for transit
system improvements.
User fees for water, wastewater, and storm water utilities. Although
customers of water and wastewater utilities are accustomed to paying for the
services they receive, user fees and charges have often fallen short of raising
adequate revenues and/or giving customers the right incentives regarding
their consumption levels (Stratton et al. 2017). The sector is characterized by
high fixed costs, and pricing structures typically rely heavily on volumetric
charges. Without significant fixed monthly customer charges in place, providers often cannot earn enough revenues to cover their fixed costs. Furthermore,
the sector is highly fragmented, with most individuals in the United States
being served by one of 50,259 community water systems. Most of the systems
are very small and serve only a few customers; the 431 largest systems, those
serving 100,000 or more, serve 142.2 million individuals.
Overall, the sector faces three key challenges. First, because users rarely
face the true marginal costs of their water use, consumption decisions are
distorted, water is directed to low-valued uses, and providers do not perceive
the true value of additions or improvements to water and wastewater infrastructure. The second challenge, mentioned above, is the sector’s significant
infrastructure needs, without corresponding sustainable revenues to pay for
them. Finally, though providers have raised rates in recent years to better
cover their costs and incentivize customers to use less water in some service
areas, higher rates have become burdensome in some communities, leading to
increased affordability concerns.
Culp, Glennon, and Libecap (2014) argue that charging water and wastewater utility customers true marginal costs of provision will increase incentives
to use water efficiently. They propose improvements in the definition and
enforcement of property rights in water to allow transfers between parties,
directing water to its more valuable uses, a key step in addressing ongoing
drought conditions in the American West. The authors also note wide variations
in water pricing across regions of the country, with agricultural use often priced
below urban use and few instances of full cost recovery. In complementary
190 |

Chapter 4

work, Ajami, Thompson, and Victor (2014) argue that full-cost, increasing block
pricing will help expand water supplies via innovation, as water suppliers will
respond to high customer valuations by increasing investments in research and
development in “smart water,” purification, desalination, conservation, and
other technologies. They further propose implementation of a usage-based
“public benefit charge” whose revenues would be directed toward innovation
and research in the sector. Furthermore, both the studies by Culp, Glennon,
and Libecap (2014) and by Ajami, Thompson, and Victor (2014) emphasize the
role played by Federal, State, and local government regulations, recommending revisions to simplify and streamline rules and to allow markets for water
rights to function more smoothly.
In practice, water and wastewater pricing structures are often variations
on two-part tariff structures. Using data from the 2014 survey conducted by
the American Water Works Association, Mack and Wrase (2017) report that
most utilities use either an increasing block structure (50 percent) or uniform
volumetric charges (29 percent), with the rest using a decreasing block structure or some other tariff structure. Water and wastewater rates have increased
significantly in recent years. The U.S. Department of Energy reported average
annualized growth rates of 4.1 percent for water rates and 3.3 percent for
wastewater rates between 2008 and 2016, compared with annualized growth
of only 1.4 percent in the Consumer Price Index for All Urban Consumers but
5.6 percent in the Consumer Price Index’s subindex for water and wastewater.
As residential rates have increased, affordability concerns have increased
as well. Mack and Wrase (2017) find that meeting the EPA’s affordability guidelines would require household income of at least $32,000, based on average
monthly water consumption of 12,000 gallons. They estimate that as of 2014,
13.8 million households, or 11.9 percent of all households, would face bills
higher than this affordability threshold. Both the Bipartisan Policy Center
(2017) and the Government Accountability Office (GAO 2016) have identified
similar patterns and concerns. In addition, the GAO found that utilities in
shrinking large and midsized cities utilities had responded to financial stress
in part by raising rates, deferring maintenance, and “right-sizing” their water
facilities to match their shrinking populations—by decommissioning plants, for
example. Such efforts to align capacity with demand may entail disinvestment
in some areas.
Many water and wastewater service providers have responded to affordability concerns by establishing or expanding a variety of customer assistance
programs. The EPA (2016b) reports that 228 of 795 water and wastewater utilities reviewed had one or more such programs in place, with wide variation in
program features such as eligibility criteria and structure of assistance. In some
cities, utilities are moving toward explicitly linking rates to income, so that lowincome users face a low or even zero marginal cost for increasing consumption
(Circle of Blue 2017; Philadelphia 2017).
Infrastructure Investment to Boost Productivity | 191

User fees and equity/efficiency trade-offs. Charging user fees linked to
income instead of marginal cost of service provision can improve equity but
comes at the expense of efficiency and, in some cases, cost recovery, as seen
above in the context of roads, transit, and water utilities. On one hand, encouraging efficiency in use requires that consumers face true marginal costs, along
with possible fixed charges to help defray fixed costs. On the other hand, high
volumetric and/or fixed charges may discourage low levels of consumption at
the intensive or even extensive margin, detracting from efficiency, equity, and
cost recovery goals. Resolving these tradeoffs can be difficult, and preferred
options may differ by the service at issue.
For example, policymakers may be willing to impose road usage charges
for their substantial efficiency and revenue effects, because there are often
close substitutes such as nontolled roads or transit that are available to serve
transportation needs, and, as Hall (2015) has argued, in some cases, time-varying road tolling does not even create a tradeoff between efficiency and equity.
Similarly, increasing transit fares would improve efficiency and cost recovery
in addition to providing valuable signals to policymakers about optimal capacity. In the water sector, some policymakers may prefer below-marginal cost
pricing for lifeline residential water consumption, giving up some efficiency
and revenue gains in exchange for increased equity; the sensitivity of users to
price will determine the efficiency “price” of achieving equity goals. On balance, policymakers wishing to maximize social surplus may wish to limit price
distortions by encouraging true marginal cost pricing and addressing equity
concerns via pro-growth policies and progressive tax and transfer programs as
needed, recognizing that residential mobility will limit the ability of local and
sometimes State governments to engage in too much redistribution.

Financing Infrastructure Investment
Once revenue sources are identified to support particular infrastructure
projects or categories, financial plans must be developed. Creative financial
structures do not negate the need to identify adequate and appropriate funding resources, but they can be used to better allocate risk, align incentives,
and lower costs of infrastructure investments and service provision. Recall
that overall, States and localities own, fund, and manage most of the Nation’s
infrastructure assets, contributing 77 percent of all public spending on transportation and water infrastructure (CBO 2015). This suggests that the Federal
role, though important, is limited. That said, Federal support for infrastructure
spending takes several forms, including grant funding for States and localities;
access to subsidized credit through direct or indirect loan programs; and the
favorable tax treatment of municipal securities. In this subsection, we briefly
discuss these three tools.
Federal grant funding for States and localities is a key financing source
for their infrastructure programs, and direct Federal spending is quite limited.

192 |

Chapter 4

For highways, most grant funds are distributed based on statutory formulas,
which can include factors such as population, lane miles, and other factors.
The 2015 Fixing America’s Surface Transportation (FAST) Act authorized $207.4
billion in grants under the Federal-Aid Highway Program for the FY 2016–20
period, all of which are apportioned by statutory formula (FHA 2017). States
must generally contribute $.20 for every $.80 provided in Federal funds, but
less ($.10 for every $.90 in Federal funds) for interstate highways. Substantially
less Federal grant funding is allocated on a competitive basis; Lew (2017)
estimates that the U.S. Department of Transportation’s (DOT’s) largest competitive grants accounted for less than 2 percent of DOT’s budget. In fact, only
$4.5 billion was authorized for FY 2016–20 for the competitive Infrastructure for
Rebuilding America (INFRA) grants program, which is intended to provide assistance for projects of national or regional significance, far less than the amount
directed to formula highway grants. Another competitive grant program, DOT’s
Transportation Investment Generating Economic Recovery program, known
as TIGER, which seeks to support projects having a “significant impact on the
Nation, a metropolitan area, or a region,” is also relatively small, with a $500
million appropriation for FY 2017.
For water and wastewater infrastructure, the Federal government’s
primary support has come through EPA grants to the States to capitalize
State-administered revolving loan funds, which in turn provide low-cost loans
to service providers for infrastructure projects. Federal appropriations for the
revolving loan funds have been essentially flat for nearly 20 years; in FY 2017,
the Clean Water State Revolving Fund allotments totaled $1.394 billion and the
Drinking Water State Revolving Fund allocations totaled $824 million. Like the
highway grant programs, these EPA programs typically require a 20 percent
match against federally provided funds (Vedachalam and Geddes 2017). The
loans themselves are repaid with revenues raised from customers along with
general tax revenues collected from local taxpayers.
Because grant funding is such a big component of resources used by
States and localities to fund infrastructure projects, the Federal government
has great opportunity and scope to shape nonfederal decisionmaking in
several ways. One obvious way is through the strategic choice of matching
requirements. Grant programs requiring a 20 percent matching of Federal
funds essentially offer cheaper funding than those requiring, say, a 50 percent
matching of Federal funds, and Federal officials can require grant recipients to
meet certain conditions—for example, a maintenance-of-effort provision—to
receive more generous matches. Alternatively, Federal officials could require
grant recipients to devote some minimum amount of resources to maintenance and repair, or to resiliency and disaster recovery planning, as conditions
of receiving Federal support.
Another option is to incentivize better project selection by grantees and
direct more grant dollars to competitive instead of formula-based programs,
Infrastructure Investment to Boost Productivity | 193

which could in principle increase the effectiveness of any given amount of
Federal grant funding. For example, the INFRA competitive grant program
requires the preparation of a cost/benefit analysis of the proposed project,
but the amount of funding at issue is relatively small. The CBO (2016a) also
highlights the importance of directing Federal dollars toward projects with the
greatest returns, as evidenced by cost/benefit analysis. In some instances, the
CBO (2016a) indicates that such a redirection would entail spending more on
major road and highway repairs, especially in urban areas, and less on overall
system expansion. Kahn and Levinson (2011) and Glaeser (2017) all emphasize
the value of maintenance spending and the importance of applying cost/benefit analyses to project selection at the State level.
Finally, policymakers should recognize the potential costs that come
with accepting Federal grant support for projects. Federally funded highway
projects, for example, come with Federal requirements related to environmental reviews, prevailing wages, and Buy America provisions, and Federal aid
dollars cannot be used on local roads (urban or rural) or rural minor collector
roads. Some States have established programs in which local governments can
exchange, at a discount, some of their Federal grant funding from the Federal
Highway Administration for less encumbered state funding. Kansas, for example, established its “Federal Fund Exchange” program in 2010, allowing local
public agencies to exchange $1 in Federal funding for 90 cents of state funding.
This gives these agencies more flexibility in project selection, and the State
uses the Federal funds for projects on State-owned roads and highways. Other
States (e.g., Indiana, Nebraska, Oregon, and Utah) have similar programs, with
exchange rates ranging from 75 cents to 94 cents on $1. The existence of these
programs and similar “after markets” for Federal grant funding indicates that
the cost of accepting Federal funds can be material and that local officials value
flexibility so they can direct funding to the projects best for local constituents.
In addition to providing grants to States and localities, the Federal
government also provides a variety of credit resources to States and localities, ranging from direct loans to loan guarantees and other instruments
intended to facilitate low-cost access to capital markets. DOT’s Transportation
Infrastructure Finance and Innovation Act (TIFIA) program provides secured
loans, loan guarantees, and/or standby letters of credit for projects of regional
and national significance. The FAST Act authorized up to $1.4 billion in TIFIA
funding over the FY 2016–20 period. TIFIA loans must be secured by “dedicated
revenue sources,” which can include tolls, user fees, TIF revenues, and other
tax revenues pledged to repayment.
The Federal government took a similar approach in the area of water
infrastructure when, in 2014, the Water Resources Reform and Development
Acts established a pilot program called the Water Infrastructure Finance and
Innovation Act. Under this program, the Federal government may provide direct
loans and loan guarantees for eligible borrowers, aiming to support larger
194 |

Chapter 4

$"0- уҊшѵ
 2
ѵѵ
// 
)
 *'
*1 -)( )/
 /
 ..0 .Ѷ
сппу–рх


Dollars (billions)
уфп

 1 )0

 ) -'
*'$"/$*)

упп
тфп
тпп
сфп
спп
рфп
рпп
фп
п

сппу

сппх

сппч

спрп

спрс

спру

спрх

*0- ѷ   -'
 . -1 
*-ѵ

projects than are usually funded by State revolving fund loans. Vedachalam
and Geddes (2017) argue that the program can lower debt service costs for
participating borrowers. Eligible projects related to drinking and clean water
must have costs exceeding $20 million for large community projects (areas
with more than 25,000 people) and $5 million for small community projects
(areas with less than 25,000 people).
The third key Federal support for infrastructure investment involves the
tax treatment of municipal debt. State and local governmental entities rely
heavily on borrowed funds to finance their public investments—and in doing
this, they benefit from the preferential tax treatment of municipal bonds issued
for governmental and qualified private purposes. In brief, the tax payments
made to owners of such debt are not taxable for Federal income tax purposes,
allowing municipal bond issuers to pay lower interest rates in equilibrium
than they would otherwise need to pay. The rationale for this exemption
is that some infrastructure provides benefits beyond the boundaries of the
jurisdiction making the investment. Without a mechanism to internalize these
externalities, States and localities could underinvest relative to efficient levels.
The tax exemption for municipal bonds cost the Federal government
$28.9 billion in forgone tax revenues in 2016 on an outstanding stock of over $3
trillion in securities issued by States and local governments (Federal Reserve
2017). Figure 4-9 shows that State and local bond issuance has risen in recent
years, reaching $431.3 billion in 2016. Revenue bonds, which are secured
Infrastructure Investment to Boost Productivity | 195

Box 4-4. Public-Private Partnerships
Public-private partnerships (P3s) allow for innovative and efficient, though
not free, procurement of infrastructure projects. When State and local
government leaders work with private partners to address infrastructure
deficiencies, there are potential synergies for both parties. Large, complex
projects with dedicated funding sources supported by tax revenues, user
charges, or other revenue sources can be provided more efficiently using P3s
rather than traditional procurement methods. Projects that offer meaningful
opportunities to decrease life cycle costs by combining design, build, operate,
maintain, and sometimes finance services into one contractual relationship
are good candidates for P3s, as private partners can contribute capital, project management expertise, and risk management in return for revenues from
the government partner.
Traditional procurement deals typically give private contractors little
incentive to consider the lifetime costs of a project, whether monetary or
opportunity. Under traditional procurement methods, for example, a design
team contracted for a project would typically not be responsible for building, operating, or maintaining the facility over its lifetime, thus would have
little incentive to consider processes that would streamline the construction
process, accelerate project delivery, or minimize maintenance needs over the
project’s lifetime. In a P3 partnership, the private partner could be responsible for designing, constructing, and maintaining the project. Therefore,
incentives are aligned for efficiencies throughout the process, for both private
and public sector parties.
P3s can also decrease risk related to uncertain future demand, cost
overruns, construction delays, and the like, though it is important to note
that reducing public sector risk will be priced into the P3 agreement. More
generally, P3s allow risk to be borne by the party best equipped to handle that
risk. For example, regulatory risk (the risk that a project may be scuttled due
to regulatory or permitting actions) is likely best borne by the governmental
partner, while the private partner likely has greater project management and
construction expertise and is therefore in the best position to manage that
risk. Demand, or revenue, risk may be shared or borne in full by one party or
the other, depending on the project’s particular features.
Despite these benefits, P3 partnerships are uncommon in the
United States. A report published by the U.S. House of Representatives’
Transportation and Infrastructure Committee finds that from 1989 to 2013,
98 highway P3 projects totaling $61 billion were completed. These projects
equal only 1.5 percent of approximately $4 trillion spent on highways during
that period by all levels of government. Currently, 34 U.S. States, the District of
Columbia, and one U.S. territory have enacted statutes that enable the use of
various P3 approaches for the development of transportation infrastructure,
as shown in figure 4-iii.

196 |

Chapter 4

Figure 4-iii. P3 Enabling Legislation, by U.S.
State and Territory
With P3 enabling legislation
Without P3 enabling legislation
WA
MT

OR

ND

ID
WY
NV

UT

CA
AZ

VT
MN

SD

IA

NE
CO

IL

KS
OK

NM
TX

WI

MO
AR
LA

NY

MI
IN OH
KY

PA
WV VA
NC

TN
MS AL

GA

SC

ME

NH
MA
CT RI
NJ
DE
MD
DC

FL

AK

HI

Puerto Rico

Source: U.S. Department of Transportation, Federal Highway
Administration.

Moreover, a 2016 report from Moody’s Investors Service finds that
though growth of infrastructure P3s in the United States has been slow and
fragmented, the market remains positioned to become one of the largest
in the world. One key provision in accomplishing this target is the recently
passed FAST Act, which created the Build America Transportation Investment
Center, intended to cultivate P3s by helping them access Federal credit and
navigate Federal permitting and procedural requirements.
Examples of P3s. In 2012, the General Assembly of the Commonwealth
of Pennsylvania amended Act 74 to Act 88, which allows private entities to
develop and operate qualifying transportation facilities and to submit solicited and unsolicited proposals; encourages investment by private entities;
and enables the procuring agency to accept offers above the lowest price
offer. Additionally, the act allows terms of up to 99 years for P3 agreements;
authorizes user fees for the subject transportation facility; and requires that
public bargaining unit covered employees displaced by the P3 project be
offered employment with the development entity on terms essentially identical to those in the relevant collective bargaining agreement for its duration.
Through this new mechanism, in 2014, Pennsylvania formed a partnership with Plenary Walsh Keystone Partners (PWKP) to replace 558 structurally
deficient bridges across the commonwealth. As of 2016, 19.8 percent of all
bridges in Pennsylvania were considered structurally deficient (compared
with 9.1 percent across the United States). The Pennsylvania Department
of Transportation (PennDOT 2014) chose the P3 structure to accelerate
the replacement of the bridges and facilitate efficiencies in design and the

Infrastructure Investment to Boost Productivity | 197

construction of bridge components; the selected bridges could be replaced
using a limited number of standardized sizes, designs, and components,
making this bundled approach an efficient one. PennDOT estimated that this
approach will speed up project completion and save 20 percent over the life of
the concession period, compared with PennDOT’s replacing the bridges itself.
The P3 agreement calls for PennDOT to make milestone payments
during the construction phase of the project and availability payments
during the concession period, with clear standards in place for keeping the
bridges in good operating condition; noncompliance with the standards
results in deductions from the payments made to PWKP. To provide the
revenues needed for these payments, the Commonwealth of Pennsylvania
laid the groundwork in 2013, when it enacted Act 89 (HB 1060). When fully
implemented, this law is intended to raise an additional $2.3 billion per year,
including $1.6 billion for roads and bridges highways and at least $476 million for transit, primarily by increasing the sales tax on gasoline as well as a
number of registration and licensing fees.
The financial structure of the agreement is depicted in figure 4-iv.
Including financing costs, the total cost of Pennsylvania’s Rapid Bridge
Replacement Program is $1.1 billion, which includes a record $721.5 million in
private activity bonds (PABs), which are discussed in this chapter’s main text.
Another noteworthy P3 has been the partnership between the City of
Phoenix and American Water Enterprises, Inc., executed to build a new water
treatment plant designed to serve 400,000 homes. The Lake Pleasant Water

$"0- 
уҊ$1ѵ  )).4'1)$
+$
-$" 
 +' ( )/ -*"-(
-*% /
*./
1$'$'$/4

+4( )/.
,0$/4

җ' )-4ҝ'.#Ҙ ‫ڦ‬тфѶчппѶппп
‫ڦ‬фшѶуппѶппп

)/ - ./
 -) 
‫ڦ‬уѶшппѶппп


.' .

+- ($0(
‫ڦ‬црѶшппѶппп

*$'$5/$*)
)

($' ./*) 

+4( )/.
‫ڦ‬ссуѶцппѶппп

*/'
‫ڦ‬рѶррчѶсппѶппп


+-* .
‫ڦ‬цсрѶфппѶппп

*0- ѷ
ѵѵ
 +-/( )/
*!
-).+*-//$*)Ѷ
  -'
 $"#24
($)$./-/$*)ѵ
*/ ѷ
'' 1'0 .
- 
$)
)*($)'
ѵѵ
*''-.ѵ

- +- . )/.
+-$1/ /$1$/4
*).ѵ

198 | Chapter 4

Treatment Plant was completed in 2007 and has a capacity of 80 million
gallons per day, with a potential capacity of 320 million gallons per day (UNC
Environmental Finance Center 2016). The P3 agreement was structured as a
design-build-operate contract, which required Phoenix to pay $228.8 million
for the design and build phases and regular service fees during the 15-year
life of the agreement. The city issued tax-exempt bonds to finance its payment to the private partner, secured by the revenues generated by the water
system from user fees and charges. Through this P3, the city largely met its
goals of reducing project risk and achieving life-cycle savings and efficiencies.
Furthermore, the city was ultimately able to renegotiate the contractual
agreement when lower-than-anticipated water demand and consumption left
the city collecting less water system revenues than planned.
The Path Forward. The future is clear with regard to P3s. There is not
one single actor; instead, the success of P3s depends on coordination and
shared responsibility among multiple entities. States and local governments
may wish to adopt broad P3 enabling legislation and establish offices to
provide technical and administrative assistance for private investors as well
as local governments. Well-structured P3s that provide incentives for efficiency, allocate and price risk appropriately, and protect the public interest
can be an effective way to leverage the skills and resources from the private
sector to accomplish public sector infrastructure goals that would benefit all
Americans.

by identified revenue streams—such as specific taxes, user fees, and other
charges—made up more than 60 percent of total bonds issued in 2016, with
general obligation bonds, backed by the issuer’s faith and credit, accounting
for the rest.
The favorable tax treatment of municipal bonds is only available to debt
that serves governmental (public) purposes or qualified private purposes.
Bonds that pass both the “use test” and the “security test” are governmental
bonds that can be issued without Federal limitation (Congressional Research
Service 2016). Municipal bonds that fail one or both of these tests are not
eligible for the Federal tax exemption. However, Congress has long recognized
that some infrastructure projects provide both private and public benefits, and
since 1968, bonds used to fund certain eligible types of projects and activities
are deemed “qualified” private activity bonds (PABs), which can and do receive
the Federal tax exemption. Currently, 22 categories of projects may be funded
with qualified PABs, and Congress caps the total amount of debt capacity
available each year, with different caps applying to different project categories.
Qualified private activities include exempt facilities projects (airports; water,
sewage, and solid waste facilities; educational facilities; and surface transportation), industrial development bonds, and student loans. In 2016, States and

Infrastructure Investment to Boost Productivity | 199

Box 4-5. Bridging America’s Digital Divide
During the past decade, high-speed Internet service has transformed the
global economy and changed how Americans live their lives. Access to broadband—defined by the Federal Communications Commission as a download
speed of at least 25 megabits per second—is increasingly necessary for modern commerce, community engagement, job creation and matching, education, healthcare, and entertainment. Today, many of even the most common
household Internet tasks require a high-speed connection, due to the rising
sophistication and heavy graphics content of many websites; paying bills,
online banking, shopping, research for homework assignments, and registering a car can be worse than frustrating for those who rely on dial-up access.
However, though just 4 percent of urban Americans lack access to
broadband speeds via fixed terrestrial service, 39 percent of rural Americans
cannot obtain it, as shown in figure 4-v. Low population density, challenging
geographic features like mountainous terrain, and exposure to harsh weather
in certain areas increase the per-customer cost of service delivery, acting as a
disincentive for broadband providers to expand service into rural communities. In addition, broadband providers often face bureaucratic obstacles to
building a network, including arduous application processes and burdensome regulatory reviews.
Even when broadband service is available, rural Americans in general
face a more limited choice set of service providers than their urban counterparts, and tend to adopt at lower rates. According to the Congressional
$"0- уҊ1ѵ
 ..
/*
$3 
 -- ./-$'
-*)
 -1$ Ѷ
спру
$/#
)*
уҝр
. -1$
$/#
)*
сфҝт
. -1$

$/#
)*
рпҝр
. -1$

Percent
уф

тп

рф

п

0-'

*0- ѷ
  -' *((0)$/$*).
*(($..$*)ѵ

200 |

Chapter 4

-)

)$/ 
// .

Research Service (Kruger 2016, table 4), though 44 percent of urban Americans
reside in areas that offer a choice between providers, just 13 percent of rural
Americans do. A Pew Research Center survey of home broadband usage
identified several persistent disparities in broadband adoption, including the
fact that rural Americans tend to adopt broadband at lower rates; 63 percent
of adults in rural areas said they have a high-speed broadband connection
at home, compared with 73 percent of Americans in urban areas. (Similar
gaps in adoption are reported by the National Agricultural and Rural Policy
Development Center, and by the Department of Commerce; Kruger 2016, 6.)
Nonadopting respondents cited cost—of computers and of the service—
as an important reason for not subscribing. And the GAO found that nonadoption is principally driven by unaffordability, a lack of perceived relevance, and
low computer skills. Interestingly, the Pew study also showed that between
2013 and 2015, the share of urban Americans with terrestrial broadband
service declined moderately. This trend toward fixed-line disadoption was
accompanied by an uptick in smartphone adoption; 13 percent of Americans
now rely on the smartphone for online access at home (Kruger 2016, 6–7).
This gap in e-connectivity not only prevents many rural Americans
from participating in the global marketplace but also restricts their ability
to improve other parts of their lives, from their job prospects, placement,
and training, to education and healthcare. Access to broadband is key for
modern private enterprise, and a lack of available infrastructure prevents
investment in rural communities. Several studies show that broadband
availability confers important economic benefits on a community (Kruger
2016, 9). Recognizing that rural America’s economic recovery from the Great
Recession has been far slower than that of the rest of the country, in April
2017 President Trump established the Interagency Task Force on Agriculture
and Rural Prosperity via Executive Order. In its final report, the task force
identified the expansion of e-connectivity as an important path to prosperity
for rural America, and prioritized identifying funding sources, streamlining
the broadband deployment process, and reducing barriers to high-speed
infrastructure buildout.
Provision of broadband in the United States is largely privately organized. Private firms of today face many of the same basic problems that
hindered infrastructure development to expand electrification and telephone
service to rural areas during the early part of the last century: challenging
geographical features and a lack of scale economies in regions with low population densities. The Federal government currently uses two vehicles to direct
funds to broadband deployment: the Universal Service Fund programs of
the Federal Communications Commission, and the broadband and telecommunications programs of the U.S. Department of Agriculture’s Rural Utilities
Service (Kruger 2016, 12).
There are many options for improving the deployment and adoption
of high-speed Internet connections to unserved and underserved areas, but

Infrastructure Investment to Boost Productivity | 201

a key consideration is striking the right balance between providing Federal
assistance where private options are unavailable or unaffordable and minimizing the detrimental effects that government intervention can have in the
private marketplace. A wide array of instruments are available to policymakers—from loans and loan guarantees, to infrastructure grants and universal
service reform, tax incentives, direct assistance to taxpayers, and regulatory
and deregulatory measures (Kruger 2016, 23). In deciding on the appropriate
method(s), however, it is important to proceed with an understanding of the
availability of next-generation and mobile broadband technologies, because
these may prove less costly and more desirable to consumers in the long run.
To advance the goal of increased access, the Federal government recently
announced that the Department of the Interior will make some of its real
property assets available for deployment of rural broadband assets.

Box 4-6. Transitioning to the 21st Century:
The Case of 5-G Wireless
Maintaining a competitive economy into the 21st century will require not only
upgrading, expanding, and enhancing conventional infrastructure assets but
also investing in new, innovative, and potentially disruptive technologies.
These technologies have the potential to profoundly alter economic relationships and increase productivity throughout America and across industries,
thereby supporting economic growth. Although the private sector is likely to
lead investments in many of these technologies, the public sector will shape
future investment choices made via its regulatory and other policies. The
example of 5G wireless technology highlights some of the risks and opportunities of these technological innovations.
Industry analysts project that the 5G market will develop rapidly.
Deloitte expects 5G trial markets to materialize by the end of 2017 and
develop into a “full, mass market” by 2020. Whereas the cellular infrastructure of the past relies in its entirety on large towers, this new 5G cellular
infrastructure will require the deployment of smaller cellular transmission
devices (often referred to as “small cells”) to augment traditional cellphone
towers (Gupta and Kumar Jha 2015). Due to the nature of wireless transmission, the addition of these smaller cellular devices will enhance the capacity
of wireless networks to transmit data. With improved capacity and speed that
improve connections of digital technologies, 5G may support the flourishing
of the “Internet of Things”—including driverless cars and high-technology
healthcare systems. Such technologies are projected to boost connectivity,
productivity, and output. By 2035, IHS (Campbell et al. 2017) projects that 5G
could support $12.3 trillion in global economic activity.

Industry analysts also expect 5G to boost high-wage employment,
lowering job search, match, and telecommuting costs, perhaps of special
value in distressed communities with limited job opportunities. In addition,
various traditional infrastructure sectors may benefit from the deployment of
5G service, including energy and utilities (e.g., energy-consuming devices in a
grid) and transportation (e.g., 5G-powered traffic management systems), as
well as public safety (e.g., integration of video surveillance).
There are two main challenges for 5G development. The first challenge is standards. Attracting private sector investments will require clarity
about the future path of the technology itself. Setting specific technological
standards for 5G wireless facilities and operations enables interoperability
and compatibility and will shape future investment choices by firms. Directly,
a country with a dominant industry share may crowd out similar telecommunications exports from other countries because of compatibility and
standards issues. Given the high fixed costs in the industry, the countries and
their companies that initiate the standard may gain first-mover advantage,
making it difficult for new entrants with different standards to enter. For
example, industry sources indicate that the adoption of China’s Polar Code
Error Connection technology for encoding 5G in November 2016 is a symbol
of China’s rising leadership in 5G technology (Rogers 2016; Lucas and Fildes
2016). Indirectly, the dominating nation may have preferential access to
foreign intellectual property that is using the 5G network, which could enable
theft of this property, an issue that is discussed in chapter 7.
Standards for 5G technologies are developed by multistakeholder
organizations, such as the 3rd Generation Partnership Project (3GPP), and are
ultimately codified at the International Telecommunications Union, a United
Nations agency that coordinates global telecommunications operations and
services. To date, the U.S. has pursued a standard-setting approach led by
the private sector, whereby product standards are generally set through
voluntary, private organizations. In contrast, many other countries engage in
active governmental direction of standard-setting activity. For example, the
Organization for Economic Cooperation and Development (OECD 2010) notes
that there is active participation by European governments in the Global
System for Communications’ mobile phone standard. Similarly, in the context
of the telecommunications industry, Linden (2004) notes that China maintains
rights “to involve government in all standard-setting decisions.” Heavy government involvement in international standard setting may be concerning if it
crowds out private actors due to governments’ larger economies of scale and
capital or if such involvement is coordinated to disproportionately benefit
particular nations.
The second challenge for 5G development is regulation. Establishing a
flexible and adaptive regulatory structure will be needed to support future 5G
deployment, with coordination across Federal, State, and local government
levels. Specifically, the April 2017 Notice of Proposed Rulemaking, issued

Infrastructure Investment to Boost Productivity | 203

by the Federal Communications Commission (FCC) regarding “accelerating
wireless broadband deployment by removing barriers to infrastructure
development,” sought comments on two sections of the Communications
Act, Sections 253 and 332. Section 253 delineates the rights of State and
local authorities to collect “fair and reasonable compensation from telecommunications providers” but also prevents any State or local government
from prohibiting “intrastate or interstate” telecommunications service. Local
authorities can stymie rapid deployment of infrastructure by delayed disposition of requests for local rights of way and siting approvals. Section 332
requires that State and local governments not discriminate between service
providers who want to site cellular infrastructure, refrain from setting prices,
and respond to such requests within “a reasonable period of time.” However,
many local authorities may not be equipped to understand the impact of
small cell deployment, which does not disturb the public rights of way as
traditional wireless infrastructure, such as cell towers. Though the FCC has
solicited input on the subject, it has yet to implement decisions about how it
will balance the interests of the different stakeholders involved in the physical
rollout of 5G.
Governments may also ensure that 5G service providers have access
to the appropriate spectrum, or the radio frequency waves over which the
signals are transmitted. Unlike the large cell towers of traditional wireless
infrastructure, 5G’s “small cells” transmit electromagnetic waves at a variety
of frequencies, ranging much higher than those on which previous wireless
data services have relied. To generate economic value from 5G infrastructure,
providers must have access to appropriate spectrum frequencies. To ensure
the availability of spectrum, the FCC voted in July 2016 to authorize the use
of spectrum bands in the millimeter wave ranges relevant to 5G. These bands
may eventually become available through overlay auctions and the secondary
market and will benefit both 5G operators and current owners of these rights.
Thus, though investment funding and asset ownership in this sector
are currently dominated by the private sector, Federal officials have opportunities to make policy decisions that will shape the environment for future
private investment in this sector, allowing the United States to take best
advantage of the benefits offered by this new technology.

localities issued $20.4 billion in qualified PABs, of which about two-thirds were
directed toward affordable multifamily housing projects (CDFA 2017, 9).
PABs have proven to be especially valuable in projects structured as
public-private partnerships (P3s). The Safe, Accountable, Flexible, Efficient
Transportation Equity Act of 2005 authorized the issuance of up to $15 billion in
PABs for use in transportation P3s. As explained in box 4-4, P3s offer an alternative to traditional project procurement, whereby a private sector entity or consortium contracts with the relevant State and/or local governmental bodies

204 |

Chapter 4

to design, build, finance, operate, and/or maintain infrastructure facilities.
Allowing private entities to issue tax-preferred PABs to finance such projects
is simply the equivalent of allowing the public sector to issue governmental
purpose bonds.
On balance, the Federal government has a key, if limited, role to play in
both funding and financing of infrastructure investments. Increasing investment to address infrastructure needs will take additional resources from
Federal, State, and local government taxpayers as well as the direct beneficiaries of the assets. On the funding side, reliance on user fees to pay for
investments has its limits, but significant efficiency gains can still be achieved
through careful expansion of their use. On the financing side, the Federal government can use grant funding as an incentive to encourage States and localities to be more efficient when undertaking infrastructure investments and can
promote the use of bonds to support additional infrastructure investment.
The Federal government can also continue to support the use of innovative
financing structures such as P3s to reduce the overall costs of infrastructure
investments.

How Core Infrastructure Ensures
a Competitive Economy
The U.S. economy of course also depends on services from assets in sectors
other than surface transportation and water and wastewater. Maintaining
a competitive and productive economy for all Americans requires a reliable, robust, and resilient energy sector, multiple transportation modes and
systems, and an advanced, productive telecommunications sector. These
infrastructure sectors support trade and economic activity and display significant economies of scale and network effects; yet infrastructure is primarily
privately owned in some instances but publicly owned in others. Therefore, as
this section explains, it is not surprising that barriers to needed infrastructure
expansion and upgrades differ across sectors, with regulatory issues appearing
paramount in some cases but funding challenges being the key issue in others.
For example, consider the telecommunications sector, for which most
infrastructure is privately owned. In some segments of the market, the key
issues are the costs of service relative to revenues collected from users, along
with regulatory concerns, and box 4-5 explores the market for rural broadband service from this perspective. In the case of 5G wireless technology and
investment, issues of regulatory barriers, technological standards, and international competition are more salient. Box 4-6 explores these issues in greater
detail, and highlights recent regulatory actions serving to simplify and clarify
regulatory roles of States and local governments in the wireless broadband
industry and to facilitate markets in which spectrum and transmission rights
can be bought and sold. Aligning regulatory policies with the Nation’s growth

Infrastructure Investment to Boost Productivity | 205

objectives will help ensure that these technologies provide the greatest possible boosts to productivity and growth for all Americans in the years ahead.
In the rest of this chapter, we discuss recent developments in the
energy sector and the inland waterways system, identifying opportunities and
challenges for getting the right infrastructure assets in the right places. We
particularly explore the roles of regulation and funding in shaping investment
decisions in these sectors—and, subsequently, America’s competitiveness and
productivity in the 21st century.

The Energy Sector
Energy infrastructure in the United States is the envy of the rest of the world,
for both fuels and power—if for no other reason than its sheer extent. The
United States has over 2.5 million miles of natural gas pipelines and 207,000
miles of petroleum pipelines, according to 2017 data from the Pipeline and
Hazardous Materials Safety Administration. By some estimations, the North
American electricity grid is the largest such facility in the world. It has 697,000
miles of high-voltage transmission lines and 6.4 million of miles of feeder and
distribution wires (Giles and Brown 2015). These giant networks have been
built piece by piece over a long period, under a range of prevailing market and
regulatory conditions. Addressing the economic and regulatory constraints on
infrastructure investment ensures that future expansions and modernizations
of theU.S. energy networks will be both prudent and timely.
Because energy infrastructure is long-lived, the United States lives with
the legacy of the past. Its electricity grid—which was built by regulated, vertically integrated utilities—differs from the grid that would be built in a restructured market that depends heavily on intermittent generation by renewable
sources, like wind and solar power. Changing market conditions, such as the
restructuring of electricity markets, are an important consideration for infrastructure investments. Restructuring has aimed at aligning investment incentives, but risks remain for new projects. For instance, when it opened in 2009,
the Rockies Express Pipeline (REX) was heralded as a bold, new 1,663-mile
link in the U.S. natural gas system, delivering abundant Western gas to hungry
Eastern markets (Carr 2013). Five years later, the flow in the Eastern reaches of
the $3 billion REX pipeline was reversed to allow newly discovered Eastern gas
to flow to the West.
The REX experience underscores the specificity problem of infrastructure—once it has been built, it cannot be moved. Specificity could lead to concern about underinvestment, but it also opens the door to natural monopoly
power. The high fixed costs and low marginal costs mean that it is socially
optimal to have a single network rather than competing ones. The natural
concern is that the operator would charge high prices to take advantage of
monopoly power, and the traditional remedy has been rate regulation—with
Federal oversight only when infrastructure crosses State lines.

206 |

Chapter 4

Most energy infrastructure—power lines and pipelines, along with the
necessary plants and terminals to serve them—is privately owned; as of 2015,
3.5 of the 6.4 million miles of distribution lines were owned by private utilities,
while the remaining 2.9 million were owned by Federal, State, and municipal
utilities (Giles and Brown 2015). Pipeline infrastructure for both gas and oil is
further skewed toward private ownership; in 2016, 91 percent of pipelines, by
capacity, were owned by corporations. Energy infrastructure is excludable—
enabling suppliers to charge customers for services and access provided; oil
producers to pay for pipeline capacity, electric consumers to ultimately pay
for power to be delivered via wires, and exporters to pay port fees and lading
charges. These user fee revenues ultimately provide the resources needed to
maintain, upgrade, and add capacity, so funding resources are rarely the limiting factor in energy infrastructure investment. Instead, regulatory oversight
has often proved to be the greater hurdle to modernizing and expanding infrastructure (Borenstein, Bushnell, and Wolak 2000).
Pipelines and transmission infrastructure. New technical abilities to
extract natural gas and oil from previously unprofitable regions and States,
such as North Dakota, have increased demand for new pipeline capacity. In
the short term, the lack of available pipeline capacity has increased demand
for alternative forms of transportation, including rail. In the electricity sector,
the falling cost of renewable generation technologies, like wind and solar
power, has increased installations and required transmission facilities that
can accommodate the intermittent nature of these technologies. For both fuel
and power infrastructure, the demand for more transmission capacity in new
regions has made issues related to gaining regulatory permission more salient.
For example, the Keystone XL and Dakota Access crude oil pipelines were
delayed, at least temporarily, by regulatory and legal challenges (see chapter 2
for a related discussion). Significant investments are currently on hold, awaiting regulatory action; at the end of October 2017, Federal approvals for new or
expanded natural gas pipelines were pending for 15 billion cubic feet per day
across a total of 1,630 miles of pipe (FERC 2017).
In the renewables segment of the sector, production and investment tax
credits as well as State-level renewable portfolio standards have encouraged
investments in solar, wind, and geothermal power. With the adoption of these
incentives, as well as improvements in generation technologies, renewables’
share of total generation capacity has risen considerably since 2005 (figures
4-10 and 4-11). Renewable growth accounted for 54 percent of new capacity
additions in 2017, and it has averaged 55 percent of all new capacity additions
since 2005. The share of electric generating capacity contributed by renewables has climbed from 12 to 22 percent since 2005, and the Energy Information
Administration predicts that this trend will continue through 2050, when
renewable capacity will exceed 35 percent of installed capacity. Falling costs
for renewable electricity generation have triggered an increase in demand
Infrastructure Investment to Boost Productivity | 207

$"0- 
уҊрпѵ
Renewable Energy’s 0-- )/
)
-*% / 

#- 
*!
ѵѵ
 ) -/$*)
+$/4Ѷ сппф–фп

Renewables' share of generation capacity (%)

уп

спфп
-*% / 

- ) 2' .Ѿ

.#-

тф
тп
сф
сп

 ) 2' .Ѿ

.#- .

рф

рп
сппф

спрф

спсф

сптф

спуф

*0- .ѷ
ѵѵ
) -"4
 )!*-(/$*)
($)$./-/$*)Ѷ
))0'
) -"4
0/'**&
)
))0'
' /-$$/4

 +*-/.ѵ

$"0- 
уҊ11. Renewable Energy’s Share of Annual Generation
$/$*).Ѷ
сппх–рц
Renewables' share of annual new capacity additions (%)
чп
цп
хп
фп
уп
тп
сп
рп
п

сппх

сппч

спрп

спрс

*0- ѷ ѵѵ
) -"4
 )!*-(/$*)
($)./-/$*)Ѷ
чхп
 /$' 
/ѵ

208 |

Chapter 4

спру

спрх

for complementary electricity transmission infrastructure. The United States
completed or began construction of 9,277 miles of transmission power lines
between November 2016 and November 2017 (EIA 2017).
Historically, the primary tax incentives for renewables have been the
Renewable Energy Investment Tax Credit, which was introduced in 1978, and
the Renewable Energy Production Tax Credit. Currently, a 30 percent tax credit
is available for investments in solar energy property, fuel cells, and small wind
systems, while a 10 percent tax credit is available for geothermal systems,
microturbines, and combined heat and power property. The Tax Cuts and Jobs
Act of 2017 limits some of the benefits of these credits, however, so their future
value is uncertain.
Port infrastructure. Increasing energy exports is an integral part of the
Nation’s energy dominance vision, but this requires additional infrastructure.
Although shipments to Canada and Mexico are possible using pipelines and
rail, port facilities are required for exporting to other countries. Port facilities
require shoreside links—pipelines for natural gas and petroleum, and rail for
coal. Also, for the pipelines and transmission infrastructure facilities discussed
above, regulatory concerns shape the investment environment.
One example is the struggle to construct a West Coast coal exporting facility so that U.S. coal producers can gain access to the Asian market.
Without such a terminal, expanding exports to Asian markets is effectively out
of reach. In a nutshell, too little of the relevant port infrastructure is on the
Nation’s West Coast and too much is on the East Coast, whose Atlantic ports
accounted for 90 percent of the coal exported by the U.S. to China through the
first half of 2017 (EIA 2017). Several coal companies have expressed interest in
sites in Washington and Oregon for a new, privately funded coal terminal. The
prospect of local tax revenues and employment from such a facility has not yet
overcome State and local opposition to the local disamenity of a coal terminal
and broader environmental opposition to facilitating increased coal usage.
This geographic misallocation of port-related infrastructure limits opportunities to expand coal exports. Figure 4-12 shows how both volume and
revenue from U.S. coal exports have declined over the past several years. No
significant expansions of coal exporting facilities in the United States are currently under construction, despite the opportunities to increase exports to
Asian markets from West Coast facilities.
U.S. natural gas producers face a similar problem in gaining access to
Pacific markets. There are currently no liquefied natural gas (LNG) export
facilities in the Northwest, despite significant commercial interest in building
such facilities. Environmental groups and landowners have opposed a proposed LNG export facility and an associated pipeline in Oregon, and although
development has continued, some of the required permits have not yet been
acquired. Thus, though natural gas exports and capacity utilization rates rose
significantly between 2007 and 2016, future growth in exports, capacity, and
Infrastructure Investment to Boost Productivity | 209

$"0- 
уҊрсѵ
*/'
*'
3+*-/.
4
 $"#/
)
 1 )0 Ѷ
сппс–рц
Millions (short tons)

Dollars (billions)

руп

спрц

рч
рх

рсп

ру
рпп

рс

чп

рп
*/'
 3+*-/.
җ' !/
3$.Ҙ

хп

ч
х

уп

*/'
- 1 )0
җ-$"#/
3$.Ҙ

сп

у
с

п
сппс

сппц

спрс

спрц

п

*0- .ѷ
ѵѵ  +-/( )/
*!
*(( - Ѷ
 ).0.
0- 0ѵ

$"0- 
уҊртѵ
ѵѵ
))0'
/0-' .
3+*-/.Ѷ
сппр–рц

 3+*-/.

$+ '$) 
".
 3+*-/.

Cubic feet (millions)

тппѶппп

0"Ҋспрц

сфпѶппп
сппѶппп
рфпѶппп
рппѶппп
фпѶппп
п
сппп

сппф

*0- ѷ ѵѵ
) -"4
 )!*-(/$*)
($)$./-/$*)ѵ

210 |

Chapter 4

спрп

спрф

capacity utilization is constrained by a lack of facilities needed to export, especially in the rapidly growing LNG market segment (figure 4-13). (The Federal
Energy Regulatory Commission has approved several LNG export facilities,
which are currently under construction, but none yet are on the West Coast.)
Export capacity utilization was at 59 percent of potential as of 2017, up 7.5
percentage points from the previous year. LNG’s share of total gas exports has
expanded rapidly in the past decade. In the 2001–10 period, LNG averaged
less than 0.5 percent of total U.S. natural gas exports; by August 2017, LNG
accounted for 20 percent of total gas exports. There remains a large international market for LNG, in which the United States has not yet carved out a share
proportional to its production capabilities. According to data from the Energy
Information Administration, the United States accounted for 1.2 percent of
global LNG exports in 2016, despite being the largest gross extractor of gas in
2015 among all nations. The U.S. is drastically underrepresented in the global
LNG market; and by expanding its LNG export capabilities, it most likely could
rapidly gain market share.

Modernizing America’s Waterways
The Nation’s inland waterways system (IWS) is a crucial component of its
transportation network, linking the producers of agricultural and energy commodities to domestic and international markets. But this system is aging, and
its users are suffering from increasing lost transportation time. Unlike other
freight modes, where the costs are mostly borne by system users, for historical
reasons the government pays almost the entire cost of operating the IWS. The
existing funding structure actually disincentivizes making timely repairs and
does not align system costs with the parties that most benefit from IWS usage.
A more robust system of user fees—possibly in the form of multipart tariffs that
include licenses, location-specific fees, congestion fees, and fuel taxes—is the
most promising approach to achieving revenue adequacy and sustainability,
facility reliability, and economic efficiency. By providing signals of system component value, such fees would also guide operators and policymakers in deciding where to focus capital expenditures and how to prioritize repair efforts.
The IWS includes more than 36,000 miles of navigable rivers, channels,
and canals across the United States, and directly serves 41 States (Clark,
Henrickson, and Thomas 2012; TRB 2015). Upstream and downstream movement of cargo is enabled by lock infrastructure managed by the U.S. Army
Corps of Engineers (USACE). Movement of goods and people over inland
waterways was an important factor in the Nation’s early economic growth, and
the system remains a small but stable part of the United States’ commercial
transportation system, accounting for between 6 and 7 percent of all ton-miles
(TRB 2015). Water transportation contributes about $15 billion in value added
to U.S. GDP, about 0.1 percent of the total size of the economy. According to
DOT, inland waterways support more than 270,000 jobs.

Infrastructure Investment to Boost Productivity | 211

For many commodities—particularly those that are heavy and transacted
at relatively low prices—the waterways system is an important component
of their transportation network, including coal, petroleum, chemicals, and
agricultural products. For example, grain is shipped via rail from the interior,
loaded onto waterside grain elevators along the Upper Mississippi River, transloaded first onto barges, and then moved downstream to southern Louisiana,
where it is then transloaded onto deepwater vessels that sail to export markets
around the globe. Compared with truck or rail, water transportation is in many
cases a less costly means of moving goods (USACE 2016).
Freight traffic across the system is highly variable; about 22 percent of
the total waterway miles account for about 76 percent of the cargo ton-miles
transported (USACE 2013). However, low-use tributaries can be critical sources
of transportation for freight systems that are organized around the low-cost
water transportation of bulk commodities on these segments; few economical
alternatives exist for these industries if low-use segments are no longer operable for commercial navigation (TRB 2015, 42). The Upper Mississippi, Illinois,
and Tennessee-Tombigbee rivers, and the Gulf Intracoastal Waterway, have
high-use locks in moderate or even low-use waterway sections, due to seasonal
peaks in the movement of certain commodities, like harvested agricultural
commodities, or because of seasonal navigation closures (due to recurring
weather conditions like ice and flooding).
The cost of poor infrastructure. According to USACE (2014), waterways’
infrastructure in the United States is operating at an overall satisfactory level.
However, the average age of system locks is increasing, even when adjusted
for date of last major rehabilitation (TRB 2015, 44). Furthermore, though
systemwide traffic is flat or declining, delays and scheduled lock outages (to
proactively address maintenance issues) are actually increasing, as shown
in figure 4-14. Shipping delays and lost service are positively correlated with
tonnage handled, indicating that investments are necessary to improve this
transportation system.
Delays are typically longer at locks with greater demand for transportation during the harvest period for U.S. agriculture, so these are in part driven
by seasonal congestion; in addition, locks experiencing the largest number of
delays are concentrated along medium- and high-use segments of the system
(TRB 2015).
Several studies have estimated significant cost effects of shipping delays
and outages. The University of Tennessee’s Center for Transportation Research
and the Engineering Center for Transportation and Operational Resiliency at
Vanderbilt University (CTR 2017) have estimated the effects of unscheduled
lock outages on additional transportation costs, and focused on four locks.
Calcasieu Lock is critical for inland navigation between Texas and Louisiana,
and the vast majority of its traffic is dominated by petroleum and chemical
products. CTR estimated that an unscheduled outage at Calcasieu would
212 |

Chapter 4

$"0- 
у–руѵ
 *./
-).+*-//$*)
$( 
-*..
/# 
 )')

/ -24.
4./ (Ѷ
сппп–спрх
# 0' 
0)1$'' 
/$(
Hours per year (thousands)
рѶпфп

).# 0' 
0)1$'' 
/$(

))0'
 '4

спрх

шпп

цфп
хпп
уфп
тпп
рфп
п
сппп

сппс

сппу

сппх

сппч

спрп

спрс

спру

спрх

*0- ѷ
ѵѵ
-(4
*-+.
*!
)"$) -.ѵ

increase transportation costs for these products by more than $1.1 billion.
LaGrange Lock and Dam and Lock & Dam 25 are both primarily dominated
by Gulf-destined, down-bound flows of corn and soybeans; 20 million tons of
farm products flow through these two locks each year, six times greater than
the volume of farm products that are moved by rail through the same corridor.
CTR estimates the cost of an unscheduled closure at either LaGrange or Lock
& Dam 25 at $1.5 billion. Yu, English, and Menard 2016 estimate that a oneyear closure of Lock & Dam 25 would reduce economic activity from corn and
soybean production by $2.4 billion, leading to the loss of 7,000 jobs and $1.3
billion in labor income. Traffic at Markland Lock is primarily composed of shorthaul coal movements, chemicals, and petroleum products; CTR estimates that
an unscheduled closure of the lock would increase the shipping costs of these
commodities by $1.3 billion.
Funding the Inland Waterways System. The Federal government’s role in
managing and funding the IWS is far larger than it is for other freight modes.
Although it is responsible for about 28 percent of highways spending and
almost none of the cost of pipelines and railroads, the federal government
contributes about 90 percent of the IWS’s cost (TRB 2015). Waterways costs
are mainly funded via the USACE budget. Operations and maintenance (O&M)
costs constitute $631 million (69 percent) of the total FY 2017 budget of $917
million, with only $243 million (26 percent) devoted to construction (USACE
2017a). The Trump Administration’s inland navigation system’s FY 2018 budget
Infrastructure Investment to Boost Productivity | 213

$"0- 
у-рфѵ
 )')
/ -24.
-0./
0)
') Ѷ сппш–рх
Dollars (millions)
хп

фп

уп

тп

сп

рп

п

сппш

спрп

спрр

спрс

спрт

спру

спрф

спрх

*0- ѷ )')
/ -24.
. -.
*-
))0'
 +*-/Ѷ
  ( -
спрхѵ

requests that 77 percent of projected expenses be devoted to O&M expenses,
which include repair costs up to $20 million and are fully funded from Federal
general revenues. Construction costs, including repairs over $20 million, are
funded through a combination of a direct tax on barge fuel and matching general funds from the Federal government.
The current funding framework presents challenges in several dimensions. On the capital front, tax revenues have not kept up with increased needs
to substantially rehabilitate facilities, echoing the situation in highway funding.
In 2015, the barge fuel tax—which is not indexed to inflation—was raised from
20 cents per gallon to 29 cents per gallon (the first increase since 1995), leading
to an increase in fuel tax revenues and reversing a several-year decline in endof-year balances of the Inland Waterways Trust Fund (Inland Waterways Users
Board 2016) (figure 4-15). Even so, the revenues generated by the fuel tax are
estimated to be only $112 million annually, compared with total annual IWS
expenses of nearly $1 billion and an estimated $4.9 billion projects backlog.
O&M spending is supported by general funds and must compete with
other priorities in the Federal budget. System managers have an incentive to
delay repairs until they reach the point of being classified as capital expenses—
because those can be covered by fuel tax revenues—reducing system reliability
by leading to delays and closures. Eliminating the funding wall between fuel tax
revenues and O&M expenditures is an obvious step that would both improve

214 |

Chapter 4

the reliability of O&M funding streams and counter the incentives problem that
now serves to exacerbate the deterioration of facilities.
Beyond improving revenue generation and permitting the system to
cover its own costs, imposing increased fees on barge operators for using the
IWS and its facilities would enhance its economic efficiency. Ideally, these fees
would be set to match the marginal costs generated by usage of the existing
system. With facilities already in place, short-run marginal costs are those
associated with operating and maintaining locks and dams, and maintaining
channel depths. However, because new construction in the system is characterized by high fixed costs, short-run marginal cost pricing would likely not be
sufficient; fees could need to be set higher to cover current and expected total
system costs.
Such charges could take the form of additional fuel taxes, lockage fees
(charges for passing through individual locks), segment fees, annual license
fees, and/or congestion fees (TRB 2015). Fuel taxes are aligned to usage, but
they apply equally across the IWS, even though some sections are more costly
to operate than others; so if used alone, they would create complicated crosssubsidies. Variable, location-specific fees can be designed to better match
actual marginal costs, but facility-based pricing by itself may not be sufficient
to cover the O&M costs of the shared components that are deemed to be essential to the national freight transportation system; in this case, systemwide
user fees or licenses can be employed. Finally, congestion charges can act as a
demand-management tool similar to peak-load prices in other settings, helping to ration access to the existing infrastructure more efficiently at times when
seasonal use of the system rises—as in the case of agricultural harvests; congestion fees also signal system operators and policymakers about the normal
and seasonal value of the facilities.
In fact, similar issues of fees, cross-subsidies, and revenue adequacy arise
in the context of maintenance and operation of the Nation’s coastal and inland
harbors. Dredging and other costs are covered by the Harbor Maintenance
Trust Fund, which is largely supported by shippers, which pay harbor maintenance taxes of 0.125 percent of the value of cargo loaded or unloaded from
commercial vessels; these taxes made up 88.5 percent of all the Trust Fund’s
revenues in FY 2017. Fund balances have risen over time, reaching $9.1 billion
by September 30, 2017, as annual appropriations have consistently fallen short
of revenues despite significant dredging needs in many harbors and ports.
Further, tax revenues generated have little connection to the costs required to
maintain the harbors, leading to concerns about the distribution of tax burdens
across harbors. As in the case of the IWS, policymakers must assess the impact
of collecting user fees or taxes on those who pay them but also on the Nation’s
transportation system as a whole.
More generally, the willingness of users to pay charges for access to
different segments or facilities of the inland waterways system can signal
Infrastructure Investment to Boost Productivity | 215

their appropriateness for investment, guiding future decisions about which
segments or facilities managers should upgrade, maintain, and/or abandon.
Indeed, given funding constraints, the Transportation Research Board (TRB
2015, 80) indicates that more consistent application of systemwide cost/benefit analysis—including ranking projects in order of urgency—would better prioritize construction projects. In practice, combining these options in the form
of multipart tariffs may be the most promising approach to achieving revenue
adequacy and sustainability, facility reliability, and economic efficiency. The
President’s FY 2018 Budget includes a proposal to “reform the laws governing
the Inland Waterways Trust Fund, including by establishing a fee to increase
the amount paid by commercial navigation users of inland waterways.”

Conclusion
Policymakers have considerable scope and opportunities to shape the Nation’s
growth prospects by improving its infrastructure. Making more efficient use
of increased, higher-quality capacity can make meaningful contributions to
economic growth. Under a range of assumptions, we estimate that a 10-year,
$1.5 trillion infrastructure investment initiative could raise average annual real
GDP growth by between 0.1 and 0.2 percentage point. Although more funding
may be needed from both public and private stakeholders to conduct such a
program, this chapter has also discussed other levers and options available to
policymakers, who must confront and manage the threats and opportunities
around conventional uses of public infrastructure and sources of funds across
varied sectors. For example, technological change and disruption in the transportation sector threaten conventional funding models for roads and transit
services, and increased congestion and overuse of some assets suggests that
the efficiency and revenue benefits of more creative and consistent implementation of congestion pricing will be considerable.
More generally, governments should be encouraged to generate needed
revenues from user charges on those who benefit from publicly provided
roads, water facilities, and other infrastructure. These user fees should reflect
the true marginal costs of service provision, serving to increase allocative
efficiency, provide signals about the value of future capacity additions and
improvements, and raise needed revenues to defray the costs of provision.
Policymakers should be sensitive to possible trade-offs among efficiency,
equity, and revenue goals but recognize that nonmarginal-cost pricing distorts
incentives and decreases overall surplus. Developing and incentivizing the use
of value capture programs where appropriate would also increase available
funding resources for infrastructure investment, as parties experiencing capital
gains (e.g., increased property values) would help pay for the costs of the infrastructure investment responsible for these gains.

216 |

Chapter 4

The Federal government also has other tools at its disposal. It can support the continued use of innovative financing options such as public-private
partnerships and private activity bonds to increase the availability of investable
dollars and lower the cost of debt service. And it can enhance the capabilities of
State and local governments to allocate scarce investment funds efficiently by
encouraging the use of cost/benefit analysis and continuing project selection
by States and local governments whenever possible, allowing local officials to
make infrastructure investments of greatest benefit to their constituencies.
Better project selection will be important in driving growth throughout the
country.
Finally, the Federal regulatory structure must adapt to ensure the pursuit of health, safety, and environmental goals without distorting investment
incentives. Conflicting, unduly complex, and uncoordinated rules and regulations can impede investments in needed infrastructure and limit the productivity of existing assets, as described above in the context of water markets, rural
broadband and 5G-wireless technologies, and the energy sector. Addressing
these issues will take time but generate significant public benefits, and several
recent Executive Orders and other regulatory actions begin that process.
On balance, with appropriate infrastructure funding, financing, and regulatory policies in place, the United States can look forward to a productive and
prosperous 21st century.

Infrastructure Investment to Boost Productivity | 217

x
Chapter 5

Enhancing U.S. Trade in
a Global Economy
Trade across international borders has motivated economic analysts since at
least the 19th century, when the economist David Ricardo invoked the example
of wine in Portugal and wool in England to illustrate the principle of comparative advantage. But the economy, both in the United States and around the
world, has changed since the days of David Ricardo and Adam Smith. Although
the economics profession has converged toward a consensus on certain principles, the Administration’s trade agenda also stands poised to update existing
trade relationships in order to maximize the benefits that America’s trade with
the world generates for our citizens in the 21st century and beyond.
The United States, for instance, faces higher barriers on its exports in markets
abroad than producers abroad face on their exports to the U.S. Nothing about
the principle of comparative advantage would lend itself to a defense of a
status quo that imposes higher barriers to exports on America’s producers than
on foreign producers.
The global trade system has come under strain due to the influence of countries, like China, that violate market principles and distort the functioning of
global markets. When America’s businesses and workers can compete in the
global economy on a level playing field, however, our underlying dynamism will
allow our economy to flourish. The Administration prioritizes its attempt to create the conditions that, according to the consensus principles in the economics
literature, would maximize the benefits accruing to the United States—and
produce gains for our trading partners as well.

219

Throughout America’s history, trade has produced costs as well as benefits. In
recent years, the economics literature has identified portions of the American
population for whom the costs of recent trade expansions have exceeded the
benefits. Even if fair and reciprocal international trade as a whole leaves the
U.S. better off in the aggregate, this does not necessarily mean that the benefits
of expanding trade flows leave all Americans better off. Indeed, new empirical
evidence suggests that certain trade flows with China may have left some
Americans worse off.
As the Administration continues to strengthen and update trade agreements
and to pursue its trade agenda, however, the United States stands poised to
capitalize on opportunities to reap the gains from trade that it has historically
enjoyed. This Administration’s focus on improving trade agreements will
benefit American businesses and American workers across a variety of sectors—in particular, the U.S. energy and agriculture sectors possess comparative
advantages and may be able to increase their exports to the rest of the world.

A

nalyzing the causes and consequences of trade across national borders has interested those who study economic activity for centuries.
In the early 19th century, David Ricardo used hypothetical trade in
English cloth and Portuguese wine to illuminate the principle of comparative
advantage—the idea that by specializing and trading, both nations could be
made better off, even if one were more efficient at making both products.
Historically, international trade as a whole has on net increased American
productivity, standards of living, and American economic growth. At the same
time, however, international trade has imposed costs on some Americans.
In contemporary economics, trade is a mainstay field. Though questions
persist about international trade, as in any other active field of economics,
when it comes to the causes and consequences of trade, the economics profession has converged toward a consensus—in certain respects. This chapter
starts by reviewing the sources of gains from trade and the central importance
of comparative advantage, which are areas of universal professional agreement. U.S. trade is then placed in the context of the global economy, including
an economic perspective on international trade balances.

220 |

Chapter 5

The chapter’s second section examines how the gains from trade are
divided between countries. Trade flows are primarily determined by economic
fundamentals like comparative advantage and geography, but policy can also
play an important role. Trade agreements are the major set-pieces of modern
trade negotiations and, as such, deserve analysis and consideration. Bagwell
and Staiger (2001) assert that trade agreements exist to allow nations to commit to a positive-sum approach to trade; Grossman (2016) enumerates the
possible political economy incentives of trade agreements, including efficiency
gains among parties, improved terms of trade between parties relative to
nonmembers, facilitation of multilateral trade liberalization, and provision of
rents to special interests. The commitment embodied in agreements is multidimensional. In addition to formal trade barriers such as headline tariffs and
import quotas, terms of contemporary international trade also depend on such
instruments as nontariff barriers. Trade agreements can have an impact on
these measures and can offer an important opportunity to create and maintain
foundational ground rules in important areas, including intellectual property
rights, labor, and environmental protections. These areas can be critical insofar as international trade can cover several areas that complement and also go
beyond the reach of national law, so trade agreements offer an opportunity for
trade to operate under rules that promote a level playing field.
In addition to the division of economic surplus between countries,
evolving trade patterns can affect the division of gains within countries. Even
if a given country has strong trade agreements and experiences positive net
gains from trade as a whole, these conditions do not necessarily imply that
all its citizens are better off from trade; for instance, though consumers may
enjoy lower prices and greater product variety, workers for firms that fail to
withstand import competition may find themselves displaced from the labor
force. This displacement can have dramatic distributional effects on workers
and geographies. The third section identifies the impact of trade on U.S. consumers and workers. A primary concern about trade and trade policy is how
the economic surplus created by trade is shared between different segments
of the population.
The fourth section examines specific existing trade opportunities for
the U.S. in the agricultural and energy sectors. U.S. farmers—benefiting from
access to plentiful land, rapid technological development, and well-functioning
capital markets—have long exported their excess production. Thanks largely to
our Nation’s world-leading technology, its refined petroleum products, crude
oil, and now increasingly natural gas are available for export around the hemisphere and the globe. The United States also has the largest coal reserves in
the world, and could increase its international shipments. Energy exports are a
crucial component of the Trump Administration’s vision of energy dominance.
The United States is now actively improving its trade agreements to
ensure that its trade with its international partners maximizes the net benefits
Enhancing U.S. Trade in a Global Economy

| 221

for the United States. The chapter’s fifth and final section thus discusses how
new and renegotiated agreements can help shape and improve trade policy.

The Economics of Trade
The international trade literature has highlighted many circumstances in
which trade leaves both nations better off. Consider the elegance of the insight
expressed in the principle of comparative advantage: People and nations are
not equally skilled at all things; but, by specializing in the activities at which
they are relatively most productive, individuals and nations can enrich themselves by trading the surplus generated from higher productivity for a share of
the other goods and services they desire.
In selecting international trading partners, firms are likely to choose their
nearest neighbors to minimize transportation costs. Larger nations also make
desirable trading partners, because they can produce a wide variety of goods
and services that might be attractive and they offer larger markets for exports.
These observations are the basis of the “gravity” model of trade—nearer and
larger objects exert greater gravitational force on each other (Melitz 2003).
Thus, nearer and larger nations are likely to have more trade with one another,
all else being equal. We may import exotic goods from the far reaches of
the globe, but most trade is likely to be with our neighbors and other large
economies.
Expanding the United States’ links abroad can offer the advantages of
competition: increased productivity, greater economic growth, increased
innovation, lower prices, and more variety. New markets may not only provide
domestic firms with more potential customers than are available in the local
market and a chance to build economies of scale; they also offer the ability to
purchase lower-cost inputs—so each can enhance their competitive position.
Consumers—and disproportionately, low-income consumers—may benefit as
import competition fosters innovation and product differentiation, as well as
drives down the prices of goods and services.
The United States has long maintained a policy of encouraging greater
economic freedom for all people of the world, with opportunities for free trade
as a central tenet of this policy. But not all nations maintain these same policies, and this can distort global markets. Today, for instance, China challenges
this U.S. policy and the world order it has engendered through its simultaneous
maintenance of a growing domestic market and market-distorting initiatives,
like industrial policies.

U.S. Trade in the World Economy
The scope for trade policy is constrained by the natural economic forces that
determine trade flows, like geography, resource endowments, and comparative advantage. Although a nation cannot choose its neighbors, its government

222 |

Chapter 5

can serve the interests of its citizens by shaping a trade policy that respects
inherent economic forces. Trade agreements play an important role in allowing
nations to commit to specific policies. Two countries might use a trade agreement to mutually obligate to lower tariff schedules, improving their joint welfare. Without the agreement, each would face a unilateral incentive to impose
tariffs on the other, potentially to prohibitive levels that would leave the gains
from trade unrealized. By the same logic, the agreement provides the partners
with opportunities to negotiate over other, nontariff measures that can be substantial impediments to trade. Agreements are legal instruments that facilitate
and establish the institutional underpinnings of exchange, including important
considerations such as respect for property rights.
The United States is endowed with factors that provide a comparative
advantage for a number of products. But this advantage can and does evolve,
and technological change is an important driver of its evolution. Consider a
hypothetical situation faced by a State supplying its own energy needs with oil
and exporting the excess. When a neighboring State discovers a new oil deposit
with lower production costs, it could make sense for the first country to quit
extracting oil and buy it from the second one (and thereby save its remaining
oil deposits for future generations).
The U.S. economy has changed markedly over time. In the early years of
the Republic, agricultural colonies struggled to establish manufacturing for a
domestic market. Later, the products of these manufacturers became exportable. In recent decades, the economy has shifted again, as figure 5-1 shows,
away from manufacturing and toward service provision industries. Buera and
Kaboski (2012) find that the rise in the services share of the U.S. economy has
been driven by greater demand for the high-skilled labor that is associated with
the country’s rising income per capita. Developing economies, including China
and India, are likewise making a rapid transition to service dependence as their
income per capita also increases. Economists sometimes call service industries
“nontradable,” because a physical presence is generally required; for example,
it is difficult to trade internationally in snow plowing. But technological change
is opening new frontiers in service trade, both through online platforms and
also via the expanding range of tradable services, like in financial management,
consulting, and engineering. Perhaps in the future, satellite-controlled drones
will plow snow, creating a new opportunity for international trade.
Traded goods and services have constituted a growing share of U.S.
gross domestic product (GDP) over time. From 1970 to 2015, the U.S. economy
grew threefold, but the total value of goods and services traded internationally increased nearly eleven times in real terms. Indeed, trade and economic
growth are strongly and positively correlated. Frankel and Romer (1999) estimate that a 1-percentage-point increase in the ratio of trade to GDP raises per
capita income by between 0.5 and 2 percent. This increasing economic reliance
on trade makes trade policy critical to the future economy. It also means that

Enhancing U.S. Trade in a Global Economy

| 223

Figure 5-1. Goods versus Services Value Added and Employment
Percentage of GDP/private nonfarm employment
100

1980

1990

2000

2010

2016

80

60

40

20

0

Goods

Services

Value added (% of GDP)

Goods

Services

Employment (% of nonfarm
employment)

Sources: Bureau of Labor Statistics; Bureau of Economic Analysis; CEA calculations.

the health of the global economy now matters more to the U.S. economy than
it did historically; rising GDP among our trading partners promotes U.S. export
growth. Although, in comparison with other major industrialized countries,
the United States has an economy that is less dependent on trade, the trends
apparent in figure 5-2 reflect a general increase in international trade among
developed economies.
The United States’ stance on trade has evolved throughout its own history. In the late 19th century, the United States relied on high tariffs by today’s
standards—in the neighborhood of 30 percent. Tariff revenue accounted for
roughly half of Federal revenue from the time of the Civil War until the income
tax was imposed in 1913 (Irwin 2010). The Tariff Act of 1930 has a legacy of
unintended consequences stemming from limited policy and monetary instruments at that time, lower economic integration than today, exchange rate
frictions, and unilateral trade policy (Irwin 2011). For the last 70 years, the
United States has advocated freer trade around the world, originally through
the General Agreement on Tariffs and Trade (GATT) framework, and later the
World Trade Organization (WTO).

Assessing the Gains from Trade
Economists agree that trade can deliver net gains for all nations involved, but
debate continues about the best way to identify and to assess these gains, as
well as to diagnose opportunities for improving trade policy. Trade affects a

224 |

Chapter 5

Figure 5-2. Trade Relative to GDP for the OECD and the
United States, 1970–2015
Trade as share of GDP (%)
70
60
50
40

Other OECD
30
20

United States
10
0
1970

1980

1990

2000

2010

Source: World Bank.

wide variety of outcomes—from national income accounts and trade balances,
to customs receipts and deliveries of foreign merchandise. One way to assess
the distribution of gains between trading partners is to examine the balance
of trade. If a country exports more than it imports, it runs a trade surplus; but
if it buys more foreign goods and services than it sells, that surplus becomes
a deficit. This is true for each bilateral trading relationship, as well as for the
aggregate across all trading partners.
Exports and imports of goods and services are components of the U.S.
balance-of-payments accounts, which are also known as international transactions accounts. There are three accounts: the current account, the capital
account, and the financial account.1 The current account includes international transactions in goods, services, and income. The income component
includes primary income transactions like investment income, employee
compensation, and income from reserve assets, as well as secondary income
transactions that consist of transfers between U.S. residents and nonresidents.
These transfers include government transfers and foreign remittances. The
current account balance of the United States reflects earnings on U.S. corporations’ foreign direct investment, whether repatriated or reinvested abroad.
1 The United States closely adheres to international standards (BPM6) for balance-of-payment
accounts, with some subtle differences. One salient difference is not adopting new guidance to
account for value added on goods in integrated supply chains as service imports, due to data
availability concerns. See BEA (2014) for the details. Some readers may recall previous standards
(BPM5 and earlier), which merged the capital and financial accounts.

Enhancing U.S. Trade in a Global Economy

| 225

The capital account includes capital transfers and transactions of nonfinancial
assets. Capital transfers involve transfers of assets (other than cash or inventories), whereas nonfinancial asset transactions largely cover intangibles such as
sales of trademarks or other contracts. The financial account includes several
types of investment and monetary reserves: direct and portfolio investments,
and currency reserves. Financial account transactions, unlike those in the current account, can change the value of a tradable asset without changing the
income or savings for either trading partner.
How can a country afford to run a trade deficit? Because the current
account balance largely reflects the trade balance, the current account deficit
largely reflects the trade deficit. To sustain a current account deficit, the country must be able to attract investment that offsets the currency depreciation
that would normally make imported goods more expensive and exported
goods cheaper in foreign markets. To sustain a trade deficit, a country must
attract foreign investment to help finance the cost of buying foreign goods.
The currency provided by foreigners for investment—for example, in the form
of government bonds—comes from net imports of goods and services, which
is equivalent to a bilateral trade deficit. The United States has been able to
sustain a trade deficit in part because of the role of the U.S. dollar in the global
economy (McKinnon 2001); foreigners are happy to hold U.S. dollars and dollardenominated assets, which they obtain by selling more goods and services to
Americans than they buy.
Through the third quarter of 2017 (hereafter, 2017:Q3, etc.), the United
States ran a current account deficit of $338 billion. The trade deficit during
the same period was $414 billion, chiefly made up of the $597 billion deficit in
goods alone; the trade surplus in services was $183 billion, and the nontrade
elements of the current account (income balances) had a surplus of $75 billion.
Figure 5-3 shows the evolution of the goods-alone trade balance since
1976, a period when the balance has always been a deficit. GDP has grown
more than trade flows over time, resulting in the merchandise trade deficit as a
share of GDP stabilizing below peak historic levels, near 6 percent.
The largest share of the current account deficit is goods trade; the United
States has imported more goods than it exported in every quarter since the
first quarter of 1976. This contrasts sharply with the period from 1960:Q1 to
1976:Q1, when the United States ran a merchandise trade surplus in 52 out of
65 quarters.
The sustainability of a country’s external position with respect to the
world, and consequently trade and foreign policy, depends on the source
of the deficit and on whether current account balances are perceived to be
sustainable. If the current account is in deficit because foreign debt is being
funneled into high-return domestic investment, then the external deficit may
not be of much concern. But an overvalued exchange rate or an excessive level
of public or private consumption—which could indicate fiscal profligacy or a
226 |

Chapter 5

Figure 5-3. U.S. Goods-Alone Trade Balance and Share of GDP, 1976–
2017

Share of GDP (%)
10

Dollars (billions, SAAR)
200

Goods-alone trade deficit as share of GDP
(right axis)

2017:Q3

0

0

-200

-10

-400

-20

-600
-800
-1,000
1976

-30

Goods-alone
trade balance
(left axis)

1984

1992

Source: Bureau of Economic Analysis.
Note: SAAR = seasonally adjusted annual rate.

2000

-40

2008

2016

-50

consumption binge—are sources of current account deficits that are of more
concern. Fortunately, policy can address the latter set of issues to bring the
resulting trade deficit more into line, because the books have to balance.
One important corrective mechanism for persistent current account
imbalances is the adjustment of exchange rates, that is, the value of foreign
currency that can be purchased for $1. Continual deficits should result in a
depreciation of the U.S. dollar relative to other currencies, all else being equal.
This depreciation would naturally make imports more expensive for U.S.
purchasers and would make U.S. exports less expensive for foreign buyers,
bringing the current account back closer to balance by increasing U.S. exports
and decreasing U.S. imports. However, the U.S. dollar is the world’s reserve
currency, which appreciates the dollar and dampens the natural correction
to trade balances (McKinnon 1982). Reserve currency status offers important
benefits that should be weighed against the monetary costs for trade policy.
The aggregate trade balance is an important component of the current
account balance, which reflects the excess of domestic savings over domestic
investment. If a country invests more than it saves, or imports more than
it exports, it finances the resulting deficit through foreign borrowing. Such
borrowing appears in the financial account, which must balance the current
and capital accounts. Because the relationship is definitional, policies that try
to affect the trade balance without considering the broader current account
balance, or vice versa, will be hard-pressed to succeed in the long run. Trade
Enhancing U.S. Trade in a Global Economy

| 227

policy focused on trade balances must be linked to current account policies
(figures 5-4 and 5-5).
A contributing factor to the U.S. goods trade deficit is the strength of the
dollar against other major currencies, which places downward pressure on
demand for U.S. exports. The contemporary large trade deficit is not entirely
unprecedented historically; as recently as the 1980s, the United States faced
“twin deficits” of concern—trade and the government’s budget—as have
weaker European economies in recent years (Trachanas and Katrakilidis 2013).
The current account reflects the balance of government, private, and corporate
deficits; as any of these components expands, the current account reflects the
change.
Earlier historic periods, including the gold specie standard era before
World War I and the times of the oil shocks of the 1970s, have been characterized by large trade imbalances around the world. Even today, the United States
is not alone in concern over structural imbalances. Fiscal and monetary policies may be more important than trade policies in determining the magnitude
of trade balances (Bracke et al. 2010). The distribution of trade balances across
trading partners is attributable to a variety of factors that are idiosyncratic to
individual countries.
Figure 5-6 illustrates the distribution of goods and services balances
across major U.S. trading partners in 2016. All countries show a services surplus
offsetting a goods deficit, with the U.S. running a net bilateral surplus only with
Canada and the United Kingdom.
The United States has a bilateral goods deficit and a services surplus
with many of its major trading partners. Overall, the United States has a goods
deficit and a services surplus with the world. The services surplus is consistent
with the structure of the private sector, which has evolved during the last few
decades toward more services output as a share of GDP. Many other advanced
economies have seen a similar evolution. Among its bilateral trade balances in
2017, the U.S. had the largest goods deficit (through 2017:Q3) with China, at
$277 billion (1.44 percent of U.S. GDP); and it had the largest services surplus
(through 2017:Q3) with the European Union, at $36.6 billion (0.2 percent of U.S.
GDP). Comparing the first three quarters of 2017 with the same period of 2016,
the bilateral U.S. goods and services trade deficit narrowed with Brazil, India,
Singapore, and South Korea.

U.S. Trade Balances in Perspective
Figure 5-7 illustrates the U.S. trade balance from 1790 to the present, expressed
as a share of GDP. From 1790 through 1873, the U.S. trade balance was volatile,
in part due to the low trade volumes (Lipsey 1994). The trade balance swung
back and forth between surplus and deficit, but was mostly in deficit. From
1873 through the 1960s, the trade balance was mostly in surplus. The largest
historic surpluses were during the years 1916–17 and 1943–44, as wartime
228 |

Chapter 5

Figure 5-4. U.S. Monthly Goods Trade with the World
Dollars (billions, seasonally adjusted)

Nov-2017
150

Exports
50

-50

Balance
-150

Imports
-250
1992

1997

2002

2007

2012

2017

Source: U.S. Census Bureau.

Figure 5-5. U.S. Monthly Services Trade with the World
Dollars (billions, seasonally adjusted)
150

Nov-2017

Exports
50

Balance

-50

Imports

-150

-250
1992

1997

2002

2007

2012

2017

Source: Bureau of Economic Analysis.

Enhancing U.S. Trade in a Global Economy

| 229

Figure 5-6. Goods and Services Trade Balance, by Country, 2016
Dollars (billions)

Goods

100

Services

Total

0
-100
-200
-300
-400

Canada

Mexico

EU28

China

Japan

U.K.

South
Korea

Sources: Bureau of Economic Analysis; CEA calculations.
Note: EU28 = Austria, Belgium, Bulgaria, Croatia, Republic of Cyprus, Czech Republic, Denmark,
Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg,
Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, and the U.K.

production and trade with allies predominated. Since 1976, the trade balance
has been continually in deficit. The largest deficit as a share of GDP was nearly
6 percent in 2006, a share exceeded in only six other years in U.S. history and
not seen since 1816.
Trade policy is not randomly assigned but evolves endogenously; the
United States is more likely to reach trade agreements with countries with
which it has large trade volumes (Baier and Bergstrand 2007, 2009). Although
trade agreements are associated with about twice as much overall trade, the
causal impact on the trade balance is unclear, in part because agreements are
more likely with countries that would otherwise have higher trade volumes.
Nor does the presence of an agreement predict the balance of trade. The
United States has free trade agreements (FTAs) with a number of countries—
some of which represent net trade surpluses for the United States (Canada and
Singapore), and some of which represent deficits (Mexico and South Korea).
This becomes immediately evident when comparing other countries’
trade balances with their volume of total trade with the United States. In 2016,
the United States ran a trade surplus of $2.6 billion with Canada on a balanceof-payments basis. The U.S. trade deficit with Mexico was $52.2 billion in 2017
through the third quarter, a deficit similar to countries further afield and without FTAs—including Japan ($42.1 billion), Germany ($51.7 billion), and India
($21.0 billion)—over the same quarters. Figure 5-8 illustrates the distribution
of trade balances between major U.S. trading partners. The horizontal axis is

230 |

Chapter 5

Figure 5-7. Trade Balance as a Share of GDP, 1970–2016
Share of GDP (%)
8

2016
6

Trade Surplus
as share of GDP

4
2
0
-2

Trade Deficit
as share of GDP

-4
-6
-8
-10
1790

1820

1851

1882

1910

1941

1971

2002

Source: Global Financial Data.
Note: Shading indicates trade deficit.

the trade balance, and the vertical axis is the trade balance as a percentage
of bilateral trade. The plotted countries account for 98 percent of the United
States’ total trade balance and 84 percent of the goods-alone trade balance.

Services Trade
In contrast to goods, the United States’ trade in services contributed a surplus of $183.1 billion through 2017:Q3. The U.S. economy has grown more
dependent on private service-producing industries, which accounted for 68.9
percent of U.S. economic output value added during the same quarters in
2017. Focusing only on the trade in goods alone ignores the United States’
comparative advantage in services, which rose as a share of U.S. exports to 33.5
percent through 2017:Q3. Travel (including that for educational purposes) has
constituted the largest share of U.S. services exports. Services trade between
countries has continued to grow—given declining travel costs, improvements
in telecommunications, and growth in online services that allow, for example,
computer coding to take place in remote locations. This allows the United
States to export high-skilled services to other countries that do not share our
expertise and training advantages.
There are important limitations in the evaluation of service trade data,
both because of the intangibility of services and data-related issues. Although
goods trade exists in a physical form that is inherently quantifiable—for
Enhancing U.S. Trade in a Global Economy

| 231

Figure 5-8. U.S. Bilateral Trade Balances with Major Trading
Partners, 2016
Balance as share of total flow (%)
80

Hong Kong

40

Singapore
Brazil
United Kingdom

0

Mexico
Japan
Germany

-40

Italy

Saudi Arabia
Canada
France
South Korea
India
Thailand

China

-80
-400

-300

-200
-100
Trade balance (billions of dollars)

0

100

Sources: Bureau of Economic Analysis; CEA calculations.
Note: Trade data are on a balance-of-payments basis. Red points indicate parties with bilateral
or multilateral free trade agreements with the United States.

example, using customs records and shipping manifests—services trade is
harder to measure because there is no similar record produced as the service
is provided. It can also be difficult to assign industries in the service sector
because services can often fall into multiple industry categories, given that
services trade data are called by the type of service traded rather than the
industry of the supplier.
The Bureau of Economic Analysis (BEA) collects services trade data,
primarily using data from surveys of U.S. companies and data from other
government agencies. BEA does not publish data for all bilateral service trade
relationships, largely because of source data limitations on geographic detail
and data confidentiality laws.
Services trade data produced in different countries are difficult to compare because the use of different definitions and estimation methods can
result in incompatible bilateral trade balances. Both the United States and the
European Union reported a trade surplus with each other from 2010 through
2015; in 2016, both reported a U.S. trade surplus, though there was a discrepancy in the level of $54.8 billion on a balance of payments. This asymmetry was
largely due to differing measure of financial and other business services. The
United States reported $67.6 billion in service exports to the United Kingdom
in 2015, while the U.K. reported $21.2 billion in service imports from the United
States. Some of the discrepancy is definitional, as the United States includes
the Channel Islands and the Isle of Man, which the United Kingdom omits.

232 |

Chapter 5

This difference on asymmetries is likely small for most services, but could be
larger for financial services, because offshore financial centers are located on
the Channel Islands (BEA 2017b). However, the Channel Islands are unlikely to
account for the difference of $46.4 billion.
The Department of Commerce, in coordination with the Office of the U.S.
Trade Representative and the International Trade Commission, is working with
America’s foreign counterparts to develop the next generation of trade statistics. This effort aims to fill the need for more complete measures of services
trade flows as services grow in relevance for the United States, complementing
other efforts to regularize services trade.

Dividing the Pie: Defining Fair
Trade between Countries
The United States is the most innovative and dynamic economy in the world,
which can provide an enormous advantage in trade. Nonetheless, the question of whether existing trade policies are beneficial to the United States is a
question distinct from the question of the underlying dynamism of the U.S.
economy. But before considering how trade agreements might be amended to
level the playing field and maximize the benefits of trade for the United States,
the first task is to identify how current trade patterns are attributable to policies that disadvantage the United States.
The United States has played a leading role in building and maintaining the modern international trade architecture that reduces barriers to
trade through both multilateral and bilateral channels. The United States is a
founding member of the World Trade Organization, which now includes 164
member economies, which together represent 98 percent of global GDP and
95 percent of global trade. Under the WTO, the United States is party to a wide
variety of agreements on trading rules regarding agriculture, antidumping,
civil aircraft, customs valuation, dispute settlement valuation, government
procurement, import licensing, intellectual property (i.e., under the Agreement
on Trade-Related Aspects of Intellectual Property Rights, TRIPS), preshipment
inspection, protocol of accession, rules of origin, safeguards, sanitary and phytosanitary measures, services (i.e., the General Agreement on Trade in Services,
GATS), subsidies and countervailing duties, technical barriers to trade, and
trade-related investment measures. WTO member countries generally set
tariffs no higher than negotiated most-favored-nation rates and apply nontariff
measures subject to these agreements.
The United States is party to two plurilateral FTAs encompassing eight
countries in total—including the North American Free Trade Agreement
(NAFTA) with Canada and Mexico; and the Dominican Republic–Central
America FTA (CAFTA-DR) with Costa Rica, the Dominican Republic, El Salvador,
Guatemala, Honduras, and Nicaragua. These plurilateral trade agreements
Enhancing U.S. Trade in a Global Economy

| 233

virtually eliminate tariffs on trade among partner countries and set standards
to reduce nontariff measures. The United States is party to a further 12 bilateral FTAs, including with Australia, Bahrain, Chile, Colombia, Israel, Jordan,
Morocco, Oman, Panama, Peru, Singapore, and South Korea. These FTAs virtually eliminate tariffs over time and set standards to reduce nontariff measures.
Among the United States’ FTA partners, Canada, Mexico, and South Korea were
the top U.S. export destinations in 2016, constituting over 29 percent of U.S.
exports, as shown in red in figure 5-9.
The United States also has in place bilateral investment treaties (BITs)
with 40 countries, which aim to “protect private investment, develop marketoriented policies in partner countries, and promote U.S. exports” (USTR
2017b). The United States also participates in two sector-specific trade agreement negotiations: (1) Trade in Services Agreement (known as TiSA), an FTA
that focuses on services, with 23 countries representing 70 percent of the
world’s services market and (2) the Environmental Goods Agreement, which
seeks to eliminate tariffs on environmental goods such as wind turbines and
solar water heaters, with 17 parties representing 90 percent of global exports
in environment goods.2
Outside the WTO, FTA, and BIT systems, the United States also provides
preferential tariff treatment to promote international development goals. The
General System of Preferences (GSP) offers duty-free access to the U.S. market
for 120 designated countries and territories, covering at least 20 percent of the
tariff schedule for all GSP beneficiaries. This access is conditional on compliance with eligibility criteria including adequate and effective protection of
intellectual property rights, taking steps to afford internationally recognized
labor rights, and assurances of access to markets for U.S. goods and services.
In 2016, U.S. imports under GSP were valued at $19.0 billion, accounting for
0.7 percent of total U.S. imports of goods and services. India, Thailand, Brazil,
Indonesia, and the Philippines exported the largest amount of GSP goods by
value in 2016, while smaller countries such as Georgia, Armenia, Lebanon, and
Tunisia relied most heavily on GSP as a share of their exports to the United
States. The list of GSP beneficiary countries periodically changes based on factors such as countries’ compliance with GSP criteria or through “graduation”
from the program as beneficiary countries achieve high-income status. For
example, in 2017, the U.S. announced its intention to terminate Ukraine’s benefits under the program and to reinstate some of Argentina’s benefits (Federal
Register 2017). The GSP program expired on December 31, 2017. As a result,
imports that were previously eligible for duty-free treatment under GSP are
now subject to regular duties.
2 The negotiating parties to the Environmental Goods Agreement are Australia, Canada, Chile,
China, Costa Rica, the European Union, Hong Kong, Iceland, Israel, Japan, New Zealand, Norway,
Singapore, South Korea, Switzerland, Taiwan, and Turkey.

234 |

Chapter 5

Figure 5-9. U.S. Exports of Goods and Services, by Destination, 2016
Total U.S. exports (%)
16

14.5

12

8

11.9

7.7
5.5

4

4.9

3.7

2.9 2.5
2.3 2.1 2.0 1.9
1.7 1.3
1.2

0

Sources: Bureau of Economic Analysis; CEA staff calculations.
Note: U.S. FTA partners in red; export levels (billions of dollars) are shown above the bar.

The GSP list is not the only trade “preference” program to benefit lessdeveloped countries. The United States also offers duty-free entry for almost
all goods to designated sub-Saharan African countries under the Africa Growth
and Opportunity Act (AGOA). These countries accounted for $20.1 billion of
total U.S. goods imports, 0.9 percent of total U.S. imports in 2016. AGOA is set
to expire in 2025. A third program, the Caribbean Basin Initiative, offers dutyfree treatment for goods imported from eligible Caribbean island and littoral
countries and territories. A fourth program for Nepal was launched in 2017.
The United States also maintains bilateral and multilateral dialogues on
trade and investment issues with trade partners under Trade and Investment
Framework Agreements (TIFAs). In the absence of FTAs with these countries,
TIFAs or equivalent agreements offer a forum to discuss trade-related issues
such as on market access, labor, the environment, intellectual property right
protection and enforcement, and capacity building. The United States currently has 18 TIFAs with Asian countries, 14 with Middle Eastern countries, 12
with African countries, five covering 19 North and South American countries,
and five with European countries.
These agreements cover a broad range of countries and issues. But what
do trade agreements do? How do they become so important to American firms,
consumers, and workers? To consider these questions, it is helpful to focus on
several dimensions of trade policy that are fundamental to trade agreements:
formal trade barriers are one important lens, but contingent antidumping

Enhancing U.S. Trade in a Global Economy

| 235

and countervailing duties are also illustrative. Increasingly important over
time are non-tariff barriers that include technical barriers to trade and other
provisions that are implemented “behind the border.” A third area of focus is
on institutional frameworks that create the “rules of the game” to bridge trade
gaps between national and international law, like intellectual property rights.

Trade Barriers
Barriers to trade block the access of foreign producers to domestic markets.
Common examples include tariffs (i.e., taxes levied on imports), quotas, import
licenses, and other mechanisms that limit access to the domestic market.
By altering the marginal costs of domestic and foreign production, trade
barriers can distort the free allocation of capital and prevent producers from
specializing in areas of comparative advantage. Under WTO rules, tariffs can be
imposed on all goods arriving in a country, regardless of where they originate,
subject to members’ commitments. Trade barriers can also be set contingently, in response to unfair trade practices. For example, dumping occurs
when a foreign firm exports at less than its manufacturing cost, undercutting
domestic producers. Government support through subsidies can provide an
unfair advantage to foreign producers and thereby place domestic producers
at a disadvantage. In response, U.S. law allows for antidumping or countervailing duties in order to address unfair trade practices like these, depending on
the cause of the injury, and they are an important tool for modern trade policy.
The United States has led contemporary multilateral efforts to promote
trade liberalization. Eight rounds of trade negotiations under the GATT and
its successor, the WTO, have contributed to the reduction in average applied
tariffs in the world’s major industrial economies from more than 20 percent in
the late 1940s to about 4 percent today.
In general, countries do not have one tariff rate for all products, instead
using tariff schedules that differentiate between products. There are exceptions. Chile and the United Arab Emirates, for example, levy the same tariff on
almost all imported products—but most countries discriminate by product.
This practice is WTO-admissible, provided that the tariff applies uniformly
to all members. The average tariff rate across thousands of products for any
given economy can be used to measure the degree of protectionism. Average
tariff rates, while often less than 5 percent in industrialized economies, are
much higher in less-developed economies. This gap may be partly attributed
to nonreciprocal preferential tariffs offered to less-developed economies and
the relatively greater importance of trade policy to generate revenue for lower
income countries (UN 2016).
Within the WTO system, tariffs can be classified into two broad categories: most-favored-nation (MFN) and preferential. MFN tariffs are levied by WTO
member countries in a non-discriminatory manner—all other member states
pay the same tariff for the product, unless they belong to a shared preferential

236 |

Chapter 5

trade agreement. Lower preferential tariffs are determined in trade agreements outside the WTO. Members of certain preferential agreements—such as
the South African Customs Union, the European Community, and NAFTA—pay
tariff rates close to zero in their joint trade of all products.
Although the United States imposes among the lowest barriers to trade
in the world, substantial formal barriers remain in place elsewhere—especially
in developing countries—despite the progress the international community
has made to open markets.3 Figure 5-10 uses estimates from Kee, Nicita, and
Olarreaga (2008, 2009) to compare the trade barriers different regions face and
impose by calculating the uniform tariff rate consistent with observed trade
flows. The indices account for both tariffs and selected non-tariff measures,
and thereby capture a richer dimension of trade policy than tariffs alone.
As shown in figure 5-10, U.S. exporters face formal barriers nearly three
times higher than those the United States imposes on importers, and nontariff
barriers imposed on U.S. exporters are 36 percent higher than those faced by
importers to the United States. The United States imposes lower trade barriers
than other high-income, emerging, and developing countries as a group, and
U.S. exporters face higher barriers to trade than exporters in those countries;
taken together, these indices summarize the prevailing imbalance the United
States faces regarding trade restrictions. The figure also shows that emerging
and developing economies in the figure impose higher tariffs than the markets
their exporters target.
Given that tariffs differ across countries, one pertinent policy question
is what is the “right” level of tariffs? The disparity between formal barriers
for U.S. imports and exports could imply that the United States should erect
higher trade barriers, or could imply that other countries should lower theirs.
Multilateral institutions have developed with the ostensible goal of moving all
countries closer to the example of the United States and other industrialized
countries. But terms of trade are not equal, and tariffs are one way to equalize those imbalances. As an example, countries impose tariffs that average 9
percentage points higher for imported products for which the quantity varies
little with price (Broda, Limao, and Weinstein 2008). In comparison with the
average rates shown in figure 5-10, 9 percentage points is a large change. The
economic intuition is straightforward—if exporters are not price-sensitive, then
they will bear a greater share of economic costs of imposing the tariff, rather
than domestic consumers. Higher barriers may be met by higher barriers in
retaliation, or they may be met by horizontal integration and less trade.

3Contemporary U.S. tariff rates are low compared with historic norms. As a point of comparison,
the average ad valorem rate for imports into the United States in 1913, the last year when tariffs
remained the major source of U.S. Federal revenue, was 41 percent; in 1931, after enactment of
the tariff-raising Smoot-Hawley Act, the average ad valorem rate was 53 percent.

Enhancing U.S. Trade in a Global Economy

| 237

Figure 5-10. Trade Barriers around the World
Tariffs–imports

NTMs–imports

Ad valorem equivalent (%)
0
5

10

Tariffs–exports
15

20

NTMs–exports
25

30

1

United States

2
3

4
Other highincome G20
5
6
7

Emerging G20

8
9

10

Developing

11

Sources: World Bank; World Trade Indicators; other high-income G20 and emerging G20
nations are listed in table 5-1.
Note: NTM = nontariff measures. Developing countries refers to selected non-G20 countries with pop
50 million and with sufficient data available to calculate trade restrictiveness: Egypt, Ethiopia,
Nigeria, Pakistan, the Philippines, Thailand, and Tanzania.

Tariff Overhangs
WTO member countries negotiate bound tariff rates, or the maximum allowed
tariff rates, across products. They can (and do) apply lower tariff rates in practice, provided that all member countries receive the same treatment. Non-WTO
members can apply whatever tariff rates they choose, but their exporters do
not benefit from the market access rights and formal dispute-settlement provisions available to members.
Table 5-1 summarizes tariff policies of selected economies; countries
are split into three groups: high-income members of the Group of Twenty,
emerging economies of the G20, and other less-developed nations with populations of more than 50 million in 2016. Table 5-1 displays the degree to which
tariff rates and trade barriers vary across individual countries and products.
Agricultural goods remain among the most protected in the world, while the
treatment of nonagricultural goods typically varies by type.
Two categories of MFN tariff rates can affect the flexibility of trade negotiations. The bound tariff rate is the maximum tariff rate allowable under WTO
agreements that can be levied on a given product. The applied tariff rate is
what is actually levied, and is generally equal to or lower that the bound rate.
In practice, applied rates are often much lower than bound rates—particularly
when members commit only to high bound rates—with the difference between
these two called the “binding overhang.” The binding overhang provides a

238 |

Chapter 5

measure of the negotiating room countries have within international trade
agreements to gain concessions through altering formal trade barriers.
As shown in table 5-1, the United States—as a founding member of both
the GATT and WTO—applies a “simple average” MFN tariff of 3.5 percent to
imports from other WTO members, almost the same as the simple average
bound rate, according to WTO tariff data. On a simple average basis, the
United States has effectively zero overhang, and therefore very little flexibility
to negotiate on trade barriers within the existing WTO architecture. (This situation is similar to that of the EU, which applies a simple average tariff rate of
5.0 percent, the same as its simple average bound rate, according to the same
WTO data.) As shown in table 5-1, binding overhangs are typically larger for
even other advanced economies, and much larger for emerging and developing countries, affording them more negotiating room. The table also confirms
that binding coverage, or the share of tariff lines that are bound at the WTO,
approaches or equals 100 percent for most of the world’s major industrialized
economies.
Observed differences in protectionism are at least in part explained by
differential rates of liberalization over time. The United States has historically
led by example, binding nearly all its tradable goods at the WTO at very low
levels. Other countries—including India and Brazil, both GATT signatories
since 1948—have been far less open, preserving high bound rates and avoiding
universal binding coverage. Another reason for heterogeneity in tariff policy
is the WTO practice of permitting developing and emerging economies to
apply higher trade barriers. Tariff liberalization has not kept pace with rapid
industrialization among emerging economies, suggesting the need for a serious reconsideration of permissible tariff differentials. More simply, it could be
that U.S. leadership has failed to convince other countries to follow suit with
respect to liberalization. Unilateral tariff reductions, used rarely in the United
States, might be a superficially good way to establish a globalized order, but it
leaves the United States with less leverage to renegotiate formal trade barriers
that remain quite high in certain corners of the world.
Trade barrier policies among high-income countries do display a considerable amount of heterogeneity with respect to tariff peaking, or instances where
MFN rates on certain products exceed 15 percent. Among other high-income
members of the G20, only Australia and Saudi Arabia display fewer instances of
applied tariff peaking than the United States, although both have higher bound
levels and tariff overhangs. In an extreme example of tariff peaking, South
Korea applied an MFN base tariff of 800.3 percent to imports of “cereals, groats,
[and] meal” before the signing the Korea-U.S. FTA (USDA Foreign Agricultural
Service Agricultural Tariff Tracker). In another extreme example, Egypt applies
a 3000 percent tariff on alcoholic beverages (U.S. Department of Commerce
2017). Binding coverage tends to be far lower for countries in the developing

Enhancing U.S. Trade in a Global Economy

| 239

Table 5-1. Tariff Policy for Selected Economies, 2016

Country

G20 high
income
United
States
Australia

MFN
applied
rate,
simple
average

MFN rate,
Bound rate,
MFN rate, MFN rate,
Binding
simple
agriculture manufacturing noncoverage
agriculture
only
average
only

3.5

3.4

2.5

9.9

97.0

4.1

6.5

99.7

5.0

5.0

100.0

4.0

4.5

2.4

3.2

1.2

1.3

2.7

15.6

2.5

2.2

11.1

2.6

4.2

99.7

13.1

1.2

2.5

11.2

100.0

6.1

4.7

5.0

South Korea 13.9
G20
emerging
Argentina
13.7

16.5

94.9

56.9

6.6

6.8

31.8

100.0

10.3

15.7

14.3

Brazil

13.5

31.4

100.0

10.0

15.3

14.1

China

9.9

10.0

100.0

15.5

12.1

9.0

India

13.4

48.5

74.4

32.7

8.8

10.2

Indonesia

7.9

37.1

96.3

8.4

7.5

7.8

Mexico

7.0

36.2

100.0

14.6

5.1

5.7

Russia

7.1

7.6

100.0

11.0

8.1

6.5

Canada
European
Union
Japan

Saudi Arabia
5.1
(2015)

99.9

5.2

South Africa 7.7

19.0

96.1

8.5

3.8

7.5

Turkey

28.5

50.3

43.2

2.6

5.5

10.9

world and is even relatively low for India. Tariff peaking likewise tends to be
higher among emerging and developing economies.

Responses to Unfair—and Not-So-Unfair—Trade Practices
Antidumping and countervailing duties (AD/CVD) are targeted measures
imposed by countries in response to unfair trade practices that lower prices
below market value, harming the domestic industry. Most recently formalized in the Uruguay Round Agreement in 1994, AD/CVD orders have been
widely applied in particular by low-tariff advanced economies like the United
States and EU member states (Prusa 2001), although India has been the most
frequent user of antidumping measures, according to the U.S. International
Trade Commission (USITC 2010). The WTO Anti-Dumping and Subsidies/
Countervailing Measures Agreements establish discipline regarding AD and
240 |

Chapter 5

Table 5-1. Tariff Policy for Selected Economies, 2016 (continued)

Country

Developing,
other*
Bangladesh
Myanmar
(2015)
DR of the
Congo (2015)
Egypt
Ethiopia
(2015)†
Nigeria

MFN
applied
rate,
simple
average

MFN rate,
Bound rate,
MFN rate, MFN rate,
Binding
simple
agriculture manufacturing noncoverage
agriculture
only
average
only

13.9

169.3

15.5

16.9

12.4

13.4

5.6

83.3

18.8

8.6

5.9

5.1

10.9

96.1

100.0

10.9

12.4

10.9

17.9

36.8

99.3

61.0

14.2

10.7

17.4

**

**

22.1

21.7

16.6

12.1

120.9

20.1

15.7

14.2

11.5

Pakistan

12.1

60.9

98.7

13.4

11.2

11.9

Philippines
Thailand
(2015)
Vietnam

6.3

25.7

67.0

9.8

4.8

5.7

11.0

28.0

75.2

31.0

9.8

7.7

9.6

11.5

100.0

16.3

9.8

8.5

Tanzania

12.9

120.0

14.5

20.5

14.8

11.6

Source: World Trade Organization.
Note: This framework follows Bagwell, Bown, and Staiger (2016). All data pertain to year 2016, except
where indicated by parentheses. *Selected other developing countries chosen as those with 2016
populations greater than 50 million. **Indicates nonuser (or unreported user) of the policy instrument.
NA = not available. G20 = Group of 20. † Indicates WTO nonmember. MFN applied rate may be higher
than the bound rate because of measurement issues arising from aggregation of HS codes and
conversion of specific duties.

CVD actions, permitting the imposition of duties when there is evidence of
dumping or unfair subsidies
Safeguard actions are likewise contingent, but do not require any allegation of unfair trade practice. Rather, safeguards—which are covered under
the WTO Safeguards Agreement and U.S. law—are meant to afford domestic
industry a period of adjustment to an environment of increased imports, via
temporary import restrictions.4 In January 2018, President Trump approved
U.S. Trade Representative (USTR) recommendations for two new safeguard
measures—one for certain crystalline solar photovoltaic products and a second
4 The Tariff Act of 1930 allows the United States to impose AD/CVD orders. In the United States,
AD duties are calculated using the difference between the actual price and the market value
price. Countervailing duties address government assistance/subsidies and are calculated to equal
the value of the subsidy (U.S. Customs and Border Protection 2017). As of January 23, 2018, the
United States had 427 AD/CVD orders in place—the largest share of them targeted China (157), or
concerned iron and steel products (222).

Enhancing U.S. Trade in a Global Economy

| 241

for large residential washing machines—the first completed safeguards investigations in the United States in 17 years.
Patterns of contingent trade cases provide insight into which countries’
practices are most at odds with established rules. Since 1995, WTO records
show that member countries have filed 344 AD and 50 CVD cases against
China; indeed, as shown in figure 5-11, from 2001 to 2015 China was named
as the respondent in over 29 percent of all such cases, the highest of any WTO
member. In addition, between 2002 and 2015 China has also been subject
to 32 specific safeguard actions at WTO. These statistics suggest that other
WTO members recognize that China often uses trade tactics that others find
objectionable.
Economic studies find that AD/CVD orders are effective insofar as targeted
imports tend to decline after measures are imposed (Prusa 2001; Besedeš and
Prusa 2017). However, despite their wide and targeted use, AD/CVD actions do
not always have the desired effect. For example, investigations into the trade
in crystalline solar photovoltaic (CSPV) cells and modules have resulted in two
sets of AD/CVD being imposed on China and one AD duty on Taiwan in the past
five years. Similarly, the U.S. government imposed import restrictions on large
residential washer imports from South Korea (2013), Mexico (2013), and China
(2017). As evidenced by the recent safeguard cases, neither set of actions sufficiently decreased import penetration. (Box 5-1 gives further details about the
solar case.) Still, AD/CVD actions can have real repercussions. In retaliation to
AD/CVD orders for solar products, China imposed a retaliatory set of duties on
imports of solar-grade polysilicon from the United States, South Korea, and
subsequently the EU; this polysilicon is a key input in the manufacture of solar
cells, and retaliatory duties directed toward U.S. exports were at least 53.30
percent.

Nontariff Measures
Even though they attract the most attention, tariffs are just one type of trade
barrier. Returning to figure 5-10, in many cases they account for less than half
of the total trade restrictiveness of a given nation. As multilateral, regional,
and bilateral efforts around the world have succeeded in lowering headline
tariff rates, other types of barriers have emerged as significant impediments
to international trade. These include a broad array of measures, interchangeably referred to as nontariff barriers and nontariff measures. According to the
United Nations Conference on Trade and Development (UNCTAD), they are
generally defined as policy measures other than customs tariffs that can have
an economic effect on international trade, a broad classification that is further
grouped into technical and nontechnical measures.
Technical measures include technical barriers to trade (TBT) and sanitary
and phytosanitary (SPS) measures. TBT measures cover technical regulations and procedures that assess conformity with technical regulations and

242 |

Chapter 5

Figure 5-11. Antidumping Measures, 2001–15
Number of measures as respondent country
500

400

300

200

100

0

China

South Korea

Taiwan

Source: World Bank, Global Antidumping Database.

United States

India

All other
countries

standards, except those covered by the SPS Agreement (UNCTAD 2012). They
establish labeling requirements, standards on technical specifications and
quality (e.g., product weight and size), other measures protecting the environment, and a conformity assessment with those measures. SPS measures seek
to protect humans and animals from risks in additives, contaminants, toxins,
and disease-causing organisms in their food; humans from plant- or animalcarried diseases; and animals and plants from pests, diseases, or diseasecausing organisms, biodiversity, and/or to prevent or restrict “damage to a
country from pests” (UNCTAD 2012).5 These also cover conformity-assessment
measures related to those restrictions including certification, testing/inspection, and quarantine. Nontechnical measures include finance measures;
special preferences affecting competition; investment restrictions including
local content requirements; distribution and post-sales service restrictions;
5 In the WTO Agreement on the Application of Sanitary and Phytosanitary Measures, an SPS
measure is defined as follows: “Any measure applied: (a) to protect animal or plant life or health
within the territory of the Member from risks arising from the entry, establishment or spread of
pests, diseases, disease-carrying organisms or disease-causing organisms; (b) to protect human
or animal life or health within the territory of the Member from risks arising from additives,
contaminants, toxins or disease-causing organisms in foods, beverages or feedstuffs; (c) to
protect human life or health within the territory of the Member from risks arising from diseases
carried by animals, plants or products thereof, or from the entry, establishment or spread of
pests; or (d) to prevent or limit other damage within the territory of the Member from the entry,
establishment or spread of pests.”

Enhancing U.S. Trade in a Global Economy

| 243

Box 5-1. Global Safeguards for Solar Products
In 2017, a Section 201 global safeguards case was brought in front of the
USITC for the first time in 15 years, asserting that imports of crystalline silicon
photovoltaic (CSPV) cells and modules injured the domestic industry. Solar
cells are assembled into modules, which are commonly used in both utilityscale and residential or commercial installations. This Section 201 complaint
followed earlier antidumping actions in 2012 and 2015 that were widely
viewed as ineffective. In September 2017 the USITC unanimously determined
that an injury had occurred. The petition was filed on behalf of Suniva Inc. and
SolarWorld Americas, two U.S.-based solar panel manufacturers. A primary
goal of the petitioners is to maintain an integrated cell and module manufacturing industry in the United States.
This Section 201 action follows on the heels of years of the AD/CVD
efforts to curtail Chinese undercutting of U.S. producers. The first investigation focused on Chinese imports during the years 2009–11, but the subsequent
investigation during 2014–15 expanded the scope to Taiwan as well. The first
investigation resulted in imposition of a dumping duty of between 14.78 and
15.97 percent on cooperating Chinese firms. All other imports were subject
to a 165 percent duty in February 2015. An additional CVD between 27 and
50 percent ad valorem was required from all Chinese importers at that time.
Antidumping duties with an average margin of 19.50 percent were applied
on all Taiwanese imports. Yet Chinese firms relocated production facilities to
other countries, including Malaysia and Vietnam, to avoid these duties. The
level of imports increased. So U.S. firms decided to pursue a global safeguard
strategy, which also does not require the allegation of unfair trade practices,
rather than a targeted AD/CVD approach.
The petitioners’ amended filing with the USITC proposed a remedy
consisting of both a quota and a tariff for imported CSPV cells and a separate
quota and tariff for imported modules. Solar cells and modules are differentiated products, with imported products dominating the low-cost segment
of the market, which is particularly attractive to utility-scale installations.
The petitioners claim that the volume and price instruments are needed to
provide relief to an integrated cell and module manufacturing industry from
cheaper imports that have captured much of the market. None of the USITC
commissioners agreed with the petitioners, and individual commissioners
made three different proposals. These included a combined quota for cells
and modules, and a tariff-rate quota for cells paired with a tariff for modules.
The different proposals have slightly different tariff levels and phase out differently over the course of the remedy.
CSPV prices have fallen markedly in the past decade, largely driven by
technological change. Module prices fell by 84 percent between 2008 and
2014, and continued to under $0.40 per watt in 2017 (figure 5-i). This price
decline has helped push the cost of installed solar generation capacity lower;
between 2015 and 2017, the cost of 1 watt of utility-scale generation capacity

244 |

Chapter 5

Figure 5-i. CSPV Module Prices, 2010–18
Module price per Watt (dollars)
2.10

Dec-2017

1.80
1.50
1.20

Approximate new import price
with $0.32 in-quota tariff

0.90
0.60

CSPV module average
price per Watt

0.30
0.00

2010

2011

2012

2013

2014

2015

2016

2017

2018

Source: Bloomberg silicon solar module index.
Note: CSPV = crystalline solar photovoltaic. Shading indicates a historic market price that is higher
than an expected new import price.

fell from $1.80 to $1.03. This cost decrease has made solar power competitive
in many markets around the country, and this has boosted demand for solar
modules.
The United States currently relies heavily on trade to purchase most
of its solar panels. In 2016, domestic manufacturers’ market share was 11
percent of the total solar generation hardware market on a value basis. After
the initial imposition of AD/CVD duties in 2013 and again in 2015, Chinese
solar manufacturing was offshored to other locations, including Malaysia and
Vietnam. As a result, the import shares by country of origin have changed
noticeably in recent years, but domestic manufacturers have consistently
been crowded out.
A factor in this rapid import penetration are allegations of forced
technology transfer by Chinese parties. Though the Section 201 process was
initiated by affected manufacturing firms, the interrelationship between the
policy instruments highlights the interconnectedness of trade issues.
The President elected to impose a four-year remedy incorporating a
tariff-rate quota for CSPV cells and a tariff for modules. The cell component
is a tariff-rate quota, with a 2.5 gigawatt quota for cells and a tariff imposed
beyond that level of imports. The tariff for assembled modules and for
imported cells above the quota will start at 30 percent ad valorem in the first
year and decrease over the duration of the remedy (USTR 2018).

Enhancing U.S. Trade in a Global Economy

| 245

and behind-the-border policies, including subsidies, government procurement
restrictions, and restrictions on rules of origin.
Some nontariff measures protect health, safety, and the environment.
In these cases, the imposition of rules helps to minimize the risks associated
with certain products, and multilateral frameworks facilitate their adoption by
governments. Others amount to deliberate, anticompetitive trade restrictions
that act to increase costs for those subject to the barrier. Measures identified as
primarily protectionist, or devised with the intent to undermine international
competition, may be challenged and addressed through the WTO. Resolving
disputes about these issues often takes substantial effort; identifying whether
nontariff measures are justified is far more complicated than simply comparing
tariff rates against an accepted international norm.
Economists have studied tariffs and formal trade barriers for decades, so
their effects are well understood. Nontariff measures are less visible and more
difficult to document, yet they are increasingly important as the WTO architecture helps erode formal trade barriers. Extra labeling, licensing requirements,
unnecessary standards compliance, or duplicative health certificates clearly
raise the cost of doing business and place foreign firms at a disadvantage
relative to domestic producers. The literature indicates that nontariff barriers
may distort trade even more than tariffs, and that as countries become richer,
the trade restrictiveness of their nontariff barriers increases relative to tariffs.
Because of their inherent flexibility, nontariff measures can present even
greater barriers in certain sectors, including agriculture. Moreover, subjecting
questionable nontariff measures to dispute resolution is in many cases a timeconsuming process, during which U.S. firms suffer the loss of revenues and
market share, as discussed further below.

Nontariff Barriers to Automobile Trade
The automobile sector is illustrative of the types of nontariff barriers faced by
U.S. companies. A variety of these barriers are imposed by U.S. trading partners
that affect automobile trade. The range and scope of these behind-the-border
restrictions in just the automobile industry provides good insight into the
importance, effectiveness, and breadth of such measures for restricting trade.
China. Since before China joined the WTO in 2001, it restricted the ability
of foreign companies to manufacture automobiles in China by requiring them
to form joint ventures with Chinese enterprises, with the foreign partner’s ownership share capped at 50 percent. New regulations in 2009 extended this same
requirement to the manufacturing of new energy vehicles (NEVs; e.g., electric-,
biofuel-, or hydrogen-powered cars). These joint venture requirements raise
concerns that foreign manufacturers are being coerced into transferring
ownership of their technologies to their Chinese venture counterparts. In the
case of NEVs, the USTR reports that China has pursued a variety of additional
policies that, among other things, appear to discriminate against imported
246 |

Chapter 5

NEVs and have generated serious concerns in light of China’s WTO obligations.
Similar concerns have been raised in connection with imported NEV batteries.
South Korea. Under the U.S.-Korea (KORUS) FTA, both parties agreed
to eliminate their respective automotive and light truck headline tariff rates,
some of which have already been eliminated, while others are currently scheduled to be phased out in future years. However, the USTR notes that there have
been significant problems in implementing KORUS, and the U.S. government
is currently seeking to improve the agreement by rebalancing it, including with
respect to the auto trade. KORUS also contained provisions nominally designed
to address nontariff barriers, most notably allowing for 25,000 cars per U.S.
manufacturer per year that are built to U.S. automotive safety standards to
be deemed as meeting South Korea’s safety standards when imported (USITC
2011). Korea also streamlined its motor vehicle taxes based on engine size so
that South Korea’s taxes did not disproportionately disadvantage larger-sized
vehicles, including U.S.-manufactured vehicles sold in Korea. Additionally,
through the Autos Working Group under the KORUS Agreement, the United
States has continued to seek progress on a range of additional issues, including
Korea’s fuel economy standards.
South Korea is considering implementing draft regulations that would
require all automobile manufacturers to report vehicle repair histories to
purchasers to account for any damage taking place between the manufacturing site and sale to customer. Although this is also regulated in some
American States, the USTR reports that these regulations differ in important
ways, so as to “create obstacles for imports, as vehicles arriving for overseas
often undergo minor reconditioning prior to sale.” Given the greater distance
that the imported vehicles must travel from the manufacturing plant to the
consumer, the USTR has urged South Korea to modify its damage disclosure
regulations so that they are reasonable and consistent with international
practices. The Korean Ministry of Land, Infrastructure, and Transportation has
been requested to draft changes that would recognize American predelivery
inspection in Korea rather than at U.S. manufacturing plants as the conclusion
of the manufacturing process, so as not to duplicate already-existing robust
inspection procedures and not unduly burden foreign automakers. In June
2016, the ministry proposed a rule imposing a floor on damage of more than 3
percent of the manufacturer’s suggested retail price for the vehicle required to
mandate reporting, which U.S. industry has argued is unreasonably low.
Japan. The United States has expressed strong concerns with the overall
lack of access to Japan’s automotive market for U.S. automotive companies. A
variety of nontariff barriers impede access to Japan’s automotive market, and
overall sales of U.S.-made vehicles and automotive parts in Japan remain low,
ultimately leading one major U.S. automotive manufacturer in January 2016 to
cease all operations in Japan.

Enhancing U.S. Trade in a Global Economy

| 247

In recent years, nontariff barriers have included issues such as unique
safety and other standards, an insufficient level of transparency, a lack of
sufficient opportunities for input in developing regulations, hindrances to the
development of distribution and service networks, and the lack of opportunities for U.S. vehicle models imported under the preferential handling procedure certification program to benefit from programs on the same terms as
domestic models. These barriers have had the long-term effect of excluding
and disadvantaging U.S. manufacturers in the Japanese market.
Colombia. Colombia’s policy concerning the registration of trucks over
10.5 metric tons, both imported and domestic, before March 2013 required
registrants to either pay a “scrappage fee” to the government or to scrap an
old freight truck of equivalent capacity on a one-for-one basis. In March 2013,
Colombia changed this policy so that companies could legally register new
freight trucks only by scrapping an old freight truck of equivalent capacity—
importers and other buyers no longer had the option to pay the fee. Within the
first year of the policy’s implementation, imports of new freight trucks fell by
65 percent and importers’ reported sales-related administration costs rose by
$60 million. In the first few years, sales of U.S. exporters fell by $600 million.
In September 2016, Colombia announced the termination of the “one-forone” scrappage policy by December 31, 2018. The existing policy remained in
place until February 2017, when Colombia announced a new interim system
that requires applying for a scrapping certificate in order to register a new
truck. Importers and other registrants must pay 15 percent of the value of the
new truck as a fee for the certificate. The number of available certificates will
depend on the number of vehicles scrapped. Although this scrapping certificate
is not a formal trade barrier, it closely resembles an in-quota ad valorem tariff
of 15 percent, subject to a variable quota determined by certificate availability.

Nontariff Barriers to Agricultural Trade
Nontariff measures, especially SPSs and TBTs, are particularly prevalent in the
agriculture and food sectors. Given the importance of domestic agricultural
production to many countries—and the many different forms of environmental,
food, and health concerns—these issues can be particularly contentious. This
subsection discusses several examples of how these measures are imposed
and become barriers to trade.
Indonesia: import licenses. Indonesia applies restrictive and complicated
import licensing to a variety of products, including horticultural and animal
products. For example, seasonal restrictions prevent the importation of
oranges during Indonesian harvest periods. Beef importer licenses are issued
only for specific countries of origin, so importers cannot change sourcing to
respond to evolving market conditions. Separately from this issue but also
relevant, meat imports are only permitted from approved facilities, but the
approval process requires on-site inspection and the inspectors do not in
248 |

Chapter 5

practice have the resources to inspect all interested U.S. facilities. Beef must
be sold at set prices in Jakarta’s traditional markets as a condition for continuation of import licenses. Indonesian regulations prevent the importation
of poultry parts, an important market for U.S. exporters. After bilateral efforts
to address the problems were unsuccessful, the United States began official
WTO dispute settlement in January 2013. In December 2016, the WTO found
in favor of the United States and New Zealand on 18 out of 18 claims against
Indonesia’s import restrictions and prohibitions. Indonesia’s appeal of that
decision was rejected in November 2017
The EU’s ban on beef hormones. The European Union bans and restricts
the importing of meat produced using various hormones, beta agonists, and
other growth stimulants. Therefore, U.S. exporters that do not participate in a
verification program for assuring the absence of such banned substances cannot enter the EU market. Scientific evidence has shown that the banned substances may not be harmful to consumers. The EU ban has led to WTO disputes.
For example, in 1996, the United States brought a WTO dispute settlement
proceeding against the European Communities by saying that the EU beef ban
did not comply with its WTO obligations. Although a WTO dispute settlement
panel determined that the beef ban was in fact inconsistent with the EU’s WTO
obligations, the EU declined to remove the ban. The United States retaliated
by imposing 100 percent ad valorem tariffs on certain imports from the EU. In
September 2009, the United States and the European Commission came to a
compromise and established a new EU duty-free import quota for grain-fed,
high-quality beef, but U.S. shipments under the quota must still comply with
the EU’s hormone ban. However, Argentina, Australia, Canada, New Zealand,
and Uruguay have begun to compete with U.S. exporters by also shipping
under the quota.
Chile’s food labeling. Chile currently requires that all prepackaged food
and beverage products display a front-of-package black octagonal “stop” sign
if the product contains levels of sodium, sugar, calories, or saturated fats that
exceed specified thresholds. The threshold is set based on quantities of 100
grams or 100 milliliters, and does not take into account the serving size of the
food or beverage product. In addition, if a product is above the threshold level
in more than one of these nutrient categories, then multiple stop signs must
be displayed. Furthermore, the law restricts the advertising of products that
require one or more stop signs, including by prohibiting the use of images
deemed appealing to children under the age of 14. This prohibition, in particular the interpretation by Chilean authorities of registered trademarks that
constitute advertising to children on product packaging, has been inconsistent. These measures have resulted in costs related to delays, shortages, and
repackaging that have proven expensive for U.S. firms. The United States has
raised concerns with these measures, both bilaterally and within the framework of the WTO Committee on Technical Barriers to Trade.
Enhancing U.S. Trade in a Global Economy

| 249

Agricultural biotechnology approval. Delays in the approval of agricultural products derived from biotechnology in China, the European Union’s
members, India, and other countries result in increased market uncertainty
among technology providers, farmers, and traders of U.S. corn, soy, cotton,
and alfalfa—leading to reduced exports of these products. For example, delays
in the EU’s approval process for biotechnology crops have prevented these
types of crops from being placed on the EU market, even though the biotechnology events have been approved (and safely grown) in the United States.
Moreover, the length of time taken for the EU to approve new biotechnology
crops appears to be increasing. The EU’s own legally prescribed approval time
for biotechnology imports is about 12 months—six months for review by the
European Food Safety Agency, and six months for the political committee
process (i.e., “comitology,” in the EU’s parlance). However, in practice, total
approval times are averaging 47 months.
India: poultry and avian influenza. Since 2007, India has restricted various
U.S. agricultural products—including poultry meat, eggs, and live pigs—supposedly to prevent the entry of avian influenza into India. The United States has
maintained that, as reflected in relevant international standards, no scientific
basis supports this wholesale ban. In March 2012, the United States initiated
a WTO dispute by requesting consultations with India. The panel found that
India’s ban was not based on science and breached several obligations of
India under the Agreement on the Application of Sanitary and Phytosanitary
Measures. The WTO Appellate Body subsequently upheld the panel’s findings
in 2015. India has failed to revise its requirements for poultry in a manner that
would allow for U.S. imports, and the case continues to be litigated in the WTO.

Defining the Rules of the Game for Trade
In addition to lowering barriers to trade, trade agreements can provide a forum
to raise and address trade concerns through dispute settlement mechanisms,
including those between one nation-state and another nation-state, and
between an investor and a nation-state. Such mechanisms can act as commitment devices that may deter parties from transgressing an existing agreement.
In doing so, these mechanisms may overcome the incentive to cheat in order to
maximize individual gains. In addition, they can help reduce the policy uncertainty associated with trade relationships by ensuring that parties to trade
agreements may formally raise concerns about trade practices. Consequently,
the United States has actively pursued cases through the WTO dispute settlement mechanism, and U.S. firms have used mechanisms between an investor
and a nation-state in trade agreements like NAFTA.
WTO dispute settlement. The WTO dispute settlement mechanism serves
as a process for addressing a variety of tariff and nontariff barriers pursuant to
a variety of agreements under the WTO, including safeguards, rules of origin,
agriculture, intellectual property, and government procurement. Dispute

250 |

Chapter 5

settlement became more legalistic when the WTO replaced the GATT procedures with a reformed dispute settlement understanding that ended the right
of a responding party to block the adoption of a report and added a standing
Appellate Body of Members to review panel decisions. Before that, under
GATT, trade disputes were at first addressed using informal consultations, or
discussions, between concerned parties; within less than a decade, however,
an independent panel process coalesced to examine the evidence and issue
rulings in response to complaints about noncompliant trade practices (Davis
2012). The GATT system’s dispute procedures were invoked in over 200 cases,
and member states generally complied with findings (Hudec, Kennedy, and
Sgarbossa 1993).
The efficacy of WTO dispute settlement mechanism remains an area of
active debate. Davis (2012) finds that the United States gets better outcomes
via formal WTO adjudication than negotiation, increasing the probability that
the complaint will be resolved and decreasing the time it takes to remove the
barrier in question. Mayeda (2017) finds that the United States has won 85.7
percent of the cases it has initiated before the WTO since 1995, compared
with a global average of 84.4 percent. In contrast, China’s success rate is
just 66.7 percent. Most U.S. WTO cases target China (21) and the European
Communities (19). When the United States is the respondent, it still wins 25
percent of the time, a rate that is better than the global average rate of 16.6
percent (Mayeda 2017). In comparison, the EU and Japan have won 0 percent
of the cases brought against them, while China has won only 5.3 percent of the
time (Mayeda 2017). Nonetheless, because countries may initiate or decline
to initiate cases based on their perceived probability of obtaining a favorable
outcome in the WTO dispute process, comparisons of WTO dispute statements
between countries should be taken with at least some skepticism.
The TRIPS Agreement sets a minimum level of intellectual protection
that each government provides and is subject to the WTO dispute settlement
mechanism (WTO 2017c, 2017d). Of the 115 cases that the United States has
initiated since 1995, 17 have been TRIPS cases, most of which have targeted
practices by European countries.
Intellectual property rights. Secure intellectual property rights foster
innovation and growth (North 1989; Mokyr 2009). Though it can take years of
numerous trials and failures to invent a product and bring it to market, duplication of an invention typically requires significantly less effort. U.S. patents,
trademarks, and copyrights typically grant the owners of intellectual property
exclusive rights only on a territorial basis—within the United States.
The United States currently leads the world in innovation, due in no small
part to a strong legal system that confers exclusive rights and privileges on the
owners of intellectual property. The World Intellectual Property Organization,
an agency of the United Nations, defines intellectual property (IP) as “creations of the mind, such as inventions; literary and artistic works; designs; and
Enhancing U.S. Trade in a Global Economy

| 251

symbols, names and images used in commerce.” U.S. IP can be voluntarily diffused overseas in cases in which U.S. firms provide their IP to overseas partners
or participate in joint ventures. In other cases, U.S. firms may agree or be pressured to submit to technology transfer conditions in order to access foreign
markets or for other reasons, such as taking advantage of tax preferences,
subsidies, and preferences in government procurement. But even firms without international operations may experience theft of their IP or technology by
or on behalf of overseas entities. Such theft represents an involuntary transfer
of IP abroad through theft, an involuntary transfer of IP that costs the United
States between $227 and $599 billion annually, according to the IP Commission
(Blair et al. 2017). By way of comparison, the OECD and the EU’s Intellectual
Property Office have estimated that the global trade in counterfeit and pirated
goods alone cost as much as $461 billion in 2013, 2.5 percent of world trade.
Virtually every industry in the United States either produces or uses IP,
but certain IP-intensive industries are an important and growing share of the
U.S. economy. These include pharmaceuticals, aerospace, computer hardware
and software, electronics, medical equipment, chemicals, and automobile
manufacturers. With their high rates of innovative research and development,
these industries are special targets for IP expropriation.
Lost IP prevents firms from generating a return on investment in their
research-and-development (R&D) costs, thus discouraging them from continuing to invest in R&D and hampering U.S. innovation. As shown in figure 5-12, the
theft of trade secrets annually represents $180 billion to $540 billion in value;
and pirated software and counterfeit goods combined account for about $47
billion of the costs of IP theft. Precise figures are difficult to come by, in part
because firms may not be aware that their property has been stolen, or may be
hesitant to publicize a theft that has been detected.
Technology transfer. Technology transfers can occur if governments
require a company to transfer its IP in order to access markets in that government’s country. For example, Chinese industrial policy includes provisions
specifically calling for the acquisition of foreign technology and innovation. In
August 2017, the U.S. government launched a Section 301 investigation into
China’s alleged acts, policies, and practices related to technology transfers.
U.S. firms may seek to enter China for reasons including taking advantage of
lower unit costs of labor, gaining access to its large domestic market, and making use of integrated supply chains for other markets. To access the Chinese
markets, however, stiff entry costs in the form of mandatory IP and technology
transfer requirements are sometimes imposed. Among other things, Chinese
policies and practices selectively require foreign firms to transfer their technologies to Chinese subsidiaries or joint venture partners to gain access.
The ongoing investigation seeks to assess whether and how four types
of Chinese practices hurt U.S. firms: (1) Chinese joint venture and other
approval processes that force U.S. companies to transfer technology or
252 |

Chapter 5

Figure 5-12. Estimated Annual Cost of IP Theft to the U.S. Economy
Pirated software

Counterfeit and pirated tangible goods

Real dollars (billions, $2015)
600

Trade secret theft
3%
5%
92%

500

400

300

8%
13%
79%

200

100

0

Lower bound, $227 billion

Source: IP Commission 2017.

Upper bound, $599 billion

otherwise compromise IP; (2) acts, policies, and practices that prevent U.S.
businesses from establishing market-based terms in licenses and negotiations
with Chinese companies; (3) Chinese government–supported acquisitions
or investment in U.S. companies to “obtain cutting-edge technologies and
generate large-scale technology transfer in industries deemed important by
Chinese government industrial plans”; and (4) Chinese government–supported
“intrusions” in U.S. computer networks or “cyber-enabled theft of intellectual
property, trade secrets, or confidential business information.”
Labor and environmental standards. Because labor and environmental
rules are such an important part of domestic regulation, the United States
has led the way in including enforceable labor and environmental standards
into its trade agreements. The bipartisan Trade Promotion Authority calls for
strong, enforceable labor and environmental provisions in all U.S. trade agreements. For example, most recently, the Trump Administration has proposed
high-standard labor and environmental chapters as part of NAFTA’s renegotiation. By including minimal environmental provisions in a trade agreement, a
country like the United States can ensure that its firms and workers compete
on more comparable terms with foreign producers.
Typically, these standards include cooperation arrangements and general commitments. For example, U.S. FTAs with Central America and the
Dominican Republic, Australia, Bahrain, Chile, Colombia, Jordan, Morocco,
Oman, Panama, Peru, Singapore, and South Korea include dedicated labor and

Enhancing U.S. Trade in a Global Economy

| 253

environmental chapters. The United States also requires GSP beneficiaries to
comply with specified labor provisions regarding workers’ rights.
Enforcement of such labor and environmental standards is difficult,
partly because of measurement issues. The Trump Administration has prioritized enforcement of these provisions. For example, in October 2017 the United
States took an unprecedented enforcement action pursuant to the Annex
on Forest Sector Governance of the United States–Peru Trade Promotion
Agreement, and blocked future imports from a Peruvian harvester because of
illegally harvested timber found in its supply chain. The Trump Administration
will continue to use trade agreements to level the playing field and to
improve labor and environmental practices, and it will continue to prioritize
enforcement.

Domestic Gains and Losses from Trade
In the pursuit of gains from trade through comparative advantage, international trade can increase total economic surplus, and trade agreements can
help determine how that surplus is divided between partners. That said, trade
agreements do not necessarily improve the lives of everyone within a country
party to that agreement; even when an increase in trade boosts national welfare in the aggregate, there is no guarantee that all residents will be better off
because of the change in trade flows.
These distributional implications of trade raise equity issues that demand
consideration in the debate over trade policy. Differing empirical findings
about the impact of NAFTA offer a helpful example. Hakobyan and McLaren
(2016) identify the concentrated job and income losses stemming from the U.S.
accession to NAFTA, and contrast those with the gains that the agreement has
delivered to the U.S. economy. Meanwhile, other studies suggest the existence
of net positive gains from NAFTA for U.S. GDP and employment (Dixon and
Rimmer 2014; Cipollina and Salvatici 2010; Hufbauer, Cimino, and Moran 2014),
even as they knowledge that there are mixed findings on industry-specific
effects. While acknowledging these dispersed benefits, Hakobyan and McLaren
(2016) note that a segment of manufacturing workers with low educational
attainment lost 8 to 16 percent in wage growth between 1990 and 2000.
Conversely, De La Cruz and Riker (2014) found small but positive effects on the
wages of skilled and unskilled workers in the years 1994–2013. Though these
empirical results differ, the point remains that trade can have differential types
of effects across workers and between workers and consumers, underscoring
the magnitude of trade’s distributional effects.

Domestic Gains
Because trade agreements can expand overseas trading opportunities, they
can raise the volume of domestic goods demanded by overseas consumers,
254 |

Chapter 5

boosting domestic output, employment, and wages in exporting industries.
In addition, domestic consumers may find they face lower prices on imported
goods and services, lowering the cost of their consumption bundles and leaving them with excess disposable income to fund additional consumption or
savings. These effects may be partially countered by a reduction in the demand
for domestically produced goods and services in which trading partners have
a comparative advantage, reducing the output of these goods and the corresponding employment and earnings in these industries.
Although empirical estimates of the effect of reducing barriers to trade on
the domestic economy are positive, they are often modest in size (USITC 2016),
because the U.S. economy is very large, and trade with any single partner (or
group of partners) is a relatively small share of GDP. Examining regional trade
agreements entered by the United States under Trade Authorities Procedures
after 1984 (including NAFTA, the Uruguay Round, CAFTA-DR agreements, and
several bilateral agreements), the USITC finds substantial increases in imports
and exports resulting from their collective implementation, so that the change
in aggregate trade flows over all agreements in their analysis sum to about 3
percent. The USITC (2016) estimated their joint effect on U.S. GDP at 1 percent,
amounting to an additional $186 billion in U.S. production, or roughly $600 in
additional wages and salaries per household, based on the current labor share.
Resulting shifts in trade volumes suggest that these agreements generated a
significant reshuffling of workers across industries, and led to net employment
gains of approximately 1 percent. The report further estimates that regional
trade agreements saved consumers up to $13.4 billion in goods pricing in 2014
while providing an expanded variety of goods for purchase.

Domestic Losses
Despite the potential of trade agreements to reduce market prices for tradable
goods and expand consumption opportunities of American consumers, workers displaced by trade may experience substantial welfare losses. Trade-based
worker displacement occurs when an industry faces new import competition
and cannot compete with the flow of international goods and services. Even
if workers hold general human capital and are readily employable in a shifting economic landscape, the transaction costs of job loss, job search, and
reemployment may be larger than the reduction in market prices for that
individual’s consumption basket. If workers have skills that are less valuable
after trade patterns shift, the welfare costs are more substantial.

Manufacturing Job Losses
A prominently cited impact of increased trade and import competition is that
manufacturing employment in the United States has fallen. As a share of total
nonfarm employment in the United States, figure 5-13 demonstrates its steady,
marked decline during the last 50 years—the result of technological change

Enhancing U.S. Trade in a Global Economy

| 255

and automation, economic recession, and evolving consumer tastes. But international trade also contributes to the demand for manufacturing workers, both
through access to foreign markets boosting U.S. exports, and foreign access to
the U.S. market increasing import competition.
Although employment in the sector has long exhibited procyclicality, it
has seen changes since 2000. In the expansionary period from the end of 2000
through 2007, the economy failed to recover the manufacturing job losses it
experienced during the previous recession years. Likewise, the current business cycle, which began at the end of 2007, corresponded to a reduction in
manufacturing employment of 2.2 million jobs at the trough. The addition of
nearly 200,000 manufacturing jobs in the first year of the Trump Administration
contributed to a partial recovery of 1.0 million jobs since manufacturing
reached its trough at the beginning of 2007.
According to evidence presented in the next subsection, manufacturing
employment declines since 2001 were related to changes in international trade
and also to the Great Recession and the macroeconomy. Regardless of the
causes, rapid losses of manufacturing jobs have large negative consequences
for the affected communities and can cause harm with long-lasting effects.
Currently, the United States coordinates worker-adjustment policies based on
the cause of displacement, though workers displaced by all causes may face
similar issues after displacement.
Trade’s effects on U.S. manufacturing employment. The adverse employment and wage implications of direct or indirect Chinese import exposure to
U.S. industries, particularly the manufacturing sector, have been well recognized (Autor 2010). The literature emphasizes the effects of China’s accession
to the WTO in 2001 and the corresponding reduction in tariff rates on Chinese
imports to the United States, the “China shock,” as a defining moment in the
history of domestic manufacturing employment. Acemoglu and others (2016)
estimate that domestic job losses from Chinese import competition over the
1999–2011 period were in the range of 2.0 to 2.4 million, with manufacturing
accounting for 41 to 49 percent of this loss. Caliendo, Dvorkin, and Parro (2015)
indicate that China’s export growth caused the loss of about 0.8 million U.S.
jobs from 2001 to 2007. Kimball and Scott (2014) find that China’s accession
to the WTO caused 3.2 million Americans to lose their jobs between 2001 and
2013.
These estimates of the effects of import competition on American manufacturing employment show the relatively weak ability of gains from export
growth to offset the China shock within a specific sector. Although the U.S. also
gained access to the Chinese market following WTO accession, relatively lower
levels of development in China at the time of its accession implied limited
Chinese consumer spending power and weaker export opportunities for the
United States. However, other export markets have opened for the United
States in recent years, and the job gains from these export opportunities have
256 |

Chapter 5

Figure 5-13. U.S. Manufacturing Employment, 1965–2017
Manufacturing share of employment (%)
100

Employment level (thousands, SA)
20,000

Oct-2017

China's
WTO
accession,
2001

18,000

16,000

80

60

Manufacturing
employment
PNTR
established,
2000

14,000

40

20

12,000

10,000
1965

Manufacturing share of
nonfarm employment
1971

1977

1983

1989

1995

2001

2007

2013

0
2019

Source: Bureau of Labor Statistics.
Note: SA = seasonally adjusted; PNTR = permanent normal trade relations.

tempered the net effect of job losses due to Chinese import competition.
Feenstra, Ma, and Xu (2017) use exogenous variation in exports to measure
the increase in employment attributable to expanded export penetration.
Weighing job gains due to increased exports to the entire world against jobs
lost due to increased import competition from only China during the period
spanning 1991 to 2007, they estimate net job losses of 0.2 million to 0.3 million;
however, these authors estimate no net change in employment (e.g., that job
gains attributable to exports to the rest of the world offset the job losses due
imports from China) when they extend this analysis from 1991 through 2011.
Permanent normal trade relations with China. Other authors focus on the
establishment in the United States of permanent normal trade relations (PNTR)
with China as an explanation for reductions in U.S. manufacturing employment
during the early 2000s. PNTR, notably, resulted in no change in U.S. tariff rates
for Chinese imports. But the status change brought certainty that potential
future tariff hikes were now low-probability events; low tariffs that were in
place and subject to annual renewal were now in place, indefinitely. Pierce and
Schott (2016) identify a significant and robust decrease in employment concentrated in industries that were the most exposed to the change in U.S. trade
policy toward China to low normal tariffs. The authors hypothesize that PNTR
gave U.S. firms the certainty they needed to increase the value of imports from
China, resulting in greater import competition facing the U.S. manufacturing
sector and a reduction in the demand for U.S. manufacturing workers.
Enhancing U.S. Trade in a Global Economy

| 257

Regardless of whether WTO accession or establishment of U.S. PNTR
with China is the proper signal event, Asquith and others (2017) demonstrate
that the China shock reduced manufacturing employment from 1992 to 2011
primarily due to firm exits in the aftermath of higher import pressures, and that
workers displaced from exposed industries were reabsorbed into sectors that
were not exposed to Chinese competition, mainly due to births of new firms.
Who are the displaced workers? A broad trade policy might consider
the fate of individuals who bear the heaviest costs of increased import competition. A supplement to the Current Population Survey (CPS) measures
the demographics and labor market fortunes of displaced workers, defined
as those workers who were rendered at least temporarily jobless in the 36
months preceding the survey because of an establishment closing, a layoff, or
insufficient work. CPS survey years 2008–16 capture the period before, during,
and after the Great Recession, without the early years of the WTO or PNTR
changes with respect to China. The sample includes wage and salary workers
not employed in agriculture. During these survey years, manufacturing workers
constituted about 18 percent of all displaced workers on average—as shown in
figure 5-14, more than in construction; finance, insurance, and real estate; or
wholesale trade; but less than in services and retail trade.
Figure 5-15 shows that displaced manufacturing workers during that
period were, on average, less educated, and older than (1) displaced workers
in other parts of the economy, and (2) nondisplaced workers. These distinctions suggest that it may be physically and psychologically more difficult for
displaced manufacturing workers to find new work opportunities. Additionally,
a substantially higher share of displaced manufacturing workers were male.
Given the declining prime-age male labor force participation rate in the United
States, this result suggests that displaced manufacturing workers may be
particularly vulnerable to leaving the labor force, a hypothesis we explore in
more detail below.
The geographic agglomeration of displacement. When trade patterns
change for exposed, geographically concentrated industries, the types of
impact—both positive and negative—can also be geographically concentrated.
This is the flip side of industry agglomeration that allows firms to take advantage of being near a skilled workforce and being able to imitate competitors.
The sharp downturn in U.S. manufacturing employment that has occurred
since 2000 largely had an impact on the Southeastern States that have a large
manufacturing presence. At the same time, some U.S. border States have
benefited disproportionately from NAFTA, given the reduced shipping costs for
their producers. Indeed, seven States—California, Florida, Illinois, Michigan,
New York, Texas, and Washington—made up more than half of all U.S. exports
in 2016.6
6 State-level trade data from the Bureau of the Census is based on origin of movement, making
the results skewed toward border States.

258 |

Chapter 5

In theory, the efficient firms that thrive in the face of import competition or take advantage of new export markets have the potential to generate
new employment opportunities for workers who are displaced due to trade.
But regional concentration presents geographic barriers that may delay this
reallocation of labor, and recent evidence on geographic mobility indicates
that Americans are less likely to move than at any point in the recent past
(see chapter 3). The CPS data show that only a small minority of displaced
workers, both inside and outside manufacturing, reported moving in response
to being displaced, while a far larger share reported finding employment in a
different, broad industry category than their predisplacement industry (figure
5-16; industries are defined in figure 5-14). On this dimension, manufacturing
workers were 9 percentage points more likely to change industries after job
displacement than workers in other sectors.
What happens to displaced workers? Displaced workers may face a variety
of obstacles to rejoining the workforce. First, because displacement may occur
in agglomerated industries, relocation may be necessary to recover previous
earnings levels. Second, displaced workers may face information asymmetries,
lacking information on job matches, both local and further afield, that are
specific to their particular skill sets. Third, displaced workers may be disadvantaged by a negative signal of quality or skills due to a layoff and therefore
face difficulty competing with other job applicants in the labor market. Finally,
displaced workers may lack appropriate skills to meet remaining labor market
demand as the economy changes. Automation and trade shifts both change
the mix of skills demanded in the United States, and the skill-biased nature of
these changes implies that less-skilled workers in particular may struggle to
find new jobs.
What can we say empirically about the reemployment prospects of workers displaced by trade? The CPS data do not allow us to specifically identify
those workers displaced due to import competition, to automation, or to
any other change in the economic fortunes of their former employers. Still,
these data give a holistic picture of the reemployment patterns of workers,
independent of the cause of their displacement, along with the challenges that
displaced workers face in returning to the labor force.
Figure 5-17 provides the reemployment prospects of private displaced
workers by industry (excluding government and military workers), highlighting
the differences in reemployment and labor market exit probabilities by sector
of previous employment. The values in figure 5-17 represent the probabilities
of reemployment and labor force exit after controlling for year fixed effects, an
important consideration, given the uneven effects of the Great Recession, and
after controlling for worker age. Each probability is relative to the outcome
for workers in finance, insurance, and real estate (FIRE), whose unconditional
reemployment rates are the highest following displacement.

Enhancing U.S. Trade in a Global Economy

| 259

Figure 5-14. Displaced Workers by Industry, 2005–15
Construction
Retail trade
Services

Manufacturing
Finance, insurance, and real estate
Other industries
12.4%

13.8%

17.7%

31.2%

17.8%
7.1%
Source: Bureau of Labor Statistics, Displaced Workers Survey.
Note: "Other industries" includes agriculture; mining; construction; transportation,
communications, and utilities; wholesale trade; government; and military workers.

Figure 5-15. Characteristics of Displaced Compared with
Nondisplaced Workers, 2005–15
Displaced: manufacturing

Displaced: nonmanufacturing

Nondisplaced

Level
80

68.9
60

56.5
44.3

41.1

52.7

42.6

40

20

0

13.0

Age (years)

13.4

13.8

Education (years)

Sources: Bureau of Labor Statistics; Current Population Survey.

260 |

Chapter 5

Gender (% male)

A striking conclusion from figure 5-17 is the substantial labor market
disadvantage faced by workers in construction, agriculture, and retail trade
industries after displacement. Agriculture and retail trade workers, relative to
workers in FIRE industries, were greater than 3 percentage points more likely
to exit the labor force following job loss. Construction workers remained in the
labor force at rates similar to FIRE workers, but their disadvantage in reemployment was high, at 8 percentage points, compared with 14.3 percentage
points in agriculture and more than 5 percentage points for retail trade workers. Again, these differences are after controlling for differences in the year of
displacement and age at displacement.
Employees in the manufacturing, wholesale trade, transportation, communication, and utilities sectors fared somewhat better than those in construction, retail trade, and agriculture, experiencing differences in labor force
participation rates after displacement of 1.1 percentage points or less relative
to FIRE workers. But their reemployment prospects were still far weaker, a
disadvantage of between 3.2 and 5.7 percentage points. At the same time, the
employment and participation outcomes for workers in services were similar
to FIRE workers, as were outcomes for mining workers.
These data are somewhat consistent with the results of Autor and Dorn
(2013), who find that workers with abstract skills are substantially more likely
to be reemployed after displacement and substantially less likely to exit the
labor force. Using years of education as a proxy for abstract skills, the industry
categories given in figure 5-17 are sorted from highest average years of schooling to lowest for displaced industry workers in the CPS sample. (FIRE has the
highest average.) Thus, the blue bars in the figure demonstrate that reemployment prospects are increasing with the educational attainment of displaced
workers. In this sense, displaced mine workers are real outliers, experiencing
higher reemployment prospects than their ordering in the average educational
attainment of industries would indicate. The ranking of labor force nonparticipation for formerly displaced workers in figure 5-17, however, does not as
clearly conform to Autor and Dorn’s predictions; no clear pattern of eventual
nonparticipation by industry average education level is apparent.
Current solutions: trade adjustment assistance. Trade Adjustment
Assistance (TAA) provides job training, job search, and relocation assistance
to U.S. workers displaced by trade. Although TAA has been in place since 1962,
analyses of the program’s efficacy provide notably mixed results. Schochet and
others (2012) note that TAA participation hurts workers’ financial prospects,
relative to similarly displaced workers who did not participate in TAA. A 2007
evaluation of the TAA program by Reynolds and Palatucci (2012) found that
only 1.4 percent of TAA participants used relocation allowances. Additionally,
a 2010 evaluation by Dolfin and Beck (2010) found that the number of TAA
participants who received relocation allowances was not significantly different
than the number of TAA nonparticipants who received reallocation allowances
Enhancing U.S. Trade in a Global Economy

| 261

Figure 5-16. Share of Displaced Workers Changing Industry or
Location, 2005–15
Percentage of workers

Manufacturing

Nonmanufacturing

70
60
50
40
30
20
10
0

Changed Industry

Moved

Source: Bureau of Labor Statistics; Current Population Survey.
Note: Industries are 11 broad categories from the 1990 Census industrial classification
codes.

Figure 5-17. Probabilities of Employment and Labor Force
Nonparticipation Transitions for Displaced Workers, by Industry
Relative to Finance, Insurance, and Real Estate, 2005–15
Not in the labor force, probability

Employment, probability

Services

-1.1

Wholesale trade

0.6

-5.7

Transportation, communications, and utilities

1.1

-3.2

Manufacturing

0.0

-5.0

Retail trade

1.1

3.4

-5.6

Mining
Construction

-8.0

Agriculture

-14.3

-20

-15

-0.7
-0.3

-0.3

3.6

-10
-5
0
5
Probability (percentage points)

Sources: Bureau of Labor Statistics; Current Population Survey.
Note: Blue and red bars represent regression coefficients from a linear probability model
of employment and labor force nonparticipation, respectively. Regressions include age fixed
effects and year of displacement fixed effects. All values are relative to employment and
nonparticipation probabilities of workers in finance, insurance, and real estate.

from sources such as unemployment insurance claims, an indicator that relocation allowances were not a highly utilized part of the TAA program.
In the newest literature on this topic, Hyman (2018) relies on quasi-random assignment of TAA cases to different investigators and employer-employee
matched Census data on 300,000 displaced workers, to assess TAA’s effects on
labor market outcomes. Hyman finds that the program fails to facilitate longterm adjustment. Although training under the program leads to large initial
returns—$50,000 in additional cumulative earnings over a decade—such gains
do not persist past 10 years. Hyman attributes the initial gains to both rising
incomes and increased labor force participation. But TAA-trained workers tend
not to participate in other Federal retraining efforts that would result in further
formal education (e.g., Workforce Innovation and Opportunity funds, or WIOA),
depressing longer-run returns. However, the nature and length of training may
be different among these programs. Geographically, depreciating returns are
concentrated in States with shorter training programs. Unsurprisingly, returns
are concentrated in the most affected regions, where workers are likelier to
switch industries and move in response to TAA training.
Other possible solutions. Beyond worker retraining efforts in TAA or WIOA,
what other possible solutions to the labor market adjustment costs facing displaced manufacturing workers are available? We do not attempt an exhaustive
list here, but we note that the Administration’s emphasis on apprenticeship
models for worker training are well positioned to assist displaced manufacturing workers.
One particular challenge for displaced workers, who are older and more
likely to face short-term household budget pressures, is funding their retraining efforts while still making household financial deadlines. Designing Federal
programs to deliver financial support to displaced workers during their retraining efforts would help relieve these pressures and allow workers to make more
strategic long-term investments in their financial security through apprenticeships (which are sometimes, but not always, paid) or other medium-length
training programs. Without financial support, some workers will find it necessary to quickly take a new job, even if it is not in their long-term best interest.
TAA Trade Readjustment Allowances are payable during approved retraining
periods for workers determined to be TAA-eligible. (For more on related issues,
in the context of the Tax Cuts and Jobs Act of 2017, see box 5-2.)

Trade Opportunities for the United States
Trade can create opportunities for both producers and merchants in the United
States to export products to the rest of the world. And U.S. policy can help to
ensure that these opportunities come to fruition and deliver economic gains
to Americans.

Enhancing U.S. Trade in a Global Economy

| 263

Box 5-2. Distressed Communities and the Tax Cuts and Jobs Act
As pressures from trade and technology have reduced the economic viability
of American manufacturers, factory closings have left communities that
relied on their jobs in distress. Job losses in distressed communities have
had obvious negative effects on income and economic well-being, but also
on health and mortality, including mortality from suicide (Sullivan and von
Wachter 2009; Classen and Dunn 2009). Effects on the children of unemployed
workers are also apparent (Schaller and Stevens 2011). Also, job loss effects
are not confined to former manufacturing workers and their households;
local aggregate demand falls with plant closings, inducing further consequent
job losses—for example, in the service sectors. Housing markets also suffer,
affecting the tax base for schools and other public goods.
In the years after the Great Recession, labor markets recovered in most
of the country, even as some communities were left behind. Uneven recovery
motivates households to consider migration as a way out of local unemployment and underemployment. But migration is not costless, and the overall
declining proclivity of Americans to move (see chapter 3) implies the need for
a rejuvenation of local job growth as one component of any policy strategy
to address local employment losses. The agglomeration of trade-dependent
industries makes local retraining and revitalization projects an important
domestic dimension of trade policy.
The Federal government has an active set of policies to encourage investment and job creation in distressed communities, including
Empowerment Zones, Enterprise Communities, Renewal Communities, and
New Market Tax Credits (NMTC). The NMTC—arguably, the most successful of
these programs—is structured to induce “patient” capital, providing substantial investment incentives if assets are held over a full seven years. As a result,
although the majority of NMTC recipients would not have otherwise invested
in the benefiting community, real estate has been the investment of choice,
both because real estate returns are naturally long-run and because these
investments clearly complied with NMTC regulations (Bernstein and Hassett
2015). But real estate is likely not the most effective tool for job growth, and
the program is reportedly difficult for entrepreneurs to navigate.
During the past two years, bipartisan support has developed for a
new program to encourage investment in distressed communities, and the
Tax Cuts and Jobs Act of 2017 greenlighted the designation of Opportunity
Zones to drive this investment. The Investing in Opportunity Act—original,
bipartisan legislation sponsored by Senator Tim Scott (R-SC) and Senator
Cory Booker (D-NJ) and Representative Pat Tiberi (R-OH) and Representative
Ron Kind (D-WI), alongside nearly 100 additional cosponsors—has three
key features. Opportunity Zones are limited to low-income areas that have
recently experienced unemployment due to business closures or relocation.
States designate particular Census tracts as Opportunity Zones, subject to
particular socioeconomic criteria. They are meant to establish an investment

264 |

Chapter 5

vehicle that pools assets from a group of investors, much like a venture
capital or mutual fund for distressed community investment, to spread risk
and increase the scale of investment. Finally, to incentivize investors to place
capital in Opportunity Funds, the Opportunity Zone program provides investors with the option to temporarily defer and modestly reduce capital gains
taxes if they keep capital in an Opportunity Fund for a period of time.

U.S. Agricultural Trade
The United States enjoys comparative advantage in many types of agricultural
production thanks to ample land, favorable climate, and smooth integration
of human, financial, and intellectual capital into production. Export markets
are crucial to U.S. agricultural producers because American farmers produce
more than American consumers can eat. The United States leads the world
in agricultural exports, and runs a trade surplus for agricultural goods. The
top three export markets for U.S. agricultural products in 2016 were China,
Canada, and Mexico; together, these three countries accounted for 44 percent
of all agricultural exports (ERS 2017a). Exports are especially important for
certain products. The Department of Agriculture (USDA 2017a) estimates that
American producers will export on average 76 percent of the cotton, and virtually half of all the wheat, soybeans, and rice they grow to the rest of the world,
from the fiscal year 2016–17 through 2026–27 crops. In general, the United
States is a major exporter of bulk agricultural commodities like coarse grains
and meat, and a net importer of consumer-oriented, processed goods that provide Americans with variety, like wine, beer, and out-of-season fresh fruits and
vegetables. The USDA (2017b) forecasts agricultural exports at $140 billion for
fiscal year 2018. This includes $33.1 billion in oilseeds and products (including
$24.1 billion in soybeans); $29.4 billion in grain and feed (including corn); $29.7
billion in beef, poultry, and animal by-products; $34.5 billion in horticultural
products; and $4.8 billion in cotton.
Export markets present a growth opportunity for U.S. agriculture.
Agricultural exports have grown substantially over time, as shown in figure
5-18, consistently contributing to a trade surplus for decades; today, they
account for about a third of American farmers’ gross cash income (Schnepf
2017), up from less than 20 percent before the mid-1970s, and less than 10
percent before the mid-1940s. By 2050, population growth and increasing
standards of living are expected to require global agricultural production to
increase 60 percent from the level of 2005–7 (Alexandratos and Bruinsma
2012). U.S. agriculture is well positioned to supply a large share of this rising
demand, provided productivity improves and trade barriers do not prevent
U.S. farmers from expanding their access to new markets and consumers.

Enhancing U.S. Trade in a Global Economy

| 265

Agricultural trade is distorted worldwide. Governments around the world
have long intervened in agricultural and food markets. Interventions distort
price signals and draw resources into inefficient forms of production, shifting
them from areas where they could be used more productively. Figure 5-19
shows that even as high-income economies have reduced government interventions in markets for nonagricultural goods, their history of interventions in
agriculture is much more heavy-handed.
Interventions in agriculture include domestic supports like production, insurance, and income subsidies, and directed trade policy like import
restrictions and export controls. Both sets of policies transfer wealth between
domestic consumers and producers. Governments offer several reasons for the
special protective treatment of agriculture:
• to make sure enough food is available to meet domestic demand,
• to shield agricultural producers from volatile weather and market conditions, and
• to preserve rural society.
But by restricting trade and warping market signals, these policies also
amplify international price spikes (Martin and Anderson 2012; Carter, Rausser,
and Smith 2011), undermining living conditions of people around the globe.
Anderson, Cockburn, and Martin (2010) estimate that agricultural market
interventions accounted for 70 percent of the global welfare cost of all goodsmarket distortions, even though the sector represents just 6 percent of world
trade and 3 percent of world GDP.
Under the WTO architecture, tariffs for nonagricultural goods have fallen
steadily worldwide, but agriculture has not liberalized at the same pace. Despite
efforts aimed at their reduction, higher barriers to trade persist, through direct
restrictions like tariffs, along with a variety of nontariff measures. Figure 5-20
recreates figure 5-10 for only agricultural products, and presents tariff and
overall trade restrictiveness for agricultural goods among the United States,
other high-income members of the G20, emerging nations in the G20, and select
countries in the developing world. Like figure 5-10, it depicts a similar overall
pattern in terms of differential trade restrictiveness for agricultural products:
the United States applies lower tariffs and trade barriers to agriculture than the
average high-income nation, emerging economy, and developing country—
however, the restrictiveness magnitudes are much higher. U.S. exporters also
face notably higher barriers to trade for their agricultural exports, compared
with exporters from the average high-income or emerging member of the G20.
While further liberalization of agricultural trade will benefit consumers around
the world, it would likely also benefit U.S. exporters.
The U.S./EU discrepancy in agricultural market protection. One example
of the discrepancy in agricultural market access between even high-income
countries is illustrated by the varying levels of import restrictions applied by
the United States and the EU, two of the world’s major producers that together
266 |

Chapter 5

Figure 5-18. U.S. Agricultural Imports, Exports, and Trade Balance by
Value, 1940–2017

Dollars (billions)
200

2017

160

Exports
120

80

Imports
40

0

-40

Trade
balance
1940

1950

1960

1970

1980

1990

2000

2010

2020

Source: U.S. Department of Agriculture, Economic Research Service.

Figure 5-19. Government Support of Agricultural and Nonagricultural
Goods in High-Income Countries
Ad valorem equivalent (%)
60
50
40

Agricultural
goods

30
20

Nonagricultural
goods

10
0

1955– 1960– 1965– 1970– 1975– 1980– 1985– 1990– 1995– 2000– 2005–
59
64
69
74
79
84
89
94
99
04
10

Source: Anderson, Rausser, and Swinnen (2013).

Enhancing U.S. Trade in a Global Economy

| 267

account for close to 40 percent of global agricultural trade. Agricultural trade
between the two has decreased over time, due in part to the relatively high
trade barriers that the EU places on U.S. agricultural exports (ERS 2016)—as
shown in figure 5-21—even as overall trade between the United States and
the EU has grown. In terms of simple averages for all traded goods, the United
States levies a tariff of 3.5 percent on EU exports, while the EU charges 5.5
percent for U.S. exports. For agricultural commodities alone, however, the
comparable rates are 4.7 percent and 13.7 percent, respectively. Many goods,
including dairy and meat products, face far higher tariffs from the EU than vice
versa (Beckman et al. 2015). These rates do not include EU nontariff measures
that combine with tariffs to present significantly higher barriers to U.S. exports
of meat products, corn, soybeans, and fruits and vegetables (Arita, Beckman,
and Mitchell 2016).
Investing in productivity and negotiating trade agreements. Increases in
American agricultural production are mainly due to innovations and advances
in technology, in areas such as crops, livestock breeding, fertilizers, pest management, farm practices, and farm equipment and structures (Clancy, Fuglie,
and Heisey 2016). Productivity-enhancing innovations rely on funding from
both the public and private sectors. Although public investment in agricultural
research has resulted in large economic benefits with annual rates of return
between 20 and 60 percent (Fuglie and Heisey 2007), real public research
investment in the United States is declining. The United States accounted
for 20 to 23 percent of global public sector funding for food and agricultural
R&D between 1990 and 2006, but fell behind China beginning in the late 2000s
(Clancy, Fuglie, and Heisey 2016) and has since remained behind. By 2013, the
United States’ share of worldwide research funding fell to just 13 percent.
Market access supports U.S. farmers. Free trade agreements help expand
U.S. agricultural exports into foreign markets. The United States’ 20 current
FTA partners represent 10 percent of the world’s GDP and 6 percent of its population, but account for 43 percent of U.S. agricultural exports, an increase of 29
percentage points since 1990 (FAS 2016). Even though farming accounts for a
relatively small share of the U.S. economy, agricultural trade makes significant
and outsized contributions to GDP. Based on 2015 data, USDA’s Economic
Research Service estimates that each $1 billion in U.S. agricultural exports
supports approximately 8,000 jobs throughout the economy, and each $1 in
agricultural exports stimulates an additional $1.27 in business activity (ERS
2017b). These indirect activities include facilitating the movement of exports
to their final destination. ERS (2017b) reports that U.S. agricultural exports in
2015 supported over 1 million jobs both in and outside agriculture, and also
generated $302.5 billion in total economic activity.

268 | Chapter 5

Figure 5-20. Comparing Agricultural Trade Restrictiveness
Tariffs–imports

NTMs–imports

Ad valorem equivalent (%)
0

10

Tariffs–exports
20

NTMs-exports
30

40

1

United States

2
3
4

Other high5
income G20

6
7
Emerging G20
8

9
10
Developing
11
Source: World Bank.
Note: NTM = nontariff measures.

Figure 5-21. Tariffs on Agricultural Imports
U.S. tariffs on imports from European Union
European Union tariffs on imports from United States

Ad valorem equivalent (%)
80

60

40

20

0

Source: U.S. Department of Agriculture, Economic Research Service.

Enhancing U.S. Trade in a Global Economy

| 269

U.S. Energy Dominance Relies on Trade
Current trends indicate that the United States may become a net energy
exporter by 2026. Historical U.S. dependence on crude oil imports makes this
vision seem farfetched—as recently as 2005, net imports accounted for 60 percent of domestic oil and petroleum product consumption. Even a decade ago,
when over 40 percent of the total trade deficit was attributable to petroleum
imports, a future where the United States was relatively independent of energy
imports was inconceivable. The outlook today is decidedly different. As shown
in figure 5-22, the petroleum trade balance was 12.4 percent of the total trade
deficit in 2016, the lowest observation since 1991.
Between 2008 and 2017, domestic petroleum production nearly doubled
and natural gas production increased by one-third, facilitated by technological
improvements that enabled profitable production from resources once considered too costly to exploit. Widespread private ownership of mineral resources
in the United States is unique in the world, and the willingness of mineral owners to form partnerships with developers has contributed to the observed rapid
increase in production. Realizing the goal of increased exports requires specific
infrastructure investments to allow for increased capacity, but also the security
provided by enhanced FTAs that expressly address energy trade.
All fuels have a role to play in U.S. energy dominance. In 2017, coal exports
expanded by nearly 60 percent year on year, and the industry’s prospects are
looking up. In natural gas, the United States is the world’s leading consumer
and among its leading producers, so gains can be had in increasing extraction
and domestic distribution. But increasing its international trade by transporting liquefied natural gas (LNG) requires more liquefaction capacity to permit
increased exports, and an expansion of pipeline links to Mexico and eastern
Canada to allow more terrestrial exports. Increasing American exports of crude
oil and refined products will further exploit the U.S. comparative advantage
and lock in the manufacturing value-added characteristic of oil refining.
Continued integration of renewable electric generation technologies into the
American grid would allow for low-cost, resilient domestic electric generation.
Finally, the United States has been a historic leader in civil nuclear technology,
and opportunities for exporting these technologies may arise in coming years.
Coal. The United States has the largest coal reserves of any country in the
world. Until the Trump Administration took office, the economic prospects for
coal in the U.S. appeared bleak in the years after 2008, when U.S. coal production reached an all-time high before steadily declining. Coal-based energy production peaked in 2007, and no new coal plants for domestic utility electricity
production have been built since 2014. Reducing reliance on coal is a long-term
trend. From the beginning of 2002 to the end of 2016, the maximum generating
capacity of coal-fired power plants in the United States fell from 306 to 270
gigawatts. Over that 14 year period, 531 coal generating units were retired,

270 |

Chapter 5

with nameplate capacity of 59 gigawatts (new units brought online explain the
difference). In 2016, coal-fired power plants made up 25 percent of nameplate
generation capacity in the United States, and supplied about 29 percent of the
electricity generated. Both figures are down, from 35 percent and 50 percent
in 2002, respectively.
From 2015 to 2016, domestic coal production declined by 18.8 percent,
to its lowest level since 1978. Contemporary employment in the coal-mining
industry fell from 69,000 in January 2015 to just 49,000 by the end of 2016, a
decline of 29 percent. Total U.S. coal distributions for 2016 were 737 million
short tons, 17 percent less than the level of distributions in 2015.
Coal-powered steam generation has declined steadily, due to the falling
price of natural gas in the past decade. Between 2008 and 2015, the use of
natural gas to generate electricity grew 51 percent. Culver and Hong (2016)
compared the monthly price of gas with that of coal from January 2012 to
January 2016 and found that for nearly 90 percent of the months in this period,
natural gas was more competitive than Appalachian coal. Natural gas was also
less expensive than Appalachian, Illinois, and Rockies coal for 57 percent of the
months studied.
Despite this long-term trend for domestic coal production and usage,
metallurgical coal exports were a bright spot in 2017. Exports of metallurgical
coal from East Coast ports expanded dramatically that year, with over 25 million short tons of coal exported through the second quarter alone—an increase
of roughly 27 percent from the same period in the previous year.
Wolak (2016) examines the potential impact on the world coal market of
increasing coal export capacity from the West Coast. The net effect increases
U.S. exports to the Pacific Basin and reduces Chinese domestic production.
Increased Chinese access to less expensive, cleaner-burning U.S. coal would
drive up coal prices and accelerate the switch to natural gas–fired generation
in the United States. The environmental dividend of this conversion is double—
U.S. emissions would fall as cleaner-burning natural gas substituted for coal,
and global emissions would also decease as higher grades of U.S. exported coal
would substitute for the poorest Chinese grades. Environmental economists
often worry about emissions leakage as energy trade increases, but Wolak’s
result suggests that leakage from U.S. coal exports to Asia could well be negative. Currently, one project expanding Pacific export capacity is involved in
litigation over rejected State permits in Washington State. The current capacity
for coal exports to the Pacific is limited—barring a new terminal in Washington,
the only outlet in the Pacific Northwest relies on transshipment through
Vancouver, in Canada.
There is unlikely to be significant short-term change in coal consumption outside the United States, due to limited opportunities for international
electricity sectors to substitute away from coal in the short and medium terms
(Wolak 2016). Power plants have life spans that last decades; once they are
Enhancing U.S. Trade in a Global Economy

| 271

Figure 5-22. U.S. Petroleum and Total Goods Trade Deficits,
1991–2016
Dollars (millions)

0

Petroleum
deficit

-200,000

-400,000
Trade deficit in goods

-600,000

-800,000

1991

1996

2001

2006

2011

2016

Source: U.S. Census Bureau.

built, they lock in fuel demand. Wolak notes that current low natural gas prices
make it unlikely that many regions of the world would expand their installed
capacity of coal-fired generation in response to increased U.S. exports. This
limited flexibility in global demand for coal may cause the potential gains to be
short-lived. U.S. exporters are likely to be inframarginal suppliers to the Asian
market as marginal Asian production is displaced. International coal trade has
been price-sensitive historically; figure 5-23 shows how U.S. coal exports have
fluctuated with coal prices since 2002.
Natural gas. Thanks to the technological transformation of the Nation’s oil
and gas sector enabling exploitation of unconventional resources (commonly
termed “fracking”), U.S. natural gas production and net exports expanded over
the past decade, as shown in figure 5-24. Between 2007 and 2016, annual gross
natural gas withdrawals expanded by 32.3 percent. Apparent consumption of
natural gas rose across all sectors during this period, with total U.S. consumption, in trillion cubic feet, rising from 23.3 in 2007 to 27.5 in 2016. The United
States is the world’s largest consumer of natural gas, and has been its largest
producer since 2006.
Thanks in part to new infrastructure investments, U.S. natural gas
exports have expanded as well; between 2007 and 2016, U.S. exports of natural
gas increased by 184 percent, totaling over 2.3 trillion cubic feet in 2016. This
has moved the United States from being a net importer to a net exporter of
natural gas. Both major forms of exports expanded—natural gas transported

272 |

Chapter 5

in pipelines, and LNG that can be transported by ship. Figure 5-25 shows the
expansion of these exports since 2000. Export capacity utilization was 59
percent in 2017, up 7.5 percentage points from the previous year. LNG exports
have the advantage in that they can be delivered to any country with an LNGunloading terminal, as opposed to pipeline connections, which are limited to
Canada and Mexico.
The United States exported nearly 410 billion cubic feet of LNG through
August 2017, doubling the quantity exported in all of 2016. Until 2016, the only
LNG capacity in the continental U.S. was configured for imports, as opposed
to exports. More than 60 percent of all current export capacity came online in
2017—a total of 2.15 billion cubic feet per day. It is particularly important that
four additional LNG export facilities are now under construction, in addition
to expansions of two existing facilities. In the years 2001–10, LNG averaged
8.3 percent of total U.S. natural gas exports because only one export facility
existed (in Alaska). As of August 2017, LNG accounts for 20 percent of total U.S.
natural gas exports.
One important question is whether U.S. natural gas producers can manage to supply exports while also satisfying domestic demand. Since the shale
revolution took hold in the mid-2000s, the U.S. natural gas market has been
isolated from the global market by limited export capacity. New LNG capacity
changes this dynamic, and U.S. exporters can sell relatively cheap U.S. natural
gas on the world market. Natural gas supply is now more elastic than it was
previously, and it responds to price movements more quickly (Newell, Prest,
and Vissing 2016). This flexibility may be used to deepen relations in existing
markets as well as in emerging markets abroad, including Latin America, Africa,
and Asia. Even though Mexico has the benefit of direct natural gas pipeline
links with the United States, it was also the largest market for U.S. LNG in 2017,
receiving 126 billion cubic feet through October. Deeper U.S. export penetration may provide these regions with increased energy security while also helping to grow U.S. exports.
A critical factor that facilitated the ascension of the United States’
natural gas market was the structuring of commodity transportation within the
country. Pipeline transporting capacity is divorced from pipeline ownership,
allowing competitive bidding for any producer to access transportation for the
fuel generated, regardless of size (Makholm 2012). When ownership of products and capacity is bundled (along with this infrastructure being vertically
integrated with larger producers), in the case of many other markets, larger,
integrated firms are often able to exercise market power in lieu of regulation.
Petroleum. The same suite of technological innovations that spurred
natural gas production contributed to higher U.S. crude oil production, which
rose by about 5 million barrels per day between 2008 and 2015. Oil production
is poised to surpass 10 million barrels per day in 2018. This windfall stands to
benefit Americans even more than abundant natural gas, because oil is more
Enhancing U.S. Trade in a Global Economy

| 273

Figure 5-23. U.S. Monthly Coal Exports, 2003–2017
Global coal price index (2012 = 100)

Metric tons (millions)

140

14

Nov-2017
12

120

Global coal price index

U.S. coal exports

2017

2016

2015

2015

2014

2013

2012

2012

0

2011

0

2010

20

2009

2

2009

40

2008

4

2007

60

2006

6

2006

80

2005

8

2004

100

2003

10

Sources: Energy Information Administration Quarterly Coal Report; CEA calculations.

transportable than natural gas. After a 2015 decision to remove a 40-year-old
crude export ban, crude oil exports ended 2017 at a monthly average of over
4,000 barrels per day. In addition to crude oil, refined product exports began
to increase before the crude export ban was lifted, and the trend continued
through 2017. For example, exports of refined products to Latin America have
grown; over the past two decades, Latin America on average accounted for
34 percent of the consumption of U.S. petroleum product exports, and for 28
percent of the growth in these exports.
Continuing this trend and expanding U.S. capacity to export crude oil and
refined products will be critical for the U.S. to reach the status of net energy
exporter by 2026. New projects to expand capacity are under way; crude oil
exports are expected to start from the Louisiana Offshore Oil Port (known as
LOOP) in early 2018, at about 2 million barrels per month.
Renewables. Electricity is a tradable product for the United States. In
2016, the United States imported a net of 60,000 gigawatt-hours of electricity,
mostly from Canada. The amount of electricity trade is small—the net imports
were about 1.4 percent of total generation in 2016. So though electricity trade
is not itself significant, the recent developments in the U.S. electricity sector
have proven disruptive and have freed up U.S. fuels for export.

274 |

Chapter 5

Figure 5-24. U.S. Monthly Natural Gas Withdrawals and Net
Imports, 1973–2017
Cubic feet (millions, 12-month moving average)
3

Oct-17

U.S. natural gas gross withdrawals
2

1

Net Imports

0

-1

1973

1982

1991

2000

Source: Energy Information Administration.

2009

201

Figure 5-25. U.S. Natural Gas Exports by Pipeline and LNG, 2000–17
Millions of cubic feet
300,000

Aug-2017

250,000
200,000

Pipeline gas exports
150,000
100,000
50,000
0
2000

LNG Exports
2002

2004

2006

2008

2010

2012

2014

2016

Source: U.S. Energy Information Administration.
Note: LNG = liquefied natural gas.

Enhancing U.S. Trade in a Global Economy

| 275

Figure 5-26. U.S. Quarterly Solar Photovoltaic Installations,
2011:Q1–2017:Q3
Residential
Non-residential

Quarterly installations (megawatts)
5,000

Utility

2017:Q3
4,000

3,000

2,000

1,000

0
2011

2012

2013

2014

2015

2016

2017

Sources: Solar Energy Industries America; GTM Research 2017:Q4 Solar Market Insight.

In recent years, renewable energy’s share of generation capacity has
grown substantially. The historic mainstay of renewable generation, hydroelectric power, has contributed to nearly constant installed capacity since the early
2000s. At the end of 2016, 8.4 percent of U.S. nameplate electric generation
capacity was powered by nonhydroelectric renewables. Wind power grew from
1 percent of the generation mix in 2008 to 5.6 percent in 2016. More than 8,000
megawatts of new wind capacity was installed in 2016, an increase in cumulative wind capacity of 11 percent from the prior year. Texas alone installed 2,611
megawatts of capacity, and in 14 States wind generation exceeded 10 percent
of total generating capacity.
Solar photovoltaic (PV) capacity has also grown rapidly since the late
2000s, contemporaneously with falling hardware and installation prices during this period. Installed capacity has grown from less than 1 gigawatt in 2008
to over 47 gigawatts in 2017. This expansion has occurred across all forms of
installations, with increased solar PV implementation in the residential, nonresidential, and utility sectors (figure 5-26). An additional 22 gigawatts are currently under contract (those with signed power purchase agreements) in the
United States for 2017, with another 36 gigawatts announced. State Renewable
Portfolio Standards, declining technology costs, Federal tax credits and subsidies, voluntary and retail procurement, the Public Utility Regulatory Policies
Act, and a new third-party ownership model for residential PV systems have all
been credited as drivers of renewable growth. According to a 2012 report on

276 |

Chapter 5

the impact of tax policies of renewables, Energy Investment Tax Credits and
Production Tax Credits combined led to more than $1.7 billion in cumulative
tax cuts toward developing PV infrastructure (Sherlock 2012).

Updating American Trade Policy
Historically, the United States has exercised leadership in pursuit of a policy
of lowered trade barriers and increased market access. The gains from these
actions have, as a whole, served to boost income in the U.S. as well as around
the world. But U.S. trade with the world has also, at least in some cases,
imposed costs on some Americans.
Congress delegates trade negotiating power to the President, which
gives the President considerable control over the outcome of a disagreement
with a trading partner. This power can be exercised without relying on a thirdparty arbiter such as the WTO, which ensures that the United States maintains
its sovereignty with respect to economic issues.
In its first year, the Trump Administration has used all available tools
to address imbalances. Section 201 of the Trade Act of 1974 seeks to protect
U.S. industries faced with serious injury from import competition through the
increasing imports of competing merchandise. Section 201 cases may be initiated by private petitioners that believe that imports have caused or are likely
to threaten serious injury. Import competition must be a “substantial cause”
of the serious injury. Two Section 201 investigations concluded in 2017 with
positive injury findings by the USITC, and both resulted in Presidential proclamations of new import restrictions—one for crystalline solar photovoltaic
products, and a second for large residential washing machines. These are the
first Section 201 cases filed in 15 years.
Section 337 of the Tariff Act of 1930 provides possible remedies against
unfair methods of competition and unfair acts in the importation of articles.
Section 337 is most commonly invoked against articles that infringe on U.S.
patents or trademarks, or that are made according to misappropriated trade
secrets; a total of 20 cases have been completed since January 2017, and an
additional 51 are under way. The primary form of remedy is an exclusion order
against imported goods.
Although the President does not have a direct role, AD/CVD proceedings,
which firms rather than the government typically initiate, are the most common trade action, with 33 completed investigations since January 2017, and
another 30 under way.

Conclusion
This chapter began by recapitulating the consensus toward which, over time,
the economics profession has converged: that trade across international
borders will virtually always generate net gains for all the nations involved. The
Enhancing U.S. Trade in a Global Economy

| 277

distribution of the gains from trade within a country are also important to consider. Although efforts to expand trade have historically provided gains in the
form of higher incomes and living standards in America (and around the world),
trade exposure has imposed costs on certain segments of American society.
The evolving U.S. and global economies provide opportunities for
the U.S. to gain from trade. The United States enjoys a comparative advantage in agricultural production, innovative goods, and many other products.
Meanwhile, U.S. energy production, boosted by technological breakthroughs
in recent years, has increased to the point where the United States could
become a net energy exporter in the coming years. The United States remains
a large and vibrant economy with many comparative advantages, and it stands
poised to gain from trade agreements that allow it to maximize the benefits of
its underlying dynamism while removing barriers to trade imposed by other
nations that disproportionately harm American producers.

278 |

Chapter 5

x
Chapter 6

Innovative Policies to Improve
All Americans’ Health
There are many determinants of health, but recent U.S. government health policy has mainly focused on one: expanding health insurance coverage through
implementation of the Affordable Care Act (ACA). The ACA expansion covered
fewer people than anticipated, and most of them gained coverage through
Medicaid, which provides small, uncertain benefits and limited access to care.
Moreover, the ACA imposed costly mandates and regulations that raised costs,
diminished people’s choices, and forced them to buy insurance they neither
wanted nor needed.
Health insurance is a positive factor in the lives of Americans, providing financial protection and peace of mind in case of serious illness. But the evidence
shows that health insurance provided through government expansions and
the medical care it finances affect health less than is commonly believed.
Determinants of health other than insurance and medical care—such as drug
abuse, diet and physical activity leading to obesity, and smoking—have a
tremendous impact and have exacerbated recent declines in life expectancy,
despite the ACA’s increased coverage.
Health policy that is predominantly focused on expanding insurance coverage
risks missing other policies that can improve the health of our citizens. This
Administration is focused on reversing the harm caused by the ACA by fostering
competition, choice, and innovation while also addressing the many factors
beyond insurance that influence health. The Administration is particularly
concerned about the opioid crisis that exploded during the ACA expansion. The
CEA finds that the economic cost of the opioid crisis—as much as $504 billion

279

in 2015, or 2.8 percent of gross domestic product—is far higher than previous
estimates. The Administration has taken substantial steps to decrease the
supply of prescription and illicit opioids and to increase treatment options.
Additional government actions and private innovation may be needed to make
further progress against opioid abuse and other behavioral health problems
such as obesity and smoking.
Innovation often reduces the price of health over time by providing previously
unavailable treatments at patent-protected prices that fall as competing
brands, and eventually generic products, come on the market. Nevertheless, as
the prices of new specialty drugs have demonstrated, the initial prices of innovative products can sometimes be so high that people struggle to pay for them.
The Administration is committed to bringing down the price Americans pay for
healthcare, especially drug prices, while preserving and improving incentives
to innovate. This chapter outlines the Administration’s efforts to move beyond
government insurance expansions that provide uncertain benefits to only a
small segment of the population and instead pursue initiatives that lower costs
and improve the health of all Americans.

H

ealth is an extremely valuable good, because it is a prerequisite for
fully enjoying life’s many activities. Although improving health is an
important goal in itself, it is also important because better population
health will increase productivity and economic growth. This chapter discusses
the Administration’s initiatives to enhance Americans’ health and how these
efforts differ from previous policies.
Over the last eight years, health policymakers focused on expanding
insurance coverage, primarily through the Medicaid expansion and exchange
subsidies in the Affordable Care Act (ACA). Health insurance is a positive factor
in the lives of Americans, providing financial protection and peace of mind in
case of serious illness. But the ACA expansion had a limited effect on health
and many downsides. The ACA only expanded coverage to, at most, an additional 6 percent of the population. In addition, most of these people gained
coverage under Medicaid, which provides limited access to care and uncertain,
and at most modest, health benefits. Moreover, the ACA imposed costly and

280 |

Chapter 6

cumbersome mandates and regulations that raised costs, decreased people’s
choices, and forced them to buy insurance that they neither wanted nor
needed. The lack of competition and choice inherent in the ACA approach was
not the most efficient way to provide coverage and medical care for the poor
and uninsured, and thus resulted in higher health insurance premiums and
spending.
By focusing primarily on expanding insurance coverage through the
ACA, the health policy community lost sight of other important policies that
can improve the health of a larger share of our citizens. Extensive economic
and medical literature shows that while public insurance coverage expansions
increase the amount of healthcare used, they generally improve health less
than is commonly believed. Indeed, there is substantial evidence that other
determinants of health outside of insurance and medical care, such as diet
and physical activity, smoking, and drug abuse, impact health enormously.
For example, for the first time in over 50 years, life expectancy in the United
States declined for two consecutive years—in 2015, and again in 2016 (CDC
2016a, 2017a). This negative health outcome occurred despite the ACA coverage expansion, suggesting that other factors are causing population health to
deteriorate.
This chapter focuses on the Administration’s goal of reversing the harm
caused by the ACA while taking a broader perspective—with a focus on fostering competition, choice, and innovation—in order to improve the health of the
entire population in the most cost-effective way possible.
We first assess the evidence about the small and uncertain, positive
effect that government insurance expansions have on health. Government
expansions in general, and the ACA in particular, often replace uncompensated
care with insurance coverage that provides limited access to low-benefit care
and imposes requirements that are ineffective. The ACA exacerbated this by
imposing mandates and price controls that resulted in an unstable market
and rapidly rising premiums. We document the policies the Administration has
adopted to restore competition and choice to the insurance market.
We next discuss three important determinants of health other than insurance and medical care—opioid abuse, obesity, and smoking—that have an outsized influence on the most common and costly illnesses. The Administration
has focused on confronting the opioid epidemic, which exploded through the
period of ACA expansion, leading to immense social and family disruption
along with loss of life and human dignity. The Surgeon General lists substance
abuse, including the opioid epidemic, among the nation’s top health priorities
(HHS 2017). The opioid problem has reached crisis levels in the United States,
and is partly responsible for declining life expectancies—in 2016, over 42,000
Americans died of a drug overdose involving opioids. We highlight a study
by the Council of Economic Advisers (CEA 2017), which finds that previous
estimates greatly underestimate the economic cost of the opioid crisis by
Innovative Policies to Improve All Americans’ Health | 281

undervaluing the lives lost to overdoses. This study found that in 2015, the
economic cost of the opioid crisis was as much as $504 billion, or 2.8 percent
of gross domestic product (GDP) that year. We enumerate the Administration’s
efforts to address the crisis and its negative health effects.
For some behaviors that drive population health, private sector medical
innovation may be a more effective way of decreasing the negative consequences of these health behaviors than traditional government interventions
like public health measures and raising prices through taxation. In particular,
the private sector has recently delivered many innovative new products and
procedures to treat everything from HIV, heart disease, cancer and hepatitis C.
These innovations can be seen as lowering the effective price of better health
by providing previously unavailable treatments—and thus, from an economist’s perspective, prohibitively expensive treatments—at patent-protected
price levels that decline over time as competitors, and eventually generic
substitutes, come to market.
But innovations that improve population health are unhelpful if people
cannot afford them. The Administration is focusing on two essential goals: first,
to decrease the price Americans pay for healthcare and especially expensive
drugs; and second, to keep lowering the effective price of better health in the
future by spurring medical innovation. We discuss how these two goals can be
achieved through a combined strategy to reduce inefficiently high prices at
home while at the same time reducing free-riding abroad.
Administration policies that promote market competition and choice will
deliver innovative solutions to these and other health problems, as opposed to
top-down, government approaches. A more holistic approach to health policy
beyond a singular focus on low-quality coverage expansions for a small part of
the population will improve health, increase productivity, and lead to greater
economic growth.

Healthcare Insurance and Spending, and the ACA
During the last eight years, the health policy community focused on the
Affordable Care Act, which set out to improve health by expanding health
insurance coverage while also decreasing healthcare spending. The ACA failed
on both fronts: it ended up insuring fewer additional people than projected
(CBO 2017)—only 6 percent of the population, most of whom gained coverage
under Medicaid (Uberoi, Finegold, and Gee 2016), a program with historically
small and uncertain effects on health outcomes and limited access to care. The
remainder of newly insured people were forced to buy more expensive and
elaborate insurance than many wanted, or else be subject to a penalty. Not
only were the ACA’s ostensible cost control features ineffective, but they also
led to market consolidation, decreased competition, higher premiums, and a
clear increase in inefficient healthcare spending.

282 |

Chapter 6

The Impact of the ACA and Health Insurance on Health
Expanding health insurance coverage provides protection against financial catastrophe in the case of serious illness. Mazumder and Miller (2016)
found that personal bankruptcies declined and credit scores improved when
Massachusetts expanded health coverage in 2006. Other research indicates
that fewer medical bills went into collections and fewer people went into medical debt after expansions of Medicaid coverage (Baicker et al. 2013; Hu et al.
2016). Insurance coverage also provides covered people with the peace of mind
that comes from alleviating the fear of possible financial distress.
Although increasing financial security is important, most initiatives to
expand health insurance coverage and access to healthcare are undertaken
with the goal of improving the health of insured people. Though the evidence is
clear that gaining health insurance increases healthcare utilization and spending, it is less clear that government coverage expansions improve people’s
health.
One of the first attempts to study the effect of health insurance on
health was the RAND Health Insurance Experiment. In the 1970s, participants
were randomly assigned to multiple levels of coinsurance, ranging from free
(no coinsurance) up to 95 percent coinsurance. The study found that as the
amount of coinsurance decreased, utilization of medical care rose. However,
with the exception of improved hypertension control, dental care, and vision
care for the poorest patients assigned free care, there were no health improvements for the average person receiving more generous insurance (Newhouse
1993; Brook et al. 2006). Hanson (2005) observes that because the researchers
conducted 80 tests by health indicators, 4 positive health results could appear
by chance alone, given a 5 percent significance level. Unfortunately, the RAND
study did not compare insurance with not having any insurance.
A careful examination of the literature by Levy and Meltzer (2008) found
that although many studies purport to find that insured people have better
health outcomes than uninsured people, most of these studies did not establish a causal relationship between health insurance and health. Observational
studies, which make up the vast majority of studies, did not adequately
address the problem of the endogeneity of health insurance—that observed
differences in health outcomes might be driven by unobserved differences
between the insured and uninsured. Results from quasi-experimental studies,
where endogeneity is less of a problem, were inconclusive. Levy and Meltzer
concluded that increases in health insurance increased the consumption of
medical care and might modestly improve self-reported health. Although
insurance can improve some health measures for some population subgroups,
especially vulnerable ones like children, there is little evidence that insurance
significantly affects the health of most people.

Innovative Policies to Improve All Americans’ Health | 283

Perhaps the best recent evidence of the effect of insurance on health
comes from the Oregon Medicaid expansion experiment (Baicker et al. 2013).
People selected at random for Medicaid coverage from a waiting list of uninsured people were compared with a control group of those who were not
selected. Both groups were followed for two years. The covered group gained
improved financial security, which was reflected in less medical debt as well as
less borrowed money to pay bills or skipped payments. They increased their
use of medical care—ambulatory care, emergency department visits, preventive visits and services, prescription drugs, and hospitalizations all increased.
Medicaid enrollees also reported an improved sense of physical and mental
health. Yet, other than improved depression outcomes, the group gaining
coverage did not show improvement in health outcomes. There was no significant improvement in blood pressure, cholesterol level, diabetes control, or
mortality.
In a recent review of the effect of insurance coverage, Sommers, Gawande,
and Baicker (2017) asserted that insurance improves health. They cited a single
quasi-experimental study showing that insurance improves health outcomes.
That study (Cole et al. 2017) only found improvement in four of eight “quality
measures,” and three of the four were process measures, not outcome measures. Blood pressure control was the only outcome improvement, a finding
that is contrary to the finding of no improved blood pressure control in the
randomized Oregon experiment. Sommers, Gawande, and Baicker (2017) also
claimed that insurance lowers mortality. They cited two quasi-experimental
studies showing a 6 percent reduction in mortality over 5 years in three States
that expanded Medicaid in the early 2000s as compared with neighboring
States that did not expand Medicaid (Sommers, Baicker, and Epstein 2012),
and reductions in mortality in Massachusetts after its 2006 health reform
as compared with mortality in demographically similar counties nationally
(Sommers, Long, and Baicker 2014). These studies, as the authors acknowledged, were susceptible to unmeasured confounding. Finally, they asserted
that the positive effect on self-reported health seen in the Oregon study
predicts reduced mortality over a period of 5 to 10 years. They relied on two
earlier studies (Miilunpalo et al. 1997; DeSalvo et al. 2006) that reported a correlation between perceived poor health at a point in time and mortality. These
studies may be accurate, but neither one sought to answer how changes in
self-reported health brought on by gaining insurance affect eventual mortality.
Multiple studies cited by Sommers, Gawande, and Baicker (2017) had
short-term follow-up. In contrast, a recent 20-year observational study of the
near elderly (age 50–61 years), which took pains to counteract the deficiencies
of earlier observational studies by using a more complete set of covariates,
found that insured people use more healthcare services, but there was little or
no effect of insurance on health and mortality (Black et. al. 2017).

284 |

Chapter 6

A study by Richard Kronick (2009)—who worked on ACA implementation
as HHS Deputy Assistant Secretary for Health Policy (2010–13) and as Director
of the Agency for Healthcare Research & Quality (2013–16)—found that on
almost every characteristic measured, uninsured people have higher risk factors compared with privately insured people; this study had a follow-up period
of 16 years. When adjustment was made for high-risk characteristics, being
uninsured was not associated with an increased risk of mortality. Although
cautioning about the difficulties of inferring causality from an observational
analysis, Kronick concluded that “there would not be much change in the number of deaths in the United States as a result of universal coverage.”

Why the ACA Expansions May Have Limited Health Effects
There are at least four reasons why health insurance, particularly through
government coverage expansions, may have a smaller effect on health than
anticipated. The first three of these reasons—that the uninsured were often
able to obtain care before coverage, access problems for patients who gain
Medicaid coverage, and mandated insurance benefits that have a minimal
impact on health—are particularly salient when examining the results of the
ACA coverage expansion.
First, care has been available for many who have no insurance coverage.
The 1986 Emergency Medical Treatment and Labor Act requires anyone coming to a hospital emergency department to be stabilized and treated, regardless of their insurance status or ability to pay. Moreover, physicians in private
practice have historically been willing to care for some uninsured or poorly
insured patients, although recent increases in operating costs and declines
in insurance reimbursements have decreased willingness to provide charity
care (Zinberg 2011). Finkelstein and McKnight (2008) explained their finding
that in its first 10 years after passage in 1965, Medicare had no discernible
impact on elder mortality, in part because before Medicare, elderly individuals with life-threatening, treatable conditions sought care even if they lacked
insurance. Finkelstein, Hendren, and Luttmer (2015) found that in the Oregon
experiment, Medicaid enrollees only valued each additional $1 of government
Medicaid spending at $0.20 to $0.40. Most of the benefit went to doctors and
hospitals, who would have otherwise provided uncompensated care to these
enrollees. Similarly, a study of how many enrollees dropped out when charged
higher premiums for Medicaid-like coverage on Massachusetts’ low-income
health insurance exchange found that most enrollees valued coverage at less
than half the cost of coverage. The availability of uncompensated care for
low-income uninsured people explained the gap between enrollee value for
Medicaid below the program cost (Finkelstein, Mahoney, and Notowidigdo
2017). In fact, the ability of low-income, uninsured people to declare bankruptcy serves as an implicit form of high-deductible insurance (Mahoney 2015).

Innovative Policies to Improve All Americans’ Health | 285

Second, about three-quarters of the ACA’s net coverage increase came
from people who were newly covered by Medicaid (Uberoi, Finegold, and Gee
2016), a type of insurance that is associated with limited access to care. When
there are public coverage expansions, demand prices (copays and premiums)
fall for those who are subsidized, but supply prices set by the government
(reimbursements) are usually lower than commercial rates. This leads to
excess demand and a limited supply of willing providers to serve those covered
by the low reimbursement rates under the public expansion. The Kaiser Family
Foundation (2016) reported that State Medicaid programs pay physicians an
average of 72 percent of Medicare fees, and Medicare fees are often lower than
commercial rates. The reimbursements are even lower (66 percent of Medicare)
for primary care physicians who are regarded as the key points of access into
the healthcare system. Low Medicaid reimbursement rates mean relatively low
physician participation in the program (Decker 2013). A study of appointment
availability found that every $10 change up or down in Medicaid fees led to a
1.7 percent change in the same direction in the proportion of patients who
could secure an appointment with a new doctor (Candon et al. 2017).
Office-based physicians’ willingness to accept new patients varies,
depending on a potential patient’s type of insurance. Nearly 85 percent
of physicians will see a new patient if they have private insurance, and 84
percent will see a patient with Medicare; but just 69 percent will accept new
Medicaid patients (CDC 2015). Moreover, the percentage of physicians accepting Medicaid varies vastly among States. Three of the most populous States
in the Nation have significantly lower percentages of accepting physicians
than the 69 percent national average—California (54 percent), New York (57
percent), and Florida (56 percent) (CDC 2015). If Medicaid coverage did not
improve outcomes in the Oregon experiment, where the proportion of physicians accepting Medicaid (77 percent) was higher than the national average,
it suggests that expansion in these States would have only a marginal effect.
Not surprisingly, Miller and Wherry (2017) found no difference in the health
outcomes of individuals in States that expanded Medicaid under the ACA compared with nonexpansion States, no significant improvement in self-reported
health status, and an increased probability of delay in obtaining care because
no appointments were available or waiting times were too long for those in
expansion States.
The third reason public coverage expansions have a small effect on
health is that government rules, associated with public expansions, may produce inefficient healthcare utilization. The ACA, for example, mandated that
preventive care and annual office visits be covered at no cost to the patient,
even though the benefits of preventive care and screening are modest. A review
by the Stanford Prevention Research Center of randomized trials and metaanalyses of the efficacy of available screening tests for diseases where death
is a common outcome found that “reductions in disease-specific mortality are
286 | Chapter 6

uncommon and reductions in all-cause mortality are very rare or nonexistent”
with these tests (Saquib, Saquib, and Ioannides 2015). Cancer screening can
generate health benefits when used for appropriate populations, but it is performed far more commonly in this country than elsewhere and is often overly
intensive, of low value, and potentially harmful (Zinberg 2016). Even Dr. Ezekiel
Emanuel (2015), one of the ACA’s architects, acknowledged that routine annual
physicals—visits for general healthcare that are not prompted by any specific
complaint or problem—do not decrease mortality, waste resources, and may
lead to harmful additional testing and unnecessary treatments. By mandating
the provision of low-value medical care measures at little or no cost, the ACA
may inefficiently allocate scarce medical resources, because low-value measures could crowd out high-value ones.
Fourth and finally, public coverage may have limited or possibly negative effects on health because of its long-run impact on innovation. Many
governments, particularly in Europe, have paired large coverage expansions
with the imposition of price and spending controls. These centralized controls
may have an adverse impact on medical innovation and make healthcare less
effective and more costly to obtain in the future. Therefore, a complete assessment of the impact of government insurance expansions on health requires a
long-run perspective. In sum, though health insurance is undoubtedly a major
positive factor in the lives of many Americans, health policy focused predominantly on expanding insurance runs the risk of crowding out other policies that
could better contribute to improved health outcomes for our citizens. Box 6-1
describes the Administration’s initiatives to restore choice and competition to
healthcare in America.

The ACA’s Impact on Spending
The share of GDP attributable to healthcare expenditures has increased sharply
over time, rising from 5 percent of GDP in 1960 to 17.4 percent in 2010, when
the ACA was enacted (CMS 2017a). This spending growth represents increased
quantities of services, increased prices for these services, or a combination of
both. Economic research finds that factors contributing to spending growth
include increasing income, the aging of the population, and the increased
extent of insurance coverage. However, new technologies have perhaps been
the central driver, contributing between 27 and 48 percent of spending growth
since 1960 (Smith, Newhouse, and Freeland 2009). The growth in real per capita
health expenditures slowed worldwide after 2002, well before the enactment
of the ACA in 2010, largely as a result of the two recessions in the last decade
(Sheiner 2014), but has recently turned upward.
Despite promises to “bend the cost curve” and purported cost control
provisions in the ACA, real national health spending per capita rose 3.0 percent
a year, on average, from 2013 to 2016, compared with 1.5 percent from 2003
to 2013, according to the Centers for Medicare and Medicaid Services (CMS

Innovative Policies to Improve All Americans’ Health | 287

Box 6-1. The Trump Administration’s Actions to
Restore Choice and Competition to Healthcare
The ACA imposed costly benefit requirements and regulations on the insurance market that limited people’s options and raised premiums. The individual mandate required people to buy insurance whether they wanted it or
not, and the ACA’s minimum essential benefit requirement mandated a more
costly and comprehensive package of benefits than what many people preferred. This requirement, combined with the ACA’s restrictions on premium
variation, has driven premiums up in the insurance market for individuals
and small groups. Far fewer people signed up on the ACA exchanges than
expected (CBO 2017). Young, healthy patients have largely shunned the
overpriced insurance coverage—in effect, refusing to subsidize lower premiums for elderly, sick patients (Antos and Capretta 2016). As a result, many
insurers have incurred substantial losses and have fled the exchanges (Cox et
al. 2016). In 2018, 51 percent of counties had only one carrier participating in
their healthcare exchange (CMS 2017d). The average number of insurers participating in each state’s ACA marketplace declined from 5.0 in 2014 to 4.3 in
2017 and to 3.5 in 2018. In 2018, eight states have only a single ACA insurance
provider (Semanskee et al. 2017). Insurers that have remained in the market
increased premiums by 25 percent, on average, for benchmark silver plans for
plan year 2017 (rates were determined during the fall of 2016), and by even
more for 2018 (figure 6-i).

$"0- 
хҊ$ѵ
  -' -& /+' 
- ($0(.
)
 ..0 -.Ѷ
')
 -.

җ.Ҙ
спру–рч
Annual percent change in premiums
уф

Benchmark silver plan (left axis)
Lowest-cost plan (left axis)
Employer-sponsored off exchange (left axis)
Total number of state issuers (right axis)

Number of issuers
300

200

тп

100

спру

спрс

сппч

рф

спрп

рпп
чп
$"0- 
хҊчѵ
…*.+$/'.
#/
Percent
хп
с
уп
…
спп
*0- ѷ ( -$)
 *.+$/'
..*$/$*)ѵ

0

п
спру

спрф

спрх

спрц

спрч

*0- .ѷ
 җспрцҘ
 '/#
')
)
#*$ 
- ($0(.
 . -#
-$ !Ѹ
 $. -
($'4

*0)/$*)
)
 '/#
 . -#
)
0/$*)'
-0./
(+'*4 -
 '/#
 ) !$/.
спрц

))0'
0-1 4ѵ
*/ ѷ
// .
0.$)"
 '/#- ѵ"*1 $)
''
!$1 
.ѵ
- ($0(. . /
$)
 +/ ( -
*!
+-$*-
4 -ѵ

288 | Chapter 6

The President has addressed the ACA’s key problems from the beginning of his tenure. Within hours of being sworn in on January 20, 2017, he
signed Executive Order 13765, directing the Secretary of Health and Human
Services and the heads of other agencies to take all actions consistent with
the law to minimize the ACA’s economic and regulatory burdens, to provide
greater flexibility to the States, and to promote the development of a free and
open interstate health insurance market.
President Trump also signed Executive Order 13813 to promote healthcare choice and competition by expanding affordable coverage options.
The order directs the Department of Labor to lower the barriers preventing
small businesses from forming Association Health Plans (AHPs), so these
firms can gain the regulatory benefits that large employers now receive. For
example, as AHP members, small businesses would gain bargaining power
to negotiate more affordable insurance and avoid some of the ACA’s costly
requirements, as large companies already do for their employees. A total of
11 million Americans lack employer-provided insurance because they or a
loved one work for a small business or sole proprietorship that does not offer
insurance. On January 4, 2018, the Department of Labor proposed a rule on
AHPs that would give employers greater ability to form AHPs and to allow sole
proprietors to join these plans.
Executive Order 13813 also calls for the consideration of new regulations to expand the availability, duration, and renewability of Short-Term
Limited Duration Insurance. These plans are exempt from the ACA’s rules,
offer greater choices of coverage, and are significantly cheaper than ACA
exchange plans. They were previously available for terms of up to one year,
but were limited to 90 days without renewability by a late 2016 regulation. The
Executive Order also seeks to expand the availability and the use of Health
Reimbursement Arrangements—employer-funded, tax-advantaged accounts
that reimburse employees for deductibles, copayments, premiums, and qualified medical expenses in plans or other arrangements that best suit them.
Finally, Executive Order 13813 directs the Secretary of Health and
Human Services to provide a report detailing how existing State and Federal
statutes and regulations limit Americans’ healthcare options, decrease competition, and raise costs. State rules and regulations—such as certificate of
need laws, narrow scope of practice rules, and restrictions on telemedicine—
are government-erected barriers to entry that benefit established providers,
allowing them to charge higher prices and reducing incentives for them to
produce higher-quality, lower-cost goods and services. This spring, HHS will
release a report detailing recommendations to reduce government-erected
barriers to entry, thereby expanding choice and competition in healthcare
markets.
On the legislative front, the Administration’s tax reform will eliminate
the ACA’s individual mandate.

Innovative Policies to Improve All Americans’ Health | 289

2017a). In part, this was due to a rapid rise in Medicaid spending, as enrollment
rose from 54 million in 2010 to 71 million in 2016 (CMS 2017a). In addition,
spending per expansion enrollee in Medicaid expansion States was far higher
than projected (CBO 2017). By 2016, health expenditures accounted for 17.9
percent of GDP (CMS 2017a).
In addition, the ACA’s signature cost control provisions were ineffective
and had unintended consequences. Accountable Care Organizations (ACOs)
were supposed to give providers incentives to become more efficient, but they
have not produced overall savings as implemented. In fact, after accounting
for bonus payments to ACOs that were awarded for keeping costs down—and
for the fact that most ACOs were in one-sided risk arrangements, whereby they
shared savings but were not liable to the government for losses—the Medicare
Shared Savings Program ACOs actually increased Medicare spending by $216
million in 2015 and by $39 million in 2016 (Capretta 2017). The initial belief that
ACOs would curb spending growth, and their subsequent failure to do so, follows a long history of such payment reforms not altering spending growth. The
market essentially was taken over by health maintenance organizations and
their more generous cousins, preferred provider organizations, while spending
kept growing at the same rate. The same was true for capitation payments and
disease management programs.
The ACA also imposed a penalty on hospitals that have high rates of readmissions within 30 days of discharge, in an effort to cut costs and improve quality. Although hospitals cut readmissions, part of the effect was due to hospitals’
decreasing admission rates for returning patients, whom they would normally
have admitted, in order to avoid penalties (Gupta 2017). A study of Medicare
patients hospitalized for heart failure found that implementing the hospital
readmission reduction program was associated with a subsequent increase in
30-day and 1-year risk-adjusted mortality (Gupta et al. 2018).
Instead of relying on consumer choice and competition to control costs,
the ACA encouraged healthcare providers to combine into larger health systems and to take on financial risk, based on the unproven assumption that
this would incentivize providers to decrease unnecessary services, cut costs,
and improve outcomes. However, excessive consolidation in the market may
enable producers to use their market power to raise prices, lower quality and
innovate less than they would in a competitive market.
ACA-mandated cuts in hospital payments and new regulatory burdens
made it difficult for smaller institutions to go it alone. Small physicians’ groups
and solo providers could not afford to purchase and maintain electronic medical records and comply with government reporting requirements. As a result,
hospital mergers are booming, leading to horizontal integration, and large
hospitals are buying up physicians’ practices and outpatient service providers to form large, vertically integrated healthcare networks. Hospital mergers
and acquisitions averaged 97.8 per year (ranging from 88 to 102 a year) in the
290 |

Chapter 6

five years after the ACA was enacted (2011–15), compared with 58.8 per year
(ranging from 38 to 83 a year) during the 10 years preceding the ACA (2001–10)
(AHA 2016). After significant consolidation, Cutler and Morton (2013) found that
almost half of hospital markets are highly concentrated, with one or two large
hospital systems dominating many regions across the country. Cooper and
others (2015) found that hospital prices in monopoly markets are 15 percent
higher than those in markets with four or more hospitals, after controlling for
several demand and cost factors. Baker, Bundorf, and Kessler (2014) found that
vertical integration led to an increase in market share, which was associated
with higher prices and increased spending. These cost increases due to consolidation are exacerbated by many State regulations, such as certificate-of-need
laws and rules about narrow scopes of practice, that serve as barriers to entry,
particularly for lower-cost alternatives.
To free the market from these mandates and constraints on competition—pursuant to the President’s October 12, 2017, Executive Order—the
Administration will release a report in the spring of 2018. It will identify Federal
and State government policies that reduce competition and increase consolidation and provide recommendations to mitigate these policies. The literature
is clear that hospital competition leads to lower prices and higher quality
(Gaynor and Town 2012). This is consistent with the Administration’s deregulatory agenda, which has already withdrawn, made inactive, and delayed hundreds of economically destructive regulations.

Improving People’s Health by Limiting
the Effects of Unhealthy Behavior
In industrialized countries, health behaviors—actions and inactions by individuals that affect their own health or the health of others—are more important determinants of health than insurance coverage and the medical care it
finances. A review of the literature has identified five key determinants of health
in industrialized countries: health behaviors, genetics, social circumstances,
environmental and physical influences, and medical care. Health behaviors
appear to be the most important, making a relative contribution of 30 to 50
percent to health, according to various studies, while medical care accounts
for only 10 to 20 percent (Gnadinger 2014). The second most important relative
contribution to health, at 20 to 30 percent, is made by genetics. The government cannot and should not directly affect people’s genomes, but here it does
have an important function: to set ethical and regulatory guardrails for the
development and use of genetic testing and therapies.
Using mortality as an indicator of health, Schroeder (2007) finds that up
to 40 percent of premature deaths in the United States are due to unhealthful
behaviors like smoking, poor dietary habits, and sedentary lifestyles. Now,
deaths resulting from the escalating opioid abuse crisis are adding to this

Innovative Policies to Improve All Americans’ Health | 291

self-inflicted toll. Although life expectancies in different geographic areas have
a negative correlation with poor health behaviors like smoking, they are not
correlated with access to healthcare (Chetty et al. 2016).
Poor health imposes economic costs in three ways: direct health care
spending; the costs of premature deaths resulting from poor health; and productivity losses from illnesses that keep people out of the labor force or cause
absenteeism and “presenteeism”—that is, decreased worker productivity
while at work—for those who are in the labor force. The Commonwealth Fund
estimated that in 2003, 18 million people between the age of 19 and 64 were out
of the labor force because of illness, and that if their lost work time was valued
at the minimum wage, the nation lost $185 billion in economic output (1.6 percent of GDP) (Davis et al. 2005). Another 69 million workers lost 407 million sick
days, which cost $48 billion, if valued at actual wages. Finally, they estimated
there were 478 million days when illness reduced workers’ productivity, resulting in a loss of $27 billion if they were working at “half capacity.” The share of
prime-age employees citing poor health as the main reason for staying out of
the labor force has increased significantly during the past two decades, and it is
higher among those with less education. During the second quarter of 2017, 5.4
percent of prime-age individuals (those age 25 to 54) reported being too sick
or disabled to work in the labor force, 1.6 percent more than two decades ago.
If this trend were reversed, it could increase the workforce by up to 4 million
people and add about 2.6 percent to GDP (Terry 2017).
In this section, we focus on three behaviors—opioid abuse, poor diets
and sedentary lifestyles that lead to obesity, and smoking—that severely exacerbate our most costly illnesses and impose enormous related economic and
social costs.

Improving Health by Combating the Opioid Epidemic
As debates focused on expanding health insurance coverage during the past
eight years, an opioid epidemic was ravaging the country, devastating the lives
of those struggling with addiction, and the lives of their loved ones. The consequences for the health of Americans—most important, a skyrocketing death
toll—have been enormous (see figure 6-1). In 2016, almost as many people died
of an opioid-involved drug overdose (42,249) as died of HIV (43,115) at its peak
in 1995 (Mendell, Cornblath, and Kissel 2001). And since 1999, over 350,000
people have died of opioid-involved drug overdoses, which is 87 percent of
the 405,399 Americans killed in World War II (DeBruyne 2017). The staggering
opioid death toll has pushed drug overdoses to the top of the list of leading
causes of death for Americans under the age of 50 and has cut 2.5 months from
the average American’s life expectancy (Dowell et al. 2017). This subsection
documents the immense economic toll the opioid epidemic has taken on the
United States, and thus, the importance of Administration actions that have
been undertaken to reduce these costs and save lives.

292 |

Chapter 6

$"0- 
хҊрѵ
+$*$Ҋ )1*'1  1 -*. 
 /#.Ѷ
ршшш–спрх
Thousands
уф

спрх

уп

тф
тп
сф
сп
рф
рп
ф

п
ршшш

сппу

сппш

*0- ѷ
 )/ -.
!*-
$. . 
*)/-*' )
- 1 )/$*)Ѷ


/. ѵ

спру

The opioid epidemic evolved with three successive waves of rising deaths
due to different types of opioids, with each wave building on the earlier one
(Ciccarone 2017). In the late 1990s, in response to claims that pain was undertreated and assurances from manufacturers that new opioid formulations
were safe, the number of opioid prescriptions skyrocketed (CDC 2017b). What
followed was an increase in the misuse of and deaths related to these prescriptions (figure 6-2). As providers became aware of the abuse potential and addictive nature of these drugs, prescription rates fell, after peaking in 2011. Deaths
involving prescription opioids leveled off, but were followed by a rise in deaths
from illicit opioids: heroin and fentanyl. Heroin deaths rose first, followed by a
rise in deaths involving fentanyl—a synthetic opioid that is 30 to 50 times more
potent than heroin and has legitimate medical uses but is increasingly being
illicitly produced abroad (primarily in Mexico and China) and distributed in the
U.S., alone or mixed with heroin. In 2015, males age 25 to 44 (a core group of
the prime-age workers whose ages range from 25 to 54) had the highest heroin
death rate, 13 per 100,000. Fentanyl-related deaths surpassed other opioidrelated deaths in 2016.
The CEA estimates that the opioid epidemic’s economic cost was $504
billion in 2015, or 2.8 percent of that year’s GDP (CEA 2017). This estimate
dwarfs estimates from previous studies for several reasons—most important,
previous studies undervalued the cost of the lives lost to drug overdoses.
For example, some studies focus mainly on healthcare costs and find that
prescription opioid abusers utilize significantly more healthcare resources

Innovative Policies to Improve All Americans’ Health | 293

$"0- 
хҊсѵ
1 -*.  /#.
4
4+ 
*!
+$*$
 )1*'1  Ѷ ршшш–спрх
Thousands
сф

спрх
сп

- .-$+/$*)

рф

-*$)

рп

/# -

.4)/# /$

*+$*$.
җ ѵ"ѵѶ

! )/)4'Ҙ

ф

п
ршшш

сппр

сппт

сппф

сппц

сппш

спрр

спрт

спрф

*0- ѷ
 )/ -. !*-
$. . 
*)/-*'
)
- 1 )/$*)Ѷ

/. ѵ
Note: “Prescription”  )*/ .
/# 
*($) 
/
!*-
( /#*) 
)
)/0-'ҝ. ($Ҋ.4)/# /$

*+$*$.ѵ

than nonaddicted peers (e.g., White et al. 2005, 2009; McAdam-Marx et al.
2010; McCarty et al. 2010; Leider at al. 2011; Johnston et al. 2016; Kirson et al.
2017). Others account for additional costs, including forgone earnings from
employment and higher costs for the criminal justice system (e.g., Birnbaum
et al. 2006, 2011; Hansen et al. 2011; Florence et al. 2016). A recent estimate by
Florence and others (2016) found that prescription opioid overdoses, abuse,
and dependence in the United States cost $78.5 billion in 2013; but they did not
account for the costs of illicit opioids.
Although previous estimates are informative about certain types of costs,
they only partially account for the damage caused by the opioid epidemic. They
do not account for the costs associated with the escalating abuse in recent
years of illicit opioids, such as heroin and fentanyl, and the resulting increase
in deaths. Evidence also suggests that fatality statistics understate the number
of opioid-related deaths (Ruhm 2017). But most important, previous studies
fail to fully account for the value of the lives lost to overdoses. Studies that
only include healthcare expenditures typically capture none of the value of the
lives lost, and studies that account for earnings losses among those who die
account for only a fraction of the loss from such mortality. Extensive research
indicates that people value fatality risk reduction far beyond the value of lost
earnings due to premature death, because earnings do not take into account
other valuable activities in life besides work. Using conventional estimates of
the losses induced by fatality routinely used by Federal agencies—in addition
294 |

Chapter 6

Table 6-1. Comparison of Estimated Costs of Opioid-involved Overdose
Deaths, CEA and Other Studies
Ratio of
CEA
Adjustment
Cost
for under- (billions, estimate to
study
counting
2015$)
estimate

Study

Study
year

Opioids
included

Fatal
costs

Birnbaum et al.
(2006)

2001

Prescription

Earnings

No

11.5

43.8

Birnbaum et al.
(2011)

2007

Prescription

Earnings

No

61.5

8.2

Florence et al.
(2016)

2013

Prescription

Earnings

No

79.9

6.3

CEA (2017)

2015

Prescription
and illicit

VSL

Yes

504.0

1.0

Sources: Birnbaum et al. (2006); Birnbaum et al. (2011); Florence et al. (2016); CEA (2017).
Note: Each of the studies listed includes healthcare, criminal justice, and employment costs in
nonfatal costs. CEA nonfatal costs are calculated by applying Florence et al. (2016) estimates of the perperson average nonfatal costs of prescription opioid disorders to individuals with prescription opioid
and heroin disorders in 2015. CEA fatal costs are calculated by applying the age-dependent value of
statistical life to drug overdose deaths involving any opioid in 2015.

to making other adjustments related to illicit opioids, more recent data, and
the underreporting of opioids on drug overdose death certificates—the CEA
study found that the overall loss imposed by the crisis is several times larger
than previous estimates (table 6-1). (The CEA uses an age-adjusted value of
statistical life measure to estimate the cost of lives lost to opioid-involved
overdoses; see CEA 2017.)
It is important to note that though the fatal costs of the opioid epidemic
($432 billion) are the major component of its total costs, its nonfatal costs ($72
billion) are also important. Florence and others (2016) estimate that in 2013,
prescription opioid misuse increased healthcare and substance abuse treatment costs by $29.4 billion, increased criminal justice costs by $7.8 billion, and
reduced productivity among those who do not die of an overdose by $20.8 billion (in 2015 dollars). To arrive at our nonfatal cost estimate of $72 billion, the
CEA adjusts these costs upward, adding in illicit opioid use and also the greater
number of opioid abusers in 2015. Other research has shown that opioid abusers miss twice as many days of work compared with other employees (Benham,
Goplerud, and Hodge 2017; Ruetsch 2010). They are also significantly less productive while at work because the drugs can induce drowsiness, cause mental
confusion, impair attention and focus, and reduce creativity or reliability.
By any measure, the opioid epidemic is exacting a massive and growing
toll on the United States. Death rates continue to skyrocket, while nonfatal

Innovative Policies to Improve All Americans’ Health | 295

Box 6-2. The Trump Administration’s Actions on Opioids
The Administration has taken a series of actions, including creating the
President’s Commission on Combating Drug Addiction and the Opioid Crisis
and declaring a public health emergency under the Public Health Service Act.
These additional actions have been taken:

To prevent prescription opioid abuse:
• The Centers for Disease Control and Prevention (CDC) launched the Rx
Awareness Campaign to increase awareness of the risks of prescription opioid
use by telling the stories of people recovering from addiction.
• The CDC awarded $28 million in new funding for prescription drug
monitoring programs, so prescribers and pharmacists can monitor how many
opioid prescriptions patients have received and prevent duplicate prescriptions, diversion, and abuse.
• The Food and Drug Administration (FDA) has worked to educate prescribers about safer pain management to reduce unnecessary prescribing.
• The Centers for Medicare and Medicaid Services (CMS) delinked pain
management scores from provider evaluations to decrease pressure to
prescribe opioids.
• The Department of Justice (DOJ) established an Opioid Fraud and Abuse
Detection Unit to crack down on prescription opioid use for nonmedical
reasons.
• The Department of Health and Human Services (HHS) and DOJ conducted the largest ever healthcare fraud enforcement action, by the Medicare
Fraud Strike Force, in 2017. More than 120 people were charged with
fraudulently billing public and private insurance programs for prescribing and
distributing opioids.

To improve access to and quality of treatment for those already
addicted:
• HHS awarded $485 million to States for prevention, treatment, and
recovery services under the newly created State Targeted Response to the
Opioid Crisis grant program.
• Another $144 million was awarded for treatment and other opioid-related
costs by the Substance Abuse and Mental Health Services Administration.
• Significant funding has also been directed to mental health and substance abuse service centers, rural health organizations, physician training
programs, and other entities that will support treatment and recovery using
evidence-based practices.
• The Administration is cutting the red tape that hinders States’ ability
to use Federal funding as effectively as possible. The CMS announced in
November 2017 that Medicaid would grant waivers from a decades-old
statute so funds can be used to pay for treatment in facilities with more than
16 beds.

296 |

Chapter 6

To encourage innovation in addressing the opioid epidemic:
• The National Institutes of Health (NIH) is partnering with innovative
companies to develop nonaddictive pain therapies, new addiction treatment
regimens, and overdose-reversal drugs.
• The Advancing Clinical Trials in Neonatal Opioid Withdrawal Syndrome
study will evaluate neonatal abstinence syndrome treatments for opioiddependent newborns.
• An $81 million research partnership between HHS, the Department of
Defense, and the Department of Veterans Affairs was announced to support
pain management research for the military and veterans.

To disrupt the supply of illicit opioids:
• President Trump signed the Interdict Act in January 2018 to equip
Border Control Agents with better technology to intercept illicit, synthetic
opioids at the border.
• DOJ shut down AlphaBay, the largest online criminal marketplace and
major source of illicit drugs and indicted two Chinese manufacturers of illicit
fentanyl.
• The U.S. Postal Inspection Service seized increased amounts of fentanyl
shipped through the mail, and the Department of State is working with
international partners in reducing the production and shipment of fentanyl
from abroad.

costs to productivity and the healthcare and criminal justice systems increasingly hurt the economy. This does not mean that opioids themselves have no
beneficial effects—they are largely effective for their main prescribed uses of
reducing acute pain and as anesthesia during surgery. But the epidemic of
misuse and abuse—along with their often deadly consequences—is a health
crisis many years in the making that requires urgent attention. Fortunately,
the Administration has taken concrete steps to begin to stem the costs of this
epidemic, and with sustained action, can make continued progress in addressing it (see box 6-2).

Obesity
Obesity has become a major health problem, leading to large direct medical
expenditures, significant premature mortality, and large productivity losses.1
It raises the risk of all-cause and cardiovascular mortality (the leading cause of
death in the U.S.) and the risk of morbidity from hypertension, dyslipidemia,
type 2 diabetes, coronary heart disease, stroke, gallbladder disease, osteoarthritis, sleep apnea and other respiratory problems, and some cancers (Jensen
1 Obesity is medically defined as the height-adjusted weight measure—Body Mass Index (BMI)—
greater than 30. BMI is defined as a person’s weight in kilograms divided by the square of his or
her height in meters.

Innovative Policies to Improve All Americans’ Health | 297

et al. 2014). Stewart and others (2009) estimated that an 18-year-old with a BMI
increasing by the historical average, 0.5 percent a year, would lose 1.02 years in
life expectancy due to obesity alone.
The Centers for Disease Control and Prevention’s National Center for
Health Statistics reports that obesity rates among U.S. adults 20 and older rose
from 22.9 percent in the years 1988–94 to 38 percent in 2013–14 (CDC 2016b).
Another 33 percent of U.S. adults were overweight (i.e., have a BMI between
25 and 30). Obese patients incur 46 percent higher inpatient costs, 27 percent
more physician visits and outpatient costs, and 80 percent higher prescription
drug spending than normal-weight patients (Finkelstein et al. 2009). In 2006,
they spent an average of 42 percent more (an average of $1,429 a year) than
normal-weight patients, resulting in a total cost of medical care associated
with obesity in the United States of $147 billion in 2008 dollars (Finkelstein et
al. 2009).
Obesity also decreases the productivity of those still in the workforce
through absenteeism and presenteeism (Goettler, Grosse, and Sonntag 2017).
An estimate of annual obesity-related absenteeism and presenteeism costs
among full-time U.S. employees in 2008 was $59 billion, in 2015 dollars
(Heidenreich et al. 2011).
Rising obesity is an unintended consequence of technological progress
(Philipson and Posner 1999). Welfare-improving technological change has low­
ered the cost of consuming calories through improved agricultural production,
while raising the cost of expending calories by making work, both on the job
and at home, more sedentary. People now need to pay for gym memberships
to exercise and forgo leisure to replace decreased physical activity at work
and in the home. Unfortunately, people do not exercise enough during their
leisure time to make up for the exercise they formerly obtained at work and
home—which, when combined with increased calorie intakes, has resulted in
rising obesity (Lakdawalla, Philipson, and Bhattacharya 2005).
The U.S. government inadvertently contributed to this problem beginning in the 1970s when it, along with major professional nutrition organizations, recommended that Americans eat a low-fat/high-carbohydrate diet.
During the succeeding decades, Americans, adhering to these recommendations, replaced fat calories with even more carbohydrate calories. Total calorie
intake increased substantially, and the prevalence of obesity rose, in part, as
a consequence. Researchers eventually recognized that fat was less of a problem, and by 2015, the U.S. Department of Agriculture’s Dietary Guidelines for
Americans essentially removed the upper limit on the recommended fat intake
(Ludwig 2016).

Smoking
Tobacco use is the leading cause of behaviorally induced disease and death in
the United States, even after recent declines in tobacco use during the last few
298 |

Chapter 6

decades. Cigarettes are the most commonly used tobacco product among U.S.
adults, partly causing more than 480,320 deaths per year in the United States,
including more than 41,000 deaths resulting from secondhand smoke exposure
(CDC 2017c). They account for about 30 percent of all cancer deaths in the
United States, including about 80 percent of all lung cancer deaths—the leading cause of cancer death for both men and women (American Cancer Society
2015). Smoking is also a risk factor for cancers of the mouth, larynx, pharynx,
esophagus, kidney, liver, bladder, and stomach. It is also strongly associated
with many significant diseases other than cancer, including cardiovascular
and respiratory diseases. The largest numbers of smoking-attributable deaths
were from lung cancer (124,800), coronary artery disease (82,000), and chronic
obstructive pulmonary disease (64,700) (Kiszko et al. 2014).
Goodchild, Nargis, and d’Espaignet (2017) estimate that the total economic cost attributable to smoking is between $418 and $514 billion. Only 40
percent of the cost is due to direct spending on healthcare, with the remaining
amount due to indirect costs of the economic loss of morbidity and mortality
due to diseases attributable to smoking. They estimate indirect economic
losses using the human capital method, which calculates the present value
of labor productivity lost due to morbidity and mortality. This estimate, like
others, suggests that health spending attributable to smoking amounts to
between 5 and 10 percent of national health expenditures. This is consistent
with earlier estimates that smoking accounted for about 7 percent of total
annual healthcare spending for noninstitutionalized U.S. adults from 2000 to
2008 (CBO 2012) and 8.7 percent of annual U.S. healthcare spending, nearly
$170 billion, in 2010 (Xu et al. 2015).
A 2014 report by the U.S. Surgeon General found that, for the years
2005–9, the value of lost productivity attributable to premature death from
smoking, based on the 19 diseases associated with smoking, was $107.6 billion annually—with cancers accounting for $44.5 billion, cardiovascular and
metabolic diseases accounting for $44.7 billion, and pulmonary diseases
accounting for $18.4 billion. Using all-cause mortality, the value would be
$150.7 billion—$105.6 billion for men and $45.1 billion for women (HHS 2014).
Additionally, the value of lost productivity due to premature deaths caused
by exposure to secondhand smoke was estimated to be $5.7 billion. Because
these figures account only for lost productivity due to premature mortality
and not for lost productivity due to morbidity that living smokers and former
smokers experience, they significantly underestimate the full value of lost
productivity from smoking.
Bunn and others (2006) found that current smokers and former smokers both had higher losses from absenteeism and presenteeism than people
who had never smoked, suggesting that former smokers have lingering health
problems. Using an average hourly rate of $34.25, the annual amount of healthrelated absenteeism was estimated to be $1,206 for nonsmokers, $1,343 for

Innovative Policies to Improve All Americans’ Health | 299

Box 6-3. The Administration’s Actions to Combat Smoking
Despite significant reductions in the number of smokers, smoking remains
the Nation’s largest behavioral source of premature death. The Trump
Administration, through the FDA, has taken the following measures:
• The FDA is exploring initiatives to lower nicotine levels in cigarettes to
nonaddictive levels (FDA 2017).
• The FDA continues to investigate the safety and health effects of
e-cigarettes and vaping (NASEM 2018).
• On December 11, 2017, the FDA announced “Every Try Counts,” a twoyear education and advertising campaign to encourage adult smokers to quit,
even if they have been unsuccessful in the past.

former smokers, and $1,836 for current smokers. Health-related presenteeism was estimated to be $1,466 for nonsmokers, $1,918 for former smokers,
and $2,620 for current smokers. In total, absenteeism and presenteeism cost
employers $2,672 for nonsmokers, $3,261 for former smokers, and $4,456 for
current smokers (see box 6-3).

Policy to Address Health Behaviors
Public health measures and higher tobacco taxes have cut the number of
smokers and decreased tobacco’s toll on health (Mader et al. 2016). Excise
taxes (both State and Federal) raised the monetary cost of smoking, and time
and place restrictions raised the nonmonetized cost by making smoking more
inconvenient. Taxes effectively raise tobacco’s price, although smuggling can
avoid some of the higher State taxes. Because many smokers were used to
smoking throughout the day, time and place restrictions could effectively limit
their consumption (Fichtenberg and Glantz 2002) and nonsmokers’ exposure
to secondhand smoke (Hartmann-Boyce et al. 2016). The health and economic
gains from reducing smoking were enormous. It is estimated that tobacco control efforts led to 8 million fewer tobacco-related premature deaths than were
expected if smoking had continued unabated. Life expectancy in the 50 years
since the Surgeon General’s 1964 report on smoking and health increased 7.8
years for men and 5.4 years for women, of which tobacco control is associated
with 2.3 years of the increase for men and 1.6 years for women (Holford et al.
2014). These gains dwarf any health improvements resulting from increased
insurance coverage. Nevertheless, although the smoking rate has plummeted
(figure 6-3), smoking remains a leading cause of death and morbidity.
It is less clear that the behaviors that are factors in causing obesity can
be addressed through taxes and public health measures. Studies of the economic and social approaches to preventing obesity that have been employed
to date—such as taxes on sugary foods, wellness programs, menu labeling,
and financial rewards for weight loss—are inconsistent and show only modest,

300 |

Chapter 6

$"0- 
хҊтѵ
(*&$)"
/ .
)
/# 
-$ 
*!
$"- // .Ѷ
ршшф–спрф
Real dollars ($2009)
х

Percentage of U.S. adults (18+) who smoke
тп

(*&$)"
-/
(*&$)"
-/
. .
җ-$"#/
3$.Ҙ
җ-$"#/
3$.Ҙ

ф
у

р

сп

 '
+-$ 
*!

+&
*!


$"- // .
җ' !/
3$.Ҙ

т

с

сф

спппѷ ѵѵ

0-" *)  ) -'

+0'$.# .

ѿ 0$)"

**
. ѿ

 +*-/

п
ршшф

сппп

рф

сппшѷ   -'
/3

-/ 
*)

+&
*!

$"- // .
-$. 

!-*(
‫ڦ‬пѵтш
/*

‫ڦ‬рѵпр
сппф

спрп

*0- .ѷ
  -/$*) *!
3
($)$./-/*-.Ѹ
!!$ 
*! $. . 
- 1 )/$*)
)

-*(*/$*)Ѹ

'0'/$*).ѵ

'/#


рп
ф
п
спрф

long-term effects on weight (Cawley 2015). A recent review of the literature concluded that “there is an abundance of evidence that suggests calorie labeling,
as it is currently being implemented, has no impact on overall food purchases
or consumption for the population as a whole” (Kiszko et al. 2014). This is not
that surprising when we consider that eating is an ingrained behavior, food is
readily available for most people, and it is hard to pass up pleasurable foods
or engage in difficult physical activity now for the promise of lower weight and
better health sometime in the future.
Wellness programs have the potential to reduce healthcare costs
and productivity losses from absenteeism (Baicker, Cutler, and Song 2010).
Unfortunately, not every employee is willing to participate, and current incentives may be inadequate. The ACA encouraged employers to offer workplace
wellness programs but limited the incentives that could be offered. There
has been rapid growth in these programs, which now cover about 50 million
people. However, a new, large randomized study found that the workers
who chose to participate in the workplace wellness programs that sprung up
after the ACA tend to be self-selected—they had lower medical expenditures
and healthier behaviors before joining the program than nonparticipants.
Moreover, the study did not find significant causal effects of participation
on total medical expenditures, health behaviors, employee productivity, or
self-reported health in the first year (Jones, Molitor, and Reif 2018). Relaxing
the limitations on incentives that can be offered may attract a wider range

Innovative Policies to Improve All Americans’ Health | 301

of participants, including those who could benefit the most, and make these
programs more effective.
It may be difficult for taxes to significantly affect opioid abuse, particularly because the recent spike in deaths is being fueled by illicit opioids (heroin
and fentanyl) that defy conventional approaches. Forty-nine States and the
District of Columbia have started Prescription Drug Monitoring Programs—
Statewide electronic databases to monitor opioid prescriptions and dispensing—which have had some success in limiting prescription opioid diversion
and abuse (Reifler et al. 2012). Unfortunately, initiatives that have decreased
prescription opioid abuse have led some abusers to turn from prescription
opioids to cheaper, more available heroin (Muhuri, Gfroerer, and Davies 2013).
For example, a 2010 reformulation of a commonly abused prescription opioid
to make it more abuse-resistant had the unintended consequence of increasing heroin deaths (Evans, Lieber, and Power 2017; Alpert, Powell, and Pacula
2017). The increased availability of cheap, ultrapotent fentanyl and fentanyl
analogues, used alone or mixed with heroin, has exacerbated this problem.
Hence, future policy will need to continue and strengthen the programs that
have limited prescription opioid abuse and add new efforts to deter and
control illicit opioid abuse. NIH and HHS have announced efforts to improve
access to opioid reversal agents, support research on pain and addiction and
on developing new addiction treatments, and improve access to treatment and
recovery services (NIDA 2017a). (Box 6-2 above outlines other Administration
initiatives.)
Intensive, State-based efforts to increase treatment availability and social
supports have ameliorated the effects of addiction (Brooklyn and Sigmon 2017;
Rembert et al. 2017), but they do not deter new abusers. Accordingly, a key
policy focus for the Administration is to interrupt the supply chain to decrease
availability and effectively increase prices in order to deter new users and make
it more difficult for current abusers to continue abuse.
On January 10, 2018, President Trump signed the Interdict Act, which will
increase the number of chemical-screening devices available to U.S. Customs
and Border Protection officers in order to intercept imports of fentanyl
and other synthetic opioids. This should decrease the supply and increase
the prices of these opioids, and, when combined with the other measures
described in box 6-2 above, will lead to decreased abuse and fewer deaths.

Improving People’s Health through
More Access to Medical Innovations
People’s health can also be improved through new technologies. Innovations
produced by the private sector, aided by public policy, may be the most efficacious way to make cost-effective progress against the behavioral determinants
of health that have resisted more standard tools like public health measures
302 |

Chapter 6

$"0- 
хҊуѵ
ѵѵ
' 
 /#.
0. 
4
 
$. . Ѷ ршчц–спрф

Age-adjusted deaths per 100,000 U.S. males
тп

ршшфѷ

++-*1 .
/# 

!$-./
+-*/ . 
$)#$$/*-ѵ

#$.
0.# -.
$)

) 2
 -

*!
#$"#'4
/$1 

)/$- /-*1$-'
/# -+$ .


сф
сп

ршшх )
ршшцѷ

++-*1 .
*(+ /$/*-


рф
рп
ф

спрф

ршчцѷ

++-*1 .
!$-./

)/$- /-*1$-'
-0"Ѷ
 3+).
 ..

/*
) 2
( $/$*).
)*/
++-*1 Ѷ

)
 ' -/ .
-0"
++-*1'
4
с–т

4 -.


п
ршчф

ршшп

*0- ѷ
/$*)'  )/ -
!*-


ршшф

сппп

сппф

спрп

спрф

'/#
//$./$.Ѷ
/$*)'
$/'
//$./$.
4./ (ѵ

and taxation. Although some raise a concern that new technologies increase
healthcare spending, it is important to distinguish the price of healthcare from
the price of health. Some new treatments have high initial prices, but they
often bring down the price of health over time.
To illustrate, consider the history of HIV/AIDS (figure 6-4). Despite much
publicity about the cause and transmission of HIV and public health measures,
the infection and death toll continued to rise until 1995, when the Food and
Drug Administration (FDA) approved the first protease inhibitor and ushered in
the era of highly active antiretroviral therapies. Before the new drugs, longer
life could not be purchased at any price. Once new innovative and effective
treatments were approved, the price of health for HIV-positive individuals
decreased to the price of the new, patented drugs. The price of health fell even
further as competing drugs and cheaper generics became available. More than
100 antiretroviral therapies have been approved since then, including generic
drugs in 2005 (HHS 2016). The lower price of health for HIV-positive individuals increased spending on healthcare, but beneficially so, because the gain in
health was much larger than the new spending.
Valuable medical innovations can reduce the current and future real
prices of health. Profitable new drugs attract competitors into the market.
Prices fall closer to costs as patents expire and generic medicines come on the
market. The FDA data show that when generics become available, prices will fall
below 50 percent of branded prices after the second generic is approved and
to 20 percent of branded prices when large numbers of generics are approved

Innovative Policies to Improve All Americans’ Health | 303

(FDA 2015). Future populations will use generic versions of today’s high-priced
therapies. The vast majority of the World Health Organization’s Essential Drugs
today are off patent, allowing poor people here and around the world to enjoy
what were innovations in the past. Generic drugs now account for 9 out of 10
prescriptions dispensed in the U.S., and saved the U.S. healthcare system $1.68
trillion from 2005 to 2014 (Woodcock 2016).
Innovations, whereby mortality and morbidity gains exceed costs, are
not always expensive. Cutler and McClellan (2001) found that the value of
decreased morbidity and mortality rates resulting from technological changes
in the treatment of heart attacks, low-birthweight infants, depression, and
cataracts far exceeded the increase in spending on these conditions. The costs
of treatment of heart attacks rose by $10,000 in real terms, but life expectancy
increased by about a year—in other words, a bargain. Similarly, survival gains
across all cancer patients in the U.S. between 1983 and 1999 cost on average
only $8,670 per year of life gained (Philipson et al. 2012).
In recent years, however, many of the breakthroughs have been specialty
drugs—large, complex molecules—many of which are efficacious but initially
very expensive. The 2013 introduction of novel drugs to treat hepatitis C
(HCV)—a chronic viral infection that leads to cirrhosis, liver failure, liver cancer,
and death—helps to further illustrate the dramatic reduction in the price of
a healthier life. In 2012, therapies to treat HCV were expensive, had low cure
rates, and resulted in various side effects. In contrast, new drugs have cure
rates well over 90 percent and fewer side effects. Although the list price for a
course of the first available treatment was $84,000, negative public reaction to
the high price led to discounting and rebates (Bruen et al. 2017). Within a few
years, several competing drugs from multiple companies came on the market,
further driving down prices (Toich 2017). Medicaid officials report discounts
and rebates of 40 to 60 percent off HCV drug list prices (Bruen et al. 2017). The
most recently approved drug is highly effective, treats all six genotypes of the
virus, and has a list price of $26,400 for a course of treatment (Andrews 2017),
less than the discounted prices of the earlier drugs. Prices will fall further when
generics become available.
Of course, not every new drug or technology is more cost-effective than
older treatments. Utilizing robotic surgery as opposed to standard laparoscopic surgery for rectal cancer surgery and kidney surgery, for example, was
not associated with any improvement in outcomes but was associated with
prolonged operating time and higher hospital costs (for the kidney surgery)
(Jeong et al. 2017). Additionally, physicians often overuse cancer drugs that
have small marginal benefit but high financial, physical, and psychological
costs (Zinberg 2015). Physicians, hospitals, and both private and public payers are increasingly recognizing the importance of evaluating the value of
treatments both new and old (Porter 2010) and utilizing these assessments in
treatment decisions.
304 |

Chapter 6

Chandra and Skinner (2012) organize healthcare technologies into three
groups: high-cost, “home run” innovations that are cost-effective for nearly
everyone, for example, the HIV medications discussed above; treatments that
are highly effective for some but have declining marginal benefits for others,
such as coronary angioplasty and stents; and “gray area” treatments with modest or uncertain clinical value, such as arthroscopic surgery for osteoarthritis
of the knee. Even home run technologies span the price spectrum; HIV medications were initially expensive, sterile surgical gloves are cheap, and antibiotics,
possibly the biggest home run of them all, range from cheap to moderately
expensive.
It often takes time to establish the value of new technologies and treatments as compared with other treatments, especially when it is in an area of
active research and product development. For instance, coronary artery bypass
grafting (CABG) was popularized in the years following 1967. Despite being
expensive ($20,000–25,000 per operation in 1983; Stason and Weinstein 1985),
CABG, as compared with the best medical therapy of the time, completely
or partially relieved angina (chest pain on exertion) in patients with severe
angina (McIntosh and Garcia 1978; Rahimtoola 1982) and improved longevity
in patients with a particular type of coronary artery occlusion—severe left main
artery disease (Takaro et al. 1982). For patients with severe angina, the estimated net cost per quality-adjusted year of life gained from CABG ranged from
$3,800 for left main disease to $30,000 for single vessel disease. For patients
with left main disease, life expectancy increased by 6.9 years (Weinstein and
Stason 1982). Not every patient with coronary artery disease benefited from
CABG. Improvements in medical therapy shortly after the popularization of
CABG meant that patients with stable angina and coronary artery disease less
severe than left main coronary artery disease (e.g., single- or two-vessel disease) had equivalent survival rates from medical and surgical treatment. The
subsequent introduction of newer techniques—such as percutaneous coronary
artery angioplasty, coronary artery stents, and drug eluting stents—made costeffectiveness determinations for coronary artery disease a moving target.
Unfortunately, policymakers and potential innovators do not know what
will prove successful in advance. It is imperative to preserve the incentive to
innovate so new treatments will become available for evaluation, and, if they
represent a good value, can be adopted. Innovators and entrepreneurs are
motivated to undertake research and development (R&D) by the potential
return on investment provided by temporary, patent-protected prices. High
patent prices are also linked to the cost of capital to fund R&D for the pharmaceutical industry. Large pharmaceutical firms do not typically borrow to finance
R&D, probably because capital markets are reluctant to invest in lengthy and
risky drug development—only 1 in 10 drug candidates are eventually approved,
the process takes over a decade, and the total cost per drug approval (inclusive of failures and capital costs) is about $2.6 billion (DiMasi, Grabowski, and
Innovative Policies to Improve All Americans’ Health | 305

Box 6-4. The Administration’s Efforts to Facilitate Innovation
The Administration is committed to rolling back regulations that impede
innovation and competition. One of President Trump’s earliest actions—on
January 30, 2017—was issuing Executive Order 13771, Reducing Regulation
and Controlling Regulatory Costs. Shortly afterward, on February 24, he
issued another Executive Order directing agencies to appoint regulatory
reform task forces to determine which rules and regulations are too expensive, burdensome, or unnecessary so that they can be simplified or repealed.
The new tax law allowing immediate expensing and lowering the corporate tax rate will increase investment and innovation, including in the healthcare field. Current profits encourage R&D by signaling future profits and by
providing the current capital on which many pharmaceutical firms rely (U.S.
Department of Commerce 2004). Chapter 1 of this Report reviews the evidence
that a decrease in taxes on corporate profits decreases the before-tax rate
of return required for the marginal product of new physical assets to exceed
the cost of producing, increasing firms’ desired capital stock. Grabowski and
Vernon (1990) found that the return on R&D for new drugs is equivalent to
the industry’s cost of capital. Lowering the cost of capital will increase R&D
intensity, given that projects would require lower return on investment. The
ability of innovative pharmaceutical, biotechnology, and medical device firms
to immediately expense equipment and pay a lower marginal tax rate will
increase spending on capital, which in turn will increase innovation.
The FDA has moved to facilitate the approval of new, innovative therapies to improve health and hold down prices by increasing competition. In
2017, the FDA approved the highest-ever number of generic drugs (1,027),
the most-ever novel drugs and biologics (56), the most-ever novel medical
devices (95), and the first-ever gene therapies (Gottlieb 2018).

Hansen 2016). Instead, they mostly rely on internal funds—in short, profits—to
finance R&D (U.S. Department of Commerce 2004) (see box 6-4).

Encouraging Innovation, and
Making It Affordable
Innovations are of limited utility if people cannot afford them. As noted above,
many of the newest treatments are high-priced specialty drugs. Table 6-2
provides a range of annual per-patient costs for treating a condition with some
of these high-priced drugs. As an extreme example, the annual cost of drugs
that treat genetic diseases can reach almost $800,000 for a single patient (AHIP
2016).
The affordability of healthcare and biopharmaceutical drugs is a top concern for Americans, regardless of political party (figure 6-5). It is often asserted

306 |

Chapter 6

' 
хҊсѵ
0((-4
*!
))0'
 -Ҋ/$ )/
-0"
3+ )$/0- .Ѷ
спрф
)" 
*!
))0'
+ -Ҋ+$/ )/

-0"
 3+ )$/0- .
җ‫ڦ‬Ҙ

*)$/$*)

*2

$"#

 ) /$
$. . .
җ$)'0$)"
# - $/-4
#4+ -#*' ./ -*' ($Ҙ

цтѶутр

цштѶхтс

) -Ѷ
# (*/'*"$'
('$"))$ .

рсѶчшц

фупѶхуч

((0) 
.4./ (
$.*- -.
җ$)'0$)"
0'/$+' 
' -*.$.Ҙ

рсѶчфх

ухсѶтчу

4./$
$-*.$.

упѶфух

тхчѶхчч

)! /$*0.
$. . .

ртѶууп

ссхѶчпп

) -Ѷ
.*'$
/0(*-.

сцѶруу

сспѶтсп

0'(*)-4
-/ -$'
 4+ -/ ).$*)

рптѶуху

ршхѶфхп

руѶсшс

шчѶпуп

) -Ѷ
.0++*-/$1 
- 
" )/.

руѶрчт

урѶфцх

-*2/#
#*-(*) 
 !$$ )4

тпѶпху

тчѶшуу

-")
/-).+')/

рфѶфсч

тчѶцхф

+#/#'($
$.*- -

ртѶтсп

сшѶсфх

- $/-4
)"$*  (

*0- ѷ
( -$Ѿ.
 '/#
 ).0-) 
').Ѷ
 $"#Ҋ-$ 
-0".ѷ
./$(/ .
*!
))0'
 -Ҋ/$ )/

3+ )$/0- .
!*-
рфп
+ $'/4
 $/$*).ѵ

Note: Based upon average wholesale prices as of September 30, 2015; the “Low” entries above
- +- . )/
/# 
( $/$*)
2$/#
/# 
'*2 ./
))0'
+ -Ҋ+/$ )/
 3+ )$/0- 
!*-
/# 
$. . 
.// 
2#$' 

the “High” entries represent the medication with the greatest annual per-patient expenditure.

that promoting innovation and affordable drugs are conflicting goals. New
innovations, however, often provide improved health that was not previously
available at any price or obviate the need for more costly care. They thereby
lower the effective price of health down to the price of the patented drugs,
and later down to the price of generic drugs. Federal policies that affect drug
pricing should satisfy two goals. First, domestic drug prices paid by Americans
should be reduced. Second, the price of better health in the future should
also be reduced by spurring medical innovation. This section considers policy
options to simultaneously advance these two seemingly conflicting goals.
Reducing the drug prices that Americans pay means recognizing that
many artificially high prices result from government policies that prevent,
rather than foster, healthy price competition. Drug prices, for example, are
sometimes artificially high due to government regulations that raise prices.
This section discusses changes to the Medicare and Medicaid programs that
could help lower domestic prices, as well as reforms to the FDA that could
encourage more robust price competition.
Preserving incentives for biopharmaceutical innovation can be achieved
while still promoting lower prices for Americans. Global financial returns
from product development drive innovation. But these returns are unfairly
low today. This is because most foreign governments, which are the primary
Innovative Policies to Improve All Americans’ Health | 307

$"0- хҊфѵ
*'$/$*
ҝ
 -1-
*''$)"
*)
( -$).Ѿ
*+
-$*-$/$ .
!*-

*)"- ..Ѷ
спрц
*2 -
+- .-$+/$*)
-0"
+-$ .
 0$)"
 !$$/
)
.+ )$)"

$.$)"
!  -'
($)$(0(
2"
 + '$)"
)
- +'$)"


Percentage of Americans responding “extremely important”
хп
фп
уп
тп
сп
рп
п
*/'

 (*-/$

 +0'$)

) + ) )/

*0- .ѷ
*'$/$*
ҝ
 -1-; Americans’ Top Priorities for Congress Through the End of 2017.
*/ ѷ
*''
*!
ѵѵ
0'/.
*+ )
0"0./
тп
/*
 +/ ( -
тѵ
 $.
/# 
!!*-' 
- 
/ѵ

buyers in their respective pharmaceutical markets, force drug manufacturers
to comply with pricing rules to gain market access. Through this leverage,
foreign governments are able to set drug prices below those that prevail in the
United States and erode the returns to innovation that manufacturers might
otherwise see from selling in their markets. Among the OECD’s members, the
United States accounts for only 34 percent of the OECD’s combined GDP (at
purchasing power parity), yet the CEA estimates that Americans pay more
than 70 percent of patented biopharmaceutical profits (CEA 2018; OECD 2016).
In short, pharmaceutical innovators—and foreign governments—around the
world rely on America’s patients and taxpayers to finance critical research and
development.
The objective of government in biopharmaceutical policy is to ensure
that the private sector competes and invests in meaningful innovations that
lower the price of healthcare, rather than incentivizing market exclusivity and
high prices for products. The two goals of reducing American prices and stimulating innovation are consistent, and they can be achieved through a combined
strategy that corrects government policies that hinder price competition at
home while at the same time limiting free-riding abroad.

Why Americans Pay High Prices for Biopharmaceutical
Products, and How to Lower Them
In a well-functioning, competitive market, the price of a good is driven down
to the cost of production of the firms producing it. This principle applies to

308 |

Chapter 6

all markets, including the market for pharmaceutical drugs. However, various
factors often preclude competition from driving down prices in U.S. pharmaceutical markets.2 In the case of patent-protected monopolies for new drugs,
the lack of competition and associated higher prices are necessary to preserve
incentives for innovation. What has been less emphasized is that government
policies and public insurance programs have unintended consequences that
prevent, rather than foster, healthy price competition and induce artificially
high prices. To promote patient welfare, government policy should induce
price competition. However, in the two main Federal insurance programs,
Medicaid and Medicare, current policies dampen price competition, thereby
artificially raising prices.
Medicaid. Manufacturers that choose to enter the Medicaid Drug Rebate
Program are required to offer State Medicaid programs their prescription medications at a price that either includes a minimum rebate or, if lower, the best
price the manufacturers offer to any other purchaser. In exchange for these
discounted rates, States are then required to cover the manufacturer’s drugs
in their Medicaid programs. The practice of tying public prices to private ones
is partly beneficial, because it allows the private market to set market prices,
based on value and competition, that then get imported into the government
reimbursement. Although this basic approach of using market prices is sound
as currently implemented, the Medicaid Best Price program can create artificially high prices in the private sector under certain conditions. If a large share
of a given drug’s market is enrolled in Medicaid (e.g., for HIV or mental health
drugs), a pharmaceutical firm has an incentive to inflate prices in the private
sector so that it can collect higher postrebate prices from its large Medicaid
customer base. Similarly, the mandated price discrimination implicit in this
program prevents price discounts for lower-income patients in the private sector. Lower-income, private patient populations cannot be charged low prices,
because they would jeopardize the Medicaid price. Reforms could help prevent
the inflated private sector prices the program induces while at the same time
allowing the government to use pricing information from the private sector to
determine value (see box 6-5).
Medicare. As the Federal program providing health insurance for the
elderly and the disabled, Medicare delivers outpatient drugs administered by
health providers through Part B and prescription drugs through Part D. In the
Medicare Part B program, drugs are reimbursed based on a 6 percent markup
(now 4.3 percent, due to the sequester) above the Average Sales Price (ASP)
that manufacturers receive, net of any price discounts. As is true in any costplus reimbursement environment, this mutes the incentive for providers to
2 It should be noted that there are various prices of pharmaceutical drugs, including the
manufacturer price, pharmacy sales prices, and patient price. Even within these categories, prices
can vary depending on rebates, markups, and insurance coverage. Each price is important, and
we focus on these different prices throughout this section.

Innovative Policies to Improve All Americans’ Health | 309

Box 6-5. The Trump Administration’s Improvements to Medicaid
The Trump Administration has proposed a demonstration to allow up to five
State Medicaid programs to use more competitive drug formularies and negotiate with drug manufacturers. Meanwhile, patients will be able to appeal for
access to nonformulary drugs when medically necessary.

prescribe cheaper drugs and, therefore, for manufacturers to engage in price
competition. Though there may be higher costs to providers for prescribing
more expensive drugs—such as storing expensive drugs and the lower probability of collecting reimbursement—these costs are routinely handled in other
healthcare markets without resorting to distorted cost-plus reimbursements.
And though some private payers have responded to this type of perverse incentive problem with alternative reimbursement procedures for drugs delivered in
clinics, similar reforms have not been made for the Medicare Part B program
(see box 6-6).
Medicare Part D has several provisions that artificially raise costs for
patients. The government has previously interpreted the Social Security Act’s
requirement to include drugs within each therapeutic category and class to
mean the inclusion of at least two drugs. This requirement eliminates the ability of Part D sponsors to negotiate for lower prices when there are only two
drugs on the market because drug manufacturers know that the CMS must
cover both. Changing this requirement could lower prices for taxpayers and
patients.
Another problem resulting from Medicare Part D is the inefficient overpricing of low-value drugs. The Social Security Act §1860D-14A stipulates
cost-sharing amounts for low-income subsidy enrollees that vary by income
and are adjusted by projected program cost growth. The use of formulary tierbased cost sharing is prohibited for low-income enrollees, which eliminates the
ability of plan sponsors to price and discount drugs according to their value to
patients. Low-income subsidy enrollees and plan sponsors should have incentives to use high-value drugs. The Medicare Payment Advisory Commission
(MedPAC 2016) has highlighted this problem by reporting that 17.3 percent of
low-income subsidy enrollees are high-cost, compared with just 2.8 percent of
other enrollees.
Another problem occurs because the Medicare Part D program breaks
payment liability into three phases that incentivize beneficiaries to utilize
expensive brand drugs over generics. In the initial phase, beneficiaries are
responsible for 25 percent of drug costs up to an initial coverage limit ($3,750),
at which point they enter the coverage gap, popularly known as the donut hole.
The Coverage Gap Discount Program requires drug manufacturers to provide
a 50 percent discount to enrollees on brand name drugs while in the coverage

310 |

Chapter 6

Box 6-6. The Trump Administration’s
Improvements to Medicare Part B
To improve Medicare Part B and lower drug prices for patients, the Trump
Administration will remove perverse incentives for prescribing higher-priced
drugs and instead provide an incentive for doctors to prescribe less expensive
drugs, putting competitive pressure on manufacturers to reduce their prices.
Thus, the Administration will:
• Provide the Secretary of Health and Human Service with the authority to
cover certain drugs in Part D that are currently covered under Part B, where
there are savings to be gained through increased price competition.
• End the “gaming” in reporting drug price data that is driving up Part
B prices for patients. This is important because not all drug makers are
required to report data on all the price concessions they offer, which results
in Medicare setting payment rates higher than would otherwise apply under
statute. For new drugs that do not have much sales data, the provider’s payment will be cut to attenuate the incentives.

gap. The plan sponsor pays 15 percent. Even though the beneficiary pays no
more than 35 percent of the price for a brand name drug, 85 percent of the
price ( 35 percent plus the 50 percent discount) is counted as a beneficiary outof-pocket payment toward reaching the out-of-pocket threshold ($5,000) and
entering the catastrophic plan phase. In contrast, beneficiaries pay 44 percent
of drug prices for generic drug prices used in the coverage gap and only the
amount beneficiaries actually pay counts toward reaching the out-of-pocket
threshold and the catastrophic phase. Once in the catastrophic phase, the beneficiary only pays 5 percent, the plan’s sponsor pays 15 percent, and Medicare
pays 80 percent. With such discounts, beneficiaries may have an incentive to
use brand name drugs and reference biologics to get through the coverage
gap to the catastrophic phase as quickly as possible, despite less expensive
generics and biosimilars being available, because the 50 percent manufacturer
discount counts toward the true out-of-pocket costs.
This Part D benefit structure creates perverse incentives for plan sponsors, which often receive large discounts from branded drug manufacturers,
to generate formularies that favor high-price, high-rebate branded drugs that
speed patients through the early phases of the benefit structure where plans
are most liable for costs. Revisions to the benefit structure that eliminate the
inclusion of manufacturer discounts from the calculation of beneficiaries’ true
out-of-pocket costs would eliminate this misaligned incentive. Additionally,
increasing plan liability in the catastrophic phase from 15 to 80 percent would
provide the appropriate incentives for plans and pharmacy benefit managers
to manage beneficiary drug costs throughout the entirety of the benefit (see
box 6-7).
Innovative Policies to Improve All Americans’ Health | 311

Box 6-7. The Trump Administration’s
Improvements to Medicare Part D
The Medicare Part D program has unintended consequences that have
resulted in higher drug prices for consumers. Solutions to overcome these
problems include:
• Requiring plans to share some of the drug manufacturers’ discounts
with patients at the point of sale. The remainder of the discount would help
offset plans’ costs and help keep premiums low, as has been the case under
the current law.
• Allowing plans to manage formularies to negotiate better prices for
patients.
• Eliminating copayments on generic drugs for low-income beneficiaries.
• Increasing Part D plan sponsors’ liability in the catastrophic phase of
coverage.
• Discouraging plan formulary design that speeds patients to the catastrophic coverage phase of the Part D benefit and increases overall spending.

Cutting High Prices Resulting from Price Manipulation
The high prices of many drugs are a result of monopoly power controlling the
production of drugs that treat severe diseases with price-insensitive demand.
Being the sole supplier allows monopolists to set high prices due to the inelastic demand often associated with more severe or life-threatening diseases.
Monopoly pricing in a class is as damaging to patients who cannot afford the
drug as having no innovation at all. The problem is not confined to branded,
patent-protected drugs. Several recent episodes have illustrated the ability
of firms to legally take advantage of their position as the sole source for old
but important drugs by rapidly increasing prices. Competitors that want to
produce drugs that are off patent still face economic, regulatory, and temporal
barriers to market entry.
Unlike most markets, where new products can enter easily and cut
prices, in the biopharmaceuticals market, the FDA acts as a strict government gatekeeper for new pharmaceutical products. The evidence is clear that
generic drug competition rapidly drives down the cost of drugs. After the 1984
Hatch-Waxman Act went into effect, researchers found that market entry of
generic manufacturers resulted in generic prices falling to between 17 and
25 percent of the pre-expiration patented price within 24 months (Caves et
al. 1991; Grabowski and Vernon 1992). A more recent study by Berndt and
Aitken (2011) estimates that U.S. generic prices fell to 6 percent of patented
prices after 24 months of generic entry. Olson and Wendling (2013) account
for the endogenous entry of new generic competitors and find even larger
decreases in drug prices after two and three competitors enter the market. Yet
even generic drugs face barriers to entry. Generic drug applications to the FDA
312 |

Chapter 6

(known as Abbreviated New Drug Applications, or ANDAs) to make the same
drug can cost applicants millions of dollars and can take several years. The FDA
reports that it is devoting more resources to lessening review times and this
backlog (Woodcock 2016).
A valuable policy option might be changing the criteria for expedited
reviews to include new molecular entities that are second or third in a class,
or second or third for a given indication for which there are no generics. This
would serve as a new pro-competition pathway that would enhance therapeutic price competition by providing expedited entry into monopoly markets. To
avoid imposing policies retroactively on the industry, this policy change could
be phased in slowly, so the current manufacturers of single-source drugs would
retain the value of their efforts to be the first in a given therapeutic space.
In later subsections, we discuss FDA reforms that may enhance faster
drug price competition. The FDA Reauthorization Act of 2017 currently authorizes the FDA to designate a drug as a “competitive generic therapy” upon
request by an applicant when there is “inadequate generic competition”—that
is, when there is no more than one approved abbreviated new drug application (ANDA) for the patented reference product (not including discontinued
products).3 This designation allows for improved communication, advice from
the FDA, and a 180-day exclusivity period with no additional ANDA approvals
available for other applicants.

Enhancing Price Competition in the Pharmacy Benefit
Manager Market
Pricing in the pharmaceutical drug market suffers from high market concentration in the distribution system and a lack of transparency. Pharmacy benefit
managers (PBMs) act as buying intermediaries between drug manufacturers
and health insurance plans and their beneficiaries. They negotiate rebates off
manufacturers’ list prices and then pass on some of the benefit to health insurance plans and beneficiaries. However, the PBM market is highly concentrated.
Three PBMs account for 85 percent of the market, which allows them to exercise
undue market power against manufacturers and the health plans and beneficiaries they are supposed to be representing, thus generating outsized profits
for themselves. More than 20 percent of spending on prescription drugs was
taken in as profit by the pharmaceutical distribution system (Sood et al. 2017).
The size of manufacturer rebates and the percentage of the rebate passed on
to health plans and patients are secret. The system encourages manufacturers
to set artificially high list prices, which are reduced via manufacturers’ rebates
but leave uninsured individuals facing high drug prices. Policies to decrease
3 ANDAs or generic drug applications generally do not have to include preclinical and clinical
data to become approved. A generic drug is one that is comparable to a novel drug product in
intended use and effectiveness.

Innovative Policies to Improve All Americans’ Health | 313

concentration in the PBM market and other segments of the supply chain (e.g.,
wholesalers and pharmacies) can increase competition and further reduce the
prices of drugs paid by consumers (Sood et al. 2017).

Raising Innovation Incentives and Decreasing Free-Riding to
Reduce the Price of Better Health
It is important to continue cutting the price of health by encouraging innovation. This can be accomplished by increasing the reward for innovation by
limiting inefficient underpricing, both in foreign countries and at home, and by
decreasing the cost of R&D through FDA policy.
Limiting underpricing and free riding in foreign countries. Worldwide
profits drive innovation incentives, but when worldwide profits are partially
determined by centralized pricing by governments, this induces unique freeriding issues (Egan and Philipson 2014). Drug prices in the United States are
less publicly financed than in other countries; governments set prices in most
foreign countries. It is in each country’s interest to have other countries provide the returns to fund innovation through generous reimbursement. Smaller
markets that do not significantly affect world returns, such as those in Europe,
have an incentive to set low prices without a discernible impact on the flow of
new products their citizens enjoy. The problem is that in the aggregate, these
free-riding actions of many small countries have a substantial negative impact
on worldwide profits. Put simply, providing innovative returns is a global
public good problem that leads to classic underprovision through government
free-riding.
The United States is the engine of worldwide pharmaceutical innovation,
accounting for an estimated 46 percent of OECD patented pharmaceutical
sales. Figure 6-6 shows patented pharmaceutical sales by country. The U.S.
market for pharmaceuticals is about three times larger than the second-largest
country for both total sales (China) and patented sales (Japan). Although
the U.S. market provides a disproportionally large share of global pharmaceutical sales, it accounts for an even larger share of world pharmaceutical
profits because profit margins are much larger in the United States. In a white
paper, “Reforming Biopharmaceuticals at Home and Abroad,” the CEA (2018)
estimates that the variable profit margins of patented pharmaceuticals sold
in the U.S. are 4.1 times higher than in other developed countries.4 Using
the estimated profit margins and sales, the CEA estimates that the share of
OECD-patented pharmaceutical profits earned in the United States is about 78
percent, despite the fact that the United States makes up only 34 percent of the
OECD’s GDP at purchasing power parity (CEA 2018; OECD 2016).

4 The developed countries included were the United Kingdom, Canada, France, Germany, Italy,
Spain, Japan, Brazil, and Mexico.

314 |

Chapter 6

$"0- 
хҊхѵ

/ )/ 
#-( 0/$'
' .Ѷ
спрх

/'4 ф‫ڿ‬

*/'
.' .ѷ

‫ڦ‬фчц
$''$*)

ѵѵ ух‫ڿ‬

*0- .ѷ
ѵѵ
 +-/( )/
*!
*(( - Ѷ 0.$) ..
*)$/*-
 )/ -)/$*)'ѵ

In a study using data from the late 1990s, Vernon (2003) found a similar
ratio of profit margins for patented pharmaceuticals between the U.S. and
other developed nations. Using a different method with novel firm-level data
that reflect the margins of the world’s top 20 firms for products sold in the
United States and the rest of world, Vernon found that drugs sold in the United
States had pretax operating profit margins that were, on average, 3.9 times
those of the more regulated markets of France, Germany, Italy, Japan, and the
United Kingdom. With this estimated profit margin ratio, the CEA estimates
that the United States has a 77 percent share of OECD profits. More recently,
Goldman and Lakdawalla (2018), using a different methodology, estimated the
U.S. share of world profits to be 71 percent (midrange), as opposed to OECD
profits. The fact that this world share estimate is only modestly smaller than
the OECD share estimated by the CEA may potentially be due to the fact that
the OECD countries account for almost all world profits. Taken together, these
studies suggest a conservative profit share of the United States among the
OECD countries above 70 percent.
Although U.S. consumers and taxpayers finance more than 70 percent of
estimated OECD profits on patented biopharmaceuticals during a single year,
this does not account for the fact that these drugs are often sold in the United
States before they enter other markets. Drug manufacturers usually pursue
market access in the United States before other markets due to the higher

Innovative Policies to Improve All Americans’ Health | 315

Box 6-8. The Trump Administration’s Steps to End
Global Free-Riding in Biopharmaceuticals
Reforms are needed to address the global free-riding that takes unfair
advantage of American innovation, through enhanced trade policy or policies
that tie public reimbursements in the United States to prices paid by foreign
governments in foreign free-riding countries. To combat this unfair free-riding
by foreign markets, the Trump Administration will:
• Analyze the drug prices in the United States compared with those in the
other OECD member countries, to better understand the unfair disparity and
support the U.S. Trade Representative.
• Make regulatory changes and seek legislative solutions to put American
patients first.
• Change the incentives for foreign, developed nations that can afford to
pay for novel drugs, to price novel drugs at levels that appropriately reward
innovation.

prices in the United States. Danzon, Wang, and Wang (2005) find evidence that
the United States gains access to drugs sooner and has earlier drug launches
compared with other developed countries. For example, the United States had
several times more drug launches than Japan in the 1990s, and those launches
that occurred in the United States were on average 19 months earlier. This
implies that if the U.S. government pays for the initial years of sales (i.e., those
discounted across fewer periods and therefore producing a higher present
value), an even larger share of the return on innovation is paid for by the United
States.
Because the OECD countries do not face a trade-off between prices and
innovation, given that innovation is not substantially affected by their pricing,
most OECD nations employ price controls in an attempt to constrain the cost of
novel biopharmaceutical products—for example, through cost-effectiveness or
reference pricing policies. In essence, in price negotiations with manufacturers,
foreign governments with centralized pricing exploit the fact that once a drug
is already produced, the firm is always better off selling at a price above the
marginal cost of production and making a profit, regardless of how small, than
not selling at all. Thus, the foreign government can insist on a price that covers
the marginal production cost—but not the far greater sunk costs from years of
research and development—and firms will continue to sell to that country (see
box 6-8).
Limiting underpricing domestically. Reducing inefficient pricing domestically could help to realign incentives for pharmaceutical firms to innovate. The
Medicaid Drug Rebate Program, discussed above, had the unintended consequence of far fewer voluntary discounts from drug manufacturers to certain
safety net health providers because the discounted price would become the
316 |

Chapter 6

Box 6-9. The Trump Administration’s
Improvements to the 340B Program
The goal of the 340B Drug Pricing Program is to provide affordable pharmaceutical drugs to low-income patients, not to pay economic rents to hospitals.
In November 2017, CMS reduced the amount Medicare pays hospitals for
drugs acquired through the 340B program. It will lower Medicare beneficiaries’ coinsurance and save them an estimated $320 million in 2018 alone (CMS
2017e). The Trump Administration will create more precise eligibility criteria
and modify hospitals’ payments for drugs acquired through the 340B drug
discount program to reward them based on the charity care they provide
and to reduce payment if they provide less than 1 percent of their operating
expenses for uncompensated care. When facilities do not provide charity care,
they should not get a discount for serving the most vulnerable populations in
society.

“best price” for all remaining Medicaid patients. Congress created the 340B
Drug Pricing Program, in part, to exempt discounts for safety net health providers from the Medicaid best price formula. The 340B Drug Pricing Program
requires manufacturers that want to have their drugs covered under Medicaid
to sell outpatient drugs at a discount to safety net healthcare providers serving
vulnerable populations (Baer 2015).
The 340B program has expanded dramatically in recent years due to
loosened eligibility requirements under the ACA (Stencel 2014). Safety net
providers and their affiliated sites spent more than $16 billion to purchase 340B
drugs in 2016, six times the amount spent in 2005 (Vandervelde and Blalock
2017). Two significant problems have emerged in the 340B program. First, the
imprecise eligibility criteria have allowed for significant growth of the program
beyond its intended purpose. Second, providers can earn significant profits by
qualifying for the program, buying heavily discounted drugs, and then selling
them to Medicare and private patients at higher prices. Furthermore, there
is no requirement that hospitals use the money they earn from the 340B discounts to benefit low-income patients. These large incentives distort provider
organizations’ decisions in attempts to qualify for the program, which is simply
a form of rent seeking (see box 6-9).
Reducing the cost of innovation through change at the FDA. The fixed costs
of developing and bringing a drug to market are typically large compared with
the small marginal costs of producing additional pills or doses. Thus, the incentive to innovate is driven by whether expected profits exceed these fixed R&D
costs, and FDA policies have a major influence on the size of these fixed costs.
It is widely accepted that the fixed cost of bringing a new, patented drug
to market has risen rapidly over time. The Tufts Center for the Study of Drug

Innovative Policies to Improve All Americans’ Health | 317

Development has estimated that the average pretax industry cost in 2013 dollars per new prescription drug approval (inclusive of failures and capital costs)
is about $2.6 billion (DiMasi, Grabowski, and Hansen 2016). Moderating this
cost growth requires an understanding of the drug development process.
The most time-intensive steps for developing drugs are clinical trials
involving human subjects (80.8 months) and the FDA review (16 months),
which has fallen with each reauthorization of the Prescription Drug User Fee
Act.5 Because clinical trials and FDA review are the most time- and resourceintensive steps, reforms that significantly reduce the fixed costs of entry must
focus on these areas. Although the FDA’s drug review and approval times have
generally been shorter than those of regulatory agencies in other countries,
over the years the FDA has attempted to speed up the process (Downing et al.
2012; CIRS 2014). In particular, the FDA has four separate programs to expedite
the development and approval process for drugs that address an unmet medical need in the treatment of a serious or life-threatening condition. Although
these programs have been put in place to speed up market entry for therapeutic drugs, there is still room for improvement—an average time of more than 10
years for the development and entry of new drugs is too long.
Reforms reducing the cost of innovation raise price competition. Drug
development reforms could lower the cost of entry and enhance price competition vis-à-vis new innovations. Reforms that lower prices after patents
expire are also important. Generic drugs have been highly successful in driving
down drug prices—as more generics come on the market, prices drop rapidly,
to nearly half after just the second generic introduction. Faster generic drug
approvals could decrease the cost of entry, and thus lower drug prices even
further. The current approval process for generic drugs is based on the 1984
Drug Price Competition and Patent Term Restoration Act (also known as
the Hatch-Waxman Act). Generic drugs go through an abbreviated approval
process, in which applicants are only required to prove bioequivalence (showing that the active ingredient is absorbed at the same rate and to the same
extent as the patented drug) to an already-approved drug. Tests to prove
bioequivalence are much less costly than tests to prove safety and efficacy.
In addition, the Hatch-Waxman Act allows applicants to start clinical testing
before the patent on the original drug expires. The result was that after the act
took effect, the lag between patent expiration and generic entry for top-selling
drugs dropped from more than three years to less than three months (CBO
1998). Nevertheless, the time until approval (and associated cost) for most
generic drugs is far greater. As several recent, well-publicized episodes have
shown—the rapid price increase for epinephrine auto-injectors, for example—
some manufacturers are willing to exploit their monopoly positions. The FDA is
undertaking needed regulatory actions to streamline and speed up the process
5 PDUFA, reauthorized in 2017, allowed the FDA to collect fees from manufacturers during fiscal
years 2018–22 to fund the new drug approval process.

318 |

Chapter 6

Box 6-10. The Trump Administration Expedites Drug Approvals
In August 2017, President Trump signed the Food and Drug Administration
Reauthorization Act (FDARA) into law, reauthorizing the Generic Drug User
Fee Amendments (GDUFA) to empower the FDA to collect user fees for generic
drug applications and consistently process applications in a timely manner. In
2017, a record number of generic drugs were approved.
The Trump Administration has prioritized the approval of more generic
drugs to bring down the cost of pharmaceuticals (FDA 2018). Last year, the
FDA announced the Drug Competition Action Plan (DCAP) to expand access to
safe and effective generic drugs. Efforts have focused on three key priorities to
encourage generic drug competition: (1) preventing branded companies from
keeping generics out of the market, (2) mitigating scientific and regulatory
obstacles in gaining approval, and (3) streamlining the generic review process. The FDA has already released guidance that outlines for companies and
FDA staff members the specific steps to reduce the number of review cycles
and shorten the approval process. Some of the actions that the FDA has taken
under the DCAP include:
• New policies to expedite review of ANDAs where there are limited
approved generics for a given product.
• Publication of a “List of Off-Patent, Off-Exclusivity Drugs without an
Approved Generic,” which will receive expedited review if an ANDA for the
product is submitted.
• New guidance for industry on submitting ANDAs.
• Product specific guidance to support generic development of smallmolecule complex generics.
• Guidance for the industry for evaluating the abuse deterrence of generic
solid oral opioid drug products.
President Trump’s fiscal year 2019 budget proposal will end so-called
first-to-file gaming. When a first-to-file generic application (and therefore one
eligible for 180 days of exclusivity) is not yet approved due to application deficiencies, the FDA would be able to tentatively approve a subsequent generic
application, which would start the 180-day exclusivity clock, rather than
waiting an indefinite period for the first-to-file applicant to fix its application.
Triggering the start of this 180-day exclusivity period for first-to-file applicants
who “park” their exclusivity would speed the delivery of generic drugs and
provide substantial cost savings to American patients.

whereby generics are approved. Inefficient gaming of the regulations that distort the industry away from the intended goals of the regulations of lowering
drug prices for patients can be further reduced.
The Trump Administration also supports increasing competition for
biologics—complex biological treatments made in living cells—by encouraging the approval of competing biologics, known as biosimilars, that have the

Innovative Policies to Improve All Americans’ Health | 319

same clinical safety and efficacy as the first FDA-approved biologic. As more
biosimilars are approved, it is estimated that the price for these advanced
treatments could decrease, saving an estimated $44 billion over 10 years
(Mulcahy, Predmore, and Mattke 2014). Unlike conventional and easily replicated small-molecule drugs, biologics and biosimilars are highly sensitive to
the living systems in which they are created, requiring significant scientific
expertise (Palmer 2013). Biosimilars can take 8 to 10 years and hundreds of
millions of dollars to gain approval (FTC 2009); see box 6-10. The FDA approved
five new biosimilars in 2017, more than the approvals during the two previous
years, which include novel treatments for cancers (FDA 2018).

The Government’s Role in Reducing Prices and Stimulating
Innovation
The two seemingly inconsistent goals of reducing American prices and stimulating innovation can be achieved only with a combined strategy to reduce high
prices at home while at the same time reducing free-riding abroad. The role of
government is to ensure that firms invest in meaningful innovations that lower
the price of healthcare, rather than provide incentives that dampen competition between firms pursuing innovations. It is also government’s role to help
solve international problems, such as global free-riding on drug innovation,
that harm U.S. citizens. The Administration’s policies as outlined in this chapter
can lower prices in the United States, foster innovation, and limit foreign freeriding on the U.S. biopharmaceutical industry and American consumers and
taxpayers. Preserving the U.S. biopharmaceutical industry and encouraging it
to innovate, while making drugs that are more available and affordable for all
Americans, is an attainable goal.

Conclusion
A vital role of government is to promote a healthy and productive society. A
main premise of the ACA—if not its central premise—was that more insurance
would mean better health. For years following the ACA’s passage, proponents
measured the success of the law by the number of people who were covered by
health insurance. Unfortunately, the ACA never expanded coverage as much as
anticipated—only about 6 percent of the population gained coverage—and its
gains largely came through expanded Medicaid coverage, which gives limited
access to care. Moreover, the ACA imposed costly and cumbersome mandates
and regulations that raised costs, decreased people’s choices, and forced them
to buy insurance that they neither wanted nor needed.
The Trump Administration is committed to reducing the harm caused
by the ACA by decreasing unnecessary regulations and mandates, and by
restoring competition and choice to the insurance markets. Instead of focusing solely on insurance expansion, which the evidence indicates has less

320 |

Chapter 6

impact on health than commonly believed, the Administration plans to take
a more holistic approach by addressing all the determinants of health, including individual behaviors that greatly influence health. Despite the increase in
insurance coverage during the last several years, life expectancy in the United
States declined in 2015, and again in 2016 (CDC 2016a, 2017a). Such a decrease
in life expectancy has not occurred since the early 1960s, suggesting that
health behaviors are causing people’s health to deteriorate. The poor health
and illnesses resulting from these behaviors impose significant economic costs
through direct medical spending, premature deaths, and reduced productivity.
Effectively addressing these behaviors can improve health, productivity, and
economic growth.
The Administration has already taken important steps to address the
opioid epidemic that has ravaged our society and led to skyrocketing deaths
while the ACA expansion was ongoing. The Administration has declared a
public health emergency under the Public Health Service Act and has taken
concrete actions to prevent prescription opioid abuse, interdict the supply
of illicit opioids, improve access to treatment for those already addicted, and
encourage innovative new treatments. Additional government actions and
private innovation will be necessary to further address the opioid epidemic and
other important problems, such as smoking and the lack of physical activity
and poor diets that lead to obesity.
Medical innovations that improve people’s health can be expensive
initially, but they have the potential to bring down the price of healthcare
over time. Yet innovations are unhelpful if people cannot afford them. The
President shares the American people’s concern over high drug prices. The
Administration has laid out a strategy to accomplish the two seemingly inconsistent goals of reducing American drug prices and stimulating innovation by
correcting government policies that hinder price competition at home while
reducing free-riding abroad.

Innovative Policies to Improve All Americans’ Health | 321

x
Chapter 7

Fighting Cybersecurity Threats
to the Growing Economy
Information technology creates enormous value for the U.S. economy. However,
it also exposes U.S. firms, the government sector, and private individuals to
new risks that originate and are often effectuated entirely in cyberspace. Due
to the difficulty of identifying and punishing malicious actors, and the evergreater interconnectedness stemming from the intensified use of the Internet,
malicious cyber activity is becoming more and more widespread. Malicious
actors range from lone individuals to highly sophisticated nation-states, and
they pose a potential threat to all Americans using any information and communications technologies.
Malicious cyber activity imposes considerable costs on the U.S. economy.
Some costs are more immediate and include the value of sensitive information
and intellectual property stolen by hackers, as well as the loss of revenues,
data, and equipment due to disruptive cyberattacks and data breaches. Other
costs are longer term, such as the slow rate of adoption of new, productivityboosting information technologies and the underinvestment in research and
development stemming from poor protection against cyber theft. The ongoing
costs could escalate considerably in the event of an attack with large-scale
consequences—for example, an attack on critical infrastructure sectors that are
crucial for the smooth functioning of the U.S. economy.
Cybersecurity is a common good. A firm with weak cybersecurity imposes negative externalities on its customers, employees, and other firms tied to it through
partnerships and supply chain relations. In the presence of externalities, firms
would rationally underinvest in cybersecurity relative to the socially optimal

323

level. Therefore, it often falls to regulators to devise a series of penalties and
incentives to increase the level of investment to the desired level.
The marketplace is responding to the growing level of cyber threats. Firms are
increasingly outsourcing cyber protection functions to the blossoming cybersecurity sector. The emergence of the cyber insurance market helps firms share
the risk of cybersecurity compromises. However, these positive developments
are hampered by firms’ reluctance to share information on past malicious cyber
activity directed at them, along with the cyber threats they currently face. This
resistance stems from a variety of concerns, such as the fact that investors will
respond negatively, causing the stock price to plunge, that the firm will suffer
reputational damage and be exposed to lawsuits and regulatory actions, or that
the revelation of potential vulnerabilities could lead to additional cybersecurity
exposure. Despite the regulatory requirement that material cybersecurity
events be reported by publicly traded firms, there is a general agreement that
underreporting is pervasive. As a result of this underreporting, the frequencies
and costs of various types of malicious cyber activity directed at firms are
largely unknown, and this lack of information hampers the ability of all actors
to respond effectively and immediately.
In addition, the scarcity of information may be slowing down the development
of the cyber insurance market. Further, the use of common technologies among
otherwise unrelated firms may impede the development of the cyber insurance
market. Common vulnerabilities in these technologies cause cybersecurity
risks to be correlated across firms in complicated and little-understood patterns, which makes it difficult for insurance companies to construct properly
diversified portfolios of insured firms.
Continued cooperation between the public and private sectors is the key to
effectively managing cybersecurity risks. The ongoing efforts by the private
sector involve making information technology more secure, providing timely
defenses to new threats, and further developing platforms for anonymous
information sharing on cybersecurity threats. The government is likewise

324 |

Chapter 7

important in incentivizing cyber protection—for example, by disseminating
new cybersecurity standards, sharing best practices, conducting basic research
on cybersecurity, protecting critical infrastructures, preparing future employees for the cybersecurity workforce, and enforcing the rule of law in cyberspace.

T

his chapter examines the substantial economic costs that malicious
cyber activity directed at firms imposes on the U.S. economy. As the
U.S. economy relies more and more on information technology (IT)
and greater interconnectedness, cybersecurity threats pose an increasing challenge. A malicious cyber activity is defined as an activity, other than one authorized by or in accordance with U.S. law, that seeks to compromise or impair
the confidentiality, integrity, or availability of computers, information or communications systems, networks, physical or virtual infrastructure controlled by
computers or information systems, or the information resident thereon.
The theft and destruction of private property are not a new problem in
economics. Economists have long understood that the effective enforcement
of property rights, for both IP and physical property, underlies economic
growth by encouraging investment in physical assets, in research and development, and in putting these assets to productive uses. A law enforcement system that efficiently identifies and punishes criminals, and also actively patrols
against criminal activity, reduces crime. Law enforcement actions to disrupt
and deter cyber-enabled crime are important components of cybersecurity.
Law enforcement has deployed massive resources towards combatting cybercrime, including an entire division of the Federal Bureau of Investigation (FBI)
and hundreds of trained Federal prosecutors. However, cybercrimes present
particular challenges for law enforcement. The identification of cybercriminals
is difficult, because the Internet presents opportunities for user anonymity.
Moreover, the proliferation and sharing of malicious computer code
intended to damage or destroy computer systems—malware—makes it difficult
to tie particular malware to particular people. Sophisticated actors are able to
obfuscate origin and pathways for malicious activities. Even when criminals
are identified, punishing them is often difficult because cybercriminals often
reside in countries with unfriendly political regimes. In fact, malicious cyber
activities are sometimes authorized by such unfriendly regimes. Nonetheless,
despite the difficulties, in a significant number of cases cybercriminals have
been arrested abroad, including in countries with unfriendly political regimes,
to face charges related to cybercrime.
The responsibility for protecting against cybersecurity threats falls
largely on individuals and economic entities and not on law enforcement—that

Fighting Cybersecurity Threats to the Growing Economy | 325

is, unless cyberattacks are directed at critically important infrastructure
sectors that are deemed to be crucial for the smooth functioning of the U.S.
economy. Firms and private individuals are often outmatched by sophisticated
cyber adversaries. Even large firms with substantial resources committed to
cybersecurity may be helpless against attacks by sophisticated nation-states.
Further exacerbating the problem, firms may be rationally underinvesting in cybersecurity relative to the socially optimal level because they
do not take into account the substantial negative externalities imposed by
cyberattacks and data breaches on private individuals and on other firms. For
example, as we show later in the chapter, a data breach experienced by Equifax
also negatively affected other similar firms, along with Equifax’s corporate
customers. The firms that own critical infrastructure assets, such as parts of
the nation’s power grid, may generate pervasive negative spillover effects for
the wider economy.
For these and other reasons, cybersecurity risks have increased significantly, and malicious cyber activity imposes substantial costs on the U.S.
economy. The Council of Economic Advisers (CEA 2018) estimates that malicious cyber activity cost the U.S. economy between $57 and $109 billion in
2016, which amounts to between 0.31 and 0.58 percent of that year’s gross
domestic product (GDP). However, this number could pale in comparison with
the potential cost that would be incurred by the U.S. economy in the event of
a large-scale cyberattack, in which IT is used to disrupt services provided by
the government to its citizens and businesses. The additional costs that malicious cyber activity imposes on economic growth are (1) underinvestment in
research and development and information assets, due to insufficient protection of property rights; and (2) the slow rate of adoption for new, productivityboosting IT, for fear that it is insufficiently secure.
One glaring problem that impairs effective cybersecurity is firms’ reluctance to share information on cyber threats and exposures. Although the
Cybersecurity Information Sharing Act of 2015 made significant progress
toward the exchange of threat and vulnerability data between the private and
public sectors, firms remain reluctant to increase their exposure to legal and
public affairs risks. The lack of information on cyberattacks and data breaches
suffered by other firms may cause less sophisticated small firms to conclude
that cybersecurity risk is not a pressing problem. In addition, insufficient data
on the frequency and costs of cybersecurity events make it difficult for firms to
determine the appropriate level of resources to manage the cyber risk. In addition, the lack of data may be stymying the ability of law enforcement and other
actors to respond quickly and effectively and may be slowing the development
of the cyber insurance market.
Another impediment to a quick development of a competitive market for
cyber insurance is insursers’ insufficient understanding of their common vulnerabilities to various types of cyber threats. These vulnerabilities could arise
326 |

Chapter 7

at the level of software, hardware, or cloud computing. Without the ability to
properly quantify how cybersecurity risks are correlated across firms, insurers
may find it challenging to construct well-diversified portfolios of insured firms.
In response to growing cyber threats, both the public and private sectors are actively working on solutions. The private sector is moving to a more
cost-efficient model for cyber protection by outsourcing it to the growing
cybersecurity sector. The private sector is also responding by developing IT
solutions and by improving information sharing. Also, the cyber insurance market is expanding to meet the growing demand. However, despite this progress,
cooperation between the public and private sectors is crucial to effectively
respond and to limit the overall risks. As the frequent target of cyberattacks
and data breaches, the government can be a valuable contributor to sharing
threat information. The government can also create educational programs to
ensure that there is a robust pipeline of domestic employees for the cybersecurity workforce. Through a system of penalties and regulations and other levers,
the government can incentivize the private sector to increase its investment in
cybersecurity to the socially optimal level. Furthermore, the government sector is nearly unmatched in its ability to identify and neutralize cyber threats.
Finally, only the government has the authority to punish cybercriminals and
thus reduce their incentives to commit future crimes.
The chapter proceeds as follows. The first section gives an overview
of cybersecurity risks and cyber threat actors. The second section estimates
the costs that cybersecurity events impose on individual firms. The third section discusses the externalities that weak cybersecurity imposes on a firm’s
customers and on other firms. The fourth section describes how firms’ use of
the same software, hardware, and cloud computing services makes seemingly
unrelated firms vulnerable to the same cyber threat vectors. The fifth section
highlights the problems imposed by insufficient data. The sixth section considers the problem of dark cyber debt. The seventh section examines the growing
market for cyber insurance. The eighth section describes the costs of malicious
cyber activity for the U.S. economy. The ninth section discusses devastating
scenarios for cyberattacks and data breaches. The tenth section explains
the risks posed by the rise of quantum computing. And the eleventh section
describes the ongoing efforts by the private and public sectors to reduce cyber
risk.

Malicious Cyber Activities and
Cyber Threat Actors
Malicious cyber activities directed at firms can take multiple forms, and
they compromise at least one component of what is known as the “CIA
triad”: confidentiality, integrity, and availability. For example, a distributed
denial-of-service (DDoS) attack—which is defined as making an online service
Fighting Cybersecurity Threats to the Growing Economy | 327

unavailable by overwhelming it with traffic from multiple sources—falls under
the “availability” category of the triad because it interferes with the availability
of a firm’s Web-based services. A theft of funds from a bank customer’s account
through cyber means violates the integrity of the bank’s transactions data. A
cyber-enabled theft of the personally identifiable information (PII) of a firm’s
customers or employees compromises data confidentiality.
We next give the definitions of the terms we use in this chapter.
According to the definition proposed by the National Institute of Standards
and Technology (NIST), a cybersecurity incident is defined as a violation of
“an explicit or implied security policy” (Cichonski ‎et al. 2012). In turn, for NIST,
cybersecurity incidents include but are not limited to (1) attempts, either
failed or successful, to gain unauthorized access to a system or its data; (2)
DDoS attacks; and (3) unauthorized changes to system hardware, firmware,
or software. We further distinguish between two types of “successful” cybersecurity incidents: a cyberattack and a data breach. As defined by the Director
of National Intelligence, a cyberattack intends to “create physical effects or to
manipulate, disrupt, or delete data.” According to this definition, a cyberattack interferes with the normal functioning of a business. Thus, DDoS attacks,
cyber-enabled data and equipment destruction, and data-encryption attacks
fall into the category of cyberattacks. In contrast, a data breach may not
necessarily interfere with normal business operations, but it involves unauthorized “movement or disclosure of sensitive information to a party, usually
outside the organization, that is not authorized to have or see the information,”
according to the Department of Homeland Security (DHS 2017d). (To draw a
parallel to the property rights terminology, a cyberattack destroys property
or makes it unavailable for use, and a data breach amounts to property theft.)
In this chapter, we also refer to cyberattacks and data breaches as “malicious
cyber activity,” “adverse cyber events,” or simply as “cyber events,” and we
sometimes refer to data breaches as “cyber theft.” When a malicious cyber
activity is attributed to a criminal group or when it is directed at private individuals, we sometimes also refer to it as “cybercrime.”
According to government and industry sources, malicious cyber activity
is a growing concern for both the public and private sectors. Between 2013
and 2015, according to the Office of the Director of National Intelligence (DNI),
cyber threats were the most important strategic threat facing the United States
(DOD 2015a)—they “impose costs on the United States and global economies”
and present “risks” for “nearly all information, communication networks, and
systems” (DNI 2017). For more on cyber threat actors, see box 7-1.
Attribution of cyber incidents is difficult, but expert analysis of the malicious code and the attack techniques combined with law enforcement and
intelligence collection can identify responsible actors. Verizon’s Data Breach
Investigations Report notes that 75 percent of recent security incidents and
breaches were caused by outsiders, while 25 percent were performed by

328 |

Chapter 7

Box 7-1. Cyber Threat Actors
Cyber threat actors fall into six broad groups, each driven by distinct objectives and motivations (CSO 2017):
Nation-states: The main actors are Russia, China, Iran, and North Korea,
according to the DNI (2017). These groups are well funded and often engage
in sophisticated, targeted attacks. Nation-states are typically motivated by
political, economic, technical, or military agendas, and they have a range of
goals that vary at different times. Nation-states frequently engage in industrial
espionage. If they have funding needs, they may conduct ransom attacks and
electronic thefts of funds. Nation-states frequently target PII in order to spy
on certain individuals. Furthermore, nation-states may engage in business
destruction involving one or more firms, potentially as a retaliation against
sanctions or other actions taken by the international community, or as an act
of war (based on interviews with cybersecurity experts). Cybersecurity experts
like to say that in an act of war or retaliation, the first moves will be made in
cyberspace. A growing consensus indicates that cyberspace is already being
used by nation-states for retaliation against policies/measures, such as sanctions, imposed on them by individual nations or the international community.
Corporate competitors: These are firms that seek illicit access to proprietary IP, including financial, strategic, and workforce-related information on
their competitors; many such corporate actors are backed by nation-states.
Hacktivists: These are generally private individuals or groups around the
globe who have a political agenda and seek to carry out high-profile attacks.
These attacks help hacktivists distribute propaganda or to cause damage to
opposition organizations for ideological reasons.
Organized criminal groups: These are criminal collectives that engage
in targeted attacks motivated by profit seeking. These groups collect profits
by selling stolen PII on the dark web and by collecting ransom payments from
both public and private entities by means of disruptive attacks.
Opportunists: These are usually amateur hackers driven by a desire for
notoriety. Opportunists typically attack organizations using widely available
codes and techniques, and thus usually represent the least advanced form of
adversaries.
Company insiders: These are typically disgruntled employees or exemployees looking for revenge or financial gain. Insiders can be especially
dangerous when working in tandem with external actors, allowing these
external actors to easily bypass even the most robust defenses.

internal actors (Verizon 2017). Overall, 18 percent of threat actors were stateaffiliated groups, and 51 percent involved organized criminal groups. The DNI
(2017) notes that Russia, China, Iran, and North Korea, along with terrorists and
criminals, are frequent cyber threat actors.

Fighting Cybersecurity Threats to the Growing Economy | 329

A PricewaterhouseCoopers (PwC 2014) report—based on a survey of more
than 9,700 C-level executives, vice presidents, other administrators, and directors of IT and security practices, with 35 percent of the surveyed firms based in
the North America—states that malicious cyber activities by nation-states are
the fastest-growing category of malicious cybersecurity incidents. Actors who
are attacking on behalf of nation-states are among the most technically skilled
actors, and attacks by nation-states often go unnoticed by firms. Although,
historically, nation-states have sought to steal IP, sensitive financial plans, and
strategic information, nation-states are becoming increasingly motivated by
retaliation goals, and thus are engaging in data and equipment destruction,
and in interrupting business (FBI 2014). The most recent publicly confirmed
attack by a nation-state was a destructive WannaCry malware attack initiated
by North Korea that is estimated to have cost the world economy billions of
dollars (Bossert 2017).
A cyber adversary can utilize numerous attack vectors simultaneously.
The backdoors that were previously established may be used to concurrently
attack the compromised firms for the purpose of simultaneous business
destruction.
Ultimately, any organization is fair game for cyber threat actors, though
at different times a different set of firms may face higher risks. For example,
corporate competitors typically target firms in their industry. So-called hacktivists, motivated by ideological considerations, may pile on to attack a different set of organizations at different times, typically because these organizations have offended hacktivists’ worldviews. We have conducted interviews
with a number of cybersecurity experts and, anecdotally, news organizations
are among hacktivists’ frequent victims. When a nation-state faces sanctions
targeting a certain industry, the nation-state may use cyber-enabled means to
target firms in that same industry in the country or countries that imposed the
sanctions. That said, any firm is a potential target, independent of its age, size,
sector, location, or employee composition.
At this time, there is no common taxonomy for categorizing malicious
cyber activities. Some cybersecurity experts believe that it is helpful to focus
on the motive and associated threat actors. For example, Verizon’s 2017 “Data
Breach Investigations Report” uses three broad classifications that encompass
both motive and threat actor categories: (1) FIG (fun, ideology, grudge, or
activist group threat actors); (2) ESP (espionage motive, or state-affiliated or
nation-state actors); and (3) FIN (financial motivation, or organized criminal
group, actors). A former special adviser on cybersecurity to the White House,
Richard Clarke, used a slightly different set of classifications: (1) hacktivists; (2)
cybercriminals; (3) cyber espionage; and (4) large-scale cyberattacks (Verizon
2017; Hughes et al. 2017). As the field of cybersecurity evolves, the Council of
Economic Advisers believes that it will be helpful to develop a common lexicon
with which to delineate categories of malicious cyber activity.
330 |

Chapter 7

The Costs of Adverse Cyber
Events Incurred by Firms
A survey of firms located in the United States and in other countries, representing different industries and firm sizes, conducted by Ponemon (2017a)
revealed that a typical firm experiences 130 security breaches each year.1 If
not addressed, a security breach may evolve into materially damaging cyber
event. Because many firms employ security procedures that help detect and
neutralize cyber threats (e.g., by employing tools for detecting and containing
security breaches as well as procedures for quick recovery), security breaches
do not necessarily result in a material impact such as a business disruption,
data theft, or data or property destruction. When a firm does fall victim to an
exploit or other attack, it may face a range of loss categories, some of which are
easy to observe and quantify, and some of which are not.
Figure 7-1 illustrates the costs associated with materially damaging
cybersecurity events. These costs vary across firms and categories of cyberattacks or data breaches. Depending on the nature of their operations, firms
are generally exposed to different cyber threats. Consumer-oriented firms
with a prominent Web presence, such as online retailers, are more likely to be
targeted for a DDoS attack, while firms engaging in research and development,
such as high-technology companies, are more likely targeted for IP theft.
To provide context for this figure, consider potential costs of a DDoS
attack. A DDoS attack interferes with a firm’s online operations, causing a loss
of sales during the period of disruption. Some of the firm’s customers may
permanently switch to a competing firm due to their inability to access online
services, imposing additional costs in the form of the firm’s lost future revenue.
Furthermore, a high-visibility attack may tarnish the firm’s brand name, reducing its future revenues and business opportunities.
The costs incurred by a firm in the wake of IP theft are somewhat different. As the result of IP theft, the firm no longer has a monopoly on its proprietary findings because the stolen IP may now potentially be held and utilized
by a competing firm. If the firm discovers that its IP has been stolen (and there
is no guarantee of such discovery), attempting to identify the perpetrator or
obtain relief via legal process could result in significant costs without being
successful, especially if the IP was stolen by a foreign actor. Hence, expected
future revenues of the firm could decline. The cost of capital is likely to increase
because investors will conclude that the firm’s IP is both sought-after and not
sufficiently protected. In addition, an adverse cyber event typically triggers a
1 In the absence of a centralized data set on cyberattacks and data breaches, many statistics
reported in this chapter come from surveys. The usual limitations of survey data apply, such
as that the set of reporting firms may not be representative, or the reported results may not be
accurate. Due to the reluctance of firms to report negative information, discussed later in the
chapter, the statistics may be biased down due to underreporting.

Fighting Cybersecurity Threats to the Growing Economy | 331

$"0- 
цҊрѵ
*./
*(+*) )/.
*!
)
1 -. 4 -
1 )/

*..

*!
 

Magnitude

*..
*!

./-/ "$

$)!*-(/$*)

)- . 

*./
*!

+$/'

 +0//$*)'

("

*..
*!

/
)

,0$+( )/

4 -Ҋ
. 0-$/4

$(+-*1 Ҋ
( )/.

*..
*!

- 1 )0



*0-/

. //' Ҋ
( )/.
о

! .
- #

)*/$!$Ҋ
/$*).

0./*( -

-*/ /$*)

 "0'/*-4

+ )'/$ . *- ).$.
3/- ( '4
$!!$0'/

Difficulty of quantifying cost

 '/$1 '4
 .4

Sources: McKinsey; CEA calculations.

range of immediate and relatively easily observable costs, such as expenditures on forensics, cybersecurity improvements, data restoration, legal fees,
and the like.
Using survey data from 254 companies, Ponemon (2017a) computes
estimates of what share of the total immediately observable, cyber-driven loss
each individual cost component represents: (1) information loss, 43 percent;
(2) business disruption, 33 percent; (3) revenue losses, 21 percent; and (4)
equipment damages, 3 percent. Moreover, the case studies provided in this
chapter’s boxes illustrate how firms, by limiting their consideration to only
immediately observable losses when evaluating the impact of malicious cyber
activity, may drastically underestimate the total losses they could suffer.

Estimating the Costs of Adverse Cyber Events for Firms
The least subjective method for estimating the impact of a cybersecurity events
on a publicly traded firm is to quantify its stock price’s reaction to the news of
such events. For a publicly traded firm, its market value reflects the sum of (1)
the value of its current assets and (2) the present discounted value of all future
cash flows that the firm is expected to earn over its life span. In efficient capital
markets, the market value will adjust quickly to reflect a new valuation following any news that affects the firm value. We use an event study methodology
to calculate how market prices react to news of cyberattack or a data breach to

332 |

Chapter 7

quantify the impact the exposure on a firm’s value. All the costs shown in figure
7-1 are automatically accounted for in this calculation, reflecting the market’s
view of how the sum of these costs lowers the firm’s value.
In this analysis, we rely on the newsfeed from Thomson Reuters for public
news of cyberattacks and data breaches suffered by specific firms. The main
readerships of the Thomson Reuters newsfeed are institutional traders and
investors, who rely on it for breaking news on firms and markets. From this
newsfeed, we separate out news of cyberattacks and data breaches suffered by
individual firms. We identify news of such events by searching news headlines
for key words such as “cyberattacks,” “hacking,” “data breach,” and the like,
including spelling and syntactic variations of these keywords. To isolate the
impact of the events on stock prices, we remove announcements of cyberattacks and data breaches that fall within seven days of a quarterly earnings
announcement. Moreover, we exclude news stories concerning cybersecurity
firms, isolating only those firms that have been victims of malicious cyber activity. Because malicious cyber activity is a relatively new phenomenon, we start
our analysis in January 2000 and run it through the last month of the available
data, January 2017.
To estimate the impact of an adverse cyber event on a firm’s value, we
estimate the reaction of its stock price over the event window that begins on
the day that the adverse cyber event was publicly disclosed in the news and
ends seven days later. We employ the methodology used in prior event studies
(e.g., Neuhierl, Scherbina, and Schlusche 2012). We consider two widely used
models, the market model and the Capital Asset Pricing Model, to estimate
baseline returns. Both models produce similar results, and we report only
results based on the market model. In the market model, the market return
is subtracted from the stock return in order to calculate the abnormal stock
return on each event day. These values are then summed over the event
window to calculate a cumulative abnormal return (CAR). Moreover, because
Thomson-Reuters frequently issues closely spaced updates on prior adverse
cyber events, we require that each subsequent news articles be at least seven
days removed from the previous news—which effectively removes updates on
a previously reported news item.
Our final data set contains news of 290 adverse cyber events committed against 186 unique firms. Because institutional customers of newsfeeds
typically trade large and liquid stocks, newsfeeds disproportionately cover
large firms. As a result, the firms in our data set have relatively high market
capitalizations. The market capitalization of a median firm in our data set is
$12 billion, which is as large as that of a firm belonging to the ninth-largest size
decile of all firms trading on the New York Stock Exchange (NYSE) (and firms
trading on the NYSE tend to be larger than firms trading on other exchanges).
The market capitalization of an average firm in our sample is even higher than
that of a median firm—equal to $65 billion.
Fighting Cybersecurity Threats to the Growing Economy | 333

We find that the stock price reaction to the news of an adverse cyber
event is significantly negative. Firms on average lost about 0.8 percent of their
market value in the seven days following news of an adverse cyber event, with
the corresponding t statistic of –2.35. This t statistic is statistically significant
and makes a researcher highly confident that the underlying stock price’s
reaction to the news of an event is negative. (Also, this t statistic implies that
there is less than a 2 percent chance that a researcher would have obtained this
particular negative estimate if stock price reactions to the cybersecurity event
were distributed around the mean of zero.) We estimate that, on average, the
firms in our sample lost $498 million per adverse cyber event. The distribution of losses is highly right-skewed. When we trim the sample of estimated
losses at 1 percent on each side of the distribution, the average loss declines
to $338 million per event. The median loss per event is substantially smaller,
and equals $15 million. By comparison, PwC (2014) reports that in 2014, the
average cost attributed to cybersecurity incidents was $2.7 million. Another
industry source, Ponemon (2017a), uses a survey sample of 254 relatively large
companies (hence, the size of the firms is closer to that in our sample) and
estimates that an adverse cyber events cost these firms $21 million per event,
on average.
The number of cyberattacks and data breaches reported by Thomson
Reuters has been increasing over the years, likely for these reasons: (1) More
firms experienced adverse cybersecurity events in later years, (2) investors
started to pay more attention to and demand reports on such events, and (3)
more advanced technology has improved breach detection and allowed for a
better deflection of DDoS attacks. Of the 290 events in our sample, only 131
were reported in the 13 years before 2014, and 159 were reported after 2014.
Previous studies and reports speculated that the market was not entirely
rational, or perhaps was too slow when evaluating the costs of adverse cyber
events because of the lack of data on past events (e.g., Kvochko and Pant 2015).
Table 7-1 presents CARs to the news of adverse cyber events, by sample period.
The table shows that though in the earlier subperiod, the average stock
price reaction is negative, the corresponding t statistic indicates that it is statistically indistinguishable from zero. In the second subperiod, the stock price
reaction is significantly negative; there is less than a 1 percent chance that
researchers would have obtained the negative CAR estimate purely because
of noise in the data if stock prices did not reliably drop in response to news
of a cyberattack or a data breach. These results suggest that the market has
gained a better understanding of the costs of adverse cyber events and thus
has started reacting to news of such events more quickly.
Our study improves on earlier ones with respect to the costs of adverse
cyber events, in that it both uses a longer and more complete data set of such
events and in that it estimates the costs from stock price reactions. We obtain
markedly more negative estimates of the impact of adverse cyber events on
334 |

Chapter 7

' 
цҊрѵ
0(0'/$1 
)*-('
 /0-).
җ.Ҙ
*''*2$)"
 2.
*!
)

Adverse Cyber Event, 2000–2017
(+' 
+ -$*

0( -
*!
 1 )/.


җ‫ڿ‬Ҙ

t 
.//$./$

2000–2013

ртр

–пѵфт

–0.8

2014– Jan. 2017

рфш

– 1.01

–3.42

*0- .ѷ #*(.*)
 0/ -.Ѹ

'0'/$*).ѵ


firm values than earlier studies (e.g., Hilary, Segal, and Zhang 2016; Kvochko
and Pant 2015; Romanosky 2016), for four reasons. First, our sample includes
a wider variety of adverse cyber events, whereas earlier studies (e.g., Hilary,
Segal, and Zhang 2016) mainly used reported data breaches that involved PII.
Second, our estimations analyze market reactions to the news of adverse cyber
events, whereas some of the earlier studies consider only a subset of measurable and observable costs that would be covered by cyber insurance. Third,
our sample extends to a more recent period, during which stock price reactions
to cyber news became more immediate. Fourth, our sample of cyber events is
newsworthy enough to warrant a report in the Thomson Reuters news feed,
and, therefore, may be worse in terms of the damage caused than cyberattacks
and data breaches that are not covered in the business press.
We next analyze whether firms of different sizes react differently to the
news of cyber events. If a cyberattack or a data breach causes the same dollar damage for two firms of different sizes, the event would have a smaller
impact on a larger firm than on a smaller firm. For example—as illustrated by
the case of SolarWorld, which is discussed later in the chapter—smaller firms,
and especially those with few product lines, can easily go out of business if
they are attacked or breached. (Note that going out of business translates into
a –100 percent return on equity.) We form firm size bins based on the NYSE
size deciles, but because our sample contains very few small firms, we further
aggregate several size deciles into a single bin for smaller firms. The results,
illustrated in figure 7-2, show a U-shaped relation between firm size and the
stock price reaction to the news.
Specifically, figure 7-2 shows that firms in the 8th NYSE size decile experience the lowest CARs in response to the news of adverse cyber events, equal
to –1.72 percent. Firms in the 9th and 10th NYSE size deciles have CARs equal
to –1.12 and –0.89 percent, respectively. We believe that the CARs associated
with such cyber events experienced by smaller firms, those in deciles 1 through
7, may be less negative, for three reasons. First, the reported events may have
been less devastating. Second, the costs may have been largely covered by
cyber insurance. And third, perhaps most important, stockholders of smaller
firms are typically retail investors rather than more sophisticated institutions,
so they may take longer than seven days to react to news about cyber events

Fighting Cybersecurity Threats to the Growing Economy | 335

$"0- 
цҊсѵ
0(0'/$1 
)*-('
 /0-)
4
$-(
$5
Percent
р
пѵф

җфчҘ

п

Ҋпѵф

җтчҘ
җртшҘ

Ҋр

җсшҘ
Ҋрѵф
Ҋс

җстҘ
 $' .
р–ф

 $' .
х–ц

 $' 
ч

 $' 
ш

 $' 
рп

*0- .ѷ
#*(.*)
 0/ -.Ѹ 
'0'/$*).ѵ
*/ ѷ
0( -
*!
*. -1/$*).
$. $)
+- )/# . .ѵ

involving firms whose stocks they hold. Hence, the full price impact of the
adverse cybersecurity events will not show up within the seven-day time frame.
Despite the small sample size, we further subdivide the adverse cybersecurity events into different categories using key word searches. We attempted
to make these categories consistent with the cybersecurity industry classifications, but because the news media use varied naming conventions, the
resulting categories are somewhat different. For example, some adverse cyber
events are only described in the news headline as having been attributed to
nation-states with no additional information on the types of events. Hence, we
include a category classified simply as “nation-state.” All categories of adverse
cyber events are made to be mutually exclusive; each incident in our data set
may have exactly one classification.
We began by identifying data breaches that may involve the theft of PII.
This category of adverse cyber events received the most attention from State
regulators, as indicated by various State laws that mandate firms to disclose
instances of PII theft. (As of April 2017, 48 States, the District of Columbia,
Puerto Rico, Guam, and the Virgin Islands have put in place legislation mandating that government organizations and/or private businesses “notify individuals of security breaches of information involving personally identifiable
information” (National Conference of State Legislatures 2017).) We identified

336 |

Chapter 7

$"0- 
цҊтѵ
0(0'/$1 
)*-('
 /0-) 4
4+ 
*!
1 -. 

4 -
1 )/
Percent
п
Ҋр

/
- #

*

'2-


/# !/

/$*)Ҋ
.// .

җтфҘ

-0

җфхҘ

/# -

җрфхҘ

җруҘ

Ҋс
Ҋт

җфҘ

җрфҘ

Ҋу
Ҋф

Ҋх

җфҘ

Ҋц
*0- .ѷ
#*(+.*)  0/ -.Ѹ

'0'/$*).ѵ
*/ ѷ
0( -
*!
*. -1/$*).
$.
$)
+- )/# . .ѵ

35 adverse cybersecurity events that fall under this classification. From the
remaining sample, we identified cyberattacks that were reported to result in
the destruction of data or equipment, ultimately finding only one attack of this
nature. Using the rest of the sample, we identified the news of DDoS attacks;
we found a total of 5 observations in this category.
Next, headlines that mentioned the use of malware, spyware, ransomware, and the like had 15 observations; we classified this category as “malware.” Of the remaining news, 5 involved espionage and and/or the theft of
IP; we classified this category as “IP theft.” Using the remaining observations,
we next searched for the mention of “nation-states,” and specifically Russia,
China, Iran, or North Korea. We were able to identify 14 attacks in this category,
and we classified them as “nation-states.”2 Of course, nation-states may also
have been involved in the previously classified four categories of adverse cyber
events. Finally, we searched for the mention of wire fraud, the type of malicious
cyber activity that predominately affects financial firms. This category has the
highest number of headlines, 56. The remaining unclassified observations were
assigned to the category “other.”
Figure 7-3 shows the average seven-day CARs associated with the various
categories of cyber events in our sample, with the number of observations per
2 It is important to note that a reference to nation-state in the news media does not necessarily
reflect the attribution made by the U.S. government.

Fighting Cybersecurity Threats to the Growing Economy | 337

each category reported in parentheses. We show only the categories with at
least five observations and, therefore, excluded the category involving destructive attacks because it had only one observation.
Although based on a small sample, the figure shows that the market perceives cyber events involving IP theft to be the most damaging, with the victim
firms losing, on average, 6.32 percent of their market value. DDoS attacks are
a distant second in terms of the damage caused, with attacked firms losing
2.41 percent of market value due to a DDoS attack. As discussed above, DDoS
attacks on those consumer-oriented firms that have a heavy online presence
have the potential to cause business disruptions that result in lost customers
and reputational damage. Moreover, according to our interviews with cybersecurity experts, while contemporaneously using a DDoS attack to distract
cyber protection resources, threat actors often engage in malicious intrusions
in the victim firm’s network. Malware attacks are a close third in harm caused,
with an associated average drop in market value of 2.37 percent. Cybersecurity
experts have related to us that a number of malware attacks in our sample had
an objective of data destruction rather than ransom, and that this destruction
of data could have been extremely damaging for the affected firms.
News of adverse cyber events that mention nation-states in the headline,
on average, led to a 1.11 percent drop in market value. “Fraud” events involving
monetary theft, which typically targeted financial firms, caused average losses
of 0.69 percent of a firm’s market value. Events that involved data breaches are
relatively less damaging for victim firms, on average causing losses of only 0.56
percent. We believe that the theft of PII data on firms’ customers and employees mainly represents an externality, for which firms are not excessively penalized by the market. Finally, the “other” catchall category of cyber events is the
least damaging on average, with the typical event resulting in a 0.33 percent
drop in a firm’s market value.
Although it may be informative to study the longer-run effect of announcements of cyberattacks and data breaches on stock prices, in case stock prices
underreact or overreact in the short run,3 such an analysis would need to be
done at the portfolio level (by combing together into a portfolio multiple firms
that experienced these adverse cyber events at about the same time) rather
than at the individual stock level and would, therefore, require more observations of news of such events than what we have in our data set in order to be

3 E.g., the academic literature on the post-earnings announcement drift has shown that stock
prices tend to underreact to earnings surprises, and the stock price drifts in the direction of the
initial reaction for up to several months in the future.

338 |

Chapter 7

convincing (for a description of this econometric approach, see, e.g., Mitchell
and Stafford 2000).4
The effect of adverse cyber events on small and medium-sized businesses.
Due to the nature of our sample, small and medium-sized firms were excluded
from our analysis. However, such events may be more devastating for smaller
firms because, for example, for a business that is focused on a single product,
IP theft could wipe out the firm’s entire livelihood. Similarly, a business disruption that lasts several days could cause customers to permanently abandon
a small firm. Finally, the fixed costs of dealing with a breach or attack—such
as the cost of cybersecurity improvements and legal fees—would represent a
larger fraction of a small firm’s operating budget. The 2015 Year-End Economic
Report of the National Small Business Association (2015) estimated that, based
on survey evidence from 884 small-business owners, 42 percent of respondents
experienced a breach or an attack. Small and medium-sized businesses are at a
high risk of being attacked by ransomware, which renders a firm’s files inaccessible until a ransom is paid, along with attacks that exploit weaknesses in email
systems in order to trick firms into transferring large sums of money into the
perpetrators’ bank accounts. According to the survey, an adverse cyber event
costs the victim company over $7,000 on average. For small businesses whose
business banking accounts were hacked, the average loss was $32,000. For the
median company in the same study, in terms of revenues, these numbers represent, respectively, 0.28 percent and 1.28 percent of firm revenue. Although
these are fairly low numbers, events are typically underreported, and the firms
in the survey likely only quantify immediate and easily observable losses.
According to anecdotal evidence and various industry sources, a nontrivial number of small businesses go bankrupt as a result of a breach or attack.
In so-called perfect capital markets, corporate bankruptcies are not costly
because the corporate assets are reallocated toward best uses. However, in
the real world, corporate bankruptcies are associated with deadweight losses;
some ongoing projects will be permanently abandoned, the output of the
research and development efforts will be lost, and firm-specific hard assets
may be abandoned or sold at deep discounts.
Case studies of various types of cybersecurity incidents. We next examine
in greater detail the various categories of cybersecurity events that occur
in the United States and abroad. Most of the firms in case studies are not in
our sample, either because the events happened outside our sample period
4 Several recent studies find that stock prices of firms that experienced a cybersecurity incident
completely recover in the long run. However, the results of these studies should be interpreted
with caution. A number of these studies lack a proper control group of otherwise similar firms
that did not experience an event. In other studies, the high longer-run returns may be explained
by positive idiosyncratic (firm-specific) news that occurred subsequent to the announcement
of the breach or attack. Interestingly, many firms affected by cyber incidents subsequently
announce increased investments in cybersecurity. Possibly, the return on this type of investment
is very positive. The return on investment in cybersecurity needs to be studied more closely.

Fighting Cybersecurity Threats to the Growing Economy | 339

or because the firms were either privately held or listed on a foreign stock
exchange. These case studies, along with cyberattacks and data breaches
experienced by specific firms described in the text, are based entirely on media
reports and our own calculations using public sources, not on an investigation
by any government agency, and this report should not be taken as an authoritative description of the events, or as an accusation of criminal conduct. These
case studies are designed to illustrate that different firms may be targeted for
different reasons, and that malicious cyber activity can easily cause substantial
material damage to firms.
The first case study is of a PII data breach at Equifax (box 7-2), which
illustrates that a breach involving PII data can be devastating for a firm if its
business model is predicated on mass collection of PII.
The second case study is an attack by a nation-state on Sony (box 7-3).
The Sony case illustrates an attack by a nation-state. It is one of the few cyberattacks or data breaches publicly attributed to a nation-state actor by the U.S.
government.
The Sony attack had adverse effects on the relationship between the
United States and North Korea, and it influenced U.S. cybersecurity policy.
In response to what it called “the Democratic People’s Republic of Korea’s
numerous provocations,” the Obama Administration filed sanctions against
various individuals and organizations tied to the North Korean military and
technology sectors, barring them from access to the U.S. financial system.
President Obama also announced additional legislative proposals in response
to the attack, highlighting the need for greater cybersecurity information sharing and a modernization of law enforcement’s response to malicious cyber
activities.
The third case study is on IP theft. According to figure 7-3, IP theft is the
costliest type of malicious cyber activity. Moreover, security breaches that
enable IP theft via cyber often remain undetected for years, allowing the periodic pilfering of corporate IP. Box 7-4 illustrates that the theft of IP and other
sensitive information can have a devastating effect on an IP-centered, narrowly
focused firm.5

5 Cyber-enabled IP theft is a subset of the pervasive problem of IP theft that imposes a substantial
cost on the U.S. economy. Frequently, IP is stolen by noncyber means. For example, pirating and
counterfeiting of IP-protected products typically involves copying an observed design. According
to the Commission on the Theft of Intellectual Property (2017), China accounts for 87 percent of
counterfeited goods sized coming to the United States. Additionally, trade secrets may be stolen
using noncyber means, such as by employee raiding. Finally, the transfer of IP may result from
unfair trade practices, and U.S. firms operating in China may be particularly vulnerable to such
practices.

340 |

Chapter 7

Box 7-2. PII Data Breach at Equifax
The September 7, 2017, public announcement that disclosed the magnitude
of the data breach experienced by Equifax came after a series of notable
events. Equifax first detected the breach that compromised over 140 million
personal records (e.g., names, addresses, and Social Security numbers) in
July 2017, and it contracted Mandiant, an independent cybersecurity firm,
to assist with forensic analysis (Equifax 2017a). Contemporaneously to
these investigations, but before the details were publicly disclosed, Equifax
executives exercised their stock options and sold shares worth nearly $2
million (Equifax 2017b). Upon finally announcing that it had been the victim
of a data breach and sharing the magnitude of the breach, Equifax’s share
price declined by 13.7 percent over the course of the following trading day.
Equifax’s executives were later formally investigated for insider trading, and
the then-CEO ultimately resigned (Equifax 2017c).
The data breach impelled calls for government action, with multiple
Federal agencies launching investigations in the weeks following the breach
(Nasdaq 2017). The breach thus put Equifax’s entire business model at risk
(CNBC 2017). The breach prompted a large downward move in the value of
Equifax stock, with share prices falling by as much as 34.9 percent of prebreach prices (CEA calculations). Cumulative abnormal returns for the seven
days after the breach totaled –41 percent, with a t statistic of –15.8 (figure 7-i).
Figure 7-i. Equifax's Cumulative Abnormal Returns After Its September
2017 Data Breach Announcement
Percent
рп
п
Ҋрп
Ҋсп

 !/ #
4

Ҋтп

Ҋуп
Ҋфп
п

р

с

т

у

ф

х

ц

Trading days after breach was announced
*0- .ѷ
'**( -"Ѹ

'0'/$*).ѵ

Fighting Cybersecurity Threats to the Growing Economy | 341

The implied volatility of Equifax’s one-year option increased by 184
percent, indicating that investors perceive the future of Equifax to be largely
uncertain over the next year (CEA calculations). This high perceived uncertainty about Equifax’s future will likely negatively affect the firm’s ability to
raise new capital and make new investments.

Box 7-3. Cyberattacks by a Nation-State:
Sony Pictures Entertainment
Sony Pictures Entertainment (SPE) is a U.S. based subsidiary of the Sony
Corporation of Japan. SPE’s global operations encompass film, television,
and digital content production. In 2013, SPE generated $7.77 billion in sales
(at end-of-period dollar/yen exchange rates), accounting for 11 percent of the
Sony Corporation’s total revenue (Sony 2014).
SPE officials and employees, and the general public, first learned of
the attack on November 24 (Richwine and Finkle 2014). Hackers identifying
themselves as the “Guardians of Peace” claimed to have gained entry to SPE’s
servers and had stolen over 100 terabytes of confidential information, including employees’ Social Security numbers and health records, private emails,
and unreleased films such as Still Alice and Annie (Ignatius 2015). At this point,
SPE executives completely shut down computer systems, communicating
solely in person or over the telephone. During the following weeks, portions
of the stolen SPE data, including personal and sensitive emails between top
executives, were repeatedly dumped on public websites and circulated by
members of the press.
On December 8, the group posted more confidential SPE data and
demanded that the company “stop immediately showing the movie of terrorism which can break the regional peace and cause the War” (Richwine
and Finkle 2014). This was widely interpreted as a reference to SPE’s The
Interview, a comedy about a journalist’s attempt to assassinate North Korean
dictator Kim Jong Un. On December 16, this threat became explicit, when
the group threatened 9/11-style consequences for moviegoers attempting to
see the film. After the threats against moviegoers, the major theater chains
announced that they would not show The Interview, and Sony canceled its
theatrical release. SPE subsequently announced that The Interview would be
made available via its online streaming platforms and would be shown in 300
small, independent theaters (Stelter 2014).
Immediately after the attack occurred, Sony officials reached out to
the FBI to determine the source of the cyberattack. On December 1, 2014,
the FBI released a Flash Alert related to the attack to a limited distribution
group (Finkle 2014). In a subsequent report released on December 19, the FBI
publicly attributed the attack to North Korean hackers (FBI 2014). According

342 |

Chapter 7

to the FBI, technical analysis of the data deletion malware used in the attack
revealed links to other malware that the FBI had previously attributed to
North Korean actors. The attack also used the same tools as previous cyberattacks on South Korean banks and media outlets, which were carried out by
North Korea. These findings were supported by a later report from a leading
cybersecurity firm, concluding that the attack had the same signatures as previous attacks on South Korean and American targets and thus were unlikely to
be the work of hacktivists or a disgruntled employee (Novetta 2016).
Although the share prices increased during the period of the attack,
SPE incurred significant costs, including those related to investigation and
remediation. Press reporting indicates that the $41 million was damage that
SPE may have incurred in March 2015 (Sony 2015), but even one such article
notes: “But there are a lot more costs to come. In addition to expenses for
investigation of the attack, IT repairs, and lost movie profits, Sony faces
litigation blaming it for poor cybersecurity that exposed employees’ private
information” (Elkind 2015).

Box 7-4. Cyber Theft of IP and Sensitive
Corporate Information: SolarWorld
SolarWorld AG is a German company that manufactures and markets products for harvesting solar energy. Between May and September 2012—at about
the same time that SolarWorld was an active litigant in trade cases against
Chinese solar manufacturers, alleging that they were dumpling products
into the U.S. market at prices below fair value—SolarWorld’s network was
the target of IP theft. In May 2014, Federal prosecutors indicted five Chinese
nationals on charges of espionage, trade secret theft, and computer fraud
for hacking the networks of six U.S. companies, including U.S. subsidiaries
of SolarWorld AG, over a period of eight years (DOJ 2014). In a series of
approximately 13 intrusions, thousands of emails and files were stolen from
seven executive-level employees. Among the stolen data was information on
SolarWorld’s financial state, production capabilities, costs, business strategy,
and strategy related to the ongoing trade litigation (United States v. Wang
Dong 2014).
By breaching SolarWorld, Chinese competitors were able to gain access
to information that provided them an unfair advantage on multiple fronts
(DOJ 2014). A stolen cash flow spreadsheet allows a competitor to know
exactly how long SolarWorld would be able to survive a shock. Additionally,
production or manufacturing information can be copied without investing
time and money into research, and the information on SolarWorld’s costs
would allow a competing firm to price its products at a rate that would
make SolarWorld financially unviable (United States v. Wang Dong 2014). The
access to the SolarWorld’s trade litigation strategy would provide an unfair

Fighting Cybersecurity Threats to the Growing Economy | 343

advantage to Chinese respondents. SolarWorld has since testified that the
cyber theft allowed Chinese manufacturers to use its proprietary research
to accelerate their own production timelines, resulting in a long-term loss of
competitive advantage and return on investment (USTR 2017). As the result
of the cyber theft, which became widely known and reported on in the aftermath of the highly publicized charges, SolarWorld AG (traded on the German
DAX) lost 35 percent of its market value (with the corresponding t statistic
of –1.9) (figure 7-ii; day 0 in the figure is the day on which the charges were
announced), which amounted to a loss of €178 million (CEA calculations).
In May 2017, SolarWorld AG filed for insolvency, and SolarWorld
America, the American subsidiary, was put up for sale to help cover the parent
company’s debt obligations (Steitz 2017; SolarWorld 2017).

$"0- 
цҊ$$ѵ
*'-*-'Ѿ.
0(0'/$1 
)*-('
 /0-).
!/ /.
/
- #
 *( 
0'$'4
 )*2)
Percent
рп

п

Ҋрп

Ҋсп

 !/ #
4
Ҋтп

Ҋуп

п

р

с

т

у

ф

х

ц

Trading days after data breach became publicly known
*0- .ѷ
'**( -"Ѹ

'0'/$*).ѵ

The Distribution of Adverse Cyber Events across Sectors
How are adverse cyber events distributed across sectors? Based on the results
of the 2014 survey of 9,700 firms, PwC (2014) reports that nation-states often
target critical infrastructure providers and suppliers in order to steal IP and
trade secrets as a means to advance their political and economic advantages
(we describe the 16 designated critical infrastructure sectors later in the
chapter). At the time of the report, cyber incidents that involve nation-states
344 |

Chapter 7

' 
цҊсѵ
0( -
*!
 0-$/4
 )$ )/.
)
- # .
4
$/$(
 )0./-4

)
-")$5/$*)
$5 Ѷ
спрх
 /**/'

)$ )/.

- # .

*/'

(''

-"

)&)*2)

*/'

(''

-"

)&)*2)

усѶпхч

хпх

ссѶсцт

ршѶрчш

рѶштф

утт

сцч

рѶссу

*(*/$*)

срф

ртр

рц

хц

спр

рсч

рс

хр

($)./-/$1

ус

х

ф

тр

сц

т

т

ср

"-$0'/0-

рр

р

р

ш

р

п

р

п

*)./-0/$*)

х

т

р

с

с

р

п

р

уфф

тц

ур

тцц

цт

рф

рф

ут

0/$*)
)/ -/$)( )/

фѶфту

ц

т

фѶфсу

рр

ф

т

т

шшч

фч

шц

чут

уцр

тш

тп

упс

'/#-

уфч

шс

рпч

сфч

сшх

фц

хч

рцр

)!*-(/$*)

црц

фц

уу

хрх

ррт

ус

ср

фп

$))

)" ( )/
)0!/0-$)"
$)$)"
/# -
 -1$ .

ч

с

т

т

т

с

р

п

хсп

х

су

фшп

рсу

т

рр

ррп

х

р

р

у

т

п

р

с

хш

сс

ф

ус

фп

ру

ф

тр

-*! ..$*)'

тѶпрх

фр

ср

сѶшуу

рпш

тц

ч

ху

0'$

рфп

срѶстш

ух

спѶцфр

уус

стш

тп

фш

 '
.//

рт

с

п

рр

рр

с

п

ш

 /$'

тсх

цп

тх

ссп

шт

ух

ру

тт

-

сп

у

рп

х

рп

т

х

р

-).+*-//$*)

хт

ф

рр

уц

ру

т

у

ц

/$'$/$ .
)&)*2)

тс

с

ф

сф

рх

р

р

ру

чѶссп

т

рѶпчш

цѶрсч

хч

с

рф

фр

*0- ѷ
 -$5*)Ѷ
спрц
/
- #
 )1 ./$"/$*).
 +*-/ѵ

were most frequent in the energy, aerospace and defense, technology, and
telecommunication sectors.
According to Verizon (2017), the finance sector, both public and private, saw the most security breaches in 2016, summarized in table 7-2.
Manufacturing, government, finance, and healthcare, which made up among
the largest shares of U.S. GDP in 2016, also saw the highest shares of security
breaches in Verizon’s sample. Like NIST, Verizon (2017) defines a security
incident as an event that compromises the CIA triad of a corporate asset, while
a breach is “an incident that results in the confirmed disclosure—not just the
potential exposure—of data to unauthorized authority.” Large companies saw
the most incidents, while small companies reported the highest number of
breaches relative to incidents, suggesting that small companies are not as well
equipped to neutralize such security intrusions as large companies. Verizon

Fighting Cybersecurity Threats to the Growing Economy | 345

$"0- 
цҊуѵ
$./-$0/$*)
*!
 0-$/4
- # .
4
 )0./-4
)0./-4
+ - )/" 
*!
спрх

)0./-4
+ - )/" 
*!
- +*-/ 
. 0-$/4
- # .
*((*/$*).
)/ -/$)( )/
0/$*)
$)$)"
/$'$/$ .

"-$0'/0)" ( )/
/# -
. -1$ .
($)$./-/$1
-).+*-//$*)
*)./-0/$*)
-
)!*-(/$*)
 /$'
-*! ..$*)'
'/#$))
 '
 .//
0'$
)0!/0-$)"

0

5

10

15

20

Industry percentage of GDP and breaches

25

*0- ѷ
0- 0 *!
*)*($
)'4.$.Ѹ
 -$5*)Ѹ

'0'/$*)..

(2017) defines large companies as those with more than 1,000 employees, and
the rest as small companies.
Figure 7-4 plots the share of total cyber breaches and the sector share of
the 2016 GDP, in the order of the declining GDP share. The figure shows that
finance, healthcare, education, and accommodation suffer a disproportionate
number of breaches relative to their contribution to GDP. These sectors are
particularly attractive to malicious cyber actors because they possess valuable
PII data of their customers.

Externalities from Weak Cybersecurity and
Underinvestment in Cyber Protection
In this section, we describe how the presence of externalities creates incentives
for private firms to underinvest in cybersecurity relative to the socially optimal
level of investment. Cybersecurity is a common good. Thus, weak cybersecurity
carries a cost not only to the firm itself but also to the broader economy through
the negative externalities imposed on the firm’s customers and employees and
on its corporate partners. When the PII of a firm’s employees and customers
is stolen, in the absence of penalties and mandatory customer protections,
the burden of the costs falls on customers. A malicious cyber activity directed
against a particular firm could also have a negative spillover effect on other
346 |

Chapter 7

firms connected to the firm through the supply chain, business partnerships,
or other firms with similar business models. Because the costs are not borne
by the compromised firm, they represent negative externalities. We describe
these externalities in detail in the next subsection.

Spillover Effects to Economically Linked Firms
Due to the immense scope of Equifax’s data breach and Equifax’s centrality in
the consumer credit sector of the economy, its data breach caused multiple
spillover effects across similar firms and firms tied to it through the supply
chain, such as companies that issue credit cards. Scherbina and Schlusche
(2015) argue that co-mentions in the news media provide information on
economic linkages between firms. By doing news searches of Bloomberg,
and by noting firm co-mentions with Equifax over the month preceding the
announcement of the breach, we determined the firms that would face the
largest spillover effects due to the economic linkages and analyzed the price
reactions of these firms to the news of the Equifax data breach.
There are at least two companies that have similar business models:
TransUnion and Experian. Contemporaneous with the ongoing Equifax breach,
representatives from these specific firms were urged to testify before Congress.
These firms were adversely affected by the attack on Equifax, most likely due
to the immediate consumer response of freezing credit across all three agencies and to common concerns about the regulatory response. In addition to
investigations currently being undertaken by the Federal Trade Commission,
the Senate Finance Committee, and other organizations, the Consumer
Financial Protection Bureau announced in September 2017 that it will implement “a new regulatory regime” for credit-rating agencies, requiring that each
firm host regulators, who would be embedded at the firm, in order to prevent
future breaches. Moreover, the data breach probably caused investors to lose
confidence in the agencies’ cyber protection to revise up the probabilities of
future data breaches. An equal-weighted portfolio of TransUnion and Experian
experienced negative CAR of over 18 percent in the seven trading days following the announcement, with a t statistic of –4.7 (figure 7-5).
We also observed the breach’s negative impact on corporate customers.
As consumers freeze credit, the data breach would have a negative impact on
firms that use the credit rating agencies’ ratings to provide consumer credit.
The economically linked firms that we identified through news searches
include Fair Isaac Corporation, Synchrony Financial, Fidelity, and Virtu. An
equal-weighted portfolio of these firms experienced a negative CAR of more
than 9 percent in the seven-day window (figure 7-6).

Attacks through the Weakest Link in the Supply Chain
A firm’s security flaw can put its customers, suppliers, and corporate partners
at risk. PwC (2014) states that “sophisticated adversaries often target small and
Fighting Cybersecurity Threats to the Growing Economy | 347

$"0- 
цҊфѵ
-).)$*)Ѿ.
)
3+ -$)Ѿ.
0(0'/$1 
)*-('

 /0-).
!/ -
,0$!3Ѿ.
/
- #
))*0) ( )/
Percent
ф

п

Ҋф

Ҋрп

 !/ #
4
Ҋрф

Ҋсп
п

р

с

т

у

ф

х

ц

Trading days after breach was announced
*0- .ѷ
'**( -"Ѹ

'0'/$*).ѵ

Figure 7-6. Portfolio of Finance Firms' Cumulative Abnormal Returns
After Equifax's Data Breach Announcement
Percent
п

Ҋу

Ҋч

 !/ #
4
Ҋрс

Ҋрх
п

р

с

т

у

ф

Trading days after breach was announced
*0- .ѷ
'**( -"Ѹ

'0'/$*).ѵ

348 |

Chapter 7

х

ц

Box 7-5. Supply Chain Attack: Home Depot
The Home Depot data breach occurred from April to September 2014, and it
compromised the information of roughly 56 million unique payment cards
and 53 million email addresses (Home Depot 2014a, 2014b). The hackers
entered Home Depot’s payment systems through the use of a third-party
vendor’s login information and then unleashed malware to gain access to the
company’s point-of-sale devices (Home Depot 2014b).
The data breach had a long-term negative impact on Home Depot, and
also on other firms that were exposed to the hacked point-of-sale devices.
Since 2014, Home Depot has incurred losses of roughly $300 million due to the
data breach (Home Depot 2017). Net of insurance payments, the company has
spent $200 million to provide credit monitoring for affected customers, and
it also had to hire additional staff for its call center, investigate and upgrade
its security network, and pay fines and legal fees related to the breach (Home
Depot 2017). The breach also affected card issuers, whose customers had
to be reimbursed for fraud and whose cards had to be reissued. The Credit
Union National Association (CUNA 2014) estimates the cost of these remedies
at $8 per affected credit card, thereby placing the direct cost incurred by the
industry as the result of the data breach at $440 million.

medium-sized companies as means to gain foothold on the interconnected
business ecosystems of larger organizations with which they partner.” This
type of breach, which is known as a supply chain attack, is one of three main
vectors whereby hackers penetrate system defenses, accounting for over 60
percent of all adverse cyber events suffered by companies in 2016 (Wired 2015;
Accenture 2016). By exploiting a weakness in a relatively small and weakly protected supplier, hackers can bypass even robust cybersecurity measures. An
advantage of this attack vector is that cybercriminals can blend in with regular
network traffic, including by using legitimate credentials harvested from the
vendor. A large-scale data breach suffered by Home Depot is an example of a
supply chain attack (box 7-5).
Realizing the importance of the safety of the entire supply chain, the
industry is finding solutions to ensure supply chain safety. McAfee (2017) notes
that multiple authentication methods—such as a second factor authentication
using a hardware token or mobile app, including for vendor access—may help
prevent cyber breaches across the supply chain. After facing a cyber breach
originating from a supplier, Target announced several supply chain security
measures in line with NIST standards, such as limiting vendors’ access to
the network and improving authentication methods, in addition to broader
cybersecurity measures, such as improving the monitoring of the cyber
network (Target 2014). As part of the conditions for its 2017 settlement with
the affected credit unions, Home Depot committed to industry standard risk

Fighting Cybersecurity Threats to the Growing Economy | 349

Box 7-6. Exploiting Cyber Vulnerabilities to
Attack a Third Party: Mirai Botnet
A high-profile example of hackers exploiting cyber vulnerabilities came in
2016, when cybercriminals began using the Mirai source code to launch
broad-ranging DDoS attacks on various targets. According to an analysis published by the Institute for Critical Infrastructure Technology, Mirai
exploited devices that work with the Internet of Things with factory default
or hardcoded user names and passwords and used them to create and build
a botnet (an army of computer devices), which then overwhelmed numerous targets with traffic (Scott and Spaniel 2016). In October 2016, the Mirai
Botnet was deployed against the Internet infrastructure company Dyn, which
provides critical technology services for websites including Twitter, Amazon,
Tumblr, Reddit, Spotify, and Netflix (Krebs on Security 2016). For much of the
day, access to each of these websites was curtailed, as Dyn’s servers were
repeatedly overwhelmed by malicious traffic launched from hacked devices
that work with the Internet of Things (Krebs on Security 2016). In a statement
made after the attack, Dyn described the Mirai botnets as the primary source
of malicious attack traffic that halted Internet use (Dyn 2016).

exception processes, as well as periodic security compliance assessments of
those vendors with access to card payment information. This reflects broader
trends within the market, such as the establishment of platforms like CyberGRX
(www.cybergrx.com), which serve as clearinghouses of information on the
risks posed to downstream firms by the underlying cybersecurity weakness of
their upstream partners (Patterson Belknap 2017). In addition, the American
Bar Association has created a Vendor Contracting: Cybersecurity Checklist to
inform information security concerns in the procurement process (ABA 2016).
As another example of reducing cyber risk in the supply chain process, a consortium of financial services companies—including Bank of America, JPMorgan
Chase, Wells Fargo, and American Express—established a company, TruSight,
to standardize the risk assessment of third-party suppliers and partners,
including of their information security (Trusight 2017).

Using Cyber Vulnerabilities to Usurp Resources and Launch
Attacks on Other Firms
A cyber threat actor may exploit inadequately protected devices to launch
external attacks against a third party. Devices that work with the Internet of
Things are notoriously unsecure, because their manufacturers aim to speed up
adoption by cutting costs, and the most commonly cut cost is that of security
protection. The Mirai Botnet attack, described in box 7-6, is an example of
a cybercriminal using an existing security vulnerability to launch an attack
against a third party.

350 |

Chapter 7

Economy-wide Spillover Effects from Firms with Critical
Infrastructure Assets
Finally, and perhaps most important, if a firm owns a so-called critical
infrastructure asset, an attack against this firm could cause major disruption
throughout the economy. The 2013 Presidential Policy Directive-21 (PPD-21),
“Critical Infrastructure Security and Resilience,” notes that 16 critical infrastructure sectors that are critically important to both the U.S. economy and
national security. These sectors include chemical, commercial facilities, communications, critical manufacturing, dams, defense industrial base, emerging
services, energy, financial services, food and agriculture, government facilities,
healthcare and public health, IT, nuclear reactors, materials, and waste,
transportation systems, and water and wastewater systems (DHS 2017b). On
January 6, 2017, DHS designated the U.S. election systems as a subsector of
the existing government facilities critical infrastructure sector (U.S. Election
Assistance Commission 2017). Insufficient cybersecurity investment in these
sectors exacerbates the risks of cyberattacks and data breaches. The economic
implications of attacks against critical infrastructure assets are described in
more detail later in the chapter.
The presence of externalities would lead firms to rationally underinvest
in cybersecurity. Left to their own devices, firms will choose their optimal level
of investment by conducting an analysis of private costs and benefits without
taking externalities into account. In light of this market failure, regulators are
likely to devise a scheme of penalties and incentives that are designed to make
firms internalize the externalities and thereby help raise levels of cybersecurity investment to the socially optimal level. For example, certain mandatory
disclosure requirements were previously shown to incentivize firms to adopt
better business practices (see, e.g., Gordon et al. 2015, who conduct an analysis
of externalities resulting from weak cybersecurity).

Common Vulnerabilities
In this section, we explore how shared usage of technologies creates common
vulnerabilities across firms. These common vulnerabilities create a high likelihood that multiple firms may be compromised by a bad actor taking advantage
of the same vulnerability in several firms. Common vulnerabilities create high
correlations in firms experiencing adverse cyber events. This matters for two
reasons. First, when news of one firm experiencing a cyberattack or a data
breach become public, very likely other firms have experienced the same
compromise, even though they may not have revealed it publicly. Second,
the high correlation in adverse cyber events creates difficulties for insurers in
constructing diversified portfolios of insured firms; we will discuss this point
later in the chapter.

Fighting Cybersecurity Threats to the Growing Economy | 351

Corporate computer systems and networks are vulnerable to compromise at multiple layers, including software, firmware, and hardware. When a
vulnerability in one of these layers is discovered and subsequently exploited by
cybercriminals or other malicious actors, it is highly probable that other firms
that use the same technology may be similarly vulnerable. Malicious actors
often target a vulnerability wherever it exists, not necessarily focusing on a
single firm or industry. In what follows, we explain how common technologies
can create common vulnerabilities across multiple firms.

Software
A computer’s software is any data or computer instructions stored on a
computer’s hardware. Software is encoded in a binary basis and forms the
tools by which computers execute tasks and manipulate information. In vulnerable systems, unbeknownst to the end user, software can be modified or
otherwise abused by malicious actors to run unwanted processes on a given
system, allowing the actors to affect adverse outcomes for a system’s users. If
undetected, these processes may allow an adversary to obtain or manipulate
information on a computer system without the end user’s permission. The
goal of these adverse actors is often to enable unauthorized access to secure
systems for the purpose of stealing, encrypting or destroying private data and
information, or for modifying industrial control processes in order to cause
harm to a company’s physical assets and/or its employees.
Software vulnerabilities often stem from simple errors in software coding. Unbeknownst to developers, innocent coding errors may make a program
vulnerable to software exploits. In a typical software code, there are an average 25 errors per 1,000 lines of code (NIST 2016). NIST has stated a goal for a
“dramatic reduction” in software vulnerabilities. The stated goal is to reduce
the error rate to 25 errors per 100,000 lines of code (NIST 2016). Systems with
near-zero errors are produced routinely today in the aerospace industry, but at
several times the cost of ordinary software. This objective will have substantial
costs associated with its implementation, but ultimately will hopefully pay off
through a sufficient reduction in software vulnerabilities.
We now discuss the particularly harmful so-called zero-day vulnerabilities, for which a security solution does not yet exist, and the “backdoor” methods that malicious actors exploit to gain entry into a seemingly secure system.
Zero-day vulnerabilities. So-called zero-day vulnerabilities are a particular subset of vulnerabilities characterized by being unknown to the hardware/
software vendor and end users prior to being discovered and/or exploited.
“Zero” days refer to the amount of time in which a producer or cybersecurity
firm has from the time of discovery to provide the users with a patch to eliminate the vulnerability.
Zero-day vulnerabilities are often exploited with the help of the so-called
exploit kits, primarily available for purchase on the so-called dark web—which

352 |

Chapter 7

refers to the large portion of the Internet whose contents are not indexed by
standard search engines. An exploit kit is a web-based application centered on
a zero-day vulnerability that streamlines the vulnerability’s exploitive application; these kits provide easy to use, replicable templates to exploit individual
vulnerabilities on a large scale. A typical kit contains mechanisms to profile
potential victims, identify compromised systems, and subsequently deliver
“payloads” (exploitative or malicious software).
Once a patch is written and released by the architects, the vulnerability
is no longer deemed a zero day. However, it is ultimately up to the end users
to update their systems in order to be considered immune to a given zero
day vulnerability. Lloyd’s of London (2017) notes that it can “take anywhere
from days to years” before a developer is made aware of the vulnerability.
This allows illicit discoverers of vulnerabilities ample time to explore angles of
compromise, develop the necessary software for exploitation, and potentially
market this exploitation technique to interested third parties.
“Backdoor” access. A backdoor is defined as a “hidden entrance to a
computer system that can be used to bypass security policies”; it may allow
one to gain access to a network a computer system or a connected device,
unbeknownst to the end user (OWASP 2006). It is common for a commercial
software package to have a backdoor to enable developers to modify the
systems they oversee. A backdoor may take the form of a hidden aspect of a
program, a separate program, a part of an operating system, or even be coded
into the firmware already installed on a system’s hardware. Threat actors may
gain access to pre-installed backdoors or install their own backdoors with the
end goal of taking control of the systems or inserting malicious modification at
any time that they wish. Many hardware products have backdoor methods of
access and may be vulnerable to security compromises using these backdoors
methods of entry, regardless of the software programs that are being run on
the hardware in question.

Firmware
Firmware constitutes the next step above hardware in a traditional system
stack. System firmware is usually software that boots or initiates systems,
along with running baseline-level tasks, such as power management and enduser controls (e.g., mice or keyboards). This software is often unique to or integrated with individual firms’ hardware, thus earning the moniker “firmware”
due to being hardware specific to a given firm’s technology. USB drives, hard
and solid-state drives, memory cards, and digital power chargers all typically
utilize firmware.
Firmware is a prime target for compromise because it resides below the
operating system and may not be protected by the security software that runs
on an operating system. These firmware vulnerabilities, which allow attackers
to take control of a system during its booting phase, have been identified in
Fighting Cybersecurity Threats to the Growing Economy | 353

USB devices (e.g., memory sticks), network cards, embedded and keyboard
controllers, baseboard management controllers, modems, central processing units batteries, home routers, office printers, IP phones, and many other
devices. McAfee has identified several instances of hacking groups, industrial
espionage teams and organized crime groups, utilizing firmware exploits in
order to commit cyberattacks and cyber theft.

Hardware
A computer’s hardware are the physical components of a computer. Hardware
components can be either active (internally powered) or passive (driven by an
external power source). Typical components include, but are not limited to,
monitors, keyboards, hard and soft drives, graphics cards, sound cards, processors, and motherboards. Although traditionally harder to attack externally,
hardware vulnerabilities can completely undermine an entire system stack’s
security. Hardware threats undermine a system’s software security measures
because software inherently assumes that hardware on which it runs is not
compromised. The discovery of a hardware-based exploitation may force system infrastructure to be replaced entirely as hardware compromises typically
cannot be fixed by software patching alone.
Hardware is a less frequent target of hackers than software for a number of reasons: hardware is typically less easily accessible, it is not as well
understood, and attacks against hardware often must be highly specialized.
However, once discovered, hardware vulnerabilities can be highly damaging:
Hardware vulnerabilities may cause compromises independent of operating
system or software security measures.
A striking recent example of a hardware vulnerability was recently discovered by the Project Zero (2018) research team at Google in certain processors
manufactured by Intel, AMD, and ARM. Specifically, Google found and reported
three unique vulnerabilities usable against these processors to the processors’
respective manufacturers on June 1, 2017.6 The vulnerabilities could allow
malicious actors to steal information stored in a processor’s memory, affecting
virtually all computing devices, such as personal computers, cloud servers, and
smartphones.

Cloud Computing
Cloud computing has allowed companies to achieve economies of scale by
outsourcing various tasks—such as data storage, services, and analytics—to
outside providers. McAfee (2017) cites that 93 percent of organizations utilized some form of cloud computing for software, platform, or infrastructure
services.
6 These vulnerabilities are registered as CVE-2017-5753, CVE-2017-5715, and CVE-2017-5754 with
the National Vulnerability Database’s Common Vulnerabilities and Exploits list.

354 |

Chapter 7

Cloud computing platforms use the virtual machine archetype; a virtual
machine simulates a physical computer system (hardware, operating system,
and applications) on top of an underlying operating system. A cloud can be
running any number of virtual machines simultaneously on top of its underlying operating system, allowing for providers to utilize the same hardware for
different customers without usage conflicts between end users. The programs
overseeing this delegation of space for different virtual machines are called
“virtual machine monitors” or hypervisors.
Cloud computing has its own inherent vulnerabilities, which can create common risks among end users. If the underlying hypervisor overseeing
a cloud network is compromised, it can be assumed that all systems being
hosted on the network will in turn be vulnerable to exploitation. This leads
to a great degree of risk correlation between firms from cyber threats that
otherwise would not exist if the firms’ data and services were located locally.
Furthermore, if a hardware replacement or hard-software update (a software
update that requires a power reset) is needed to resolve these problems,
computing jobs need to be interrupted, which upsets customers and in turn
discourages hosts from running these time-consuming updates or patches.
Managed service providers (MSPs) are similar to cloud computing providers, but they typically provide additional IT services, such as network connectivity, data security solutions, and general IT strategy management. According
to a 2017 report by PwC, multiple MSPs were targeted from 2016 onward by a
single adverse actor, APT10 (PwC 2017). (According to FireEye, a cybersecurity
firm, APT10, is a Chinese cyber espionage group that FireEye has tracked since
2009.) PwC (2017) further states that as a result of its activities, APT10 has
potentially gained access to “the intellectual property and sensitive data of
those MSPs and their clients globally” (PwC 2017).

The Problem of Insufficient Data
In today’s data-driven world, important investment decisions are based on
sound empirical analysis. However, the field of cybersecurity is plagued by
insufficient data, largely because firms face a strong disincentive to report
negative news. Cyber protection could be greatly improved if data on past data
breaches and cyberattacks were more readily shared across firms.
There are multiple reasons for insufficient disclosure. To being with,
many cybersecurity breaches go undetected by firms. Citing data from cybersecurity firms, PwC (2014) reports that as many as 71 percent of cyber compromises go undetected. Furthermore, according to industry reports, the U.S.
government can frequently observe an attack. For example, the Center for
Strategic and International Studies (2014) reports that in 2013 U.S. government
notified 3,000 companies that they had been hacked. Even when a firm is aware
that it had experienced an adverse cyber event, it would frequently refrain

Fighting Cybersecurity Threats to the Growing Economy | 355

from reporting the event for fear of negatively affecting its market value and
its relationships with corporate partners. For example, the Center for Strategic
and International Studies (2014) reports that when Google was hacked in 2010,
another 34 Fortune 500 companies were hacked at the same time (that fact
eventually became public knowledge through WikiLeaks), but only one of these
companies reported publicly that it had been hacked.
Data on adverse cyber events that involve breaches of PII and a subset
of other security breaches are slowly becoming available, partly due to disclosure requirements. Countries around the world are adopting mandatory
data breach disclosures, for compromised PII on firms’ customers (though at
different levels of coverage), such as the General Data Protection Regulation
(GDPR) in the European Union. The U.S. government also imposes sectorspecific cyber disclosure legislation. The Health Insurance Portability and
Accountability Act (HIPAA), pursuant to Public Law 104-191, sets disclosure
requirements on personal data protection, though studies have raised concerns about compliance with, exemptions to, and the lack of, “standardized
technology requirements” in the regulations (Chang 2014; Koch 2017). Banks
and certain financial institutions are subject to regulatory examinations that
include review of their safeguards for protecting the security, confidentiality,
and integrity of consumer information, which include disclosure requirements
in the event of a breach. The Department of Energy also requires disclosure of
events—including those that are cyber-related—that may have an impact on
the electricity system, through the OE-417 Electric Emergency Incident and
Disturbance Reports, pursuant to Public Law 93-275. These reported incidents
are posted on the Department of Energy’s website, which gives information on
the event’s date, date of restoration, areas affected, alert criteria, event type,
demand loss, and number of customers affected. Of 141 events reported in
2016, 5 were cyber-related. Of the 127 events reported in 2017, two were cyberrelated, though these events were not reported to affect customers or result in
the loss of demand (DOE 2017).
For publicly traded firms, public disclosure of materially important
adverse cyber events is mandated by the Securities and Exchange Commission’s
(SEC) 2011 Guidance, and also by the requirements that trigger the filing of the
SEC’s Form 8-K. Specifically, the 2011 Guidance mandates that publicly traded
firms disclose “material” cybersecurity risks and cyber incidents. However, the
effectiveness of the SEC’s 2011 Guidance is frequently questioned. There are
concerns that companies underreport events due to alternative interpretations of the definition of “materiality” (Gordon et al. 2006, 2015). There are also
concerns that the disclosure requirements are too general and do not provide
clear instructions on how much information to disclose, and that they therefore “fail to resolve the information asymmetry at which the disclosure laws are
aimed” (Ferraro 2014). For example, according to the 2017 survey of 2,168 individuals who were involved in both cyber risk and enterprise risk management
356 |

Chapter 7

activities in their firms, 36 percent of survey participants said that a material
loss of information assets does not require a disclosure on the firm’s financial
statements. At the same time, 43 percent of respondents stated that their firm
would disclose a loss of property plant and equipment on its financial statements (Ponemon 2017b). According to these studies, more comprehensive
and mandatory disclosure guidance, such as through legislative endorsement
(Ferraro 2014) or endorsement by the SEC (Gregory 2014), may help overcome
these issues.
If, between quarterly reports, a cyberattack or a data breach triggers an
event that would mandate the filing of Form 8-K (e.g., bankruptcy, departures
of corporate directors, entry into or a termination of a “material definitive
agreement”), then victims must disclose the cyber event under the requirement that the firm file the form within four business days of the event. If a
materially important cyber event is privately disclosed by the affected firm
to a financial intermediary—such as a buy- or sell-side analyst, an investment
manager, a broker dealer, or an investment adviser who could generate a
profit for themselves or their clients from having this informational advantage—Regulation Fair Disclosure requires that the event must be disclosed to
the public promptly.
Other countries also mandate disclosures of cyber breaches, and some
countries have stricter disclosure requirements than the United States does.
For example, in April 2016 the European Union adopted GDPR, which becomes
effective in May 2018 and mandates companies to disclose data breaches. This
regulation expands the scope of the EU’s 1995 data protection regulation to
all companies that process the data of EU-based subjects, regardless of the
company’s location. Past regulations only applied to companies based on their
physical location, and the new regulation will also affect United States–based
firms as long as they have European customers. Companies subject to this
regulation must notify their customers and other affected parties of breaches
where “a data breach likely to result in a risk for the rights and freedoms of
individuals” (GDPR 2017). The breach must be disclosed to the government,
customers, and controllers within 72 hours of the firm’s becoming aware of
the breach. This new rule will further increase the number of publicly reported
data breaches.
Even if cyber events are not being disclosed by firms, the news media
can find out about such events through journalist investigations. For example,
Verizon (2017) reports that 27 percent of data breaches were discovered by
third parties. These third parties may notify the news media in addition to notifying the affected firms, creating another channel for the spread of information.
The lack of a representative data set for cybersecurity incidents poses a
number of challenges to firms and policymakers. For policymakers, it makes it
next to impossible to accurately measure the cost of cybersecurity incidents
for the U.S. economy and to determine whether more active government
Fighting Cybersecurity Threats to the Growing Economy | 357

involvement is needed to limit cybersecurity risk. Likewise, for firms, the lack
of data makes it difficult to correctly assess the expected costs of cybersecurity
exposure and to determine the optimal level of investment in cybersecurity.
Moreover, when negative information is underreported for incentive reasons,
agents may erroneously assume that the negative information/events simply
do not exist (see, e.g., Scherbina 2008). In case of adverse cyber events, underreporting may lead the less sophisticated managers to assume that the risk
is not significant and consequently to underinvest in cybersecurity. Industry
sources speculate that less sophisticated smaller firms underinvest in cybersecurity for this reason.
Unlike firms and private individuals, cyber insurance and cybersecurity
providers have the advantage of being able to collect data on cyberattacks
and data breaches through their business operations. However, these entities
are reluctant to share their data with the public because of privacy concerns
for their clients and also because these data represent a source of competitive
advantage in providing security services for cybersecurity companies and in
pricing cyber insurance products for insurance companies.
A more robust data set on cyber incidents and cyber threats that could be
updated in real time would greatly help firms improve their cybersecurity. And
still another negative effect of the paucity of publicly available data is that it
may slow the development of a more competitive market for cyber insurance.

Dark Cyber Debt
As discussed above, firms are reluctant to reveal cyber breaches to the public
for fear of lowering their valuations; even when a firm’s management does
report a breach, it often underreports its scope. Most likely, the information
about the breach will eventually become public, at which point the value of
the firm will decline to reflect the resulting monetary losses. In this section,
we introduce the concept of “dark cyber debt” to describe the future negative
valuation impact of a breach that a firm hid from the public. It is “dark” because
it is currently hidden, and it is a “debt” that eventually would need to be paid
before investors are paid.
Consider the latest illustration of the concept. In October 2016, the personal data of approximately 57 million customers and drivers was stolen from
Uber Technologies Inc. (Newcomer 2017). The data were then ransomed back
to Uber in exchange for an illicit payment of $100,000 to the hackers by Uber’s
security chief and one of his deputies (Newcomer 2017). The compromised
data included some 600,000 driver’s license numbers for Uber’s drivers, which
were linked to their identities (Newcomer 2017). Though Uber has admitted it
had a legal obligation to disclose the attacks on a timely basis to regulators and
also to the drivers whose identities were compromised, it instead chose to hide
the news and to pay the perpetrators to delete the stolen sensitive information

358 |

Chapter 7

(Newcomer 2017). Further attempts to conceal the damage manifested themselves through Uber’s executives writing off the $100,000 as a “bug bounty,” a
practice whereby technology companies hire external parties to attack their
software in order to test for vulnerabilities (Isaac, Benner, and Frenkel 2017). It
is now clear that these breaches were the work of criminals rather than firms
hired to test Uber’s cybersecurity. The timing of Uber’s hack was particularly
unfortunate, because the firm had been planning to go public. In the aftermath
of the news, SoftBank, a Japanese firm, and a group of Uber’s shareholders
agreed to a deal valuing the company at $48 billion, a notable decline in the $70
billion that Uber commanded just over a year ago (Reuters 2017b). Although
not all the decline in value can be attributed to the data breach, given that Uber
also faced other types of negative publicity, offers following the hack were
substantially lower than pre-breach figures.
This particular nondisclosure is far from the only example of dark cyber
debt. For example, in 2016 Uber faced a $20,000 fine for its failure to disclose a
2014 breach (New York State Office of the Attorney General 2016).7

Cyber Insurance
The rise in malicious cyber activity directed at firms over the past decade gave
an impetus to a quickly growing market for cyber insurance. The global cyber
insurance market is estimated to be worth roughly $3.5 billion today, up from
less than $1 billion in 2012, and is projected to grow to $14 billion by 2022
(Lloyd’s of London 2017; Allied Market Research 2016).8 North America, particularly the United States, accounts for roughly 90 percent of the global cyber
insurance market. In 2016, property and casualty insurers wrote $1.4 billion
in direct premiums for cyber insurance, up 35 percent from the previous year
(Lloyd’s of London 2017). This figure, however, is only a miniscule fraction of
the roughly $530 billion in premiums for the entire insurance market. Although
the U.S. market is more developed than markets in other countries, only about
a third of U.S. companies have purchased some sort of cyber insurance, with
large variation across sectors (Romanosky et al. 2017). Though supply and
demand for cyber insurance continues to grow, the cyber insurance market
faces a number of challenges that slow down its pace of development.
Compared with other risks covered by insurance—such as wind, flood,
and fire—cyber risk is perhaps the fastest-evolving and least understood. A
big challenge faced by the market is the scarcity of data on past incidents. The
importance of modeling cannot be understated when it relates to pricing risk,
7 Attorney General of the State of New York, Internet Bureau, Assurance No. 15-185. We must note
that even when a company takes all reasonable cybersecurity measures and makes appropriate
disclosures, its stock price will likely decline when a data breach becomes public.
8 Our discussion of cyber insurance and cyber insurance policies only includes specialized cyber
insurance policies marketed as such. It does not include other broader policies that may cover
losses from a cyber event.

Fighting Cybersecurity Threats to the Growing Economy | 359

and the lack of historical data and unpredictability of risk make it difficult to
model and therefore underwrite. There is a significant qualitative aspect to
pricing that insurers rely on when pricing policies. As a result, policies for cyber
risk are more customized than other risk insurers taken on, and, therefore,
more costly (NAIC 2017). Insurance firms also need to be able to assess the correlation in risks and losses across firms in order to form diversified portfolios
of insured firms. Cyber insurance, like most insurance products, distinguishes
between two loss categories, first party and third party. First-party losses are
those that directly harm the insured, while third-party liability relates to claims
undertaken by external parties who experience losses due to the insured’s
actions. Without good data, it is very difficult to quantify the potential spillover
effects to third parties. Firms that got into the cyber insurance market early
clearly have a data advantage over new entrants by having collected historical
data of past cyber insurance claims from client firms. However, the “data barrier” makes the market less competitive by deterring new entrants.
Some insurers utilize information from policyholders’ self-assessment
forms to place a firm’s risk into a generic high, medium, or low level. The varying levels of information that insurers possess on each of their clients determines the size and sophistication of their respective policies and explains the
differences in coverage among firms. Some insurers have admitted to relying
on other insurance companies’ premiums to determine pricing, due to a lack
of their own data. The adverse effect of insufficient data is that insurance firms
struggle to price cyber risk. Underpricing cyber risk could result in insurers
being unable to cover claims. Overpricing cyber risk could lead to underinsurance on the part of the firms.
A systematic collection of data on past cyberattacks and data breaches
would be a big push for a quicker development of the cyber insurance market.
Currently, companies are required to publicly report data breaches that expose
personally identifiable information, payment data, or personal health information. However, there are no mandatory reporting requirements for other types
of cyber events, such as those involving IP theft, ransoms, data and equipment
destruction, or business disruption. Though publicly traded firms are required
to report cyber events that have the potential to materially affect the firm’s
value, there is substantial underreporting, as firms are free to determine themselves whether an event is “materially important” (Jin 2015). The absence of
data on cyber risk and the difficulty of monitoring firms’ behavior have resulted
in insurability challenges for the cyber insurance market. Because of this, general challenges to all insurance markets (adverse selection and moral hazard)
exist as companies that have been victim to a serious malicious cyber activity
are more likely to buy insurance, and having insurance fosters a lack of incentive to invest in self-protection measures.
Another major reason the cyber insurance market is relatively small
in size is due to the relatively high premiums for relatively limited coverage.
360 |

Chapter 7

Box 7-7. Target’s Cyber Insurance Policy
In 2013, Target experienced a data breach in which payment information
and customer data were stolen. Between 2013 and 2015, Target’s cumulative losses for this incident were $290 million (Target 2017). These expenses
included legal and other professional services related to the data breach, but
did not include insurance compensation for the potential reputational damage. Insurance coverage offset only $90 million of the losses, resulting in a net
pretax loss of $200 million (Target 2017). Thus, Target’s cyber insurance only
provided coverage for about 30 percent of the easily quantifiable data breach
costs. Of course, because the out-of-pocket costs were used to lower Target’s
tax liability, the after-tax losses were somewhat lower (SEC 2017).

Typically, only costs that can be easily quantified are covered by insurance. The
most frequent type of cyber insurance coverage today relate to data breaches,
including first-party coverage for costs such as crisis management and identity
theft response and third-party coverage related to privacy liability. Other common types of first-party loss coverage include costs related to investigations
of the attack, restoring business services, credit-monitoring services, notifying
affected parties, ransom payments, and other losses associated with business
disruption. It is uncommon for cyber insurance to cover reputational harm,
loss of future revenue, costs of improving cybersecurity systems, losses from
IP theft, and nation-state attacks.
The ambiguity of coverage for cyber insurance products, which is also
limited, coupled with the heterogeneity of offerings across insurance firms, is
another challenge faced by the burgeoning cyber insurance market. The lack
of standardization of insurance products makes it difficult for firms to compare
coverage across insurance firms. This, in turn, is another reason for the slow
adoption of cyber insurance coverage.
As a result of underreporting of cybersecurity incidents and the associated costs, firms may underestimate their own risks, and the demand for
insurance may be lower than optimal. Moreover, insurance firms themselves
may be unwilling to provide sufficient limits for cyber insurance. The case study
given in box 7-7 describes how only a small portfolio of losses stemming from
Target’s 2013 data breach was covered by insurance (Naked Security 2015).
Luckily, the passage of time will allow insurance companies to collect
sufficient data from their clients to better price insurance products and to
expand coverage. A competitive cyber insurance market that offers a wide
array of efficiently priced products would become an important contributor
to economic growth. Though publicly traded firms enjoy a diversified set of
investors, who do not demand to be compensated for the idiosyncratic risk
associated with cybersecurity breaches and attacks, the cyber insurance

Fighting Cybersecurity Threats to the Growing Economy | 361

market should allow private firms to cross-insure their cyber risk and lower
their cost of capital. Cyber insurance will help reduce the deadweight losses
that are associated with corporate bankruptcies driven by cyberattacks and
data breaches. Another advantage of the cyber insurance market is that
through the underwriting process, it may encourage the adoption of better
cybersecurity practices. The insurance provider’s ability to assess whether
a potential policyholder has adequate cybersecurity incentivizes the firm to
undertake cybersecurity investment. Premiums would generally be higher for
firms that do not have any substantial cybersecurity measures. This mitigates
moral hazard in the marketplace.
However, the cyber insurance market will continue to face challenges.
Cyber threats are ever evolving. Increasing reliance on information, increasing
interconnectedness, and the adoption of new technologies will bring about
new cyber threats. Thus, even after collecting sufficient data on past cybersecurity events, predicting risks and correlations will remain a challenge.
By offering protection against theft, cyber insurance will reinforce firms’
incentive to invest in IP and proprietary data. However, it will be difficult to
achieve complete protection, for two reasons. First, as discussed above, firms
may often be unaware that they have been breached. Second, coverage of
third-party losses will continue to be limited, due to firms’ reluctance to admit
that they have been breached.

The Costs of Malicious Cyber
Activity for the U.S. Economy
The losses suffered by the corporate sector as a result of cybersecurity
breaches and attacks extend beyond the direct losses suffered by firms that are
targeted. These additional costs arise from (1) spillover effects to economically
linked firms, (2) the ever-increasing expenses for cybersecurity, and (3) the drag
on economic growth caused by cyber threats. We describe these costs in more
detail in this section.
A cyberattack or data breach experienced by a firm is likely to have
significant spillover effects on its corporate partners, employees, customers,
and firms with a similar business model. As we highlighted in the case of the
Equifax attack, stock prices of firms that have a similar business model and
of firms that rely on Equifax data also declined in response to the news of the
data breach.
Firms also incur nonnegligible costs associated with preventing cyber
incidents, and they must acquire security products (spam filters, antivirus
protection), offer services for consumers (training), and engage in other fraud
detection / tracking efforts (Anderson et al. 2012). The investment bank Morgan
Stanley (2016) estimates that the global IT security product and services market will grow by 18 percent each year between 2015 and 2020, to become a
362 |

Chapter 7

$128 billion market by 2020. We estimate that the Equifax data breach resulted
in significant share price increases for cybersecurity firms. This implies that
market participants revised up their expectations of the cybersecurity firms’
future revenues. We are reluctant to ascribe the cost of cybersecurity protection to a deadweight cost to the U.S. economy. Employment and output in the
cybersecurity sector contribute to economic growth. Innovative technology
solutions developed by the sector may generate positive spillover effects elsewhere in the economy. A sophisticated cybersecurity sector could become a
reliable source of exports for products and services many years to come.
Finally, malicious cyber activity imposes a drag on economic growth
by enabling theft of IP and by slowing the speed of the adoption of new
technologies. Lacking sufficient protection, firms will underinvest in research
and development, slowing the pace of innovation. Additionally, ever-evolving
cyber threats slow down the rate of development and adoption of new types of
information and communications technology, and thereby lower the efficiency
gains that can be achieved with these new technologies (for a detailed discussion and analysis of this and related issues, see Hughes et al. 2017).
When estimating the total economic costs of cybersecurity incidents
against the U.S. economy, one should not overlook the substantial direct
cost imposed on the government sector. Using a data set of cyber incidents
from Advisen, a for-profit organization that collects and resells data from the
commercial insurance industry and public news sources, Romanosky (2016)
estimates that government agencies are at a highest risk for a cyber incident
(risk is defined as the number of cyber incidents divided by the number of
firms/agencies in a sector).
According to the Government Accountability Office (GAO 2017), Federal
agencies reported a 58.9 percent increase in the number of cyber incidents
between fiscal years 2012 and 2015 (the most recent year available). In a
highly publicized incident, between 2014 and 2015, the Office of Personnel
Management (OPM 2015) suffered a system breach, in which security data from
submissions of Form SF-86 were breached for 21.5 million individuals, including 5.6 million sets of fingerprints. Another separate cyber incident involving
personnel records occurred in 2015, which affected 4.2 million individuals
(OPM 2017).
The government incurs substantial, though not easily quantifiable, costs
of IP theft and theft of information pertaining to national security. The case
study of the IP theft for the F-35 fighter plane described in box 7-8 illustrates a
very costly cyber theft from the U.S. government (Capaccio 2017).
Evidence from State and local governments suggests that cyber risks are
also pervasive at these levels. Data breaches or compromises have the potential to affect thousands or even millions of individuals through the release
of personal or sensitive information or disruptions of government service
provision. Responses to a 2013 survey of State and local government officials
Fighting Cybersecurity Threats to the Growing Economy | 363

Box 7-8. The Theft of U.S. Military Secrets
through Cyber Means: the F-35
The F-35 is a single-seat, single-engine fighter aircraft that was developed
primarily by Lockheed Martin to be used by the U.S. armed forces, as well as
allied countries. The plane is optimized for use as a multirole fighter, with the
ability to perform air-to-air; air-to-ground; and intelligence, surveillance, and
reconnaissance missions. Program development officially launched in 2001,
and deliveries began in 2011. The program’s cost to complete is estimated at
more than $400 billion (Wall Street Journal 2014).
It has since been verified that these malicious cyber activities were carried out by foreign agents, with the Chinese national Su Bin pleading guilty in
2016 to stealing data related to the F-35 and seeking financial gain by selling
the illegally acquired data (DOJ 2016c). As noted by Department of Defense
Undersecretary Frank Kendall, these breaches could “give away a substantial
advantage” and “reduce the costs and lead time of our adversaries to doing
their own designs” (DOJ 2016c). This appears to have been the case, because
observers have noted that the J-31, a Chinese stealth fighter introduced in
2014, appears to have been modeled on the F-35 (Weisgerber 2015). If the
Chinese did use designs stolen from U.S. contractors, it could have allowed
them to cut down significantly on the $350 billion spent by the United States
through fiscal year 2017 on the F-35’s development and production (DOD
2015c).

suggested that officials often underestimate the prevalence and potential
severity of adverse cyber events (Center for Digital Government 2014), and
Security Scorecard’s 2016 Cybersecurity Report ranks government (Federal,
State, and local) at the bottom of 18 major industries in terms of cybersecurity
(Security Scorecard 2016). Data on the number of data breaches at government
entities do not show rates of increase that give particular cause for concern,
but trends in the affected numbers of individuals could potentially be quite
different. According to a recent survey of IT and security management professionals in State and local government, 40 percent of respondents indicated
that the number of cyber incidents associated with malware had increased
over the preceding year (Center for Digital Government 2014).
Cyber threats impose significant costs on private individuals. Cyber
intrusions that steal PII from the corporate and government sectors generate
welfare losses for those uninsured individuals whose private information is stolen. Attacks against State and local governments, furthermore, have a negative
impact on households that rely on the services provided by the government
entities. Finally, individuals are frequent direct targets of cybercrimes committed via email and the Internet. The FBI’s Internet Crime Complaint Center
provides the public with a mechanism to report Internet-facilitated criminal

364 |

Chapter 7

activity. In 2016, this center received nearly 300,000 individual complaints
of cybercrimes, with an estimated total cost in excess of $1.3 billion. Among
the most costly crimes targeted at individuals were confidence and romance
frauds. These attacks cost victims $220 million in 2016, and were carried out by
criminals posing as a close family member or romantic partner for the purpose
of convincing victims to send money or personal information. Moreover, the
agency also estimates that only 15 percent of cyber-related criminal activity is
reported each year, so actual damages are likely significantly higher.
It is difficult to estimate how much malicious cyber activity costs the
U.S. economy because, as discussed above, many events go undetected—and
even when they are detected, they are mostly unreported or the final cost is
unknown. After accounting for the negative spillover effects, the CEA (2018)
estimates that breaches and attacks cost the U.S. economy between $57 billion and $109 billion in 2016, which amounted to almost 0.31 percent to 0.58
percent of that year’s GDP. The Center for Strategic and International Studies
(2014) computes the global cost of malicious cyber activity as between $375
billion and $575 billion. The report further estimates that the cost of malicious
cyber activities directed at U.S. entities was $113 billion in 2013, which represented 0.64 percent of GDP that year. Aggregating information from a variety of
industry studies, MIT (2015) comes up with an estimate of a similar magnitude
for the global cost of adverse cyber events, about $400 billion a year.

Devastating Cyberattack Scenarios
Cybersecurity professionals, in both the private and public sectors, stress that
the potential costs of malicious cyber activity could far exceed the ongoing
costs suffered by the U.S. economy. After the worst terrorist attack in U.S.
history, the 9/11 Commission (2004) concluded that the attacks revealed a
failure of imagination—stating that “it is therefore crucial to find a way of
routinizing, even bureaucratizing, the exercise of imagination.” Much effort
is being expended by the cybersecurity community to proactively anticipate
the most devastating vectors for cyberattacks. Government agencies are
particularly concerned about cyberattacks on the 16 critical infrastructure
sectors, described earlier in this chapter. Attacks on these sectors would cause
considerable hardships for U.S citizens, and would create significant spillover
effects to multiple sectors of the U.S. economy. Of these 16 sectors, we focus
in detail on the financial services and energy sectors—more specifically, on the
power grid. These sectors are the most internally interconnected and interdependent with other sectors as well as most robustly connected to the Web, and
are thus at risk for a devastating cyberattacks that would ripple through the
entire economy. In this section, we describe the current concerns and ongoing
efforts to secure these sectors.

Fighting Cybersecurity Threats to the Growing Economy | 365

The Financial Sector
Attacks on the financial sector can reduce confidence in the financial system
and affect a great number of public and private entities, which rely on the
smooth functioning of financial markets and global payment systems for the
supply of capital and the transfer of funds. In recent years, certain aspects
of the global financial system have proven to be vulnerable to cyber threats.
For example, the Bank of Bangladesh reported that over $81 million had been
stolen from its account at the Federal Reserve Bank of New York, and more
than half the stock exchanges worldwide have reported experiencing breaches
and attacks (Anand 2017; Lema 2017). Moreover, in 2011 and 2012, 46 American
entities, primarily in the U.S. financial sector, faced DDoS attacks by Iranian
individuals in Iran-based computer companies that conducted work “on behalf
of the Iranian Government.” The attacks resulted in as much as 140 gigabits
of data per second and hundreds of thousands of customers preventing from
online access to their bank accounts (DOJ 2016b).
A number of attempts have been made to exploit vulnerabilities in cybersecurity infrastructure in order to create desired movements in stock prices. To
be clear, an attack does not need to target the financial sector to have financial
market effects. The majority of these incidences have been small in scale and
directed at specific companies. For example, in 2015 actors posted a fraudulent
story that Twitter was in talks to be acquired for $31 billion. This story, posted
on a website designed to mirror Bloomberg, drove Twitter’s share prices up by
over 8 percent before further investigation revealed that the story and website
were fraudulent.
False news reporting has also moved the broader market. In 2013,
members of the Syrian Electronic Army gained access to the Associated Press’s
official Twitter account, and subsequently tweeted that the President had been
injured in two explosions targeting the White House. This tweet caused the
Standard & Poor’s 500 Index alone to lose $136.5 billion in market capitalization; however within 6 minutes, the losses were erased when the Associated
Press and other sources noted that the tweet was a hoax (Domm 2017). The
three members of the Syrian Electronic Army were ultimately charged with
multiple conspiracies related to computer hacking by the Department of
Justice (DOJ 2016a), with the hack of the Associated Press’s Twitter account
used as evidence.
Cyberattacks on the financial sector could impose substantial costs on
the U.S. economy. If investors could no longer trust that traded securities were
priced efficiently, financial assets would lose their attractiveness as investment
vehicles. In turn, firms would no longer be able to rely on the stock market as a
reliable means for raising capital. As a result, the cost of capital would increase,
reducing economic growth. Investors, having moved into other investment
assets, would likely incur higher costs associated with information gathering,

366 |

Chapter 7

and would lose the benefits associated with liquidity and risk sharing facilitated by well-functioning financial markets.
The Defense Advanced Research Projects Agency (DARPA 2017), a part of
the U.S. Department of Defense, runs a pilot program to identify and help mitigate the risks to the financial sector that could be posed by cyber threat actors.
So far, DARPA has identified several areas of concern. Among them is the risk
of so-called flash crashes, named after the 2010 Flash Crash. To achieve flash
crashes, sell orders can be manipulated to cause a rapid decline in the stock
market index. Though the mispricing corrects quickly, it creates economic
costs for market participants, because wealth is being redistributed across
traders in an arbitrary manner, and it causes investors to lose trust in the stock
market. If flash crashes become a frequent occurrence, high-frequency traders
could be forced to exit the market, potentially leading to lower liquidity levels.
Another area of concern is an attack on the order-matching system,
which would cause a random fraction of trades to be left unmatched and would
result in unwanted exposures to risk factors that the trader tried to hedge with
a combination of long and short positions in securities. Manipulations of data
feeds and news feeds, on which the automated trading systems employed by
institutional traders frequently rely without human input, could pose another
set of challenges to price efficiency. If the intrusions in the data feeds were
small in scale and in scope, they would make it difficult to verify the starting
and ending times of an intrusion in order to eventually certify that the data
feeds are no longer contaminated. DARPA’s efforts focus, among other things,
on constructing simulated trading environments and then attacking these
environments with various attack vectors in order to evaluate which defense
solutions work best.

The Power Grid
A cyberattack on the power grid could have devastating consequences for
firms and private citizens. Lloyd’s of London and the University of Cambridge’s
Centre for Risk Studies lay out a scenario for how hackers could attack power
grids with malware that could lead to large-scale blackouts in the United
States. At the basis of this scenario are real-world examples of attacks on
power grids (Lloyd’s of London 2015). Such examples include the December
2015 and 2016 attacks that cut power in Ukraine. Cybersecurity companies
involved in the investigation of the Ukraine attack found a piece of software
capable of ordering industrial computers to shut down electricity transmission
(Reuters 2017a). As electrical systems become more intelligent, they become
an easier target for cyberattacks and data breaches.
A cyberattack on the electrical grid would have large-scale economic
effects, because infrastructure damage, loss of output, delayed production,
spoiled inventory, and loss of wages all decrease productivity and earnings
for the duration of the blackout. In addition to the economic effects of a

Fighting Cybersecurity Threats to the Growing Economy | 367

large-scale power outage, there are concerns related to health and safety and
national security. DARPA is performing a large-scale study of how to best prevent and mitigate cyberattacks on the power grid. Among other things, DARPA
is building grids that are isolated from the power grid network, and it is using
various attack vectors and other methods of defense to determine the most
effective form of defense against possible cyberattack scenarios.
A cyberattack launched against the electric grid could affect large swaths
of the U.S. economy, because most economic activity is dependent on access
to electricity. Economic analysis conducted by various industry studies estimates that cyberattacks against critical infrastructure assets could cause up to
$1 trillion in damage (Tofan 2016; Lloyd’s of London 2015, 2017). The tail risk
scenarios described in this section indicate that cyberattacks on critical infrastructure sectors could result in escalating cyber costs that eclipse the ongoing
costs of doing business in an interconnected world.

The Rise of Quantum Computing and
the Need for Better Encryption
A final potential threat to the existing cyber infrastructure is the rise of quantum computing and the possibility that a malicious actor may have access to
a powerful quantum computer. Cybersecurity depends to a large extent on
public key encryption technologies, such as the widely used algorithm named
RSA. With RSA encryption, a message is encoded using a public key, and then
decoded using a key known only to the private user. The connection between
the two keys often is a complicated and time consuming math problem, such
as prime number factorization, which could in principle be solved, but may
take hundreds or even thousands of years of computing time to do so.
A traditional digital computer relies on bits, which can take on a value of
0 or 1. Newly developed quantum computers instead rely on quantum bits, or
“qubits,” that are not constrained to a binary nature. A quantum computer can
take advantage of the possibility that a quantum variable, such as the spin of
an electron, can probabilistically occupy more than one state simultaneously
(i.e., be both 1 and 0 at the same time). Different particles can have states
that are correlated with one another, allowing small numbers of correlated
quantum bits to express distributions that would require far more normal
bits to express. In some cases, algorithms, such as that introduced by Shor
(1997), have been developed that utilize these processes to allow the computer
qubits to multitask. This increased processing efficiency could allow quantum
computers to develop solutions to problems in much less time than would be
necessary for a traditional computer.
A world with quantum computing would not necessarily be less secure
than today’s world. Cryptographies that rely on alternative approaches such
as lattices, multivariate polynomial equations, or even those that use current
368 |

Chapter 7

approaches but rely on larger prime factorizations may well be secure in the
quantum computing world.
The problem lies in the transition to that world. If quantum computing
advances faster than anticipated, a large amount of data could be incorrectly
believed to be secure, when in fact it is not. A malicious agent who moves
the fastest in this space could potentially subject the economy to large-scale
security breach. Moreover, the agents who anticipate the eventual emergence
of powerful cyber computers that can decrypt currently safe files may have an
incentive to steal and archive today’s encrypted files so they can be decrypted
in the future. Things that are safe today may not be in retrospect, which raises
a host of pecuniary and nonpecuniary concerns. Thus, the economic costs
of cyberattacks and data breaches may well increase sharply relative to our
current estimates if malicious actors gain an upper hand during the transition
period.
Given such costs, cybersecurity will continue to be a high-priority
national-security issue for many years to come. The private sector and the U.S.
government have a number of ongoing efforts to reduce the cyber risk. We
describe these efforts next.

Approaches to Reducing Cyber Risk
Defending against cyber threats requires building effective and evolving cybersecurity capabilities that span all entities in the U.S. economy, and no single
solution is expected to permanently resolve the cyber problem. Current efforts
across the public and private sectors to address cyber concerns are already
under way. They are a step in the right direction, but the ever-evolving nature
and scope of cyber threats require continued efforts. Effective cyber protection
will require the cooperation of the private and public sectors to report and
mitigate cyber threats.

Public Sector Efforts
As discussed earlier in the chapter, cybersecurity is a common good, and
therefore government involvement in cybersecurity efforts may be beneficial.
The U.S. government is actively facilitating cybersecurity solutions on multiple
fronts. Because no single private entity faces the full costs of the adverse cyber
events, the government can step in to achieve the optimal level of cybersecurity, either through direct involvement in cybersecurity or by incentivizing
private firms to increase cyber protection. When the adversary is as formidable
as a nation-state, the government may be the only defender with the adequate
resources and technology to meet this challenge. Additionally, as a frequent
target of attack, the government sector is already actively involved in cyber
protection. For example, one of the tasks of the Department of Homeland
Security is to protect the dot-gov domain (U.S. Congress 2017b), and a number

Fighting Cybersecurity Threats to the Growing Economy | 369

of other government agencies are tasked with protecting various critical infrastructure sectors against cyber threats. Because the government is able to
achieve economies of scale in its responses to cyber threats, it is cost-efficient
for the government to take an active role in aggregating information, monitoring cyber threats, engaging in defensive action, disseminating knowledge, and
devising effective policies.
Information sharing and transparency. Information sharing is crucial
for coordinating cybersecurity efforts, informing public and private entities
of cyber vulnerabilities, determining appropriate levels of defense investments, and facilitating the effective functioning of the cyber insurance market.
However, private sector firms face a strong disincentive to voluntarily disclose
cyber vulnerabilities because of business and reputational concerns. To overcome these dynamics in the market, the government may facilitate information
sharing through a variety of channels. For example, government-monitored
information-sharing platforms for anonymous disclosures of adverse cyber
events are designed to increase the real-time awareness of cyber vulnerabilities and facilitate timely and publicly shared security solutions. The Automated
Indicator Sharing (AIS) Program at DHS (2016a) facilitates the sharing of “commercial data feeds, internally generated analytic products, analytics tools,
threat indicators and warnings, real time incident, and continuous monitoring
data.” Additionally, the FBI issues Joint Indicator Bulletins, Joint Analysis
Reports, Private Industry Notifications, and FBI Liaison Alert System (FLASH)
reports that inform public and private entities about cyber threats (FBI 2017c).
Other mechanisms to increase transparency about cybersecurity breaches
that were discussed earlier in the chapter are the SEC’s 2011 Guidance, the
HIPPA disclosure requirements and the Department of Energy’s OE-417 Electric
Emergency Incident and Disturbance Report. U.S. Congress has also proposed legislation on cybersecurity disclosure; for example, the Cybersecurity
Disclosure Act of 2017 (proposed by the Senate Committee on Banking,
Housing, and Urban Affairs), which seeks to ensure disclosure on cybersecurity
expertise and measures taken by qualifying firms (U.S. Congress 2017a).
Cyber protection investments. Basic research on cybersecurity generally
underlies investment in cybersecurity. Though this research may benefit from
economies of scale if data and resources are pooled across organizations,
companies generally do not have incentives to share this basic research with
each other, and this may result in duplicative investment efforts across companies. Therefore, direct government investment in this research may be a
way to leverage economies of scale that ultimately benefit private firms across
industries. Firms may then take the responsibility to adapt this research to the
needs and risks of the companies in question. Also, it is often argued that market forces provide firms with little incentive to invest in basic research because
the nontrivial nature of knowledge makes it difficult for firms to appropriate
the resulting returns. Government support for basic research can overcome

370 |

Chapter 7

the lack of incentives and generate critical discoveries that will benefit society
writ large.
Indeed, the Federal government has made investments in cybersecurity
basic research and threat analysis, particularly through DARPA. In fiscal year
2018, DARPA’s budget allocated about 10 percent ($41.2 million) to research in
the cyber sciences, most of which went to the Transparent Computing program,
which seeks to create technologies that will allow for better security policies in
distributed systems, such as distributed surveillance systems, autonomous
systems, and enterprise information systems (DARPA 2017). Such government
investment in basic cyber research benefits from economies and scale and
may reduce private firms’ duplication of research efforts. For example, DARPA’s
basic research investments in unmanned aerial vehicles have spurred innovation in the private aerospace industry (DARPA 2015). According to the Office of
Budget and Management, Federal IT spending, which includes cybersecurity,
has been on an upward trajectory since 2013 (OMB 2017).
The public sector may also incentivize private sector investment in
cybersecurity firms to increase the availability and growth of cybersecurity
products and services. For example, Maryland’s government provides a refundable income tax credit to qualified Maryland cybersecurity companies. These
companies receive a credit of 33 percent of an eligible investment, though the
credit is limited to $250,000 for each investor during each fiscal year (Maryland
Department of Commerce 2017).
Cybersecurity standards. Standards for cybersecurity are also important
ways to ensure that companies are aware of proper cyber practices. Standards
are effective to the extent that they enable a risk-based approach to cybersecurity, which naturally varies across sectors and firms. For example, such cybersecurity standards may create a common lexicon for cybersecurity, including the
definition of what constitutes a cyberattack or a data breach, which currently
is not standardized across government and private organizations. The 2013
U.S. Executive Order 13636 “Improving Critical Infrastructure Cybersecurity”
encourages the U.S. government’s IT agencies to adopt “The Framework for
Improving Critical Infrastructure Cybersecurity,” as developed by NIST, in
order to enhance risk management. For example, the Financial Services Sector
Coordinated Council (FSSCC 2017), which seeks to “strengthen the resilience of
the financial services sector against attacks and other threats,” in collaboration
with the Department of Treasury, has developed an automated cybersecurity
assessment tool. Though standards can be beneficial for addressing cyber
threats, if they are not properly coordinated across government agencies and
are too prescriptive, they could be very costly to implement and thus lead
companies to use a compliance-based rather than risk-based cybersecurity
approach.
The NIST cybersecurity framework is an example of a standards tool that
was originally targeted for critical infrastructure, then adopted by the broader
Fighting Cybersecurity Threats to the Growing Economy | 371

government community (both inside and outside the United States, e.g., in
Italy), and increasingly by the private sector (NIST 2017b). It is a voluntary,
broad-based set of standards that seeks to “identify effects of cybersecurity on business; align and de-conflict cybersecurity requirements; prioritize
cybersecurity outcomes; organize, authorize, task and track work; express risk
disposition; and understand gaps between current and target.” The framework
is made up of five functions: to identify, protect, detect, respond, and recover
from cyber risks. It also establishes a common lexicon to discuss cybersecurity
issues across stakeholders, and it is meant to be adaptable to changing cyber
technologies and threats. All 16 critical infrastructure sectors adopted NIST’s
cybersecurity framework in fiscal year 2016 (U.S. Department of Commerce
2017; NIST 2017a).
Industry-wide cybersecurity standards may also ensure the security of
supply chains by providing a baseline of risk management and other security
mechanisms across firms. NIST has established standards for a risk management process to ensure the cybersecurity of supply chains, known as the Cyber
Supply Chain Risk Management process (NIST 2018).
International efforts. Cyber risks may also result from foreign government
actions or weak cyber defenses across countries, which may be addressed
through international diplomatic and enforcements efforts. For example, the
United States initiated the “Section 301 Investigation of China,” pursuant to
the Trade Act of 1974, to assess Chinese practices, including cyber practices,
which may weigh on U.S. commerce. Also, the United States annually discusses problematic cyber practices that could put IP property in the country
at risk and affect U.S. commerce in the annual Special 301 reports. The World
Trade Organization’s Agreement on Trade-Related Aspects of Intellectual
Property Rights (known as TRIPS) could also be a tool for addressing unfair
cyber practices abroad; however, at this point, it seems to be primarily focused
on other, non-cyber-related forms of IP theft. International bodies—such as
the G-7, G-20, and Financial Stability Board—have also provided forums to
address cybersecurity issues in the financial sector. For example, in October
2017, the G-7 adopted the “Fundamental Elements of Cybersecurity for the
Financial Sector” which sets “non-binding, high-level fundamental elements”
for private and public actors in the financial sector to customize to their specific regulatory landscapes and cyber risks (Department of Treasury 2017). In
the same year, G-20 Finance Ministers and Central Bank Governors released
the “Roadmap for Digitalization: Policies for a Digital Future,” which included
a provision to “strengthen trust in the digital economy,” including through
“exchanging experiences” on “guidelines and best practices to identify, assess,
and manage security risks” (German Federal Ministry for Economic Affairs and
Energy 2017). Finally, the international Financial Stability Board has facilitated
communication between international public and private sector actors on
cybersecurity in the financial sector (FSB 2017).
372 |

Chapter 7

Governments can also pursue bilateral measures. For example, in
September 2015 the U.S. and China signed the nonbinding Cyber Agreement
(referred to as the Xi Agreement), whereby the two countries agreed to (1)
“timely responses” regarding “malicious” cyber activities; (2) cooperation
on cybercrime investigations, and provision of updates, “consistent with
their respective national laws and relevant international obligations”; (3)
ensuring that neither government would “conduct or knowledge support”
cyber-enabled theft of IP, including “trade secrets or other confidential business information for commercial advantage”; (4) promoting norms for nationstates’ cyber behavior within the international community; and (5) creating a
high-level joint dialogue to “fight” cybercrime and related issues (White House
2015; CRS 2015). It has been noted that the number of suspected network compromises by 72 China-based groups in the U.S. and in 25 other countries has
declined since mid-2014, since the Xi agreement, from a peak of over 70 compromises in August 2013 to fewer than 5 in May 2016.9 In his 2017 Worldwide
Threat Assessment, the Director of National Intelligence confirmed that the
volume of cybercrimes committed by China has declined since the September
2015 commitments, but noted that China continues to actively target U.S. firms
and government for cyber espionage (DNI 2017).
Developing a cybersecurity workforce. There is a significant skills gap
in the cybersecurity field, which is reflected in a shortage in the number of
American workers left to fill cyber positions. Almost 210,000 cybersecurity jobs
went unfilled in the United States in 2015 alone (McAfee 2016). Projections
estimate that the global shortage will increase to 1.5 million unfilled positions
by 2020.
Cybersecurity jobs are a subset of IT jobs; however, the number of cyber
job occupations is expected to grow by 28 percent between 2016 and 2026,
“much faster than the average for all occupations.” In comparison, the growth
rate for all computer-related jobs is projected to be 13 percent, while the
growth rate for all other occupations is projected to be only 7 percent during
this same period (BLS 2017).
One possible source of the cybersecurity workforce shortage is that
access is lacking to education in science, technology, engineering, and mathematics (STEM), and particularly to computer science (CS)—a field within
which cybersecurity falls—for schools from kindergarten through grade 12.
For example, more than half these schools do not offer computer programming coursework (Gallup 2016; Code.org 2016), and almost 40 percent of high
9 The 25 other countries and economies, in order of the frequency of incidents, are the United
Kingdom, Japan, Canada, Italy, Switzerland, Germany, the Netherlands, India, Australia,
Denmark, Philippines, Sweden, Taiwan, Brazil, China, Colombia, Colombia, Egypt, France, the
Hong Kong Special Administrative Region, Israel, South Korea, Norway, Saudi Arabia, Singapore,
and Tunisia. Of the 262 compromises that occurred during the 2013–14 period, 182 (69.5 percent)
occurred on U.S. entities’ networks, and 80 (30.5 percent) occurred on networks in the other 25
countries.

Fighting Cybersecurity Threats to the Growing Economy | 373

schools do not offer physics (U.S. Department of Education 2016). Greater
access to STEM and CS programs for younger students would likely increase
the number choosing to pursue these fields at a higher level. This is especially
important for the cybersecurity labor force, given that many of the available
cyber positions require significant educational credentials and experience.
About 84 percent of cybersecurity postings require a bachelor’s degree at a
minimum, and 83 percent ask for at least three years of previous experience
(Burning Glass Technologies 2015). Meanwhile, almost 79 percent of students
in the United States pursuing a master’s degree in CS are citizens of other
countries (NFAP 2017).
These numbers indicate a dependence on foreign workers and foreign
companies to help meet much of the United States’ domestic cybersecurity
needs. An example portraying the necessity of decreasing our dependence
on foreign cybersecurity expertise is the case of Kaspersky Lab, a prominent
Russian cybersecurity research firm founded in 1997 and headquartered in
Moscow (Subcommittee on Oversight 2017). Kaspersky’s antivirus software
has been sold throughout the United States, and was even being used in the
computer systems of some two dozen Federal agencies. As with most security
software, Kaspersky’s antivirus products require access to everything stored on
a computer, which allows it to search for viruses or other malware. By conducting scans for malicious software, the program removes any risks and sends
back a report to the company. Though this is a routine procedure, in 2017,
suspicion grew that the software was in fact providing an all-too-perfect tool
for Russian intelligence to access content of interest on American computers,
especially those utilized by the government.
DHS issued a directive on September 13, 2017, for Federal Executive
Brand departments and agencies ordering them to remove and discontinue
use of Kaspersky products (DHS 2017c). In the press release DHS stated that
“the risk that the Russian government, whether acting on its own or in collaboration with Kaspersky, could capitalize on access provided by Kaspersky
products to compromise Federal information and information systems directly
implicates U.S. national security.” The case of Kaspersky demonstrates the
critical need to increase the domestic supply of cyber workers, and reduce
American dependence on foreign cyber products, which cannot always be
trusted, and instead develop our own cyber expertise.
Meanwhile, another source for the cyber workforce shortage is the lack of
diversity in the field, particularly the underrepresentation of women. The share
of CS degrees awarded to women by higher education institutions has fallen
over the past 30 years. In 1985, the proportion of women earning CS degrees for
all levels of education was approximately 36 percent. By 2015, this share had
dropped to about 22 percent (NCES 2016). The reasons for the decline in female
CS degree enrollment are not well understood, though there are a number of

374 |

Chapter 7

competing explanations (e.g., Roberts 1999; Irani, Kassianidou, and Roberts
2002; Wang et al. 2015).
Corresponding to the low percentage of women studying CS, the share
of women in CS occupations was only 24.5 percent in 2015 (NSF 2017), well
below the rate of 34 percent in 1990 (U.S. Census Bureau 2013). For comparison, in 1990 women represented 45 percent of the labor force, compared with
47 percent in 2015 (DOL 2017). For cybersecurity positions specifically, which
typically require a CS background, a 2017 study conducted by the Center for
Cyber Safety and Education found that the number of women working in the
cybersecurity field in the United States is a mere 14 percent. Other studies suggest that women make up as little as 10 percent of the U.S. cybersecurity labor
force. Increasing the domestic cybersecurity workforce will crucially rely on
attracting more U.S. women to CS coursework—and thus to the cybersecurity
profession.
Although many technology companies already offer STEM-related scholarships to women, the government should continue to promote grants offered
to women studying CS (and cybersecurity) through various avenues, such as
the National Science Foundation. It is equally important to provide female-tofemale mentoring to help encourage women to study CS. By offering structured
opportunities for mentorship, women can better understand the field while
interacting with female leaders and role models (Bohnet 2016).
Additionally, universities need to consider the impact of professors’
genders on the gender gap in the sciences (Carrell, Page, and West 2010; Vilner
and Zur 2006). Though a professor’s gender has little effect on male students, it
can play an important role in the performance of female students in math and
science, and can affect their likelihood of pursuing STEM degrees. Research
suggests that the gender gap in academic performance and STEM majors can
be eliminated, specifically for high-performing female students, when introductory math and science courses are taught by women.
Overall, the Administration is playing a leading role in the STEM movement, attracting both women and men to the cyber field by offering more
opportunities for exposure to STEM concepts earlier in life. The Administration
is already spearheading a movement to promote greater access to computer
science education in elementary and high schools (see White House 2017a),
directing the Department of Education to invest at least $200 million in annual
grants to help fund the expansion of STEM and CS in schools across the country. Additionally, Executive Order 13800: “Strengthening the Cybersecurity of
Federal Networks and Critical Infrastructure” aims to foster the “growth and
sustainment of a workforce skilled in cybersecurity and related fields” in the
public and private sectors, beginning with an assessment of current cybersecurity workforce development programs in the U.S. and in cyber peers.
Raising cybersecurity awareness. Governments may inform consumers
of cyber risks to ensure that demand-side factors internalize cybersecurity
Fighting Cybersecurity Threats to the Growing Economy | 375

risks. For example, governments—at the Federal, State, and local levels—may
also educate consumers about cybersecurity, including current vulnerabilities
and best practices, to ensure that consumers demand secure products and
therefore incentivize businesses to supply such products. For example, DHS
initiated the Stop.Think.Connect Campaign to increase public awareness
of cyber threats—which includes toolkits customized for students, parents,
young professionals, the elderly, government, industry, small business, and
law enforcement—discussing topics such as reporting a cybercrime complaint,
recognizing and reducing cybersecurity risk, online privacy, and phishing.
The campaign also disseminates instructional videos and audio materials on
cybersecurity (DHS 2017a).
Protecting critical infrastructure. Cyber protection is particularly important for critical infrastructure, given the potential for both physical and virtual
damage to systems that may affect many people and organizations at once—
for example, in the case of a disruption in the electricity grid or power plant.
The public sector is particularly important in preventing and addressing such
breaches because of the magnitude of negative externalities that are possible
with such a breach. In line with this, Executive Order 13800, “Strengthening
the Cybersecurity of Federal Networks and Critical Infrastructure,” seeks to
strengthen cybersecurity risk management in critical infrastructure sectors.
Several executive orders over the last several years have addressed protection and coordination concerns about critical infrastructure cyber networks.
The 2013 Presidential Policy Directive-21 (PPD-21), “Critical Infrastructure
Security and Resilience,” notes the above-mentioned 16 critical infrastructure
sectors that are important for both the U.S. economy and national security,
for which cyber protection is particularly essential. That same year, Executive
Order 13636, “Improving Critical Infrastructure Cybersecurity,” expanded the
Enhanced Cybersecurity Services program that enables real-time sharing of
cyber threat information, and ordered NIST to develop a cybersecurity framework. Most recently, in May 2017, Executive Order 13800, “Strengthening the
Cybersecurity of Federal Networks and Critical Infrastructure,” includes efforts
to improve cybersecurity risk management across the government (White
House 2017b). It is too early to assess the effectiveness of these orders, but
their implementation is an important step toward limiting cyber risks.
Law enforcement in cyberspace. Effective law enforcement is critical for
discouraging cybercrime, and its continued success is predicated on coordination among various law enforcement agencies. The FBI Cyber Shield Alliance,
initiated by the FBI’s Cyber Division, engages in partnerships with U.S. State,
local, territorial, and tribal law enforcement agencies to synchronize efforts
against cybercrime. Law enforcement agencies and private entities may report
cyber incidents through the FBI’s online portal system.
Law enforcement has had major successes bringing charges against
criminals in cyber space and helping dismantle their criminal operations,
376 |

Chapter 7

Box 7-9. Law Enforcement’s Role in Mitigating
Cyber Threats: The Kelihos Botnet
The Kelihos botnet was a malicious operation that started in 2010 as a global
network of tens of thousands of infected computers running on Microsoft’s
Windows operating system (DOJ 2017a). The botnet was used to steal the
login credentials of infected users and send hundreds of millions of spam
emails that included malicious software and ransomware (DOJ 2017a). At
its peak, the botnet grew to over 100,000 infected computers, ordering them
to carry out various cybercrimes including password theft, pump-and-dump
stock schemes, and the advertisement of counterfeit drugs.
The DOJ led the effort to free the infected computers from the botnet
(DOJ 2017a). Specifically, the DOJ obtained warrants under Rule 41 of the
Federal Rules of Criminal Procedure, allowing it to establish substitute servers into which to redirect Kelihos-infected computers, and also to collect the
Internet Protocol addresses of the computers that connected to the servers
(DOJ 2017a). This was done to provide these addresses to those who can
assist victims in removing the malicious software from their computers (DOJ
2017a). And this operation also blocked all commands sent in an attempt to
regain control of the victimized computers.
The DOJ, in partnership with a private security firm who provided
technical analysis and aid, provided the legal means necessary for successful
execution. In addition to liberating the already infected devices, the DOJ
pledged to continue to share samples of the Kelihos software with all the
major players in the cybersecurity industry, thereby training the antivirus
software to detect and remove the botnet, should it resurface (DOJ 2017a).
Microsoft’s Safety Scanner is one example of an antivirus software now programmed to do this, thanks to the efforts of the DOJ.

including many that were located abroad. For example, in April 2017, DOJ
played an active role in disrupting the Kelihos botnet and later extradited
him to the U.S. This operation is described in detail in box 7-9. This and other
successful operations demonstrate the important role of law enforcement in
reducing cyber threats and discouraging future cybercrime.

Private Sector Efforts
Although the government can help address some elements of cyber protection,
ultimately, the most direct cybersecurity actions are in the hands of the private
sector. These include direct investments in cyber protection, emergency
preparedness, and information sharing, among others. Together, these efforts
strengthen a firm’s ability to prevent, address, and recover from security
breaches.

Fighting Cybersecurity Threats to the Growing Economy | 377

Investments in cyber protection. Although the public sector may have
a comparative advantage in basic research that depends on economies of
scale, cyber risks for particular industry-specific factors are most efficiently
addressed by private firms because they own and operate most critical infrastructure. One indicator of cybersecurity investment is venture capital funding
for and major industry spending on services from cybersecurity firms. This
funding has recently more than doubled, from $108 billion in 2010 to $336 billion in 2015 (Nasdaq 2016). And Morgan Stanley (2016) estimates that spending
on cybersecurity products and services will again more than double, from $56
billion in 2015 to $128 billion in 2020, though spending on these products will
remain below spending on other IT hardware, software, equipment, and services (figure 7-7). Moore (2016) notes that in a survey of 40 executives, mostly
at the level of chief information security officer, most respondents (88 percent)
reported that their cybersecurity budgets have increased. The survey revealed
that frameworks, compliance obligations, and direct engagement with business units on cyber threats were common ways for executives to gain greater
budgets for cybersecurity. The survey noted that the most frequently cited
response for investment was “perceived risk reduction,” followed by compliance and industry best practices.
Private investments in cyber protection can come in the form of cyber services and technologies. At the service end, Ernst & Young (EY 2014) emphasizes
the importance of strong security operations centers, aligned with business
concerns, that stay informed about impending threats. Such a center could
embody a “cyber threat intelligence capability” addressing questions such as
“What is happening out there that the organization can learn from [the experience]? How can organizations become “target hardened,” and is this required?
How are other organizations dealing with specific threats and attacks? How
can the organization help others deal with these threats and attacks? Which
threat actors are relevant?” (EY 2014). According to EY’s survey, 36 percent of
respondents attested to not having a threat intelligence program, suggesting
that some companies either perceive low cyber threats or are underfunding
cybersecurity.
There are also a variety of security technologies that may be used to
reduce exposure to cybersecurity risk. Ponemon (2017a) notes that the most
common ones in its sample of 254 companies are security intelligence systems
(67 percent), which also have the highest reported costs savings and returns
on investments ($2.8 million and 21.5 percent, respectively); these are followed
by advanced identity and access governance, advanced perimeter controls,
extensive use of data loss prevention, and deployment of encryption technologies, among others (figure 7-8). Morgan Stanley (2016) projects that companies
will move away from “a la carte solutions” to “more-efficient platforms,” resulting in greater consolidation in the cybersecurity industry, with the five largest

378 |

Chapter 7

$"0- 
цҊцѵ
 )1 ./( )/
-*% /$*).
$)
4 -. 0-$/4
/# -
 
#-2- 
)
.*!/2 ' *((0)$/$*).
 ,0$+( )/

/# -
 
. -1$ .
4 -. 0-$/4
+-*0/.ҝ. -1$ .

Dollars (billions)
рѶчпп
рѶхпп
рѶупп
рѶспп

рѶппп
чпп
хпп
упп
спп
п

спрф

спсп

*0- ѷ
*-")
/)' 4
җспрхҘѵ

security vendors growing from 26 percent market share to 40 percent in the
short to medium terms.
The use of distributed ledger technology to ensure data integrity.
Distributed ledger technology (DLT) is an innovative technology-based solution to address cyber threats and provide data integrity. DLT entails having
a database of transactions decentralized across multiple sites in order to
eliminate the need for an intermediary to process, validate or authenticate
peer-to-peer transactions. Blockchain is a well-known example of DLT that
creates a historical record of ledgers that containing every transaction that has
taken place among users.
Third-party institutions, such as banks or credit card companies, historically have helped to validate transactions and establish identities. To establish
identities, third parties require that users divulge significant, confidential information, which is then stored in a centralized database. As discussed above,
centralized PII repositories pose significant risks that the PII data will be stolen
by cybercriminals. The additional contribution of DLT to improved cybersecurity is that it is better able than traditional record keeping to ensure data
availability and integrity by recording transactions in multiple cryptographically secured public ledgers that are verified in large peer-to-peer networks
(Tapscott and Tapscott 2016). The ledgers are distributed around the world

Fighting Cybersecurity Threats to the Growing Economy | 379

$"0- 
цҊчѵ
- ,0 )4
)
*./
1$)".Ѷ
4
 #)*'*"4
Deployment frequency (bottom axis)
Dollars (millions) п

Annual cost savings (top axis)
р

с

т

сф

фп

цф

0/*(/$*)
)
(#$) 
' -)$)"
4 -
)'4/$.
0/*(/ 
+*'$4
()" ( )/
*1 -)) Ѷ
-$.&Ѷ
)
*(+'$)
)-4+/$*)
/ #)*'*"$ .
/
'*..
+- 1 )/$*)
1) 
+ -$( / -
*)/-*'.
1) 
$ )/$/4
ҝ
 ..
"*1 -))
 0-$/4
$)/ ''$" ) 
.4./ (.
Percent п
*0- .ѷ
*) (*)
җспрцҘѸ 
'0'/$*).ѵ

on computer servers supported by volunteers. Therefore, if any location that
holds a copy of the ledger is compromised, other uncompromised ledgers may
be used.
Bitcoin is the most popular form of cryptocurrency that uses its own DLT
protocol called blockchain. Box 7-10 gives an example of how this technology
is implemented.
Better authentication procedures. The relevance and importance of creating and utilizing better authentication methods has intensified in response
of the PII incident on Equifax in 2017. Better authentication procedures may
prevent cyberattacks and data breaches by ensuring that proper personnel
are operating cyber networks. McAfee (2017) notes that “legitimacy tests for
every transaction” may identify improper use of network systems. There are
other ways of enhancing cybersecurity through authentication improvements.
Beyond usage of one-time passwords, individuals and private firms can employ
biometrics or two-factor authentication. Two-factor authentication provides
an additional layer of security and makes it harder for cybercriminals to gain
access to another’s account, because knowing the victim’s password alone is
not enough to pass the authentication check.
In addition, biometric authentication (e.g., using fingerprints or retina
scanners) could enhance security as verification is determined by an individual’s unique characteristics that are extremely difficult to fake. Network

380 |

Chapter 7

Box 7-10. Bitcoin
Bitcoin is a piece of open source software that allows for the creation of a
secure public ledger of transactions that keeps track of how much bitcoin
(the unit of account on the ledger) different users of the system own. At a very
high level, there are two key components that allow the system to function.
First, Bitcoin uses public/private key cryptography to ensure that transfers
of bitcoin recorded on the ledger have in fact been authorized by the owner
of the relevant account (Nakamoto 2008). Second, Bitcoin uses blockchain
technology to achieve and record consensus on the order and legitimacy of
transactions so as to prevent double spending (Nakamoto 2008). These two
mechanisms as implemented through the Bitcoin software seek to allow a
secure decentralized peer to peer digital payment system to function.
The key technological advance underlying Bitcoin stems from its ability
to achieve consensus in a decentralized system. This occurs through a six-step
process that was outlined in the original Bitcoin paper (Nakamoto 2008), and
is given here, with additional explanations added in brackets:
1. New transactions are broadcast to all nodes. [Nodes can be thought
of as individuals or businesses running the Bitcoin software.]
2. Each node collects new transactions into a block. [A block can be
thought of as a collection of transactions.]
3. Each node works on finding a difficult proof-of-work for its block.
[The proof-of-work is the solution to a computationally intensive mathematic
problem that is generated using information embedded in the last accepted
block; see step 6.]
4. When a node finds a proof-of-work, it broadcasts the block to all
nodes.
5. Nodes accept the block only if all transactions in it are valid and not
already spent. [This involves checking the proof of work, verifying the public/
private key cryptography to be sure transactions were authorized by the relevant account holders, and verifying that the sender both previously received
the funds and has not already spent them.]
6. Nodes express their acceptance of the block by working on creating
the next block in the chain, using the hash of the accepted block as the previous hash. [The hash can be thought of as the solution to the computationally
intensive math problem, and by using the solution to the last problem to help
generate the next one, blocks are linked together in a chain.]
The process described above creates consensus on a decentralized
system and de-incentivizes fraud (such as creating multiple accounts or
double-spending) by increasing computationally costs (via the proof-of-work
requirement). Specifically, fraud requires consistently providing proofs of
work faster than the rest of the network, which in turn requires having a
majority of the processing power on the network. Bitcoin incentivizes users
to participate in the network and produce the proofs-of-work that make the
system secure by rewarding users that produce a proof-of-work for a block on

Fighting Cybersecurity Threats to the Growing Economy | 381

the blockchain with newly issued bitcoin. As time passes, the system increasingly relies on transaction fees to incentivize computational work. Either way,
by incentivizing users to do computational work, and by tying the viability of
fraud to the amount of computational work being done on the network, the
system makes it extremely difficult for fraud to take place. This in turn creates
confidence on the part of users that the system and the ledger of transactions
it creates is in fact secure and can be used as a payment system.
The bitcoin protocol specifies that only a finite amount of bitcoin will
ever be issued (21 million), a significant proportion of which have already
been issued (over 16 million). The inherent scarcity of bitcoin is important
as it has helped many people to see bitcoin not only as a unit of account and
medium of exchange in the decentralized payment system described above
but also as a store hold of value. In fact, many see bitcoin as a potential competitor to gold, whereby it serves as an inflation hedge asset that can also be
relied upon in a time of crisis. However, similarly to other investment assets,
especially those with short trading histories and unclear valuation models,
bitcoin may be vulnerable to price bubbles driven by investor sentiment.
Looking forward, bitcoin faces a number of challenges, some stemming
from the underlying technology itself and some from the regulatory environment. On the technological front, the bitcoin protocol over time will have to
adapt to allow more transactions to go through the system more quickly and
with lower average transaction costs. Many competitor crypto-currencies
have sprung up seeking to make technological improvements, and this too
poses a risk to bitcoin. Additionally, the work of transaction validation is
energy-intensive: Böhme and others (2015) estimate that blockchain proof-ofwork calculations require more than 173 megawatts of electricity, equivalent
to about $178 million per year at average U.S. residential electricity prices.
On the regulatory front, a number of regulatory ambiguities also will need to
be addressed and the development of broader market infrastructure such as
exchanges and ETFs will be important.
In addition, the potential use of bitcoin for illicit transactions—such as
money laundering, terrorist financing, tax evasion, and fraud—raises additional regulatory concerns. Some governments like Japan have taken steps
to embrace bitcoin, while others like China have sought to limit its use, while
many more have taken a piecemeal approach to regulating the technology by
applying existing law to entities engaged in regulated activities like running
an exchange. For example, in the United States, bitcoin exchanges have to
register with the Financial Crimes enforcement Network (FinCEN) as a money
services business. Finally, the digital nature of the underlying technology
presents cybersecurity risks for bitcoin (Böhme et al. 2015).

segmentation may also reduce unauthorized access to sensitive information
on networks. Multiple authentication methods—such as a second factor using

382 |

Chapter 7

a hardware token or mobile app, including for vendor access—may help to
prevent cyber breaches across the supply chain.
Facilitating information sharing. As mentioned above, information sharing is critical for raising awareness of rising cyber threats and solutions across
industries. In addition to government-led efforts, industry-led channels may
also enable the dissemination of information across private firms, despite
reputation and competitor concerns. These channels have the potential to
be particularly relevant in addressing industry-specific risks that broader
government-enabled channels may not be able to isolate. For instance, across
sectors, there are now industry-led information sharing and analysis centers
(ISACs), which “collect, analyze, and disseminate actionable threat information” on cyber threats to their members (National Council of ISACs 2017). ISACs
span a variety of sectors—including automotive, aviation, communications,
defense industrial base, defense security information exchange, downstream
natural gas, electricity, emergency management and response, financial services, healthcare, IT, maritime, multistate, national health, oil and natural gas,
real estate, research and education networks, retail cyber intelligence, supply
chain, surface transportation, public transportation and over-the-road buses,
and water.
In general, ISACs have been an important step in promoting information
sharing in the industry, partly overcoming the disincentives to share information on vulnerabilities with competitors. However, there is room for improvement to make the information shared in these units actionable by ensuring
that cyber breaches are shared in real time. Complementary to ISACs, though
not critical infrastructure sector specific, Information Sharing and Analysis
Organizations (ISAOs) are facilitated by DHS pursuant to 2015 Executive Order
13691, “Promoting Private Sector Cybersecurity Information Sharing.” They
are intended to create “transparent best practices” addressing the needs of
all industry groups through an “open-ended public engagement” led by the
Standards Organization. ISAOs may share information with ISACs (DHS 2016b).
Emergency preparedness and risk management. Companies may minimize the costs associated with cyberattacks and data breaches by ensuring
that their organizations have proper response mechanisms in place to recover
from attacks, which requires understanding and managing cyber risks. Risk
management assessments are a critical way to determine whether systems are
protected against cyber threats, and they allow firms to determine whether
their level of investment is proportionate to the cyber threats facing them. A
risk-based approach takes a customized account of a firm’s specific factors—
associated with the supply chain, industry, product, region, and the like—to
determine the level of cyber threat. Such an approach differs from what some
consider a “compliance-based” approach, which involves only following basic
guidelines set by regulators.

Fighting Cybersecurity Threats to the Growing Economy | 383

Once risks are determined, the next step is to ensure that firms are
prepared to address and recover from potential cyber breaches. The Federal
government lays out several guidelines that may inform such emergency preparedness, including checking whether systems are infected through security
scans, disconnecting devices if problems are found, and reporting incidents.
Third-party cybersecurity services may also offer risk and emergency preparation tools—although, ultimately, to be effective such preparedness must happen autonomously at the firm level (FireEye 2017).
Outsourcing cyber protection to domestic cybersecurity firms. Some firms
may choose to hire security companies to manage cybersecurity risks, both
preventing cyberattacks and data breaches and mitigating successful attacks.
For prevention, cybersecurity companies offer risk assessments, red-teaming
exercises, mergers-and-acquisition risk assessments, and security program
assessments, among others. For detection, cybersecurity firms offer incident
response services and compromise assessments (FireEye 2017). Outsourcing
cybersecurity may be especially valuable for small firms, which typically cannot afford to hire a security professional on staff.
Employee training. Training for employees may be a useful preventive
mechanism—for example, training employees on filtering emails and reporting
“phishy” emails, and also deterring shared logins (Verizon 2016). Such types
of training may build general awareness of the cybersecurity risks associated with daily tasks, such as password protection and information sharing.
The NIST Framework includes “awareness and training” in the cybersecurity
framework, which requires ensuring that all users are informed and receive
training on cybersecurity risks and that stakeholders all understand their roles
and responsibilities—including privileged users and third-party stakeholders,
which include suppliers, senior executives, and physical and information security personnel.
Other methods. Another way for private firms to improve cybersecurity
is by increasing the cost of cyberattacks and data breaches to deter future
attacks. One proposal has been to deploy “honey pots” to attract adversaries and distract them from more valuable assets (McAfee 2017). Monitoring
internal networks, devices, and applications also may improve detection and
recovery from future attacks; this may be done through account monitoring,
audit log monitoring, and network / intrusion detection system monitoring
(Verizon 2016).

Conclusion
Cyber threats are likely to remain a reality that has an impact on individuals,
firms, and the government. Cooperation across the public and private sectors
on cybersecurity is ultimately critical as the economy advances to a new era

384 |

Chapter 7

of technology. Therefore, comprehensive approaches that pool the resources
of the private and public sectors are necessary to address the evolving nature
of cyber threats. The public sector may enable the reduction of cyber risks by
supporting basic research, overseeing cybersecurity standards, engaging in
cyber education and awareness, protecting the critical infrastructure sectors,
devising methods to overcome barriers to information sharing, and incentivizing private investment in cybersecurity. Meanwhile, the private sector has a
comparative advantage in devising technology-based solutions, information
sharing, emergency preparedness, and employee training. Effective cybersecurity solutions will contribute to the growth of the U.S. economy.

Fighting Cybersecurity Threats to the Growing Economy | 385

x
Chapter 8

The Year in Review and
the Years Ahead
The U.S. economy experienced a strong and economically notable acceleration
in 2017, with growth in real gross domestic product exceeding expectations
and increasing to 2.5 percent, up from 1.8 percent during the four quarters of
2016, and the unemployment rate falling 0.6 percentage point to 4.1 percent,
the lowest since 2000. Over the course of 2017, the economy added 2.2 million nonfarm jobs, averaging 181,000 per month, with particular strength in
the manufacturing (+189,000 jobs) and mining (+53,000) sectors, which had
lost 9,000 and 98,000 jobs, respectively, in 2016. Challenges remain, however,
because the combination of strong employment growth and modest output
and real earnings growth in recent years reflects low labor productivity growth,
due in part to historically low levels of capital deepening—an issue that the
recently enacted Tax Cuts and Jobs Act was intended to address (see chapter 1
of this Report). In addition, long-term downward trends in labor force participation due to the aging Baby Boom generation will require fresh policy ideas to
offset (see chapter 3).
Acknowledging these challenges, the Administration’s baseline forecast for
the longer term is for output to grow by an overall average annual rate of 2.2
percent through 2028, excluding the effects of the December 2017 Tax Cuts and
Jobs Act. The full, policy-inclusive forecast, however, which assumes implementation of the Administration’s agenda, is for real gross domestic product
to grow by 3.0 percent a year, on average, through 2028. We expect growth
to moderate slightly after 2020, as the capital-to-output ratio approaches its
new, postcorporate tax reform steady state, and as the pro-growth effects

387

of the individual elements of the Tax Cuts and Jobs Act dissipate (though
the level effect remains permanent). The current Administration’s long-run,
policy-inclusive forecast is conservative relative to those of previous administrations, and is in fact slightly below their median of 3.1 percent. Moreover, the
baseline forecast is exactly in line with the long-run outlook given in the 2017
Economic Report of the President, reflecting our view that nonimplementation
of the Administration’s policy objectives would simply result in a reversion
to the lower growth expectations of recent years. But if these objectives are
implemented, the expected contribution to long-run growth of the resulting
deregulation and infrastructure investment will offset the declining contribution to growth of corporate and individual tax cuts and reforms near the closing
of the budget window. On average, through 2028, the Administration expects
deregulation and infrastructure investment to each contribute to GDP growth
beyond 2020 (as discussed in chapters 2 and 4, respectively). Consistent with
growth slowing slightly in the latter half of the budget window, from 3.2 percent
in 2019 to 2.8 percent in 2028, the Administration expects unemployment to
gradually return to its natural level, which will also stabilize inflation.

F

rom 2010 through 2016, the United States’ real gross domestic product (GDP) grew at an average annual rate of 2.1 percent. The pace of
economic recovery was slow by historical standards; indeed, recent
research finds that deep recessions are typically followed by strong recoveries in the United States, and that the correlation is in fact stronger when the
contraction is accompanied by a financial crisis (Bordo and Haubrich 2017).
Since 1882, there have been only three exceptions to this observation—the
Great Depression, the recession of 1990–91, and the aftermath of the Great
Recession of 2007–9.
As discussed in chapter 1, on tax reform, factors in this historically weak
recovery included the coincidence of high and rising global capital mobility
with an increasingly internationally uncompetitive U.S. corporate tax code and
worldwide system of corporate taxation. This combination had the effect of
deterring U.S. domestic capital formation, thereby restraining capital deepening, labor productivity growth, and, ultimately, output and real wage growth.
The five-year, centered-moving-average contribution of capital services per

388 |

Chapter 8

hour worked to labor productivity actually turned negative in 2012 and 2013 for
the first time since World War II. Moreover, as demonstrated in chapters 2 and
3 on, respectively, deregulation and labor, an increased regulatory burden on
businesses and recession-era labor policies had additional negative effects on
productivity growth, particularly through the effect of regulation on firm entry
and exit, and labor force participation among prime-age workers.
During the four quarters of 2017, real GDP grew by 2.5 percent; and annualized quarterly growth rates rose from 1.2 percent in the first quarter of 2017
(hereafter, 2017:Q1, etc.) to, respectively, 3.1, 3.2, and 2.6 percent in 2017:Q2,
2017:Q3, and 2017:Q4. The year 2017 therefore constituted a positive surprise
to expectations. After four-quarter growth of 2.0 and 1.8 percent in 2015 and
2016, expectations for 2017 were subdued. The 2017 Economic Report of the
President anticipated four-quarter real GDP growth of 2.4 percent, while in
January 2017 the Blue Chip panel of professional forecasters predicted fourquarter growth of 2.3 percent (figure 8-1). With 2017 thus surpassing expectations, the January 2017 Blue Chip consensus forecast of 2.3 percent for 2018
was revised upward in February 2018 to 2.7 percent, while “nowcasts” from the
Federal Reserve Banks of New York and Atlanta have projected 2018:Q1 growth
above 3 percent.
Consider the expenditure-side components of GDP in turn: During the
four quarters of 2017, real consumer spending growth of 2.8 percent exceeded
real disposable income growth, such that the personal saving rate fell, which
could in part reflect the effect of a $7.7 trillion rise in household wealth during
the four quarters of 2017, driven primarily by rising equity prices. Business
fixed investment grew 6.3 percent during the four quarters of 2017, up from
only 0.7 percent during 2016, the strongest four-quarter growth rate since
2014:Q3. Residential investment increased 2.3 percent during the four quarters
of 2017, similar to the 2.5 percent growth in 2016; this was a somewhat surprisingly modest rate, given the solid fundamentals: low mortgage interest rates,
rising real income, and rising house prices. Inventory investment—one of the
most volatile components of GDP—subtracted 0.3 percentage point from average growth during the four quarters of 2017, and accounted for much of the
quarterly fluctuations. Government purchases were roughly neutral in their
effect on overall GDP during the four quarters of 2017. Exports contributed 0.6
percentage point to real GDP growth during the four quarters of the year, a
notable increase from the average contribution of –0.1 percentage point in the
years 2015–16, which partly reflected emerging economic growth among the
United States’ trading partners.
Over the course of 2017, the U.S. economy added 2.2 million nonfarm
jobs, averaging 181,000 per month, thereby extending the streak of positive
nonfarm employment growth to 87 consecutive months (or 88 consecutive
months, through January 2018). There was particular strength in the manufacturing (+189,000 jobs) and mining (+53,000) sectors, which lost 9,000 and
The Year in Review and the Years Ahead

| 389

$"0- 
чҊрѵ
1*'0/$*)
*!
'0 
#$+
*). ).0.
*- ./. !*-
 '


-*2/#
0-$)" /# 
*0-
0-/ -.
*!
спрц
)
спрч
спрц
!*- ./

спрч
!*- ./

Percent change (Q4-to-Q4)
сѵч

 Ҋспрч

сѵх

сѵу

сѵс

сѵп
 +Ҋрх

 Ҋрх

-Ҋрц

0)Ҋрц

 +Ҋрц

 Ҋрц

Month forecast published
*0- .ѷ
'0 
#$+
*)*($
 )$/*-.Ѷ +0'$.# 
 +/ ( -
спрх
/*
 -0-4
спрчѵ


98,000 jobs, respectively, in 2016. By the end of 2017, the unemployment rate
had fallen to 4.1 percent, the lowest rate since 2000 and down 0.6 percentage
point since December 2016 (figure 8-2). The unemployment rate for African
Americans was down 1.1 percentage points during the 12 months through
December 2017, to 6.8 percent—the lowest rate recorded in a series that
began in 1972. Meanwhile, after falling from 2008 through 2015 due to factors
including the retirement of Baby Boom generation cohorts, an atypically slow
recovery from the 2007–9 recession, and additional supply-side factors (see,
e.g., Mulligan 2012), in 2016 and 2017 the annual average labor force participation rate edged up slightly as a tightening labor market offset negative demographic trends.
The conjunction of a falling unemployment rate and output growth
below historical averages reflects slow labor productivity growth during this
business cycle’s expansion relative to its long-term average. The slow growth
of labor productivity, along with falling participation rates, have accounted for
a decline in growth of real GDP per capita from a 1953:Q2–2007:Q4 average of
2.1 percent a year to 0.7 percent a year since the business cycle peak in 2007. As
discussed in this chapter and chapter 1, on tax reform, one factor holding down
labor productivity growth has been slow growth in business fixed investment
during the 10 years since the 2007 business cycle peak (1.7 percent at an annual
rate, down from 4.8 percent during the preceding 54 years since the 1953 business cycle peak). Notably, business fixed investment picked up to 6.3 percent
and labor productivity in the nonfarm business sector picked up to 1.1 percent

390 |

Chapter 8

$"0- чҊсѵ
) (+'*4( )/
/ Ѷ
ршцф–спрц
Percentage of labor force
рр

 Ҋспрц

рп
ш
ч
ц

х
ф
у
т
ршцф

ршчп

ршчф

ршшп

ршшф

сппп

сппф

спрп

спрф

*0- ѷ
0- 0
*!
 *-
//$./$.Ѷ
0-- )/
*+0'/$*)
0-1 4ѵ
*/ ѷ
#$)"
 )*/ .
-  ..$*)ѵ

during the four quarters of 2017, and the reduction in the cost of capital by the
Tax Cuts and Jobs Act (TCJA) was specifically conceived to promote increased
investment and capital deepening.
Despite low and declining unemployment, inflation remained low and
stable, with consumer price inflation, as measured by the consumer price index
(CPI), only 2.1 percent over the 12 months through December, the same pace as
a year earlier. Low import prices continue to restrain overall inflation. The core
CPI, which excludes volatile food and energy, increased 1.8 percent over the
12 months through December. Over the same period, personal consumption
expenditure (PCE) inflation increased 1.7 percent, remaining below the Federal
Reserve’s 2 percent target. Real average hourly earnings of nonfarm private
sector employees rose 0.6 percent during the 12 months through December, as
nominal wage growth continued to exceed the subdued pace of price inflation,
building upon the 0.6 percent gain during 2016.
Challenges remain for 2018 and the longer term, including increased
opioid dependence (see chapter 6 on health), the recent low rate of labor
productivity and real wage growth, and downward pressure on the labor force
participation rate from demographic shifts. However, these challenges are not
policy-invariant, as is discussed in chapters 1 and 3, on tax reform and the labor
market, respectively. Capital deepening, a key driver of labor productivity, has
been lackluster in recent years, in large part due to the combination of increasing capital mobility and an internationally uncompetitive corporate tax code,
headwinds that the TCJA was designed to counteract. Meanwhile, though

The Year in Review and the Years Ahead

| 391

demographics are a principal determinant of long-run trends in labor force
participation, this participation is not policy-invariant, and demographics are
not destiny. Policy has affected participation in the past and can do so again.
For example, as demonstrated in chapter 3, policies designed to mitigate the
demand-side effects of rising unemployment during the Great Recession and
other structural factors—such as regional variation in unemployment and
geographic immobility (see chapters 2 and 4, on regulation and infrastructure,
respectively)—have had persistently negative effects on the labor supply of
both prime-age and young adults. Chapter 1 also demonstrates that the supply-side effects of reductions in marginal income tax rates through the TCJA
may be concentrated at the upper end of the age distribution, with important
implications both for labor force participation and labor productivity, as highly
productive, near-retirement workers with higher elasticities of labor supply are
incentivized to defer retirement.
Accordingly, the Administration expects real GDP to grow at 3.1 percent
during the four quarters of 2018, and 2.8 percent in the long term, a forecast
that assumes enactment of the President’s policy proposals. In 2018, consumer
spending and exports are expected to continue to sustain solid growth, with a
lift from a pickup in business fixed investment and a sharp decline in corporate
profit shifting. The unemployment rate is expected to fall slightly from its
projected 2017:Q4 rate of 4.1 percent to 3.7 percent in 2019:Q4. Inflation, as
measured by the price index for GDP, is expected to rise from its projected rate
of 1.6 percent during the four quarters of 2017 to 2.0 percent during the four
quarters of 2021, and to continue at that pace thereafter. Yields on 10-year
Treasury notes are expected to edge up from their projected rate of 2.3 percent
during the four quarters of 2017 to 3.6 percent in the mid-2020s, due partly to
increasing inflation and partly to term premiums returning to more historically
normal levels.

Output
Real GDP grew by 2.5 percent during the four quarters of 2017, somewhat
above its pace in recent years. Real gross domestic output—an average of
GDP and gross domestic income that is generally a more accurate measure
of output than GDP—grew at a similar 2.4 percent annual rate during the first
three quarters of 2017, up from 1.2 percent during the four quarters of 2016.
The overall composition of demand during the four quarters of 2017 shows that
most of the growth was accounted for by strong increases in consumer spending and business fixed investment. These were partially offset by a decline in
inventory investment, while contributions from net exports and other sectors
were generally small.
Real GDP accelerated to a 2.5 percent rate of growth during the four
quarters of 2017, from a 1.8 percent pace during 2016. The 2017 acceleration

392 |

Chapter 8

$"0- 
чҊтѵ
 '
-*2/#
$)
Ѷ
-$1/ 
*( ./$
$)'
0-#. .

җҘѶ
)
-*..
*( ./$
0/+0/
җҘѶ
сппп–спрц
Four-quarter percent change
х



җспрцѷуҘ

у



җспрцѷтҘ

с
п
Ҋс
Ҋу
Ҋх
сппп



җспрцѷуҘ
сппс

сппу

сппх

сппч

спрп

спрс

спру

спрх

*0- ѷ
0- 0
*!
*)*($
)'4.$.Ѷ
/$*)'
 )*( 
)
-*0/
*0)/.ѵ
*/ ѷ
#$)"
 )*/ .
-  ..$*)ѵ

$.
/# 
1 -" 
*!

)
"-*..
*( ./$
$)*( ѵ


was largely accounted for by business investment and exports. Within business
investment, the step-up in growth was due to equipment investment, which
grew 8.8 percent (contributing 0.5 percentage point to overall real GDP growth)
during 2017, after shrinking 3.7 percent a year earlier. The faster pace of export
growth during 2017 reflects the emerging strength among our trading partners,
as discussed below. Real consumer spending grew 2.8 percent during the four
quarters of 2017, the same pace as 2016.
The aggregate measure of consumption and private fixed investment,
known as private domestic final purchases (PDFP), rose 3.3 percent during the
four quarters of 2017 (faster than overall output), up from 2.5 percent in 2016
(figure 8-3). Real PDFP growth is typically a better predictor of future real GDP
growth than is real GDP growth itself (possibly because it aggregates the bestmeasured components of demand). The strength of PDFP suggests somewhat
better growth prospects for 2018 than might be inferred from GDP alone.

Consumer Spending
Consumer spending, which constitutes 69 percent of GDP, was the major
demand-side contributor to real GDP growth during 2017. Real consumer
spending grew faster than disposable income (1.8 percent), so the personal
saving rate fell from 3.6 percent in 2016:Q4 (and an average 4.9 percent during 2016) to 2.6 percent by 2017:Q4 (and an average 3.4 percent during 2017)
(figure 8-4). Real consumer spending outpaced real income, probably because
of large increases in household net worth attributable to the substantial gains

The Year in Review and the Years Ahead

| 393

$"0- 
чҊуѵ
 -.*)'
1$)"
/ Ѷ
ршшп–спрц
Percentage of disposable personal income
рп

спрцѷу

ч

х

у

с

п
ршшп

ршшф

сппп

сппф

спрп

спрф

*0- ѷ
0- 0
*!
*)*($
)'4.$.Ѷ
 -.*)'
 )*( 
)
 /.
$.+*.$/$*)ѵ
*/ ѷ
#$)"
 )*/ .
-  ..$*)ѵ


in the stock market, continued increases in housing wealth, and a stable
household debt-to-income ratio. Low interest rates and improving access
to credit also supported consumer spending. In general, real consumption
growth and the wages-and-salaries component of real income growth tend
to track one another well, as was the case in 2017 (figure 8-5). The low saving
rate is consistent with the elevated level of employment expectations, which
in 2017:Q4 was at the 87th percentile of its historical distribution (according to
data from the University of Michigan’s Surveys of Consumers), indicating that
consumers are optimistic about the economic future and thus are not engaging
in precautionary “buffer stock” savings (Carroll 1997).
During the past four quarters, growth was strong for real household
purchases of durable goods (7.4 percent) and nondurable goods (3.2 percent),
while services purchases grew moderately (1.9 percent). Light motor vehicles
sold at an annual rate of 17.1 million units during the 12 months of 2017, down
slightly from the 17.5 million units sold during 2016, which was the strongest
selling pace on record. Domestic automakers assembled 10.8 million light
vehicles in 2017, down from 11.8 million units, on average, during the two
preceding years.
Consumer sentiment increased in 2017, due in part to a strong labor
market, low inflation, and the stock market highs (figure 8-6). In 2017:Q4,
the two major indices of consumer sentiment—the University of Michigan’s
Index of Consumer Sentiment, and the Conference Board’s Index of Consumer
Confidence—reached their highest quarterly averages since 2000:Q4.

394 |

Chapter 8

$"0- 
чҊфѵ
 '
*).0( -
+ )$)"
)
 '
*(+ )./$*)Ѷ
ршцф–спрц
Four-quarter percent change
ч

 '
+ -.*)'

*).0(+/$*)

3+ )$/0- .

х

спрцѷу

у
с
п

 '
2" .
)

.'-$ .Ѷ

*( ./$

(+'*4 -.

Ҋс
Ҋу
Ҋх
ршцф

ршчп

ршчф

ршшп

ршшф

сппп

сппф

спрп

спрф

*0- .ѷ
0- 0
*!
*)*($
)'4.$.Ѷ
/$*)'
 )*( 
)
-*0/
*0)/.Ѹ

'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .
-  ..$*)ѵ
" .
)
.'-$ .
*!
*( ./$
 (+'*4 -.
- 
 !'/ 
0.$)"

/# 
+ -.*)'
*).0(+/$*)
 3+ )$/0- 
#$)
+-$ 
$) 3ѵ


$"0- чҊхѵ
*).0( -
 )/$( )/Ѷ
ршчп–спрц
Index (1985 = 100)
рфп
рсф

спрцѷу

*).0( -
 )/$( )/
 ) 3Ѷ

)$1 -.$/4
*!
$#$")


рпп
цф
фп

*).0( -
*)!$ ) 
 ) 3Ѷ

*)! - ) 
*-

сф
п
ршчп

ршчф

ршшп

ршшф

сппп

сппф

спрп

спрф

*0- .ѷ
)$1 -.$/4
*!
$#$")Ѹ
*)! - ) 
*-Ѹ

'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .
-  ..$*)ѵ

The Year in Review and the Years Ahead

| 395

$"0- 
чҊцѵ
 *0. #*'
 /
 '/$1 
/*
$.+*.' 
 -.*)'
 )*( 

җ ҘѶ
ршшф–спрц

Years of DPI
рѵф

спрцѷт

*/'
#*0. #*'
 /

рѵс

пѵш

*-/"" 
 /
пѵх

*).0( -
- $/

пѵт

пѵп
ршшф

сппп

сппф

спрп

спрф

*0- .ѷ
*-
*!
*1 -)*-.
*!
/# 
  -'
 . -1 
4./ (Ѷ
$))$'
*0)/.
*!
/# 
)$/ 

// .
') 
# /.Ѹ
0- 0
*!
*)*($
)'4.$.Ѷ
 -.*)'
 )*( Ѹ

'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .
-  ..$*)ѵ

Meanwhile, U.S. household debt relative to income was stable during the
first three quarters of 2017 (figure 8-7). Before the financial crisis, household
debt relative to income rose dramatically, largely due to rising net mortgage
originations, and then declined sharply after the crisis through 2016, a pattern
known as “deleveraging,” which entailed declines in new mortgage originations and less consumer borrowing. Charge-offs of delinquent mortgage debt
also played a role (Vidangos 2015). The level of mortgage debt relative to
income edged down during the first three quarters of 2017, while the level of
consumer credit (including credit cards, automobiles, and student loans) relative to income increased slightly.
Moreover, with low interest rates, the amount of income required to
service these debts has fallen dramatically. Still, it needs to be noted that
estimates based on aggregate data could mask higher debt burdens for some
families, because the health of personal finances varies substantially among
households. Nonetheless, in the aggregate, deleveraging has brought levels of
debt relative to income and ratios of debt service to income back to the comfortable proportions of the early 2000s.
Gains in nominal household net worth (i.e., assets less debts, also
referred to as household wealth) during 2017 were accounted for by increases
in equity wealth of about 19 percent and in real estate wealth of roughly 9
percent. These gains in wealth were only slightly offset by increases in liabilities that were in line with income. By the end of 2017, household wealth had
reached a value equivalent to seven years of income, the highest ratio of wealth

396 |

Chapter 8

$"0- 
чҊчѵ
*).0(+/$*)
)
 '/#
 '/$1 
/*
$.+*.' 

 -.*)'
 )*( 
җ ҘѶ
ршфс–спрц
Ratio to annual DPI
рѵрф

Years of DPI

спрцѷу

ч

*/'
) / 2 '/#
җ-$"#/ 3$.Ҙ
рѵпф

пѵшф

х

*).0(+/$*)
җ' !/
3$.Ҙ

у

 / #*0.$)"
2 '/#
җ-$"#/
3$.Ҙ

пѵчф

пѵцф
ршфп

/*&
(-& /
2 '/#
җ-$"#/
3$.Ҙ

с

п
ршхп

ршцп

ршчп

ршшп

сппп

спрп

*0- .ѷ
*-
*!
*1 -)*-.
*!
/# 
  -'
 . -1 
4./ (Ѷ
$))$'
*0)/.
*!
/# 
)$/ 

// .
') 
# /.Ѹ
0- 0 *!
*)*($
)'4.$.Ѷ
 -.*)'
 )*( Ѹ

'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .
-  ..$*)ѵ
'0 .
$(+0/ 
!*-
спрцѷу
4
ѵ


to income since records began to be kept in 1947 (figure 8-8). These gains in
wealth supported consumer spending growth in 2017 because the consumption share of income tends to increase along with increases in net worth.
Because the saving rate is the inverse of the consumption-to-income ratio, high
levels of the wealth-to-income ratio can support a low saving rate, such as the
2.6 percent rate in 2017:Q4.
Three caveats need to be added about this correlation between net
worth and the consumption rate. First, changes in net worth have been spread
unevenly across households, and these disparities may have implications
for families and macroeconomic activity. Second, in the Federal Reserve’s
Financial Accounts of the United States, the household sector includes the
endowments of nonprofit institutions. And third, stock market wealth is volatile, so this picture could change quickly.

Investment
The main types of investment include business fixed investment, residential
investment, and inventory investment. This subsection considers each in turn.
Business fixed investment. Business fixed investment was a bright spot
in 2017, after two years of little growth. It grew by 6.3 percent during the four
quarters of 2017, an acceleration from an annual rate of 0.5 percent during the
years 2015–16. And its three major subcomponents also grew during the four
quarters of 2017 (figure 8-9): Equipment investment grew by 8.8 percent, up
from a 3.7 percent contraction in 2016; intellectual property products grew by
The Year in Review and the Years Ahead

| 397

$"0- 
чҊшѵ
*(+*.$/$*)
*!
-*2/#
$)
 '
0.$) ..
$3 

)1 ./( )/Ѷ
спрп–рц
Percent change or percentage points (annual rate)

рп

*/'
0.$) ..
!$3 

$)1 ./( )/
җ‫ڿ‬Ҙ

ч

)/ '' /0'
+-*+ -/4

+-*0/.
җ+ѵ+ѵҘ

х

/-0/0- .
җ+ѵ+ѵҘ

у

,0$+( )/
җ+ѵ+ѵҘ
с
п
Ҋс

спрп–рр

спрс–ру

спрф–рх

спрц

*0- .ѷ
0- 0
*!
*)*($
)'4.$.Ѷ
/$*)'
 )*( 
)
-*0/
*0)/.Ѹ

'0'/$*).ѵ
*/ ѷ
*(+*) )/.
(4
)*/
.0(
/*
/*/'
0 
/*
-*0)$)"ѵ
-*2/# -/ 
*(+0/ 
0.$)"

уҊ/*Ҋу
#)" .ѵ

4.8 percent, down slightly from 5.2 percent growth in 2016; and nonresidential
structures grew by 3.7 percent, similar to 3.5 percent growth in 2016.
Net investment (gross investment less depreciation) is required to
increase the capital stock. It is a matter of concern that net investment has
been generally falling as a share of the capital stock during the past 10 years,
which limits the economy’s productive capacity. In 2016, net investment as a
share of the capital stock fell to a level previously seen only during recessions
(figure 8-10). The Tax Cuts and Jobs Act is designed to increase the pace of net
investment. As discussed further below, the slowdown in investment has also
exacerbated the slowdown in labor productivity growth.
The rate of payouts to shareholders by nonfinancial firms, in the form
of dividends together with net share buybacks, has been gradually trending
higher for several decades, although it fell in 2017 (figure 8-11). Nonfinancial
corporations returned nearly half the funds that could be used for investment
to stockholders in 2017. In a well-functioning capital market, when mature
firms do not have good investment opportunities, they should return funds to
their stockholders, so the stockholders can invest these funds in young and
growing firms. Although it may be admirable for individual firms to thus return
funds to their shareholders, the rising share of paybacks to shareholders suggests that investable funds are not being adequately recycled to young and
dynamic firms. Gutiérrez and Philippon (2016) find that firms in industries with
more concentration and more common ownership invest less.

398 |

Chapter 8

$"0- 
чҊрпѵ
 /
 )1 ./( )/
.

#- 
*!
/# 
+$/'
/*&Ѷ -$1/ 

*)- .$ )/$'Ѷ
ршуф–спрх
Percent
х

спрх

ф

у

ршуф–спрх
1 -"

т
с
р
п

ршуф

ршфф

ршхф

ршцф

ршчф

ршшф

сппф

спрф

*0- .ѷ
0- 0
*!
*)*($
)'4.$.Ѷ
$3 
.. /. *0)/.
/' .
рѵрѶ
рѵтѶ
)
рѵфѸ


'0'/$*).ѵ
*/ ѷ
 / )*($)'
$)1 ./( )/
.
.#- 
*!
/# 
0-- )/Ҋ*./
+$/'
./*&ѵ


$"0- 
чҊррѵ
*/'
4*0/.
/*
#- #*' -.Ѷ
*)!$))$'
*-+*-/ 

0.$) .. .Ѷ
ршфс–спрц
Percentage of operating cash flow
цф

спрцѷт

хп
уф
тп

рф
п
ршфп

ршхп

ршцп

ршчп

ршшп

сппп

спрп

*0- .ѷ
*-
*!
*1 -)*-.
*!
/# 
  -'
 . -1 
4./ (Ѷ
$))$'
*0)/.
*!
/# 
)$/ 

// .Ѹ

'0'/$*).ѵ
*/ ѷ
#$)"
 )*/ .
-  ..$*)ѵ
 -$ .
.#*2)
$.
/# 
!*0-Ҋ,0-/ -
(*1$)"
1 -" 
*!
/# 
-/$*

*!
$1$ ).
+'0.
.#- 
04&.
- '/$1 
/*
+-*!$/.
+'0.
 +- $/$*)
($)0.
/3 .ѵ


The Year in Review and the Years Ahead

| 399

Residential investment. Residential investment was one of