View original document

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

Productivity Growth in the Advanced Economies:
The Past, the Present, and Lessons for the Future
Remarks by Jason Furman
Chairman, Council of Economic Advisers
Peterson Institute for International Economics
July 9, 2015
As prepared for delivery
It is great to be here with my close colleagues to discuss productivity growth—among the more
vexing challenges that economists and policymakers face today. Indeed, productivity poses
perhaps the largest disparity between its substantive importance and how much we understand it.
Productivity growth is central to a range of economic questions from the slowdown in middleclass incomes in recent decades to the outperformance of employment over output in the current
recovery. Looking forward, productivity growth is essential to understanding how quickly wages
can grow, how fast the economy can grow, and the magnitude—and potentially even the
existence—of a long-term fiscal gap.
Our understanding has often been clouded, not just by the inherent difficulties of the subject, but
also by the smoke thrown up by the often heated debates about its future. The steps we take
toward a fuller understanding of productivity growth will necessarily be incremental and will
require as much patience as analysis. The OECD’s recent work on “The Future of Productivity”
is an important step. It is one entry to the recent debate that sheds more light than heat. With a
study of firm-level data that permits analysis of productivity differences within industries, the
OECD has moved us closer toward understanding the path forward.
Today, I will reflect on the evolution of productivity in the United States and globally—focusing
on its volatility through the postwar era and its recent slowdown. Both time periods are relevant
to productivity’s importance to aggregate output, middle-class incomes, and the trajectory of our
fiscal position. While projecting the future of productivity is too ambitious for a twenty minute
speech—or even a twenty hour speech—I am encouraged by a number of lessons we might draw
from the past and present.
First, the slowdown in productivity growth across the advanced economies in the wake of the
crisis offers very little predictive information about the future. Second, the slowdown has been
driven by reduced post-crisis investment, a less persistent factor than the sort of actual
innovation shifts that have affected productivity in the past. And third, evidence from the
technology sector suggests some reasons for optimism about innovation in the United States and
around the world—although I would caution anyone against too much confidence with respect to
how quickly we will develop new ideas in the future. Finally, I will review aspects of the
President’s agenda most relevant to productivity growth. Much of my discussion will emphasize
the United States, but I will try to put it in a broader global context because, as the OECD has
emphasized, the challenges of productivity are in large part common to the advanced economies.

The Mysteries of Productivity
Productivity is important, but it is also mysterious—on three levels. In the first place, labor
productivity data are notoriously volatile, as shown in Figure 1. This is intuitive and is to be
expected, in part because of the compounding of measurement error in both the numerator and
the denominator.
But it means that very little information is contained in each quarterly productivity release—
denying economists like us our usual pleasure of analyzing high-frequency data. Indeed, it takes
years for clear long-term trends in productivity growth to emerge. Even though the longer-term
trends do not suffer from the same volatility as quarterly data, they too can be affected by more
systematic measurement error over time as statisticians struggle to keep up with the innovations
associated with quality improvements, new products, and goods with low or zero marginal cost.
Figure 1

U.S. Labor Productivity Growth, 1950‐Present
Percent, Annual Rate
15
5‐Year
Moving
Average

10

5

0
Quarterly
Growth

‐5

‐10
1950

1960

1970

1980

1990

2000

2015:Q1

2010

A second level of mystery is that total factor productivity (TFP)—an important contributor to
labor productivity, which I will discuss more in a moment—cannot be directly observed. It is
measured as the residual between output and the weighted growth of inputs, and is subject to
considerable measurement error.
The third level of mystery is explaining the conceptual drivers of productivity growth. Even if
we agreed on the facts of historical productivity growth, explaining those facts is more difficult
still. Moses Abramovitz famously called TFP a “measure of our ignorance,” the unexplained gap
between input and output.1 And a rigorous conceptual understanding of that gap continues to
elude economists. Intuitively, TFP tells us how efficiently and intensely inputs are used. This is
easily mapped to innovation of the technological and managerial sorts. But periods of outsized

1

Moses Abramovitz. 1956. “Resource and Output Trends in the United States Since 1870.” NBER Occasional Paper
52. Available at http://www.nber.org/chapters/c5650.pdf.

2

productivity growth do not necessarily align with periods normally associated with technological
innovation.
And if understanding productivity’s past or present were not challenge enough, understanding its
likely future path poses a staggering challenge for researchers and policymakers, a challenge that
sometimes exceeds the degree of humility that some have brought to question. Personally, I find
the more I think about productivity growth the more questions I have.
Productivity in the Past: A Driver of Middle-Class Incomes
As shown in Figure 2, the evolution of labor productivity growth in the United States since
World War II can be partitioned into three regimes.
Figure 2

U.S. Labor Productivity Growth

15‐Year Centered Moving Average of Annual Percentage Growth
3.5
1948–1973:
2.9 percent per year
3.0
1995–2014:
2.2 percent per year
2.5

2.0

15‐year centered
moving average

1.5
1973–1995:
1.5 percent per year
1.0
1950

1960

1970

1980

1990

2000

2010

Labor productivity in the private nonfarm business sector rose by an average of 2.9 percent per
year between 1948 and 1973.2 Beginning in the earlier 1970s, though, productivity slowed
sharply, averaging only 1.5 percent growth between 1973 and 1995. Several factors can help
explain the downshift. First, growth in the immediate post-war era benefited from the
commercialization of numerous innovations made during World War II, including the jet engine.
The early 1970s marked the point at which the wartime innovations became exhausted. Public
investment also slowed, and the 1970s oil shocks and collapse of the Bretton Woods system
caused dislocations that weighed on growth.

2
Figure 2—and all subsequent references to annual U.S. labor productivity in these remarks—references real output
per hour worked in the private nonfarm business sector (excluding government enterprises) as reported by the
Bureau of Labor Statistics (BLS). In other contexts, I have referenced the BLS’ labor productivity series for the
nonfarm business sector (including government enterprises). The two series are closely correlated and exhibit the
same trends, but excluding government enterprises permits the analysis of total factor productivity (TFP) that
follows.

3

Productivity growth did not rebound meaningfully until the mid-1990s, when the new economy
improved the production and use of information technology at a startling rate. Productivity
growth surged, rising 2.2 percent at an annual rate between 1995 and 2014. I will return to the
question of whether the recent period should be considered part of the post-1995 resurgence or a
new regime of slower productivity growth.
The international story is somewhat different. In the immediate postwar period, our G7 partners
generally experienced much faster productivity growth than the United States as they rebuilt
from the devastation of the war and moved closer to the American-led technological frontier.
While all of our G7 partners experienced the productivity slowdown that began in the 1970s,
most did not see the 1990s rebound. As shown in Figure 3, the largest European nations have
seen a relatively consistent decline since the 1970s. The somewhat more positive situation in the
United States likely reflected the U.S. concentration of high-tech innovation over the past twenty
years.
Figure 3

Labor Productivity Growth, 1950‐2014
15‐Year Centered Moving Average of Annual Percentage Growth
9

France
Italy
United Kingdom
Canada
Japan
United States
Germany

8
7
6
5
4
3
2
1
0
1950

1960

1970

1980

1990

2000

2010

Productivity is also central to a much longer-term challenge facing our economy and many of
yours in the OECD: the multi-decade stagnation in middle-class incomes. As shown in Figure 4,
real median disposable income in the United States was lower in 2012 than it was in 2000,
suffering from a longer-term trend of slow growth compounded by the huge recession. The story
varies across the G7 but the U.S. experience is broadly in the middle of the range.

4

Figure 4

Real Median Disposable Income
Index (1995=100)
140
United Kingdom
130
Canada
120
France
110

United
States

Germany

100

Italy

90
Japan
80
1995

2000

2005

2010

In fact, a simple thought experiment provides a sense of how important productivity is to
incomes: what if productivity growth from 1973 to 2013 had continued at its pace from the
previous 25 years? In this scenario, incomes would have been 58 percent higher in 2013. If these
gains were distributed proportionately in 2013, the median household would have had an
additional $30,000 in income. Had income inequality and labor force participation not worsened
markedly, middle-class incomes would be nearly twice as high.3
Productivity in the Present: An Investment-Driven Slowdown
I will now turn to the present and toward the global concerns over productivity growth in the
wake of the global financial crisis. Many economists have noted the appearance of a disconnect
between labor market performance and economic output in 2015 so far, as employment growth
in the first quarter remained at its 2014 pace while GDP was markedly lower. If output
consistently rises more slowly than aggregate hours worked, productivity is falling by definition.
No one is suggesting that we face an environment of persistently negative productivity growth,
of course, and the well-known volatility of productivity in the short-term thankfully insulates
economists from having to contend with that question on a regular basis.
But the recent disconnect between employment and output is only a more extreme version of the
disconnect we have seen in recent years. Since mid-2010, output has increased at a 2.1 percent
annual rate, which is below what most economic forecasters expected, while the unemployment
rate has fallen by an average of 0.8 percentage point at an annual rate, which is faster than most
forecasters expected. These two facts are reconciled by the fact that labor productivity has grown
only 0.7 percent per year since 2010, well below the 2.3 percent average from 1948 to 2007.

3

For more detail on these thought experiments, see Chapter 1 of the 2015 Economic Report of the President.

5

Moreover, this is not just an American phenomenon—in fact it has been even more extreme in a
number of other OECD countries. For example, the United Kingdom has had comparable
employment growth combined with slower output growth than the United States—a fact that is
manifested in the negative productivity growth experienced by the United Kingdom since 2011.
As shown in Figure 5, other G7 economies have also experienced similarly low productivity
growth recently—in fact, the United States is in the middle of the pack.
Figure 5

Labor Productivity Growth in the G‐7
Percent Change, Annual Rate
5

4.5

1950‐2007
2010‐2014

4.0
4

3.5

3.2

3
2

1.1
1

2.5

2.2

1.9

0.8

0.7

0.6

0.4

0
0.0

0.0

‐1
Canada Germany United
States

France

Japan

United
Kingdom

Italy

Some have suggested that the disconnect between output growth and employment growth is an
artifact of the data, claiming that the official statistics reliably record employment growth but
miss much of the output growth that is accounted for by quality improvements, new goods, and
goods with low marginal cost but high consumer surplus.4 This is an important issue that merits
considerable additional work. There is certainly reason to believe that the true growth rate
exceeds what is shown in the official statistics. But there is also reason to believe this was true in
the past as well so that it is implausible that any bias in the measurement of GDP has grown so
quickly that it can fully account for the slowdown in reported productivity growth, although it
may explain part of the slowdown.
Explaining the post-crisis productivity slowdown can be informed by a bottom-up decomposition
of the factors contributing to productivity growth. Because labor productivity reflects the amount
of output generated per unit of only one input to production, it is positively related to increased
quantities of other inputs. So when the capital stock rises, output and therefore labor productivity
also tend to rise. Accordingly, one can decompose growth in labor productivity into three
components: growth in investment per hour worked (or “capital deepening”), the quality of labor
writ large, and total factor productivity (TFP).
4

See, for example, Martin Feldstein. 2015. “The U.S. Underestimates Growth. Wall Street Journal. (May 18);
Goldman Sachs. 2015. “U.S. Economics Analyst: 15/21 – Productivity Paradox v2.0.” (May 23); David M. Byrne,
Stephen D. Oliver, & Daniel E. Sichel. 2015. “How Fast are Semiconductor Prices Falling?” NBER Working Paper
21074 (April).

6

Figure 6 shows this decomposition for the United States applied to labor productivity growth in
each of the three historical periods I considered earlier. Notably, increases in capital intensity and
in labor quality have been roughly constant, on average, across the three periods. Virtually all the
variation in labor productivity growth is accounted for by variation in TFP.
Figure 6

Sources of Productivity Growth Over Selected Periods
Percentage Points, Annual Rate
3.5
3.0

Total Factor Productivity
2.9

Capital Intensity

2.5
2.0

Labor Composition

2.2

2.2

1.1

1.2

0.9

0.9

1.9
1.5

1.5

0.4

1.0
0.5

0.9

0.8

0.0

0.2

0.2

0.3

0.2

1948‐1973

1973‐1995

1995‐2014

1948‐2014

Since the crisis, the story has been quite different. Total factor productivity growth has been
lower since the recovery began, rising 0.6 percent per year since 2010 compared with an average
of 1.2 percent per year between 1948 and 2007. But that difference only accounts for a fraction
of the lower post-crisis growth in labor productivity, as shown in Figure 7, with most of the
slowdown explained by a reduction in the contribution of capital deepening. Capital intensity
actually declined between 2010 and 2014, despite growing at a consistently positive rate since
the crisis.
Figure 7

Sources of Productivity Growth, 1948‐2007 vs. 2010‐2014

Percentage Points, Annual Rate
3.0
2.5

2.3
0.2

2.0
1.5

Labor
Composition

1.0

Capital
Intensity

0.5

Total Factor
Productivity

0.9

1.2

0.7
0.2
0.6

0.0

‐0.2

‐0.5
1948‐2007

2010‐2014

7

The story is similar across some—but not all—of the other G-7 economies. In the United States,
Canada, Germany, and Japan, more than half the slowdown in productivity growth between the
pre-crisis period and the recovery is accounted for by reduced capital deepening, as shown in
Figure 8a-c. But in the United Kingdom, France, and Italy, most of the decline is attributable to
lower TFP, as shown in Figure 8d-f.
Figure 8
b) Sources of Productivity Growth: Germany

a) Sources of Productivity Growth: Canada

Percentage Points, Annual Rate
2.5

Percentage Points, Annual Rate
1.5

2.1

1.2

2.0
0.7

0.9

1.0
0.6

0.5
0.6

1.5

Capital
Intensity
Total Factor
Productivity
& Labor
Composition

1.0

0.6

0.5

0.0
1985‐2007

2.2

2.0

2.5

1.0
1.2

2010‐2013

Percentage Points, Annual Rate
2.5

2.5

1.3

0.9

d) Sources of Productivity Growth: United Kingdom

Percentage Points, Annual Rate
3.0

1.5

Total Factor
Productivity
& Labor
Composition

1985‐2007

2010‐2013

c) Sources of Productivity Growth: Japan

2.0

1.0

0.1

1.4

0.0

0.5

Capital
Intensity

0.3

0.8

Capital
Intensity

1.4

Total Factor
Productivity
& Labor
Composition

1.5

Capital
Intensity

1.0

Total Factor
Productivity
& Labor
Composition

0.5

0.8

0.0

‐0.1
0.2
‐0.3

‐0.5

0.8

‐1.0

0.0
1985‐2007

1985‐2007

2010‐2013

e) Sources of Productivity Growth: France

2010‐2013

f) Sources of Productivity Growth: Italy

Percentage Points, Annual Rate
2.5

Percentage Points, Annual Rate
1.5

1.2
2.0

2.0
1.0
0.7

1.5

0.8

1.0

0.5

1.2

Capital
Intensity
Total Factor
Productivity
& Labor
Composition

0.5

0.8

0.6

Capital
Intensity
Total Factor
Productivity
& Labor
Composition

0.1

0.5

0.0

0.4

‐0.3
0.4

0.0

‐0.5
1985‐2007

1985‐2007

2010‐2013

2010‐2013

Productivity variation in the sixty years following the end of World War II was almost entirely
attributable to TFP variation. But, since the global financial crisis, U.S. labor productivity has
been driven down by negative capital deepening, the principal source of the shortfall. This has
8

been generally true across advanced OECD economies where investment growth in the wake of
the crisis has been slower relative to previous business cycles, as shown in Figure 9. The OECD
and the IMF have argued that this can largely be explained by a classic accelerator model where
aggregate demand growth was not sufficiently strong to justify strong investment. This effect
was likely compounded by a combination of capital overhang going into the crisis and
widespread deleveraging in the wake of the financial crisis.5
Figure 9
Capital Deepening in the G‐7
Percent Increase in Capital Intensity, Annual Rate
5.0
4.5
4.0

1985‐2007
2010‐2013

3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Canada

France Germany

Italy

Japan

United United
Kingdom States

A different way to visualize this pattern is to look at the historical evolution of labor productivity
along with its two most influential components—TFP and capital intensity—as shown in Figure
10. Labor productivity and TFP have generally moved together throughout most of the post-war
era, but diverged around the financial crisis.
Figure 10
Labor Productivity and Major Components
Percent Change, Annual Rate (5‐Year Centered Moving Average)
6
5

Labor
Productivity

4

Capital
Intensity

3
2
1
0

Total Factor
Productivity

‐1

2014

‐2
1950

1960

1970

1980

5

1990

2000

2010

Organisation for Economic Co-operation and Development. 2015. “OECD Economic Outlook, Chapter 3: Lifting
Investment for Higher Sustainable Growth.” (May 2015); International Monetary Fund. 2015. “World Economic
Outlook: Uneven Growth, Short- and Long-Term Factors, Chapter 4. Private Investment: What’s the Hold-up?”
(April 2015).

9

So the post-crisis productivity slowdown has been no normal shift by historical terms. I
personally find this encouraging in thinking about the future of productivity growth. It suggests
to me that much of the recent weakness in productivity growth across the advanced economies is
attributable to the macroeconomic circumstances coming out of the crisis—a consequence of a
unique cyclical moment rather than some new structural omen. Therefore, it is not the sort of
shift that one would extrapolate forward over the next decade or two. Indeed, TFP—the principal
source of historical variations in labor productivity—has slowed much less than labor
productivity.
Looking forward, the evidence shows that TFP growth is more inertial than capital deepening.
The story is confirmed and strengthened somewhat by looking at “utilization-adjusted total factor
productivity” (UATFP) developed by the economists at the Federal Reserve Bank of San
Francisco.6 UATFP models and removes changes in labor effort and capital workweek in an
attempt to isolate the “innovative” elements of productivity growth, rather than capturing
changes in the intensity of capital and labor usage that show up in TFP.
Recently, UATFP has been growing at roughly the same rate as TFP (0.5 percent per year
between 2010 and 2014, compared with 0.6 percent per year TFP growth). This is relevant
because the historical data suggest that the correlation of productivity growth over the next five
years against the past five years is strongest for UATFP and is in fact negative for capital
deepening, as shown in Figure 11, suggesting that measured “innovation” has some inertial
tendencies while capital investment tends to be mean reverting.
Figure 11

Correlation of Five‐Year Productivity Growth
with Prior Five Years' Growth in Potential Predictors

Correlation Coefficient
0.50
0.40

0.33

0.30
0.20

0.16

0.15

0.10
‐0.17
0.00
‐0.10
‐0.20

Labor Productivity
Growth

Total Factor
Utilization‐Adjusted
Productivity Growth
TFP Growth

Capital Deepening

Potential Predictors (5‐Year Periods)

Of course, a correlation coefficient of 0.33 explains only about 10 percent of the variance in
productivity growth, so I do not claim to have unlocked the mysteries of productivity by
observing this relationship. But the comparative correlation of lagged UATFP versus broader
measures suggests that the “pure innovation” measured by TFP is a core, persistent component
6

Susanto Basu, John Fernald, and Miles Kimball. 2006. "Are Technology Improvements Contractionary?"
American Economic Review.

10

of productivity growth. It is conceptually akin to the relationship between private domestic final
purchases (the sum of personal consumption and fixed investment) and gross domestic product,
which I have discussed in other contexts.
Productivity in the Future
Some economists have argued that the slower productivity growth across the advanced
economies in recent years is evidence of a shift towards sustained sluggish productivity growth
in the future. John Fernald, for example, has argued that total factor productivity growth slowed
prior to the Great Recession and that potential GDP is lower than commonly estimated.7 Robert
Gordon has argued that faster productivity growth in the 1950s, 1960s, and 1995-2005 was the
result of special, one-time factors and we are unlikely to be able to continue to innovate at that
pace going forward.8 Some commentators have seized on the last few years of productivity
growth to argue that potential growth rates will be much lower going forward.
But, as I have discussed, there is evidence that the post-crisis slowdown was a special, one-time
result that is more akin to a cyclical feature of the economy than a structural one. Moreover,
forecasting productivity based on just the past few years is an unproductive exercise as a general
matter. In Figure 11, I showed that the labor productivity growth over five-year periods is a
virtually useless predictor of labor productivity growth over the subsequent five. Goldman
Sachs9 has argued that if one is going to forecast productivity only by looking at its historical
average, the forecast error is smallest when the historical window is largest, as shown in Figure
12.
Figure 12
Average Absolute Forecast Error from Using Trailing Average
Productivity Growth to Forecast the Next Five Years
Percentage Points
1.25
1.00

0.96
0.87

Labor Productivity
Total Factor Productivity
0.87
0.68

0.75

0.84

0.83
0.69

0.73
0.64

0.60

0.59
0.48

0.50
0.25
0.00

1

3

5

10

Trailing Average Horizon (years)

7

20

40

John G. Fernald. 2014. “Productivity and Potential Output Before, During, and After the Great Recession.” NBER
Working Paper. June.
8
Robert J. Gordon. 2012. “Is U.S. Economic Growth Over? Faltering Innovation Confronts the Six Headwinds.”
Centre for Economic Policy Research. September.
9
Goldman Sachs Research. 2014. “US Daily: Trend Productivity Growth: 2% Still Seems About Right
(Mericle).” November.

11

But the choice of a long historical window is only a way to optimize a less-than-ideal forecasting
methodology: assuming mean reversion in the rate of productivity growth. To truly understand
productivity mechanics, we will need some model that can identify regime changes and respond
to macroeconomic fundamentals.
But as I pointed out when I began my talk, this is quite difficult. The OECD’s recent
innovation—using firm-level data to identify productivity differences within industries—takes a
step in the right direction.10 The OECD has identified that despite the slowdown in average
productivity growth in recent years, the most productive firms—those at the “global frontier”—
have actually seen continued robust productivity growth, as shown in Figure 13a-b. To the extent
that economic policies do not stand in the way, diffusion of productivity from the frontier can
provide some upside potential for aggregate productivity.
Figure 13a
Labor Productivity: Manufacturing Sector

Index (2001=0)
0.5

Index (2001=0)
0.5

Figure 13b
Labor Productivity: Services Sector
2009

0.4

0.4
2009

0.3

Frontier firms
(3.5% per year)

0.3

All firms
(1.7% per year)

Frontier firms
(5.0% per year)

0.2
All firms
(0.3% per year)

0.2
0.1
0.1

Non‐frontier firms
(0.5% per year)

0.0

0.0

Non‐frontier firms
‐(0.1% per year)

‐0.1
2001

2002

2003

2004

2005

2006

2007

2008

2001

2009

2002

2003

2004

2005

2006

2007

2008

2009

The productivity frontier thesis embedded in the OECD’s work ties in well with the observation
that the post-crisis slowdown is mostly a product of reduced investment rather than reduced
innovation. To the extent that more productive firms have prospered in this recovery, they are
also likely to have suffered the least from reduced investment.
While the data do not support a strong presumption of slowing productivity growth in the future,
there is certainly substantial reason to be uncertain about the future of productivity growth
which, unlike demographic factors, can be very difficult to predict. There are, however, some
reasons to be optimistic—including the potential in a variety of areas including cloud computing
combined with mobile devices, biotechnology and personalized medicine, advanced materials,
and clean energy research—some of which may not be fully reflected in the productivity data.11
But the degree to which we make advances in these areas—and productivity growth more
broadly—will depend on the policy choices we make.

10

Organisation for Economic Co-operation and Development. 2015. The Future of Productivity.
Erik Brynjolfsson & Andrew McAfee. 2014. The Second Machine Age: Work, Progress, and Prosperity in a Time
of Brilliant Technologies.
11

12

The Productivity Policy Agenda
Encouraging productivity growth is critical to boosting middle-class incomes. The President’s
economic agenda recognizes that we must not only improve the transmission of productivity to
typical households, but we also must increase the pace at which productivity grows. Today, I will
outline four major elements of the President’s agenda designed to do just that.
Expanding International Trade
I will begin with the importance of expanding free trade for boosting both the level and the
growth rate of productivity. The traditional economic understanding of trade rests on the
argument that it raises the level of output and incomes. But recently, economists have developed
a greater appreciation of the ways in which trade can increase innovation and therefore economic
growth. As Nobel Prize-winning economist Robert Solow wrote, “[r]elatively free trade has the
advantage that the possibility of increasing market share in world markets is a constant incentive
for innovative activity.”
Trade can increase growth in a number of ways. First, trade fosters greater specialization in
research and development—which in turn can increase innovation. For example, if engineers at
one firm focus on improving memory chips, and engineers at another firm focus on improving
microprocessors, the R&D productivity of each firm may be higher, leading to better and cheaper
computers than if each company had to improve both components simultaneously.
Trade also helps firms become more productive by accelerating the global flow of ideas. Both
exporters and importers are frequently exposed to new ideas and novel tools, materials, or
techniques that make them more productive. For example, many multinational companies have
systems and standards to promote the diffusion of “best practices” within their global supply
chains. Learning also occurs when a firm adapts novel ideas to suit its own operating
environment, leading to both new goods and greater productivity.
Moreover, a larger market can increase the incentives for innovation. International trade allows
companies to access a larger market, which yields more profit for a given level of innovation,
boosting the incentives to innovate. For example, the global reach of the “App Stores” managed
by Apple and Google provides a larger market—and greater reward—for developers seeking to
market their software.
Finally, even holding market size constant, increased trade can promote innovation by
strengthening competition. More than fifty years ago, the Nobel Prize-winning economist
Kenneth Arrow pointed out that a monopolist may have relatively weak incentives to innovate,
because its innovations do not allow it to “steal” business from competitors. A similar idea
appears in more recent “Schumpeterian” models of innovation and economic growth, where
competition can promote growth by increasing the expected payoffs of successful innovation.
When combined with smart intellectual property laws, our trade policies can promote
harmonization around a set of rules that strike the appropriate balance for promoting long-run
growth and job creation.

13

Public Investment in Infrastructure
Investments in infrastructure—such as those the President proposed in his FY2016 budget—can
have a substantial positive impact on productivity and long-run economic performance. Strong
infrastructure facilitates the more efficient exchange of goods, labor and ideas. Investments by
previous generations of Americans—from the Erie Canal in 1807, to the Transcontinental
Railroad in 1869, to the Interstate Highway System in the 1950s and 1960s—were instrumental
in putting the country on a path for sustained economic growth. Continued investment in
infrastructure is essential to supporting productivity and economic growth in the future.
For example, strong transportation and communication infrastructure reduces commuting times,
making it easier for workers to move between jobs and expanding the labor force. Businesses are
also able to manage their inventories more efficiently and transport goods faster and more
cheaply, helping them access new suppliers and markets. Faster information flows and increased
access to knowledge also facilitates greater innovation. Concentrated investment in infrastructure
may even draw market participants together and build new cooperative efficiencies, multiplying
the productivity gains.
The benefits of investing in infrastructure are especially high when there are underutilized
resources in the economy. In addition to the long-run effects on economic growth and
productivity, investments in infrastructure can have short-run benefits by supporting employment
in construction and in the production of materials. Moreover, increased spending by the workers
hired in these sectors can have positive ripple effects throughout the economy.
Business Tax Reform
Reforming our nation’s broken business tax code has important implications for productivity
growth. In particular, business tax reform has the potential to increase productivity growth by
improving the quality rather than the quantity of overall investment. By the “quality” of capital, I
do not mean more expensive or higher-end equipment, which, after all, would show up in the
dollar value of investment. Instead, I mean better choices about what to invest in, where to make
those investments, how to finance them, and so on.
To be a bit more concrete, let me offer three examples of ways in which business tax reform has
the potential to improve the quality of investment. One is by shifting capital from less productive
areas of the economy to more productive areas of the economy. The tax code has numerous
benefits for particular industries as well as a complex set of rules around depreciation that do not
match the actual economic depreciation of assets. As a result, there is wide variation in effective
tax rates by industry, ranging from 14 percent for utilities to 31 percent for construction and
wholesale and retail trade, as shown in the Treasury Department estimates in Figure 7. These
differences can potentially lead to too much capital in industries that are tax preferred and too
little capital in industries that are tax-disadvantaged. This misallocation of capital reduces
productivity. To the extent that reform can remove these distortionary incentives—ensuring that
business decisions are made for business reasons and not tax reasons—we can help resolve the
problem.

14

A second way in which tax reform can improve the quality of capital is by ensuring that capital is
distributed in a globally efficient manner, including not over-allocating capital overseas and
having more economically productive global sales and supply chains. Differences in
international taxation can encourage locating profits abroad even when it is not otherwise
productive to do so. For example, the fact that in 2010, U.S. controlled foreign corporation
profits represented 1,578 percent of Bermuda’s GDP and even 15 percent of the Netherlands’
GDP does not simply reflect business decisions made for purely business reasons. The
President’s framework for business tax reform would establish a hybrid international system with
a minimum tax on the earnings of foreign subsidiaries. This system could effectively deter some
tax-based decisions on the location of production, while also having the potential to improve the
global competitiveness of U.S. corporations.
A third example is encouraging investment in projects with externalities, like research and
development in general, or clean energy in particular. Such tax-encouraged investments in
innovation can help boost aggregate productivity. The key test for any incentive is whether it is
motivated by a positive externality such that the underlying good or service will be
underprovided by the private economy. The framework singles out three incentives as passing
this test: the research and experimentation tax credit because the social returns to R&D are
roughly twice the private returns, the Production Tax Credit because of the negative externalities
associated with carbon emissions, and the manufacturing deduction because of the broader
spillovers in the manufacturing sector. While one can and should debate what provisions should
be added to or subtracted from this list, the key is orienting the argument around the principle of
economy-wide spillovers.
Continued Investment in Technology
Innovation is the core driver of productivity growth in advanced economics. The U.S.
government has a long tradition of supporting the research and development driving technical
change, and the President’s economic agenda advocates expanding these important investments.
The long-term benefits of this investment are evident in the way technology is creating new
industries and transforming old ones throughout the United Sates.
Some of the most visible changes are occurring in the information and communications sector,
where a combination of smaller, more powerful computing and communications devices as well
as improvements in mobile broadband connectivity have unleashed a new wave of invention.
While most of us feel the impact of the computing revolution every day, many scientists say that
we are also on the cusp of a revolution in life sciences. The first complete human genome was
sequenced in 2003 at a cost of roughly $3 billion—today it can be done for as little as $1,000 per
person.
Transformations like these are occurring throughout the rest of the economy as well. Progress in
nanotechnology has the potential to tremendously improve the efficiency of energy consumption
and production through the use of new materials for light bulbs, wire insulation, combustion
engines, and photovoltaic cells, to name just a few.

15

The President’s technology policies are contributing in all of these areas. We are working to
nearly double the amount of spectrum available for mobile broadband. We have made openness
and interoperability the new defaults for government data. And we are working to ensure that our
patent and copyright laws are well-suited to the modern age.
While economists debate the extent to which innovation has shown up in productivity statistics,
the idea that the next twenty years will look more like the late 1990s because of continued
innovation seems entirely plausible to me.
Conclusion
The combination of productivity’s importance and its murkiness ensures that it will remain a
focus of macroeconomic research for decades to come. As I have emphasized today, the past is
rarely prologue when it comes to productivity—but despite the difficulty of prediction, we
should resist the pessimistic case that the recent slowdown is persistent. Indeed, it looks more
like yet another result of post-crisis deleveraging than a regime change in the pace of innovation.
Going forward, it will remain important for the United States and its partners to ensure that
economic policies promote private innovation while targeting public investments toward highreturn projects—helping to foster productivity growth in a shared and sustainable manner.

16

Notes to Figures
Figure 1
Note: Data reflect quarterly releases of labor productivity for the nonfarm business sector.
Shading denotes recession.
Source: Bureau of Labor Statistics.
Figure 2
Note: This figure—and all subsequent references to U.S. labor productivity reported by the
Bureau of Labor Statistics in these remarks—references real output per hour worked in the
private nonfarm business sector (excluding government enterprises). The dotted lines divide the
last 60 years into three periods that broadly reflect three "episodes" in productivity growth for the
private nonfarm business sector.
Source: Bureau of Labor Statistics.
Figure 3
Source: Conference Board; CEA calculations.
Figure 4
Note: Median disposable incomes in local currencies at current prices are reported by the OECD,
and are then deflated by each country’s consumer price index. All series are indexed to 1995,
except Italy (indexed to 1996) and the United Kingdom (indexed to 1994) due to data
availability.
Source: Organisation for Economic Co-operation and Development; national sources via Haver
Analytics; CEA calculations.
Figure 5
Note: For all nations except the United States, productivity growth rates are those reported in the
Conference Board Total Economy database; for the United States, the Bureau of Labor Statistics’
productivity series for the private nonfarm business sector is used.
Source: Conference Board; CEA calculations.
Figure 6
Note: Displayed series are the contributions to labor productivity growth in the private nonfarm
business sector.
Source: Bureau of Labor Statistics; CEA calculations.
Figure 7
Note: Displayed series are the contributions to labor productivity growth in the private nonfarm
business sector.
Source: Bureau of Labor Statistics; CEA calculations.
Figure 8a-f
Note: Total labor productivity growth and total multifactor productivity growth (including labor
composition changes) for the displayed nations are reported by the OECD. The contribution of
capital deepening is inferred as the difference between the two.
17

Source: Organisation for Economic Co-operation and Development.
Figure 9
Note: Capital intensity is defined as capital services per hours worked of all persons. Capital
deepening is the percent increase in capital intensity.
Source: Organisation for Economic Co-operation and Development; CEA calculations.
Figure 10
Note: Capital intensity is defined as capital services per hour worked.
Source: Bureau of Labor Statistics; CEA calculations.
Figure 11
Note: The displayed correlation coefficients result from the comparison of a five-year moving
average of labor productivity growth with a five-year lag of five-year moving averages of the
potential predictors. Accordingly, they reflect the ability of the predictors averaged from years t
to t+4 to predict labor productivity growth from years t+5 to t+9.
Source: Bureau of Labor Statistics; Federal Reserve Bank of San Francisco; CEA calculations.
Figure 12
Note: This analysis follows the general methodology adopted in Goldman Sachs Research. 2014.
“US Daily: Trend Productivity Growth: 2% Still Seems About Right (Mericle).” The displayed
values are the average absolute difference between trailing averages of growth in the given
productivity series (for the given horizon) and averages of the next five years’ growth in the
given series, from 1988 to 2008. This is the longest period over which the calculation can be
performed for a forty-year time horizon.
Source: Bureau of Labor Statistics; CEA calculations following Goldman Sachs (2014).
Figure 13a-b
Note: These figures are adapted from the OECD’s 2015 work The Future of Productivity.
“Frontier firms” corresponds to the average labor productivity of the 100 globally most
productive firms in each 2-digit sector. “Non-frontier firms” is the average of all other firms.
“All firms” is the sector total.
Source: Organisation for Economic Co-operation and Development. 2015. The Future of
Productivity. Andrews, D., C. Criscuolo and P. Gal (2015). “Frontier firms, technology diffusion
and public policy: micro evidence from OECD countries,” OECD Mimeo.

18