Full text of Business Review (Federal Reserve Bank of Philadelphia) : Second Quarter 2003

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

Corporate Governance and Responsibility
Based on a speech by President Santomero at the Corporate Governance & Responsibility Seminar,
Wilkes University, Wilkes-Barre, PA, February 24, 2003

C

BY ANTHONY M. SANTOMERO

orporate governance problems have recently
gained the spotlight. What is the root of
these problems? The mega-merger frenzy
of the 1980s? Overly optimistic forecasts of
earnings? Innovations in the financial services industry?
Although these explanations are plausible and these
factors undoubtedly played some role, President
Santomero views governance problems as having deeper
roots. The central dilemma is one of conflicting
interests in organizations—what economists call “the
principal-agent relationship.”

Recent headlines have brought
the issue of corporate governance to
everyone’s attention. We have all seen
the many stories about Enron,
WorldCom, Adelphia, and other
companies that were once mainstays of
our economy and our business
community. Television has brought us
images of corporate executives, not
being recognized for their civic
contributions but being led away in
handcuffs on allegations of malfeasance.
A common refrain in all these
stories is that company executives were
not acting in the best interest of their
organizations, their shareholders, and
their employees. A combination of
inadequate monitoring, a breakdown in
internal controls, and the systematic
failure of both outside directors and
outside auditors appear to have led to a
less than desirable outcome.
WHY HAS THIS HAPPENED?
Several reasons have been
www.phil.frb.org

offered to explain why these corporate
governance problems have recently
gained the spotlight. Some have argued
the origin of these problems was the
mega-merger and takeover wave of the
1980s, when innovative compensation
programs for top executives were
established — including a significant
increase in the use of stock options.
While these programs were supposed to
improve management’s incentives to
increase shareholder value, some see
them as the seeds of our current
problems.
These compensation programs
expanded and covered more companies
during the 1990s. For example, Harvard
law professor John Coffee, Jr., points out
in a recent paper that in 1990, equitybased compensation for CEOs was 5
percent of total compensation. By 1999,
it was 60 percent. Stock options rose
from 5 percent of outstanding shares in
U.S. companies in 1991 to 15 percent a
decade later. Meanwhile, the value of

stock options in the largest 2000
companies in the U.S. more than tripled
between 1997 and 2000.
Detractors argue that these
changes in executive compensation
tended to place more emphasis on shortterm gains in a company’s stock price,
rather than on long-term performance.
Then, the 1990s brought about rapid
changes in technology, greater
deregulation, and increased
globalization of markets. This placed
more pressures on companies’ cash flows
and made it more difficult to raise share
valuations. The innovative
compensation programs encouraged
executives to take greater risk or to
engage in more creative accounting to
improve reported earnings. In effect,
corporations shifted their business
standards and were not held in check
by either their corporate directors or
others charged with guarding
shareholder interests.

Anthony M. Santomero, President,
Federal Reserve Bank of Philadelphia
Business Review Q2 2003 1

Another explanation of recent
events focuses on the fact that longterm earnings forecasts for many
companies were overly optimistic during
the decade of the 1990s and generated
unrealistic expectations. In a speech last
year, Fed Chairman Alan Greenspan
noted that three- to five-year earnings
forecasts for S&P 500 companies
averaged almost 12 percent per year
between 1985 and 2001. However,
actual earnings growth over that period
was 7 percent.
Some blame analysts and Wall
Street for these overestimates of
earnings. They argue financial firms
promoted and retained those analysts
with the most optimistic forecasts of
companies’ earnings. Interestingly, the
bias to do so was especially pronounced
among analysts employed by the
underwriting firm.
Still another explanation of
these scandals lays blame at the foot of
innovations in the field of finance. By
this view, these innovations outpaced
the ability of traditional accounting and
auditing standards to monitor many
corporations’ activities. They allowed
some executives to engineer creative
accounting techniques to obfuscate
earnings and conceal negative results.
Consequently, investors and outside
parties had more and more difficulty
understanding the financial statements
and the risk positions of these large,
complex organizations.
Of course, all these suggested
explanations received little attention
when stock prices were rising rapidly
during the bull market. The sharp
declines in stock prices have led to
greater awareness and concern.
In essence, the foundation of
trust was breached between corporations
and shareholders with regard to the
meaningful disclosure of corporations’
financial information. The outcome was
a break between executives’ pay and
the corporations’ performance. The

2 Q2 2003 Business Review

whole process of reporting earnings and
financial statements became tainted.
AN OLD PROBLEM
WITH DEEP ROOTS
Although the problems
outlined in these explanations certainly
played some role in recent events — or
even a major role — focusing on them
gives the impression that corporate
governance problems are a relatively
new issue. I disagree. Rather than a
recent development, such issues have
deep roots. They are inherent in what
economists call “the principal-agent

The oversight of the
firm falls directly on
the company’s board
of directors.
relationship” in organizations.
The central dilemma here is
one of conflicting interests. Much
research has been devoted to how to
provide incentives to the agent — or
executive management of a firm, for the
purposes of our discussion — to act in
the best interests of the principals — the
owners of the firm. In essence, the
challenge of our form of capitalism is,
and has always been, to construct a
system of corporate governance so that
company management acts in the best
interests of shareholders.
This is what we have
attempted to do in our structure of
corporate governance. The oversight of
the firm falls directly on the company’s
board of directors. In the end, the board
bears ultimate responsibility for the
company’s performance. The board is
supposed to implement methods to
monitor and control management so
that abuses are prevented or at least
minimized. To do this, some members of
the board of directors are outsiders —

people who are not part of the
management team of the company.
These directors are supposed to act as
an independent check on corporate
management to ensure they act in the
shareholders’ best interest. American
capitalism relies on the fiduciary
concept to protect those who entrust
their money to large — and often
distant — corporations.
THE IMPORTANCE
OF CORPORATE GOVERNANCE
Yet today, there is a sense that
the model just described is not working
well enough. A crisis of confidence in
corporate America has resulted. Recent
scandals have generated a lingering
sense of uncertainty and vulnerability
among investors. This has put pressure
on companies’ management, corporate
directors, and regulators to address
problems of accountability and control.
To restore public confidence
in the integrity of corporate America,
companies must demonstrate a strong
commitment to the development and
enforcement of rigorous standards of
corporate governance. These standards
must encompass the relationship
between a company’s board of directors,
its management, and its shareholders.
They must require corporate leaders to
be faithful to shareholder interests and
act with both competence and integrity.
At a very basic level, trust is at
the heart of the free enterprise system.
But the current state of public trust in
American corporations is not good.
According to a recent poll, 77 percent
of the public believes CEO greed and
corruption caused declines in the stock
market. In addition, polls suggest much
of the public rejects the view that the
scandals were isolated incidents within
a system in which most corporate
leaders are good and honest people.
Yet, the continued success of
our economic system requires the
confidence and trust of investors,

www.phil.frb.org

employees, consumers, and the public at
large. In short, there is much work to be
done.
WHAT IS BEING DONE?
We know that as corporations
grow larger and more complex, it
becomes more difficult for boards of
directors to monitor activities across the
company. Directors cannot be expected
to understand every nuance of or
oversee every transaction. They should
look to management for that.
Nonetheless, the role of a
corporate board of directors is quite
substantial, and directors are required to
be highly knowledgeable. They must
know — understand — the nature of
the firm’s business, its financial
performance, and the nature of the risks
facing the firm’s strategic plan.
Collectively the board should have
knowledge and expertise in areas such
as business, finance, accounting,
marketing, public policy, manufacturing
and operations, government, technology,
and other areas necessary to help the
board fulfill its role. They must set the
tone for risk-taking in the institution and
establish sufficient controls so its
directives are followed. They also have
the responsibility to hire competent
individuals who possess integrity and the
ability to exercise good judgment.
Members of the audit committee, in
particular, must be independent and
have knowledge and experience in
auditing financial matters. This is no
small task.
Recent events have been a
loud wake-up call, focusing attention on
the need to heighten our commitment
to proper corporate governance and
improve both accountability and control.
In response, a number of measures have
been taken or proposed by various
groups to bolster confidence in our
corporate system.
Recognizing that boards have
come under increased scrutiny, the New

www.phil.frb.org

York Stock Exchange has appointed a
Corporate Accountability and Listing
Standards Committee. The committee
has come up with a number of
recommendations to improve corporate
governance. One proposal is to increase
the role and authority of independent
directors by having them make up a
majority of a company’s board. The
committee also recommends that
companies adopt corporate governance
guidelines and a code of business ethics
and conduct. In addition, the
committee has suggested shareholders
be given more opportunity to monitor

growing concerning some needed
changes to certain underlying
accounting standards and their
application. The U.S. Financial
Accounting Standards Board (FASB) is
considering how to improve accounting
standards for special-purpose entities.
This is in response to the growth of
securitization and the added complexity
securitization has introduced into
financial reporting.
A pilot program is under way
to standardize financial reporting data
and make them available to investors via
a web site hosted by Nasdaq. In the

The U.S. Financial Accounting Standards Board
(FASB) is considering how to improve
accounting standards for special-purpose
entities.
the governance of their companies.
They must vote on all equity-based
compensation plans and have access to
the company’s corporate governance
guidelines.
The Conference Board
Commission on Public Trust and Private
Enterprise has also made
recommendations on best practices. It
recommended that executive
compensation be performance-tied and
stressed the importance of independent
directors being able to retain outside
consultants. The commission also
suggested that the Federal Accounting
Board (FAB) and the International
Accounting Standards Board (IASB)
come up with standardized definitions
of revenues in order to achieve true
parity in determining executive
compensation based on company
performance. Finally, it recommended
that America’s senior executives should
be subject to much longer-term holding
periods for company stock and higher
ownership requirements.
A consensus is now also

future, technology will be instrumental
in improving transparency in financial
reporting by making corporate financial
information easily available.
Congress has also responded.
The newest legislation about corporate
governance, the Sarbanes-Oxley Act,
addresses the wave of recent events that
shook public confidence. The act seeks
to protect investors by improving the
accuracy and reliability of corporate
disclosures. Among its major provisions is
the establishment of a new privatesector regulatory regime in which the
SEC handpicks an oversight board to
monitor standards and conduct in the
accounting industry. In fact, my
colleague Bill McDonough, president of
the New York Fed, has been tapped to
head this new board.
Sarbanes-Oxley also
emphasizes the need for a wall of
independence between auditors and
firms. In addition, and perhaps most
controversially, the act seeks to
strengthen corporate responsibility by
creating a structure for holding

Business Review Q2 2003 3

individuals and companies criminally
and/or civilly accountable for their
actions. CEOs and CFOs are now
required to certify quarterly and annual
reports to ensure proper disclosures.
While some may disagree with
any of these proposals, it is important to
realize that those involved and
responsible have begun to take action to
address the perceived problems of
corporate governance.
CORPORATE GOVERNANCE
IN BANK REGULATION
The nonfinancial sector is not
alone in its search for better corporate
governance. The requirement of trust
and confidence in corporate America is
analogous to the trust and confidence
issues that the Federal Reserve faces in
its role as the regulator of the U.S.
banking system. Let me touch on some
of the parallels and briefly describe how
we have addressed them.
Of course, banks have
shareholders, too. Their business
involves making loans to customers who
are expected to repay. Bank
management has a good deal of
information about the quality of the
loan portfolio. The question facing bank
managers and their directors is how
much information to provide to
shareholders. As stewards of the public
trust, bank regulators and supervisors ask
the same questions.
But beyond this, a bank’s
relationship with its depositors is another
example of the principal-agent problem.
Depositors depend upon bank regulators
and supervisors, as well as deposit
insurance, to keep their money safe in
spite of the opaque nature of bank
assets. As a result, we have substantial
interest in the ways in which corporate
governance is performed in the
regulated banking sector, and we have
incorporated these concerns into our
regulatory and supervisory model.
The primary focus of the

4 Q2 2003 Business Review

Federal Reserve’s approach to
supervision and regulation is ensuring an
institution’s safety and soundness. The
Federal Reserve’s examiners also ensure
compliance with banking laws and
regulations, including consumerprotection laws and regulations.
Historically, a major focus of
banks and their regulators has been on
whether they accurately report their
financial condition and appropriately
assess the quality of their assets. Beyond
this, supervisors have long been
concerned about the quality of internal
controls. During the past 15 years, the
Fed’s supervisory program has been
broadened to focus on banks’ overall

In 1991, Congress
broadened the
scope of banks’
assessments of
risks and controls.
risk-management systems, comparing
them against both regulatory standards
and industry best practices.
The Fed’s risk-assessment
process analyzes the nature and extent
of risk to which a financial institution is
exposed and assesses how well the
institution is identifying, controlling, and
managing risks. It requires integrated,
enterprisewide risk management that
considers all areas of risk, including
credit risk, market risk, liquidity risk,
operational risk, legal risk, and
reputational risk. The idea is to identify
not only the type of risk and its level but
also its direction and whether the bank
has means to effectively control each
risk.
The Fed also wants to ensure
that the bank has a strong internal audit
function and that it also receives a
thorough, complete, and independent
external audit. To accomplish all this,

Fed examiners conduct on-site
examinations and provide institutions
with continual off-site monitoring and
analysis as economic conditions and the
bank’s financial condition change.
In 1991, Congress broadened
the scope of banks’ assessments of risks
and controls. Since then, bank
managers are required, at least annually,
to step back from other duties and
evaluate risks and internal controls. In
addition, external auditors must attest to
management’s results of this selfassessment of risks and internal controls.
The results are reported to the audit
committee of the bank’s board of
directors. Incidentally, the audit
committees of banks’ boards have been
required to be independent of
management for a long time —
something that is now being stressed for
all corporations. In fact, this approach to
risk-assessment and internal controls is
also the one followed by all of the
Federal Reserve Banks for several years.
Ensuring a broad-based
assessment of risks and internal controls
has served the banking industry well in
recent years. For instance, despite the
economic downturn in 2001, most banks
continue to be in good health.
BUT THERE ARE LIMITS
The Fed’s experience,
therefore, suggests some success with
the evolving model of better corporate
governance. Nonetheless, it is important
to remember that the process is still
evolving. Much work remains to be
done. It is important to remember that
proposals to improve corporate
governance must take into account not
only the expected benefits of new
standards and regulations but also their
expected costs to both the corporations
and the economy as a whole. Good
intentions do not always prevent
unintended consequences. Some
unintended side effects might include
high compliance costs, ambiguous

www.phil.frb.org

liabilities, or reduced innovation. This is
particularly true when various proposals
about corporate governance have been
arising at both the state and federal
levels.
Disclosure should never be so
onerous as to make the cost of compliance prohibitive or impractical.
Regulations and standards of any sort —
whether by regulators, government,
trade groups, or the companies themselves — should not excessively impede
the ongoing process of innovation.
Rather, we must ensure an environment
conducive to markets that are effective
and efficient, safe and sound.
IS THERE A BETTER WAY?
One criticism of the past
approach to corporate governance is that
it tended to focus on the development
of fairly specific rules of behavior rather
than insisting on adherence to certain
principles of behavior. Most of the
proposals we now debate to improve
corporate governance are new rules.
However, the problem with
rules, particularly accounting-based
rules, is that innovations in the financial
system can open loopholes in rules.
When loopholes open, inappropriate or
unethical actions that are not
specifically prohibited by the rules can
take place. Basically, this is what
happened in many of the corporations
that made news headlines in the past
year or two.
A credible case can be made
that we should focus on principles
instead of rules. That is, we should
establish key principles against which
corporate decisions should be held
accountable, regardless of whether a
certain type of behavior is prohibited.
This way, when innovations make old
rules obsolete, corporate leaders and
their financial executives would have to
consider not only whether some action
would violate a rule but whether it
would violate a principle.

www.phil.frb.org

There are strong arguments for
developing principles-based standards —
in addition to our reliance on traditional
rules-based standards. However, the
challenge is to establish a set of
principles that are sufficiently clear and
concise. This may not be an easy task,
and the result may be substantial
litigation, rather than simplification and
clarity. It is important to consider both
the costs and the benefits of new

The crisis of confidence in
corporate America has been created by
recent scandals that have generated a
sense of uncertainty and vulnerability
among investors. These events have put
pressure on regulators, corporate
directors, and management to address
problems of corporate accountability
and control. Changes are in the works
and appear to be in the right direction.
To a large extent, this direction is where

Companies should take action to ensure their
financial statements divulge what is truly
essential for investors to understand the
business and make informed decisions.
standards, as well as the unintended
consequences that may result. In the
end, rules cannot replace ethics and an
exemplary “tone at the top.”
Nonetheless, whichever way
regulation evolves, disclosure and
transparency are imperative to adequate
corporate governance. Such disclosure
need not be identical across all industries and companies. The information
available to the public should be what is
necessary for them to evaluate a particular firm’s risk profile. That is why
principles-based accounting has some
appeal. Companies should take action to
ensure their financial statements divulge
what is truly essential for investors to
understand the business and make
informed decisions.
CONCLUSION
Good corporate governance is
critical to the health of the corporate
system, our financial system, and our
economy. Our economy will be stronger
if corporate decisions are made with
competence and integrity, and if shareholders and the public can appropriately
assess the profitability and riskiness of
corporations’ business activities.

banks and bank regulators have gone
before.
Many ideas to improve
corporate governance are being offered
by a variety of parties. Adopting a
system of principles-based accounting
standards, rather than primarily
amending the current rules-based
standards, may be useful in ensuring
that our accounting rules do not
become quickly out of date in the face
of rapid financial innovation. But given
the wide range of ideas being offered,
the challenge will be to move forward
and implement those proposals most
likely to be effective in yielding benefits
at a reasonable cost of compliance and
to do so without generating unintended
consequences. The challenge is in the
implementation, but the challenge is a
noble one. We must proceed.
In the end, however, we must
bear in mind that the core principles of
ethical behavior and sound business
practices are the keys to any real
success in this arena. This tone is set at
the top. Without these values we will
never really succeed in conquering the
problems and conflicts that arise in
corporate governance. BR

Business Review Q2 2003 5

U.S. Coins: Forecasting Change
BY DEAN CROUSHORE

A

lthough the government annually produces
about 70 new coins for every man, woman,
and child, the economy’s need for coins can
vary from year to year. So how do the U.S.
Mint, which makes the coins, and the Federal Reserve,
which distributes them, decide how many coins the
economy needs? Dean Croushore highlights some facts
about coins and describes how demand for change is
forecast.

Every year, the U.S. government produces about 70 new coins for
every man, woman, and child in the
country, or about 20 billion coins. In
recent years, the program to produce
new state quarters, plus the introduction
of the golden dollar, increased total
demand for new coins. When the
demand for the new quarters and
dollars became surprisingly strong in
1999 and 2000, shortages of some coins
developed in different parts of the
country.
To help prevent such shortages,
a team of economists and analysts at the
Federal Reserve are developing new
models to forecast demand for coins.
This article describes some of the work

Dean Croushore
is vice president
and economist in
the Research
Department of
the Philadelphia
Fed.

6 Q2 2003 Business Review

the Philadelphia Fed has undertaken
since the project began in 2000.
Let’s see how the Federal
Reserve and the U.S. Mint decide on
the number of coins to be produced and
distributed and how difficult it is to
forecast demand for coins. We’ll begin
by looking at some basic facts about the
institutions involved and how demand
for coins is calculated.1
THE FACTS ABOUT COINS
The U.S. Mint is in charge of
producing coins. The Mint must obtain
the raw metals to be used in production,
procure the equipment, and hire

1
The author thanks David Griffiths of the
U.S. Mint for comments on an earlier draft of
this article, and Brian Coulter of the Federal
Reserve Cash Product Office, Geoffrey
Gerdes of the Board of Governors, and
Christopher Sims of Princeton University for
consultation on the coin forecasting models.
The author further thanks the team of
economists and analysts at the Federal
Reserve Bank of Philadelphia who worked on
this project, including John Chew, Brian
DiCiurcio, James Gillard, James Sherma,
Keith Sill, and Tom Stark.

workers to produce the coins needed for
the economy. Once coins are produced,
the Mint sells them at their face value to
the Federal Reserve. The Mint makes a
considerable profit (called seignorage)
on the sales of coins. For example, one
of the new Sacagawea golden dollars
costs about 12 cents to produce and
yields one dollar in revenue. The
resulting profit of 88 cents goes to the
U.S. Treasury. The Mint’s profits
increase the government’s revenue by a
billion dollars or more each year.
The Federal Reserve distributes coins from 37 coin offices, mainly
Reserve Banks and their Branches, and
more than 100 coin terminals operated
by armored carriers such as Brinks and
Loomis-Fargo. The armored carriers
hold inventories of coins for the Fed and
for banks and other financial institutions
(hereafter just called banks). The
carriers move coins between their
terminals, banks, and Federal Reserve
coin offices.
The U.S. Mint generally
produces coins for circulation according
to orders from the Federal Reserve. The
Federal Reserve, in turn, generally
orders an amount of coins based on the
expected demand from banks. And, of
course, those banks want coins to satisfy
the demands of their customers. So,
ultimately, the amount of coins produced depends on the demand for coins
by people and businesses. Let’s take a
look at how that demand is measured.
The main concept in analyzing
the demand for coins is called net pay.
Net pay is an unusual economic concept
because it represents the change in
banks’ demand for coins, which can be
either positive or negative: It’s positive
www.phil.frb.org

when banks ask the Federal Reserve to
deliver coins because demand has
increased; it’s negative when those
institutions return coins to the Federal
Reserve because demand for coins has
declined.2 Because of the durable
nature of coins and because they can be
returned to the Federal Reserve, coins
(as well as paper money) are different
from all other goods and pose unique
challenges.
Coins flow between people or
businesses and banks (Figure 1). When
people and businesses want more coins,
their banks order more from their local
Federal Reserve offices, which then ship
coins to the banks, either from inventories of coins located at armored carriers
(the line labeled A in the chart) or
directly from the offices’ own inventories (line B). Occasionally, the U.S.
Mint ships coins directly to a bank (line
C). Each of these shipment methods to
banks results in positive net pay, since
banks’ demand for coins is positive.
However, if people begin
turning in more coins than banks want
to keep on hand, the banks may return
the extra to the Federal Reserve, either
directly (line D) or through armoredcarrier terminals (line E). So lines D and
E represent negative net pay because
demand for coins is negative.
In any given month, some
banks may have positive net pay and
others may have negative net pay. Net
pay is the sum of the amounts shown by
lines A, B, and C minus the sum of the
amounts shown by lines D and E:
Net pay = (A+B+C)-(D+E).
Net pay is calculated separately for six
different denominations of coins: penny,
nickel, dime, quarter, half-dollar, and

Note that net pay differs from the change in
the demand for coins in the rare instance in
which there is a shortage of coins. Because
such cases are rare and we do not have
reliable data on the amount of shortages, our
models of coin demand ignore such instances
and assume that net pay equals the change in
demand.

dollar. Total net pay is the sum across all
banks.
Let’s take a look at the data on
net pay for each denomination to see
how each has changed over time. In
doing so, we will look at the net pay for
each denomination in units of millions
of coins each month. The data run from
1957 to 2002, except for half-dollars and
dollars. For each denomination, the
black line shows the actual monthly
amount of national net pay (summed
across all 37 Federal Reserve offices).
The green line is the average volume
over the past 12 months (called a 12month moving average), which is shown
to help illustrate the long-term trend in
the data (Figures 2a to 2f).
The charts show some
interesting patterns. First, you can see
that month-to-month seasonal fluctuations in net pay are huge. For some
denominations, national net pay is even
negative in some months. The net pay
of different denominations swings
dramatically from one month to the
next, mostly because of changes in
people’s spending patterns. We need a

lot more change in the summer months
for parking meters at the shore and for
soda machines. We use more change
around holidays at the end of the year,
as well. But we need much less in the
middle of the winter.
About half of all coins
produced are pennies. Net pay of
pennies has averaged over 800 million
coins per month in the last five years,
while the sum of all other denominations has been about 700 million coins
per month. The net pay for pennies has
been fairly constant since 1980, perhaps
trending down slightly (Figure 2a). For
other main denominations (nickels,
Figure 2b; dimes, Figure 2c; and
quarters, Figure 2d), the trend over time
has been slightly upward, which suggests
that these other denominations may be
gradually replacing pennies in terms of
quantity used for making change.
There are some interesting
variations in net pay for those coins,
especially in the 1960s when the value
of silver, which was a major component
of dimes and quarters, increased sharply.
Demand for those coins declined

FIGURE 1
Net Pay
People
and
Businesses

Banks

B

C

A

E

D
Federal
Reserve
Office

Armored
Carriers

2

www.phil.frb.org

Mint

Business Review Q2 2003 7

sharply when the coins were redesigned
with no silver in them. The other two
denominations, half-dollars and dollars,
had net pay near zero for much of the
1980s and 1990s. The introduction of
the Sacagawea golden dollar in 2000
caused a sharp increase in the net pay of
dollars that year.
Given these trends in demand
for different denominations, what can
we say about overall demand for coins?
To investigate this issue, we’ll examine
total demand for coins in terms of
numbers of coins, adding up the net pay
for all six denominations. Also, to avoid
confusion arising from seasonal fluctuations, we just look at the total net pay in
each calendar year (Figure 3). We look
at the total number of coins rather than
their dollar value, in part because the
Mint’s ability to produce enough coins to
meet demand depends on the number
of coins rather than their dollar value.
In the graph, you can see that
overall net pay generally increased over
time, from under 2 billion coins in 1957
to a peak of 23 billion in 1999 and 2000.
But the increase was not steady. From
one year to the next, sometimes net pay
rose and sometimes it fell.
We might expect a correlation
between net pay and the strength of the
economy because it seems likely that
people will use more coins if they’re
buying more goods and services.
However, there does not appear to be a
strong correlation between net pay and
economic activity. For example, while
net pay fell when the economy weakened, as in 1990 and 1991, it fell even
when the economy was strong, as it was
in 1996 and 1997.
Special events raised net pay to
very high levels in 1999 and 2000. First,
beginning in 1999, the Mint (directed by
laws passed by Congress) rolled out the
first quarters in the state commemorative program. The demand for these
new quarters turned out to be significantly stronger than anticipated, thus

8 Q2 2003 Business Review

www.phil.frb.org

causing net pay to rise sharply (Figures
2d and 3). Then, in 2000, the
Sacagawea dollar was introduced to
much fanfare. Initial demand for the
new coin was also strong, and the Mint
produced over 1 billion of them. At the
same time, the demand for the new
state quarters increased 50 percent from
the year before, so again net pay was
much higher than expected. At the
same time, the demand for nickels and
dimes also rose substantially (Figures 2b
and c).
FORECASTING COIN DEMAND
Sharp, unexpected increases in
net pay during 1999 and 2000 led the
Federal Reserve Bank of Philadelphia to
investigate ways to improve forecasts of
demand for coins. Developing a
forecasting model involves several steps:
choosing among different types of
models, testing the different models to
see how well they perform, seeing how
they deal with changes, such as the
introduction of the new quarter and
dollar coins, then running forecasts in
real time and investigating the quality of
the forecasts. Because demand for each
coin denomination seems to behave
differently from that of the other coin
denominations, the models we examine
will contain a separate forecasting
equation for each denomination, rather
than modeling overall coin demand in a
single equation.
Four Models of Coin
Demand. We considered four different
types of models: (1) a structural model;
(2) a time-series model; (3) a vector
autoregression (VAR) model; and (4) a
Bayesian vector autoregression model.
Brief descriptions of each model follow.3
Structural Model. Adapting the
work of earlier researchers who had
modeled coin demand, we first examAdditional details about each model can be
found in the research paper that I wrote with
Tom Stark. The paper is listed in the References section at the end of this article.
3

www.phil.frb.org

Business Review Q2 2003 9

FIGURE 3
Annual Net Pay of Coins
25

Billions of coins per year

20

15

10

5

2001

1998

1995

1992

1989

1986

1983

1980

1977

1974

1971

1968

1965

1962

1959

0

Date

ined a structural model of net pay for
each denomination. In a structural
model, economic theory dictates which
variables should affect demand for each
denomination. We then develop a
forecasting equation based on that
economic theory.
Because many economic data
are quarterly, we took quarterly
averages of the monthly coin data.
Economic theory suggests that demand
for coins depends on economic activity,
interest rates, and the inflation rate. We
experimented with a number of
measures of economic activity, including
consumer spending on services (used in
older models), retail sales, industrial
production, personal consumption
expenditures, and payroll employment.
Payroll employment gave the best
results in our tests, so we used that
variable to represent economic activity.
For an interest rate, we used the federal
funds rate, which is the interest rate that
banks charge each other on overnight
loans. Since the fed funds rate is also the
main variable the Federal Reserve
targets with its monetary policy, it is a
good indicator of the overall level of

10 Q2 2003 Business Review

short-term interest rates. For inflation,
we chose the inflation rate as measured
in the Consumer Price Index. Again,
based on an older model, the inflow of
coins from banks to the Federal Reserve
is modeled separately from the payout of
coins from the Federal Reserve to banks.
The forecast for the inflow is then
subtracted from the forecast for the
payout to generate a forecast for net pay.
The forecasting model for each coin
denomination also includes a seasonal
variable for each quarter of the year to
account for the seasonal pattern in coin
demand.
Time-Series Model. The second
forecasting method we tried was a timeseries model, which uses data only from
the past and data only on the variable
being forecast. For example, the model
for net pay of pennies assumes that net
pay of pennies in the future depends
only on past movements of net pay for
pennies; it does not depend on the net
pay of any other coin denomination or
on any macroeconomic variable.
Though such a model is very
simple, we still had to make choices
about the forecasting equation of the

model: how far back to go in determining the forecast, whether to model the
level of net pay or the change in the
level of net pay from one month to the
next, and how to deal with the seasonal
fluctuations — seasonally adjust the
data before they go into the model or
account for seasonal fluctuations within
the model. Experimentation suggested
that the best model was arrived at by
using 14 months of lagged data for each
forecast, modeling the change in net pay
from one month to the next, and
adjusting the data for monthly fluctuations before running the forecasting
equation.4
Vector Autoregression (VAR)
Model. In the past 20 years, many
economists have stopped using structural
models for forecasting because these
models require more precise economic
theory than we usually know; economists have also moved away from timeseries models because such models use
no economic theory at all. A vector
autoregression (VAR) model is a mixture
of a structural and a time-series model.
The VAR uses economic theory to tell
the researcher which variables should be
included in the model, but it also
incorporates time-series techniques by
including past data on each variable in
the model. A VAR is useful because it
allows us to conduct “what-if” experiments, such as: “What will happen to
demand for coins if the economy goes
into a recession?”
For our VAR model, the net
pay of each coin denomination depends
on past data on economic activity, the
interest rate, and the inflation rate, just
as in the structural model. But the
equation for each coin denomination
depends on that denomination’s own
history, just as in the time-series model.
In the VAR, the economic data

The best model is determined on the basis of
the root-mean-squared forecast error, which
is described in detail later.

4

www.phil.frb.org

influence forecasts for coin demand, but
coin demand is not allowed to influence
forecasts of the economic data. Experimentation showed that the best model
came from using data on the logarithm
of payroll employment as the variable
related to economic activity; that
variable proved better in our tests than
the growth rate of employment and was
also superior to other variables, including
industrial production, retail sales, and
personal consumption expenditures. We
also found it was best to use employment
data that were not seasonally adjusted
but to account for seasonal fluctuations
within the forecasting model. Using 13
months of past data also provided the
best results for this model.
Bayesian VAR. Bayesian
techniques basically involve a
researcher’s beliefs (for example,
concerning seasonality) about the
outcome of an empirical investigation:
The researcher examines the data in
light of those beliefs, then sees if his
beliefs change after he has examined
the data. Essentially, Bayesian techniques help us rule out certain outcomes
that differ so much from economic
theory that we do not believe them. The
techniques help to keep the estimated
forecasting model within certain
bounds. Economists use Bayesian
methods because research has found
that such methods often generate
superior forecasts and may handle
monthly fluctuations in the data better.
Because of the large seasonal fluctuations in the coin data, Bayesian
techniques may be very fruitful.
To implement Bayesian
methods for forecasting net pay for
coins, we applied them to the VAR
model described above. The main
differences between the VAR and the
Bayesian VAR are in the amount of past
data used (24 months in the Bayesian
version versus 13 months in the nonBayesian version) and in the coefficients
of the forecasting equation, which are

www.phil.frb.org

fixed in the VAR but allowed to change
over time in the Bayesian VAR. In the
Bayesian VAR, some key coefficients
are chosen to make the model perform
well, the most important being those
that concern the seasonal patterns in the
data. The Bayesian VAR is also not as
restrictive as the VAR because it allows
the data on coins to affect the macroeconomic variables (perhaps because
people’s spending habits are reflected in
coin demand, which then helps predict
the macroeconomic variables) and it

we calculated summary statistics on how
well each forecast did. The most
appropriate statistic is the root-meansquared forecast error (RMSFE), which
quantifies deviations — either positive
or negative — of the actual from the
forecast, where larger errors are
penalized more. The RMSFE is calculated by taking a forecast for each year,
calculating the forecast error (actual
minus forecast) for each month,
squaring each error, adding up the
squared errors, dividing by the number

Because of the large seasonal fluctuations in
the coin data, Bayesian techniques may be
very fruitful.
allows data on one coin denomination to
affect the forecasts for other coin
denominations.
Comparing the Models.
After we built the models, we tested
their performance over several periods.
To avoid being unduly influenced by
the introduction of the new state
quarters and by the new dollar coin, we
chose 1990 to 1998 as our main testing
period. To see how the models would
have performed over that time, we
generated forecasts at the start of each
three months, as if we were at that date
and did not know what was to come.
That is, we first used the coin data from
January 1957 through December 1989,
which would have been known to a
forecaster making a forecast in January
1990, and generated a forecast for the
next 12 months for each coin denomination. Then we stepped forward three
months, as if we were in April 1990.
Then, using the coin data through
March 1990, we generated a forecast for
the next 12 months. We continued this
process until January 1998, at which
point we generated forecasts through
the end of 1998 (just before the start of
the new state quarters program).
With these forecasts in hand,

of forecasted values, then taking the
square root. Researchers have found
that the RMSFE has a number of
desirable properties and gives them a
general guide to using forecasts: The
best forecasts are those with the lowest
RMSFEs. By squaring the forecast errors
in calculating the RMSFE, forecasts that
are far from actual are penalized more
heavily than if we just calculated the
average error (Table 1).
As you can see from the table,
the time-series model, which was
originally proposed as a benchmark
model, proved very difficult to beat. In
fact, only the Bayesian VAR did better,
and its improvement was only slight.5
Using these models, we began
to generate forecasts periodically, as
requested, first by the Fed’s Cash-Fiscal
Product Office (located at the Federal

Of course, all the forecasting models did
much worse in forecasting coin demand in
1999 and 2000 because nothing in the models
accounted for the introduction of new coins.
But a researcher could have used these
models to forecast net pay for purposes of
determining how many coins were needed for
circulation, then added a projection for
demand for new coins that would not
circulate because people would keep them as
collector’s items.

5

Business Review Q2 2003 11

Reserve Bank of Philadelphia), and
then the Fed’s Cash Product Office
(located at the Los Angeles Branch of
the Federal Reserve Bank of San
Francisco) when that office took over
the responsibility for coin issues in spring
2001. Because the structural model
performed so poorly in our tests, we
stopped generating forecasts with it in
early 2001. Instead, we added the
Bayesian VAR to our process in
September 2001.
Now, after the Federal Reserve
coin offices calculate data on net pay at
the end of each month, we generate
new forecasts for net pay at the national
level using the time-series, the VAR,
and the Bayesian VAR models. Because
no one knows how long or how large the
increased demand for state quarters or
the demand for the new dollar is likely
to be, the best forecast is likely to be one
that simply tracks the overall trend but
does not generate forecasts that make
strong assumptions about that future
demand. All the models we use have
that feature. For example, in 2001,
demand for coins slowed substantially.
Although the models did not predict the
slowdown, the forecasts adjusted fairly
quickly after the slowdown began.
How Have the Forecasts
Performed So Far? The key question
for any forecasting method is: How well
does it work? Unfortunately, we have
been forecasting demand for coins only
for about two years, so we cannot
answer that question very well. Table 2
shows the forecasts made every three
months from February 2001 to November 2002, along with the actual values in
2001 and 2002. As you can see in the
table, the initial forecasts for 2001 and
2002 were fairly high. Given what had
happened in 1999 and 2000, with coin
demand rising, the forecasting models
predicted continued strong demand in
2001 that did not materialize. Instead,
coin demand began declining substantially, and it took the forecasting models

12 Q2 2003 Business Review

several months to adjust fully.
So far, it appears that the timeseries model has done the best job of
forecasting because it was the quickest
to lower forecasts for 2001 and 2002 as
net pay fell. But the period is much too
short to favor the use of that model over
the others. In a few years, we will have
much more data on the forecasts and
the errors made by each model, and we
will be able to undertake a more
complete examination.
USING THE FORECASTS
How can the Federal Reserve
use these forecasts for coin demand?
First, the national coin forecasts can
help the Mint in planning its production.
The Mint needs to schedule workers
and to purchase enough equipment to
produce the right amount of coins.
Improved forecasts will allow the Mint to
reduce production costs by getting a
better idea of how many coins it will
need to produce. In addition, the Mint
will be able to order the appropriate
amount of raw materials needed for
production.
The forecasts can also help the
Federal Reserve in ordering coins. Each
Federal Reserve office must order coins

every month, and improved forecasts
could give them an additional tool for
deciding how much to order. Those
offices maintain inventories of coins in
case of sudden changes in demand, so
improved forecasts can help them
maintain appropriate levels of inventories. Improved forecasts can also help
these offices reduce costs by keeping
inventories from becoming too large or
too small, since shipping coins between
offices is costly.
Because the time-series models
performed so well at the national level,
we began forecasting net pay for each
office based on such models in spring
2002. Because there are 37 offices and
six coin denominations, we generated
222 forecasts (37 x 6), each running
monthly for the next 30 months. These
forecasts are distributed to each office
for its use in ordering coins and for
planning. Coin offices must also take
into account changes in local and
national economic conditions that may
not be captured in the time-series model
that forecasts net pay.
SUMMARY
Forecasting the demand for
coins is difficult because of seasonal

TABLE 1
RMSFE for Different Coin Models
Model
Structural Model
Time-Series Model
VAR
Bayesian VAR

RMSFE
2.61
1.75
2.01
1.72

Note: Figures shown are the root-mean-squared forecast error (RMSFE) for each
model over the testing period from 1990 to 1998, in billions of coins. A forecast is
more accurate if it has a smaller RMSFE. Models for half-dollars and dollars were
not run because we have insufficient data.

www.phil.frb.org

TABLE 2
Annual Real-Time Net Pay Forecasts
Forecast Date

Feb 2001
May 2001
Aug 2001
Nov 2001
Feb 2002
May 2002
Aug 2002
Nov 2002

Actual Data Through

Jan 2001
Apr 2001
July 2001
Oct 2001
Jan 2002
Apr 2002
July 2002
Oct 2002

Actual

Calendar Year Forecasts
2001
2002
21.4
20.8
18.1
16.7

21.7
20.5
18.1
15.2
14.0
17.1
17.0

17.1

15.2

Note: Amounts in billions of coins per calendar year.
Numbers shown for forecast dates from February 2001 to August 2001 are the
average forecasts from the time-series model and VAR; numbers shown from
November 2001 on are the average forecasts from the time-series model, VAR,
and Bayesian VAR. Each forecast is a projection for the calendar year shown
in the column header for the last two columns.

fluctuations in net pay and the introduction of new coins. By using some
standard types of forecasting models, we
have attempted to improve on existing
forecasts of net pay. Whether these
forecasting models will perform well in
practice will require several years of
observations. The models we use
depend on the stability of historical
relationships. As such, changes in how
people use coins could cause the models
to make large forecast errors in the
future. If the models do not do well, we
may be able to modify them so that they
forecast better in real time.
Our hope is that we will be
able to forecast coin demand well
enough to prevent any shortages of coins
in the future, without the expense of
piling up large inventories of unused
coins. BR

REFERENCES
Croushore, Dean, and Tom Stark.
“Forecasting Coin Demand,” Federal
Reserve Bank of Philadelphia Working
Paper 02-15/R, September 2002.

www.phil.frb.org

Roseman, Louise L. “The Golden Dollar
Coin: Testimony Before the Subcommittee
on Treasury and General Government of the
Committee on Appropriations, U.S.
Senate,” May 17, 2002.

United States Mint, 2001 Annual Report.

Business Review Q2 2003 13

Antitrust Issues in Payment Card Networks:
Can They Do That? Should We Let Them?
BY ROBERT HUNT

I

n the United States, payment card networks
coordinate the activities of thousands
of financial institutions that issue cards,
millions of retail locations that accept them,
and several hundred million consumers that use them.
This coordination may include the collective setting
of certain prices and other controversial network rules.
Such practices have recently come under the scrutiny
of antitrust authorities in the U.S. and abroad. In this
article, Bob Hunt describes the economics of the
payment card industry and explains how it differs
from the textbook model of competitive markets. He
argues that these differences should be reflected in the
antitrust analysis of payment card networks.
In the United States, generalpurpose payment cards — Visa or
MasterCard, ATM cards, or debit cards
— are ubiquitous and easy to use. In
2000, there were about 900 million
general-purpose payment cards in the
U.S., or about four for every adult.
These cards were used in 28 billion
transactions worth $1.9 trillion. Indeed,
payment cards are displacing the paper

check at the point of sale — the number
of consumer checks written peaked
during the 1990s and is now in decline.1
In this article, I explore how
payment cards work and explain why
we need to think a little differently
about the market for consumer payment
methods than we do for most other
markets. This has implications for
when, why, and how the rules of

These statistics are from the Bank for
International Settlements’ Statistics on
Payment Systems in Selected Countries (2002),
Table 6. They exclude store cards. The
decline in checks’ share of consumer
transactions — relative to credit and debit
card transactions — is documented in the
article by Geoffrey Gerdes and Jack Walton.
1

Bob Hunt is an
economist in the
Research Department
of the Philadelphia
Fed.

14 Q2 2003 Business Review

antitrust law — which regulate how
firms may exercise market power —
should be applied to this industry. This is
not just an academic question: In the
U.S. there are currently two important
antitrust cases involving payment cards.
Australia recently introduced new
regulations for the payment card
industry in that country while the U.K.
and the European Union have contemplated similar measures.
Payment cards have two
distinguishing features that lead us to
think differently about this market.
First, payment cards exhibit what
economists call network externalities —
for example, payment cards are more
valuable to consumers when more
merchants accept them. Second, in the
U.S. at least, thousands of banks and
other firms provide payment card
services to millions of cardholders and
millions of merchants who accept cards.
In an environment with network
externalities and so many participants,
economic theory suggests that some
form of coordination is beneficial,
possibly essential. In the U.S., this is
usually done by forming a payment card
network.
Payment networks coordinate
the behavior of banks, merchants, and
consumers by setting certain prices and
rules. In many other contexts, such
practices might be considered anticompetitive. It is also possible they can
have anti-competitive effects in the
market for consumer payments. Yet a
careful examination of economic theory
tells us this is not always the case.
The challenge to policymakers
is to decide, based on the available
information, whether a network’s
www.phil.frb.org

pricing strategy and rules are likely to
advance or retard economic efficiency.
Such conclusions are complicated by
dynamic considerations — a network
that exercises market power may spur
the development of competing networks
with superior technology.
THE ORGANIZATION AND
ECONOMICS OF PAYMENT
CARD NETWORKS
The U.S. payment card
industry involves thousands of banks
participating in a number of networks,
millions of consumers who find it
valuable to use a payment card, and
millions of merchants who find it
valuable to accept those cards.2
Pricing and Rulemaking in
Payment Card Networks. Banks
engage in two types of activities within a
payment card network.3 Card issuers
are banks that offer cards to consumers
and determine the level of any fees or
finance charges their customers see on
their regular statements.
Merchants also have banks,
called acquirers, that process card
payments on their behalf.4 Merchants
pay their acquirer for these services by
accepting a merchant discount — when a
consumer makes a $1 purchase using a
payment card, the acquiring bank pays
the merchant slightly less than $1 for
that transaction (Figure 1).
An open payment network, like
the bankcard associations Visa and
MasterCard and most electronic funds

The organization and development of the
U.S. payment card industry is described in
the book by David Evans and Richard
Schmalensee and the book by Lewis Mandell.

transfer (EFT) networks, allows many
banks to participate. The association
builds and maintains much of the
infrastructure: the lines and switches
required to route transaction information between different acquiring and
issuing banks. The associations specify
that, for each transaction, an interchange
fee be paid to the bank issuing a card by
the bank acting as the acquirer for the
merchant.5 In the U.S., about 1.5
percent of the value of all generalpurpose-card transactions flows to issuers

in the form of interchange fees — about
$23 billion a year.6
An interchange fee is one way
to ensure that network participants are

5
In a closed network, the card issuer also acts
as the acquirer. Such networks carry a
merchant discount but not an interchange
fee. Examples include American Express and
Discover.
6
This estimate is from The Nilson Report,
February 2002 (No. 758).

FIGURE 1
Flow of Payments in a Stylized Card Network*

2

In this article, I will focus on generalpurpose credit cards, such as Visa or
MasterCard, and debit cards. I will not
discuss department store cards, oil company
cards, or bank cards when they are used at
ATMs.
3

They are called acquirers because they
acquire transactions for the network.
4

www.phil.frb.org

* This figure is an illustration; it is not a precise description of any actual network.
In particular, it does not reflect the timing of actions required to authorize or settle a
transaction.
In the figure, the card issuer makes a payment directly to the merchant’s bank. In some
networks, this is not done directly, but instead is done via payments to and from the
network itself.
The merchant discount and interchange fee are simply illustrative. Other fees (e.g., switch
fees) are not included in the figure.

Business Review Q2 2003 15

able to recover their costs. But, as we
will see, it can also be used to coordinate
the activities of banks that issue
payment cards and acquiring banks that
process transactions on behalf of
merchants. Even though consumers do
not directly pay the interchange fee, it
often affects the cost and benefits of
using a payment card. For example,
issuing banks that receive higher
interchange fees will have an incentive
to reduce fees that cardholders pay
(annual fees, transaction fees, or interest
rates). Or they may offer incentives
such as cash back or frequent flyer
miles. The interchange fee also affects
the acquiring bank because the bank
must cover that fee through the
discount it charges merchants. That, in
turn, influences merchants’ willingness
to accept cards.
The bankcard association also
acts as the rule-making body for the
network. In recent years, two of these
rules have received a great deal of
attention. First, the honor-all-cards rule
says that merchants wishing to accept a
card brand must accept all cards issued
under that brand. For example, a
merchant who accepts a Visa card
issued by ABC Bank must also accept
Visa cards issued by XYZ Bank. In
addition, merchants must accept all
types of a particular brand — from
platinum to plain vanilla cards.7
Second, the no-surcharge rule
says that merchants may not charge
customers more for a transaction using
one brand of card than for a transaction
involving any other brand. Taken
together, these rules require that
merchants treat all cards issued under a
given brand equally and must not favor
another card brand by offering its users
better prices.

The New Kid on the Block:
The Debit Card. Debit cards allow
customers to pay for goods and services
at the point of sale by authorizing a
withdrawal from their checking or
savings account. In the U.S., debit
transactions at the point of sale only
became common in the 1990s, but they
have increased extremely rapidly. In the
20 years ending in 2000, consumer
purchases made via debit cards rose
from essentially zero to account for

In the U.S., debit
transactions at the
point of sale only
became common in
the 1990s, but they
have increased
extremely rapidly.
nearly 12 percent of all noncash
consumer transactions. During this same
period, the share of these transactions
paid via credit card increased from 14 to
21 percent; the share paid by check fell
from 86 to 59 percent.8
Most ATM cards can be used
at the point of sale as debit cards. Such
transactions are called PIN debit
transactions because the cardholder
must enter her four-digit PIN to
authorize the transaction.9 Funds are
then immediately withdrawn from the
associated bank account. The transaction itself is routed through an electronic
funds transfer (EFT) network, for
See the article by Gerdes and Walton.
These statistics refer to the number of
transactions, not the value of those
transactions.
8

A PIN, or personal identification number, is a
four-digit number entered on a keypad at an
ATM machine or point-of-sale terminal.
Many networks require a PIN because, as
long as it remains confidential, a PIN can
verify that the card is being used by the
authorized cardholder.

9

Bankcard association rules require
merchants to accept their brands of debit
cards as well.
7

16 Q2 2003 Business Review

example, Star, NYCE, and Pulse.10 But
in order to accept PIN debit transactions, the merchant must first install a
PIN pad and have a contract with one
or more EFT networks. Today, about 1.3
million different store locations can
accept a PIN debit transaction.
Visa and MasterCard offer
their own brand of debit card, but they
function differently. These cards can
typically be used for PIN debit transactions, but they can also authorize
transactions with just the signature of
the cardholder. A purchase paid for in
this manner is often called a signature
debit transaction. Unlike a PIN debit
transaction, a signature debit transaction
does not immediately remove funds
from the cardholder’s account; it
typically takes a day or two for the
transaction to clear. This delay creates
some credit risk for the issuing bank
because the cardholder may have
insufficient funds in her account at the
time the transaction clears. So, unlike
with ATM cards, banks offer signature
debit cards only to account holders that
meet minimum credit standards.
Another important difference
between the two types of debit transactions is that a signature debit transaction
can be carried out with the same
equipment used to authorize credit card
transactions. In fact, under the honorall-cards rule, stores that accept Visa or
MasterCard must also accept the
comparable brand of debit card.
Currently, 4.9 million store locations in
the U.S. accept one or both of these
cards, offering a huge merchant base for
signature debit cards.11 Signature debit
transactions are routed over the card
associations’ network, and the card

These networks are also the backbone of
the 350,000 ATM machines around the
country.
10

11
Information is from The Nilson Report, June
2002 (No. 765), in a column entitled
“Retailer/Bank Card Lawsuit.”

www.phil.frb.org

issuer receives an interchange fee
comparable to the interchange fee on a
credit card transaction.12
Network Effects and Fixed
Costs in Payment Card Networks.
Payment networks function differently
from most markets, in part because of
network effects.13 A payment card is
more valuable to consumers when more
merchants accept the card. At the
same time, merchants are more willing
to accept a card if they know many
consumers use it. Every consumer who
obtains a card and every retailer who
accepts a card increase the value of the
network to all other cardholders and all
other merchants who accept it. Such
decisions create externalities that have a
number of implications for the evolution
and efficiency of payment networks.14
First, consumers and merchants are unlikely to take network
effects into account unless such effects
are reflected in the prices they pay or
the benefits they receive.15 If these
effects are ignored, the payment
network is likely to be too small or
underutilized.
Second, payment networks
exhibit increasing returns to scale. If a
network is introduced on a small scale,
there is inertia — consumers and

A $40 signature debit transaction generates
an interchange fee of about 60 cents (1.5
percent) while a comparable PIN debit
transaction generates an interchange fee of
about 18 cents (0.5 percent). See the August
8, 2002 edition of ATM and Debit News.
12

A nontechnical discussion of this subject
can be found in the “Symposium on Network
Externalities.” For applications of the theory
to financial networks, see the article by
Nicholas Economides and the one by James
McAndrews.
13

An externality exists when the decisions or
activities of one entity affect, positively or
negatively, the environment of another.
14

For example, consumers are sometimes
offered cash back or frequent flyer miles as an
incentive to use their cards.
15

www.phil.frb.org

merchants have little incentive to join.
Since such a network would clearly be
unprofitable, it would never be
launched. But if a network is launched
on a large scale, it’s possible that many
consumers will carry the card and many
stores will accept it. If that happens,
even more stores are likely to accept the
card, which may induce even more
people to carry the card and so on.
Third, network effects suggest
that there could be significant barriers to

cost of fraudulent transactions only if
the network’s antifraud technology is
sufficiently effective.17
Network effects and fixed
costs at the network level may explain
why bankcard associations and EFT
networks use many of the strategies
described earlier. Setting an appropriate
interchange fee is one way to ensure
that network members take into
account network effects, presumably
increasing card usage. This in turn

Modern payment card networks require large
investments in communications and computing
facilities in order to make card transactions
convenient to customers and merchants and to
minimize fraud.
entry, and those barriers may permit
established payment networks to
maintain prices above costs for some
time. The reason is that in order to
successfully enter the market, a rival
network must do so on a large scale. But
the market may not be large enough to
support more than a few networks at
such a scale.
A second factor that distinguishes payment cards from many, but
certainly not all, industries is the large
fixed cost associated with establishing a
viable payment network. Modern
payment card networks require large
investments in communications and
computing facilities in order to make
card transactions convenient to
customers and merchants and to
minimize fraud. The latter is especially
important to card issuers because
network rules typically promise to pay
merchants for fraudulent transactions as
long as the network’s procedures are
followed. If networks did not make this
promise to merchants, fewer merchants
would be willing to accept payment
cards.16 But card issuers will accept the

reduces the unit cost of card transactions, making payment cards more
attractive to use or accept.
Setting the fee at the network
level eliminates costs associated with
bargaining between individual card
issuers and acquirers and uncertainty
about the actual costs of a card transaction.18 Consumers are more likely to use

An exception to this rule for credit cards is
card-not-present transactions, such as
Internet or telephone orders, where network
rules stipulate that fraud losses are borne by
the merchant.

16

As with network effects, large fixed costs
imply economies of scale. This could explain
why we did not observe an increase in the
number of payment networks in the 1990s (in
fact, there was a significant decline), even
though technological advances significantly
reduced the merchant’s cost of accepting a
new payment card.

17

Why doesn’t the collective setting of the
interchange fee — an obvious example of
price fixing among competitors — violate U.S.
antitrust law? Federal courts have recognized
there are situations where such arrangements
may promote rather than retard competition.
See the Broadcast Music, Inc. and NaBANCO
cases.

18

Business Review Q2 2003 17

a payment card if they know where it
will be accepted and on what terms.
The honor-all-cards rule and the nosurcharge rule reduce the uncertainty
consumers would otherwise face. This

was especially important in the late
1960s and 1970s, when the card
associations were trying to build nationwide acceptance of credit cards issued
primarily by small banks. But do these

mechanisms remain essential after a
payment card network becomes well
established?
There could be a dark side to
all this coordination: These rules might

Legal and Regulatory Challenges to Payment Card Networks
United States. In October 1996, Wal-Mart and
other retailers filed an antitrust suit against Visa and
MasterCard.a This suit later became a class action,
representing several million retail locations. In April 2003,
the district court ruled on the pre-trial motions, reaching a
number of conclusions favorable to the plaintiffs’ tying
claim.b The case was settled shortly thereafter. The
bankcard associations agreed to revise their honor-all-cards
rules so that merchants can separately decide whether to
accept their brands of credit and debit cards, to reduce
interchange fees charged on signature debit transactions,
and to pay $3 billion in damages over a 10-year period. The
reduction in interchange fees in 2003 alone is expected to
save merchants $1 billion.c
In 1998 the U.S. Department of Justice (DOJ)
filed a separate antitrust suit against Visa and MasterCard.
Among other things, DOJ objected to the associations’
exclusivity rules, which prevent banks that issue Visa or
MasterCard credit cards from simultaneously issuing a
Discover or American Express card. In October 2001, the
trial court invalidated these rules.d The case is currently
under appeal.
Europe. In July 2002, the Competition Directorate of the European Commission announced a settlement
with Visa that addresses multilateral interchange fees levied
on certain credit and debit transactions that involve banks
in more than one member state.e Under the terms of the
agreement, Visa pledges to reduce those fees gradually over

the next five years and to keep them below a cap that will
be calculated each year on the basis of card issuers’ costs.
Allowable costs include transaction processing, financing
the interest-free period enjoyed by cardholders, and certain
payment guarantees provided to merchants. Visa also
agreed to amend its rules so that its interchange fee can be
disclosed to merchants.
In a separate decision published in November
2001, the Commission concluded that Visa’s honor-all-cards
rule did not restrict competition even when applied to
different types of cards (for example, credit and debit)
within the same brand (for example, Visa).f
Australia. In August 2002, the Reserve Bank of
Australia (RBA) announced regulations that apply to
domestic credit card transactions using Visa, MasterCard, or
Bankcard credit cards.g As of January 2003, merchants are
permitted to surcharge transactions using these cards. In
October 2003, credit card interchange fees will be capped
according to a cost-based formula that will be revised every
three years.h Allowable costs include authorizing and
processing transactions, financing the interest-free grace
period enjoyed by cardholders, and costs resulting from
card fraud and its prevention. The card associations must
provide RBA with audited data on these costs each year.
RBA also invalidated certain card association rules that it
concluded were inhibiting entry by monoline credit card
banks and merchant acquirers.

a

In re Visa Check/MasterMoney Antitrust Litigation, N0. 00-7699 (2d Cir 2001).

b

In re Visa Check/MasterMoney Antitrust Litigation, 96-CV-5238 (E.D.N.Y. 2003).

c “MasterCard, Visa to Pay $3 Billion to Resolve Card Suit; Will Modify Debit Card Policy, Fees,” BNA Banking Report, Vol. 80 (May 5,

2003), pp. 739-40.

d

U.S. v. Visa U.S.A., Inc. 163 F. Supp. 2d. 322 (S.D. NY 2001).

e Case No. COMP/29.373 — Visa International-Multilateral Interchange Fees.

002).

f Case No. COMP/29.373 — Visa International.
g Reserve Bank of Australia.

August 2002.

Official Journal of the European Community (July 24,

Official Journal of the European Community (November 10, 2001).

“Reform of Credit Card Schemes in Australia IV: Final Reforms and Regulation Impact Statement,”

h RBA is imposing the cap against the volume weighted average interchange fee levied on card transactions rather than specifying caps
for different kinds of transactions. RBA expects that once implemented, the caps will reduce average interchange fees about 40 percent.

18 Q2 2003 Business Review

www.phil.frb.org

be used to enhance a dominant
network’s market power. Such allegations form the basis of an important
antitrust case in the U.S., regulation of
the payment card industry in Australia,
and far-reaching inquiries in Europe
(see Legal and Regulatory Challenges to
Payment Card Networks). In the
following sections, I’ll examine in greater
detail the role of interchange fees, the
no-surcharge rule, and the-honor-allcards rule in consumer payment
networks.
THE PROS AND CONS OF
INTERCHANGE FEES
An interchange fee can be
used to solve a seemingly intractable
problem: how to maximize the value of a
payment network while ensuring that
retailers and banks are able to cover
their costs. Economic theory offers some
intuition about solving this problem. All
other things equal, prices should be set
lower for customers who are more price
sensitive, that is, more likely to switch to
another form of payment in response to
a small change in price. Conversely,
prices should be set higher for those
customers who are not as sensitive to
price differences. Economists refer to
this strategy as Ramsey pricing, in honor
of the mathematician and economist
Frank Ramsey.19 Intuitively, this rule
leaves consumers as close as possible to
the consumption choices they would
make if they were able to purchase
goods and services at a price equal to
their marginal cost of production — the
competitive ideal.

Ramsey solved the following problem: A
fixed amount of revenues must be raised by
charging prices for goods in excess of their
marginal cost. But higher prices will reduce
consumption and therefore welfare. Under
these circumstances, the best that can be
done is to charge higher markups on those
goods with less elastic demand curves and
lower markups on those goods with more
elastic demand curves. See Ramsey’s 1927
article.
19

www.phil.frb.org

But open payment networks
do not actually control all the prices that
consumers and retailers pay for a transaction. Instead, they influence those
prices by setting the interchange fee.
For example, suppose the network raises
the interchange fee so that each card
transaction is more profitable for card
issuers. Card issuers will seek out more
cardholders either by offering them
more benefits or by reducing cardholder
fees. Merchants will observe more
cardholders using the card. But there is
a trade-off to raising the interchange fee

particular payment network need not be
the best from the standpoint of consumers.20 The economic literature has
explored a variety of reasons a privately
determined interchange fee may not
correspond to the fee that maximizes
social welfare, but in this article, we’ll
focus on just one.21
Suppose that a merchant
charges the same prices regardless of the
manner in which consumers pay. For
example, customers who pay with a
credit card pay the same price as
customers who pay with cash. While

An interchange fee can be used to solve a
seemingly intractable problem: how to maximize
the value of a payment network while ensuring
that retailers and banks are able to cover their
costs.
because it raises costs for the acquiring
banks, and at least some of that cost is
passed on to merchants. The higher
cost of card transactions may cause
some merchants to stop accepting the
card.
To summarize, economic
theory suggests two factors that are
likely to influence the size of a privately
optimal interchange fee, that is, one that
maximizes the value of a payment network. The first is the relative size of costs
borne by issuing and acquiring banks —
banks will not willingly participate if
they cannot recover their costs. The
second is the degree of price sensitivity
exhibited by cardholders on the one
hand and merchants on the other. In
order to maximize the value of a payment card network, it is necessary to
impose more of the costs on those participants who are least likely to stop
using or accepting the card.
Networks May Not Choose
the Interchange Fee Best for Society.
But the best interchange fee for a

different payment instruments mean
different costs for the merchant, those
costs are not reflected in the prices paid
by every customer. Users of the cheaper
form of payment bear some of the costs
created by customers who use a more
expensive form of payment. In economic terms, there is a cross-subsidy.
Consumers tend to overuse the more
expensive form of payment because they

A payment network that enjoys market
power has some freedom to choose the
amount of resources to raise through some
combination of fees and discounts to
merchants and cardholders. When there is
more than one payment network, it is not
necessarily efficient to encourage the growth
of a network if it is at the expense of a less
costly one.
20

See the articles by William Baxter; Dennis
Carlton and Alan Frankel; Sujit Chakravorti
and Ted To; Sujit Chakravorti and William
Emmons; Howard Chang and David Evans;
Joshua Gans and Steven King; Jean-Charles
Rochet and Jean Tirole; Richard
Schmalensee; and Julian Wright.
21

Business Review Q2 2003 19

enjoy all the benefits but do not bear all
the costs. At the economywide level,
this could mean a payment network will
grow too large because purchases
outside the network are subsidizing
purchases made within the network.
It’s possible a network can
exploit this cross-subsidy by raising
interchange fees while reducing
cardholder fees. Merchants would pass
on these costs to all their customers,
increasing the subsidy noncard users pay
to card users. If some of the increased
interchange revenues are passed on to
cardholders (through lower fees or more
perks), this would, in turn, increase the
number of cardholders. Merchants may
not like this outcome, but they may be
reluctant to stop accepting the card if
they think cardholders will take their
business elsewhere.
The actual outcome depends
crucially on how consumers react to
changes in prices and the nature of
competition among merchants. If
customers who do not use the card
respond to small price increases by
switching to merchants that accept only
cheaper cards, or just cash, any crosssubsidy must be small. Thus, an
important insight gleaned from theoretical models of payment networks is that
the effects of interchange fees depend
on the extent of market power enjoyed
by retailers.
THE PROS AND CONS OF
SURCHARGES
So far, we’ve assumed that
merchants do not set different prices for
different card transactions. What
happens if payment networks permit
merchants to add a fee to transactions
when consumers use a more expensive
payment method? In principle, merchants could pass on any difference in
their cost of using different payment
cards to the customers using those cards.
This would eliminate any cross-subsidy
between customers using different

20 Q2 2003 Business Review

payment methods and encourage
consumers to use the most efficient
forms of payment. So if we think that a
costly payment instrument is being used
too much, allowing merchants to
surcharge may be a useful remedy.
But permitting surcharges may
have a second effect. If merchants are
willing to pass on costs in this way, an
open payment network cannot use an
interchange fee to influence the prices
paid by merchants and consumers. To
see this, imagine what would happen if
the network raised the interchange fee

An important insight
gleaned from theoretical models of payment
networks is that the
effects of interchange
fees depend on the
extent of market
power enjoyed by
retailers.
and card issuers passed on the additional
revenue to cardholders via lower fees or
additional perks. Suppose also that card
acquirers pass on the higher interchange
fee to retailers by raising the merchant
discount. In turn, merchants could
increase the surcharge on card transactions, essentially negating the increased
benefits provided by card issuers. So
with surcharging, raising interchange
fees in order to stimulate card use will
not be very effective. If network effects
are important and cannot be taken into
account by some other means, the result
could be underutilization of payment
cards.
But so far we have assumed
that given the opportunity, merchants
actually would impose surcharges.
There is reason to doubt much surcharging would occur. In the U.S., federal law

has permitted merchants to offer a
discount for cash purchases since 1975.
Yet, in 1983, less than 10 percent of
retailers offered cash discounts. Most of
this activity occurred at gas stations and
even this became less common once
these stores began to accept bank-issued
credit and debit cards.22 In Sweden
and the Netherlands, two countries that
banned no-surcharge rules during the
1990s, less than 10 percent of retailers
report surcharging their customers and
there has been little change in the
merchant discount.23 In practice, then,
permitting surcharging may have little
effect.
THE PROS AND CONS OF AN
HONOR-ALL-CARDS RULE
The question regarding an
honor-all-cards rule is not whether
networks should be permitted to have
such a rule, but rather how broadly it
can be applied. Suppose a network
issues two types of cards — a credit card
and a debit card — under the same
brand. Should a merchant be required
to accept both types of cards even if it
prefers to accept only one type?
This was the central argument
in the Wal-Mart antitrust case: The
plaintiffs argued that the bankcard
associations were using the honor-all-

Statistics on cash discounting are from
Credit Cards in the U.S. Economy. Evans and
Schmalensee document the subsequent
decline in cash discounts in their 1999 book.
Edmund Kitch argues that regulatory barriers
made it relatively costly for merchants to
offer cash discounts. Alan Frankel argues
that merchants believe any benefit of
charging different prices is not worth risking
a negative reaction from customers. He also
wonders whether consumers react more
strongly to fees charged at the point of sale
than to fees that appear later on their bank
statement.
22

See the November 2001 European
Commission decision. In his 2001 report,
Michael Katz notes that among retailers in
the Netherlands who were aware of the legal
change, 20 percent surcharged.
23

www.phil.frb.org

cards rule to impose an illegal tying
arrangement, forcing merchants that
accept a brand of credit card to accept
the same brand of signature debit
card.24 Because card issuers receive
more interchange revenue from
signature debit transactions, they have
an incentive to subsidize consumers’ use
of these cards. This, in turn, influences
consumers’ choices about signature vs.
PIN debit transactions. Merchants pay
these higher fees and pass at least some
of the cost on to consumers via higher
prices. During the 1990s, the use of both
types of debit cards grew immensely
(Figure 2). But the share of all debit
transactions using a PIN fell from about
60 percent in 1993 to about 38 percent
in 2002.
If we view credit and debit
cards as distinct products, it seems
reasonable that an honor-all-cards rule
should be enforced separately for each
type of card. In other words, merchants
could be allowed to refuse the debit
card but still be required to accept all
credit cards issued under that brand.
Similarly, merchants could separately
decide whether to accept a debit card
brand but would then be required to
accept all debit cards issued under that
brand.
The likely result would be that
merchants would pay different fees for
credit card and signature debit transactions. If signature and PIN debit
transactions offer merchants the same
benefits, the interchange fee for a
signature debit would have to fall in
order for these cards to remain competitive with a PIN debit, since the cost of
processing a PIN debit is less than the
cost of processing a signature debit.
Users of signature debit cards would

FIGURE 2
PIN Debit vs. Signature Debit
Transaction Volume
(billions per year)
9
8

In a tying case, the plaintiff tries to prove
that the defendant is using the market power
it enjoys in one market to extract profits from
another, typically more competitive, market.

www.phil.frb.org

Signature

6
5
4
3
2
1
0
86

Sources:

88

90

92

94

96

98

2000

2002

EFT Network Data Book, Debit Card and POS Market Data Book,
and Card Industry Directory, various years.

presumably pay higher fees or enjoy
fewer benefits, since the use of the cards
25
could not be subsidized as heavily.
Do Credit and Debit Cards
Compete in the Same Market?
Applying the honor-all-cards rule to
both credit and debit cards of the same
brand seems more reasonable if these
products actually compete in the same
market. But do they? Economists
typically define the boundaries of a
market based on how consumers
respond to price changes. If a change in
the price of one good induces consumers
to switch to another good, we say these
goods are substitutes. If only a small
change in price is sufficient to cause
consumers to switch to the substitute
product, we say they compete in the
same market.
What can we say about
consumer substitution between credit

cards and debit cards? There is at least
some evidence that consumers do not
use credit and debit cards in the same
way. For example, consumers are more
likely to use debit cards in drug and
grocery stores than they are at department stores (Table). In addition,
consumers can often initiate a PIN debit
transaction larger than their purchase
and receive the difference in cash, a
feature not available in a credit card
transaction. While credit cards provide
an explicit line of credit, debit cards do
not. Such differences suggest that credit
and debit cards are not pure substitutes
as a means of payment.
In 2001, a federal court, in a
separate antitrust case against the
bankcard associations, reached just this
conclusion.26 But in a number of
previous antitrust decisions, the courts

The Wal-Mart case was settled in April
2003. (See Legal and Regulatory Challenges to
Payment Card Networks.) Industry analysts
predict it will change the debit card market
in precisely the way described in the
preceding two paragraphs.

26

25

24

PIN

7

This case was initiated by the U.S.
Department of Justice in 1998. A similar
conclusion was reached in a ruling on pretrial
motions in the Wal-Mart case (see Legal and
Regulatory Challenges to Payment Card
Networks).

Business Review Q2 2003 21

TABLE
Data on the Use of Debit
and Other Forms of Payment (for 1999)
Stores with
PIN Pads
(percent)

Percentage of Store Sales Paid via
Cash

Check

Credit Card

Debit Card

All Stores

50

35

21

25

8

Discount

43

47

17

27

3

Drug

73

41

17

26

14

100

44

32

11

12

20

29

15

26

2

7

21

27

26

6

38

28

19

32

10

Supermarket
Department Store
Home Center
Apparel

Source: “Survey of Retail Payment Systems,” Chain Store Age (December 1999)
Note: These statistics are derived from a survey of large retail chains. It is not a
representative sample of the retail sector.

have defined the relevant market more
broadly to include cash, checks,
department store cards, and ATM
cards.27 Even in the 2001 decision, the
judge recognized that the emergence of
all-in-one cards — a single card that can
be used for credit, signature debit, or
PIN debit transactions — may increase
consumers’ willingness to substitute
between these different forms of
payment.
YESTERDAY, TODAY,
AND TOMORROW
At the end of the day,
policymakers need to know the answer
to the following question: Does the
conduct of a payment network benefit

27

See the NaBANCO decisions.

22 Q2 2003 Business Review

or harm consumers? In antitrust cases,
judges are often forced to weigh the
static costs of certain conduct against
any dynamic benefits it may offer. This
is not easy to do when it is not clear how
a market would have developed in the
absence of the conduct under scrutiny.
Suppose we return to the
1980s, before a thriving debit card
market developed. How might such a
market be developed? The method
chosen by Visa and MasterCard was to
graft debit cards on to the existing credit
card networks. Using their honor-allcards rule, the associations ensured that
millions of merchants would accept
signature debit cards. Using their nosurcharge rule, the associations ensured
that these cards would be accepted on
terms equal to those of any other debit
card. Even with these advantages,
signature debit cards were not immediately successful. Their success occurred

only after credit cards were commonly
accepted in more price-sensitive retail
segments, and virtually all merchants
were using modern electronic terminals
to authorize transactions.28
When network effects and
dynamic issues are both important, as
they appear to be in this industry,
policymakers face a difficult problem in
deciding what remedies, if any, will
benefit consumers in the long run. On
the one hand, network rules and pricing
strategies may be essential elements in
the successful launch of a payment card
network and its subsequent expansion.
On the other hand, once a payment
network is well established, it is possible
the same rules can lead to their
overutilization and to pricing well in
excess of costs. A further complication is
that the pricing strategy of existing
payment networks affects how and
when newer and presumably better
forms of payment emerge. New forms of
payment must overcome any subsidies
consumers receive when using today’s
payment instruments, and this may
require offering subsidies of their own.29
Policymakers should take all of these
factors into account when examining
competition among consumer payment
networks. BR

Both of these developments were promoted
by offering interchange fees below the
standard rate. For a more detailed description of the evolution of the debit card market
in the U.S., see the book by Evans and
Schmalensee and the article by Steven
Felgran and the one by Felgran and R.
Edward Ferguson.
28

John Caskey and Gordon Sellon argue that
the adoption of debit cards in the U.S. was
delayed in part by subsidies resulting from the
pricing of consumer check transactions.
Today, higher interchange fees paid on debit
transactions coincide with debit cards’
displacement of checks in many consumer
transactions.
29

www.phil.frb.org

REFERENCES
Baxter, William F. “Bank Interchange of
Transactional Paper: Legal and Economic
Perspectives,” Journal of Law and Economics,
26, 1983, pp. 541-88.

Felgran, Steven D. “From ATM to POS
Networks: Branching, Access, and Pricing,”
New England Economic Review, May/June
1985, pp. 44-61.

BNA Banking Report, “MasterCard, Visa to
Pay $3 Billion to Resolve Card Suit; Will
Modify Debit Card Policy,” Vol. 80, May 5,
2003, pp. 739-40.

Felgran, Steven D., and R. Edward
Ferguson. “The Evolution of Retail EFT
Networks,” New England Economic Review,
July/August 1986, pp. 42-56.

Carlton, Dennis W., and Alan S. Frankel.
“The Antitrust Economics of Credit Card
Networks,” Antitrust Law Journal, 63, 1995,
pp. 643-68.

Frankel, Alan S. “Monopoly and Competition in the Supply and Exchange of Money,”
Antitrust Law Journal, Vol. 66, 1998.

Caskey, John P and Gordon H. Sellon, Jr.
.,
“Is the Debit Card Revolution Finally
Here?” Federal Reserve Bank of Kansas
City Economic Review, Fourth Quarter
1994, pp. 79-95.
Chakravorti, Sujit, and William R.
Emmons. “Who Pays for Credit Cards?”
Federal Reserve Bank of Chicago Public
Policy Series, February 2001 (EPS-2001-1).
Chakravorti, Sujit, and Ted To. “A Theory
of Credit Cards,” mimeo, Federal Reserve
Bank of Chicago, 2002.
Chang, Howard, and David S. Evans. “The
Competitive Effects of the Collective
Setting of Interchange Fees by Payment
Card Systems,” Antitrust Bulletin, 45, 2000,
pp. 641-77.
Credit Cards in the U.S. Economy: Their
Impact on Costs, Prices, and Retail Sales.
Washington: Board of Governors of the
Federal Reserve System, 1983.
Debit and Credit Card Schemes in Australia:
A Study of Interchange Fees and Access.
Reserve Bank of Australia and the
Australian Competition and Consumer
Commission, 2000.
Economides, Nicholas S. “Network
Economics with Application to Finance,” in
Financial Markets, Institutions & Instruments, Vol. 2, 1993, pp. 89-97.
Evans, David S., and Richard Schmalensee.
“Some Economic Aspects of Antitrust
Analysis in Dynamically Competitive
Industries,” NBER Working Paper No.
8268 (2001).
Evans, David S., and Richard Schmalensee.
Paying with Plastic: The Digital Revolution in
Buying and Borrowing. Cambridge, MA:
MIT Press, 1999.
Evans, David S., and Richard Schmalensee.
“Economic Aspects of Payment Card
Systems and Antitrust Policy Toward Joint
Ventures,” Antitrust Law Journal, Vol. 63,
www.phil.frb.org
1995, pp. 861-901.

Gans, Joshua, and Steven P King. “The
.
Neutrality of Interchange Fees in Payment
Systems,” mimeo, University of Melbourne,
2001.
Gans, Joshua S., and Steven P King.
.
“Regulating Interchange Fees in Payment
Systems,” mimeo, University of Melbourne,
2001.
Gerdes, Geoffrey R., and Jack K. Walton II.
“The Use of Checks and Other Noncash
Payment Instruments in the United States,”
Federal Reserve Bulletin, August 2002, pp.
360-74.
Katz, Michael L. “Reform of Credit Card
Schemes in Australia II: Commissioned
Report,” Reserve Bank of Australia, August
2001.
Kitch, Edmund W. “The Framing Hypothesis: Is It Supported by Credit Card Issuer
Opposition to a Surcharge on a Cash Price?”
Journal of Law, Economics and Organization,
Vol. 6, 1990, pp. 217-33.
Mandell, Lewis. The Credit Card Industry:
A History. Boston: Twayne Publishers,
1990.
McAndrews, James J. “Network Issues and
Payment Systems,” Federal Reserve Bank of
Philadelphia Business Review, November/
December, 1997, pp. 15-25.
Nilson Report, The. February 2002 (no.
758) and June 2002 (No. 768).
Ramsey, Frank. “A Contribution to the
Theory of Optimal Taxation,” Economic
Journal, Vol. 31, 1927, pp. 47-61.
Reserve Bank of Australia. “Debit and
Credit Card Schemes in Australia: A Study
of Interchange Fees and Access,” October
2000.

Rochet, Jean-Charles, and Jean Tirole.
“Cooperation Among Competitors: The
Economics of Payment Card Associations,”
Centre for Economic Policy Research,
Discussion Paper 2101, 1999.
Schmalensee, Richard. “Payment Systems
and Interchange Fees,” Journal of Industrial
Economics, Vol. L, 2002, pp. 103-22.
“Study Regarding the Effects of the
Abolition of the Non-Discrimination Rule
in Sweden,” IMA Market Development,
AB (February 2000).
“Survey of Retail Payment Systems,” Chain
Store Age (December 1999).
“Symposium on Network Externalities,”
Journal of Economic Perspectives, Vol. 8,
1994, pp. 93-133.
Wright, Julian. “Optimal Card Payment
Systems,” mimeo, University of Auckland,
2002.
Wright, Julian. “The Determinants of
Optimal Interchange Fees in Payment
Systems,” University of Auckland,
Department of Economics Working Paper
220, 2001.
Cases
Broadcast Music, Inc. v. Columbia Broadcasting Co., 441 U.S. 1 (1979)
Case No. COMP/29.373 — Visa International. Official Journal of the European
Community (November 10, 2001)
Case No. COMP/29.373 — Visa International-Multilateral Interchange Fees.
Official Journal of the European Community
(July 24, 2002)
In re Visa Check/MasterMoney Antitrust
Litigation, No. 00-7699 (2d Cir 2001)
In re Visa Check/MasterMoney Antitrust
Litigation, 96-CV-5238 (E.D.N.Y. 2003)
National Bancard Corp. (NaBANCO) v. Visa
U.S.A., Inc. 596 F. Supp. 1231 (S.D. FL
1984), 779 F2d 592 (11th Cir 1986)
U.S. v. Visa U.S.A., Inc. 163 F. Supp. 2d.
322 (S.D. NY 2001)

Reserve Bank of Australia. “Reform of
Credit Card Schemes in Australia IV: Final
Reforms and Regulation Impact Statement,”
August 2002.
Business Review Q2 2003 23

Should Philadelphia’s Suburbs Help
Their Central City?

T

BY ROBERT P INMAN
.

he United States is unique in its
commitment to local government as the
primary provider of essential public services
and in its use of local taxes as the primary
means for paying for these services. The Philadelphia
metropolitan area is typical of the U.S. pattern. But
the city of Philadelphia faces the burdens and
responsibilities of all older central cities, including a
higher proportion of poor residents than its surrounding
suburbs. Such circumstances lead the city to impose
higher taxes on city residents, workers, and businesses.
Raising revenues through higher taxes, however,
becomes self-defeating when tax rates drive people and
businesses away. The result is a weaker city and
regional economy. How can Philadelphia strengthen its
finances? Bob Inman proposes a targeted program of
suburban assistance to lower the commuter wage tax
and presents evidence that such a program is likely to
benefit city and suburban residents alike.

Should the residents of
Philadelphia’s suburbs — Bucks,
Chester, Delaware, and Montgomery
Bob Inman is
Miller-Sherrerd
Professor of Finance
and Economics,
Wharton School,
University of
Pennsylvania. When
he wrote this article,
he was a visiting
scholar in the
Research Department of the Philadelphia Fed.

24 Q2 2003 Business Review

counties — contribute to the financing
of services provided to city residents and
businesses? The United States is unique
in its commitment to local government
as the primary provider of essential
public services and in its use of local
taxes as the central means for paying for
these services. The Philadelphia
metropolitan area is typical of the U.S.
pattern. In the five counties that
comprise the Pennsylvania portion of the
Philadelphia metropolitan area, there
are 243 municipal governments and 62

separate school districts servicing a
combined population of 3.85 million
residents.1 Local property taxes and,
particularly in the case of the Philadelphia region, local resident wage and
income taxes are the primary sources of
locally raised revenues. In the Delaware
Valley, property taxes account for 58
percent and wage/income taxes an
additional 28 percent of all locally raised
revenues. Locally raised revenues pay
for 77 percent of local government and
school district spending in the fivecounty Philadelphia area. Fiscal
transfers from the state cover most of the
remaining 23 percent.
While only one of many local
governments in the metropolitan area,
the city of Philadelphia is arguably the
region’s economic, cultural, and
entertainment center. The city has 34
percent of the five-county region’s jobs.
There are 12 Fortune 500 corporate
headquarters in the five-county area,
and eight of those headquarters are
located in Philadelphia. Four of the
nation’s 100 largest law firms have their
home offices in Philadelphia. Philadelphia’s four medical schools are
national leaders in patient care and
medical research. Together, the four
schools currently receive more than
$550 million a year in National Institutes
of Health funds for faculty research.
Higher education is a major industry for
the region, and 46 percent of the

There are also four New Jersey counties
included in the official definition of the
Philadelphia metropolitan statistical area
(MSA). For the reasons noted in footnote 9,
these counties are not included in this paper’s
policy analysis.
1

www.phil.frb.org

region’s college and graduate school
enrollees attend Philadelphia universities. The Philadelphia Orchestra, the
Curtis Institute of Music, the Philadelphia Museum of Art, the Franklin
Institute, and the Philadelphia Zoo are
world-recognized centers of arts and
science education. Philadelphia has the
area’s four major league sports franchises. There are 13 professional
theaters providing full seasons, two
professional dance companies, and nine
music venues featuring artists from
major record labels. Of the region’s 318
restaurants rated excellent by the 2002
Zagat’s Guide, 220 are in Philadelphia.
Though it is the economic and
entertainment center of the region,
Philadelphia also faces the burdens and
responsibilities of all older central cities.
Philadelphia is home to most of the
region’s poor and elderly households.
While the city has 39 percent of the
region’s population, it has 70 percent of
the region’s poor. Philadelphia residents
also face significantly higher rates of
crime. In 1998, the rate of violent crime
was 1465 per 100,000 residents in
Philadelphia, yet only 286 per 100,000
residents in the suburbs; the rate of
property crime was 5855 per 100,000 in
Philadelphia compared with 2503 per
100,000 in the suburbs. These higher
service burdens from poverty and crime
necessarily translate into higher city tax
rates. The overall tax burden for a typical homeowner in Philadelphia is 14.4
percent of family income but only 9.5
percent for an identical family living
and working in the suburbs. A similar
differential tax burden holds for a
typical city firm. Leaving the city for a
suburban location will lower a firm’s
effective state and local tax burden on
profits from 16.5 percent to 13.2
percent.2
The estimates of local tax burdens are
available from my biennial report on local
taxation, “Local Taxes and the Economic
Future of Philadelphia: 2002 Report.”
2

www.phil.frb.org

The unique fiscal burdens that
Philadelphia faces contribute importantly to these tax differentials. The
danger is that these added burdens —
and resulting higher tax rates — will
undermine the city’s economic, cultural,
and entertainment advantages to the
detriment of all residents in the fivecounty region. Large cities often have
significant cost advantages, known as
agglomeration economies, in producing
and providing goods and services.
These agglomeration economies arise
when firms, retail stores, or cultural
activities are concentrated in common
and usually small geographic areas

city and suburban homeowners both
gain; regional home values are predicted
to rise by $2.1 billion, or about $2250 per
family.
THE ECONOMIC CONNECTION
BETWEEN CITY AND SUBURBS
There are two potential links
from the city economy to suburban
residents’ economic welfare: jobs and
wage income for suburban commuters
and the market price for city-produced
goods and services purchased by
suburban firms and residents. For a
typical suburban resident, the primary
economic advantage of a strong city

Though it is the economic and entertainment
center of the region, Philadelphia also faces
the burdens and responsibilities of all older
central cities.
within the city. High tax rates and low
quality public services, however, may
drive firms and middle and upper
income households from the city. As
firms and families exit, city agglomeration economies are lost. The loss of
agglomeration economies leads to higher
prices for city-produced goods and
services. Population and income grow
more slowly or decline, and house values
in the city and suburbs fall. In the end,
the region as a whole loses, not just the
central city.
The solution is to strengthen
city finances. To this end, suburban
residents might wish to make a contribution, most realistically done through
adjustments in state assistance to local
governments. If done correctly,
suburban fiscal assistance to the central
city can be a high-return investment in
suburban jobs, growth, and house values.
Here I offer one proposal for
such assistance, a targeted program of
suburban assistance to lower the
commuter wage tax. With this reform,

economy lies in the ability of city firms
to provide goods and services at prices
lower than (or, equivalently, at a quality
higher than) what might be available
from suburban firms or from firms
outside the metropolitan region.3 The
central city’s economic advantage can
arise from either of two sources: natural
advantages because of the city’s
proximity to an important production
input such as power or raw materials
(e.g., Pittsburgh’s history as a steel
production center), or agglomeration
advantages facilitated by the density of
firms and households within the city. For
U.S. cities today, the likely source of any
advantage is agglomeration economies.
Agglomeration economies
benefit producers and consumers of cityproduced goods and services. A high

See the article by Richard Voith for more
about the advantages commuter suburbanites
derive from a strong city economy, with
particular reference to Philadelphia
commuters.
3

Business Review Q2 2003 25

density of firms within the same industry
— called Marshallian agglomeration in
honor of Alfred Marshall’s initial analysis
— leads to lower shipping costs for firms’
inputs when there are economies of
scale in transportation (e.g., coal and
iron ore to the steel mills of Pittsburgh
and Gary). The density of firms may
also lower labor costs in industries when
laid-off workers from a declining firm
are quickly hired by an expanding firm.
This will be the case in industries where
brand loyalty is weak and current fads
define consumer demands, for example,
the fashion industry in New York, the
entertainment industry in Los Angeles,
and the “dot.com” industries of Silicon
Valley. Having many firms in the same
local labor market reduces the unemployment risk to workers with unique
talents and therefore allows all firms in
the market to pay a lower wage. For
much the same reason, a high density of
firms in the same industry may also
encourage supplier innovation and
specialization, again lowering firms’
production costs. Furthermore, low-cost
production technologies are likely to be
more quickly copied when firms and
workers are in close proximity. These
idea “spillovers” may occur within or
across industries, an advantage called
Jacobian agglomeration for Jane Jacobs’
insightful analysis of growing city
economies. Finally, a high density of
households and firms gives rise to
agglomeration advantages in retailing
and consumer services, for example,
dining, specialty shopping, and entertainment. In Philadelphia, South Street,
Rittenhouse Square, and the Avenue of
the Arts are lively examples.
A growing body of economic
research has demonstrated the presence
and importance of agglomeration
economies in regional economies. This
research has found that concentration
of industry employment has a statistically significant and quantitatively
important effect on plant productivity.

26 Q2 2003 Business Review

For example, Mark Beardsell
and Vernon Henderson found that
doubling the number of computer firms
in a given location increases the
productivity of those firms as much as 17
percent; little wonder, then, that the
Silicon Valley has become the world’s
leader for computer industry production
and innovation. Though not as
dramatic, the findings of Antonio
Ciccone and Robert Hall showed that
overall employment concentration also
improves worker productivity; doubling
county-level employment density within
a state improves all worker productivity
in that state by 6 percent. Ciccone has
found a similar effect of employment
concentration on worker productivity in
European firms.4 Finally, Stuart
Rosenthal and William Strange find that
the benefits of employment density
occur within small geographical areas
and are typically exhausted beyond a
distance of five miles. The RosenthalStrange results suggest, importantly, that
the spatial reach of agglomeration
advantages will typically be confined
within a political jurisdiction. For
historical reasons, that political jurisdiction is most often the region’s central
city.
The gain to city and suburban
residents of living in or near a productive central city comes largely from their
ability to buy city-produced goods and
services at comparatively low prices.

In addition, there is preliminary evidence
that an increased density of firms not only
increases the initial equilibrium level of
regional production and incomes but it may
also stimulate higher economic growth as
well, primarily through the sharing of ideas
and innovation. See the article by Edward
Glaeser, Hedi Kallal, Jose Scheinkman, and
Andrei Shleifer and the one by Vernon
Henderson, Ari Kuncoro, and Matt Turner. In
his Business Review article, Gerald Carlino has
provided a valuable survey of the theory and
evidence on agglomeration economies and the
economic performance of cities and regions.
4

This advantage is larger the more goods
and services suburban households and
firms buy from city producers and the
greater are the cost and shipping
advantages those city producers have
over their closest competitors. For
suburban residents in the Philadelphia
metropolitan area, it is cheaper to buy
expert legal advice, accounting services,
life-saving medical care, or first-run
professional entertainment from
Philadelphia businesses and venues than
from those in New York, Baltimore, or

The gain to city and
suburban residents of
living in or near a
productive central city
comes largely from
their ability to buy
city-produced goods
and services at
comparatively low
prices.
Washington, D.C. James Rauch has
shown that the ultimate beneficiaries of
access to low-cost goods and services
due to agglomeration economies are the
region’s workers and home owners.
Wages and house prices are higher in
economically more efficient regions.
Low quality public services or
high city taxes are one important factor
that might undo this economic advantage, however. On this point the
evidence is clear. My research for
Philadelphia —now confirmed in
companion studies for Houston,
Minneapolis, and New York City —
shows unequivocally that high city taxes
unmatched by compensating service
benefits will drive middle- and upperincome taxpayers and businesses from

www.phil.frb.org

the city, first to the surrounding suburbs,
but just as likely, to other regions of the
country as well.5 For example, I
estimate that Philadelphia lost 207,000
jobs over the past 30 years solely because
of increases in the city’s wage tax rate.
Most damaging for the location of
businesses in Philadelphia is the city’s
nonresident portion of the wage tax, a
tax whose burden is largely shifted back
onto city businesses as workers with
suburban job alternatives require a wage
increase to compensate for the city’s tax.
The recent wage tax cuts proposed by
former Philadelphia mayor Ed Rendell
and the current mayor, John Street, are
estimated to have restored approximately 12,000 of the jobs previously lost.
Similar adverse effects of high taxes on
city property values and city jobs have
shown up in Houston, Minneapolis, and
New York City.
The loss of city jobs due to
high city taxes will mean a reduction in
the city’s production advantage because
of lost agglomeration economies. In our
2002 study, Andrew Haughwout and I
tested empirically the argument that
weak public finances in the central city
can undermine private-sector economic
performance, for both city and suburban
residents. Our study seeks to explain
the growth in city and suburban
incomes, populations, and house values
over the decade 1980 to 1990 for the 195
largest metropolitan areas in the United
States. Weak city finances are measured
in this study by four separate indicators
of budgetary pressure: the share of city
spending paid for through taxation of

businesses and middle-class residents;
whether the city is required by state law
to bargain with its public employee
unions; whether the city lacks mayoral
veto control over city budgets; and the
fiscal burden of city poverty (Table 1).
Columns (1) and (2) of Table 1
show what happened to house values in
the average U.S. central city and its
surrounding suburbs because of weak
city finances during the 1980s, the most
recent decade for which we have
complete information.6 On average, the
share of city taxes borne by middle and
upper income residents and firms rose 1
percent over the decade (Change in
SHARE). The consequence was to
depress house values in the average city
by $3638, or about 7 percent of the
initial 1980 value. Importantly, the
house values in the average suburb fell
too, by $2468, or about 4 percent of their
initial 1980 values. (See Understanding
the Economics Behind Table 1.)
Cities that are required by
state law to hire their labor services
exclusively from public employee unions
(BARGAIN) typically face higher labor
costs, and this raises city taxes. Again,
higher taxes without compensating
service benefits drive households and
firms from the city, and city and
suburban house values decline from
what they might have been had the city
retained the right to hire nonunion
public employees — that is, to contract
out. The average city house loses $5358,
or 11 percent of its initial 1980 value, by
being in a strong union city. Suburban

Our full study reports the effects of weak
city finances on city and suburban incomes
and population as well; see Haughwout and
Inman (2002). Elsewhere we have shown that
changes in house values are the best single
predictor of changes in the economic welfare
of residents in a metropolitan area, so I shall
focus on these results here; see also the 2001
article by Haughwout and Inman.
6

See the article by Andrew Haughwout,
Robert Inman, Steven Craig, and Thomas
Luce. The conclusion that high taxes and
low services depress local economies is now a
well-established fact more generally, and
Timothy Bartik provides an excellent review
of this research.
5

www.phil.frb.org

house values fall too, now by $4047, or 8
percent of their 1980 values.7
The same adverse effects on
city and suburban house values, though
not as large, are seen when city budgets
are decided by a majority of wardelected city council members, unchecked by a constitutionally strong
mayor with veto powers (GOVERNANCE). The budgets of such
governments typically favor neighborhood services, with tax burdens
allocated toward business. We should
expect to see an exit of businesses from
the city, lost agglomeration economies,
higher prices for city goods, and for this
reason, lower city and suburban house
values. This is what we find. City house
values are lower in weak governance
cities by an average of $1948 (4 percent
of 1980 values) while house values in
suburbs surrounding a weak governance
city fall an average of $3035, or 6
percent of 1980 values (see Table 1).
Interestingly, in cities with weak fiscal
governance, we see suburban house
values falling more than city house
values, but this too makes sense if city

While the typical city or suburban resident
loses by living in or near a strong union city,
unionized city workers are net winners.
Although the value of a unionized city
employee’s house falls like the house values of
all residents, the worker is compensated by
his or her gain in personal income. Estimates
of the premium unionized city employees earn
above what they might earn in their next best
private-sector job are typically 3 to 8 percent
of the worker’s wage; see the article by
Richard Freeman and Robert Valletta. For a
unionized city employee earning $30,000 a
year, this is an annual wage premium of $900
to $2400 per year. Unionization reduces
house values by $5358 in the city and $4047
in the suburbs, implying an equivalent
annual loss in value of $270 a year for a city
house and $200 per year for a suburban house
when interest rates are 5 percent. Conservatively, then, the typical unionized city worker
in a strong union city gains $900 to $2400 in
wages each year and loses from $200 to $270 a
year in declining house value. On balance,
city employees gain when working in cities
with strong public-sector unions.
7

Business Review Q2 2003 27

TABLE 1
City Finances and Metropolitan Area Home Values
Estimates from a regression model
based on data from 195 U.S. metro areas

Estimates from a simulation model
of the Philadelphia area economy

Estimated Change in
Average U.S. City
Home Value
(1)

Estimated Change in
Average U.S. Surburban
Home Value
(2)

Estimated Change in
Average Philadelphia
City Home Value
(3)

Estimated Change in
Average Philadelphia
Suburban Home Value
(4)

Change in SHARE

-$3638
(974)

-$2468
(873)

-$265

-$1224

BARGAIN

-$5358

-$4047

-$7184

-$5902

(1739)

(1563)

-$1948

-$3035

(1052)

(946)

—

—

-$12,345

-$6696

-$410

-$15

(2460)

(2212)

GOVERNANCE

Change in POVERTY

Columns 1 and 2: Source: Table 6, Haughwout and Inman (2002).
The standard error of the estimated change in the city median house
value is presented in parentheses below each estimate. Columns 3
and 4: Estimates of changes in the Philadelphia and suburban median home values are computed using an equilibrium political
economy model of the Philadelphia MSA, calibrated to match the
Philadelphia MSA for the decade 1980-90. The simulation model is
described in an appendix to this article. Estimated changes in city
and suburban home values are computed for each of the following
four changes in the underlying structure of Philadelphia’s public
finances: Change in SHARE, allowing for the increase in the middle
and upper income families’ share of city taxes for 1980-90 because

residents gain at the expense of city
businesses. Unfortunately, what city
residents seem to gain from their budget
is more than offset by what they lose
from the decline in the city’s economy.

28 Q2 2003 Business Review

of the change in the percent of Philadelphia’s population who are
poor or over the age of 65; BARGAIN allowing for a 20 percent
increase in the real cost of city services for 1980-90 because of the
32.9 percent increase in the real cost of city workers’ compensation
over the decade; the constitutional form of city GOVERNANCE
remained constant over the decade as the city retained its strong
mayor form of government and thus there is no impact on city
budgetary costs nor city and suburban home values; Change in
POVERTY, allowing for the decline in the city’s rate of poverty
from 20.6 percent to 20.3 percent and the mandated increase in the
city’s real (inflation-adjusted) contribution for services for lowincome households of approximately $200 per poor household.

The most damaging change
for our nation’s largest cities during the
1980s came from the growth in the rate
of urban poverty (Change in POVERTY) and associated increases in city

spending and taxes. If the city’s rate of
poverty increased — and this was the
case for most U.S. cities — the resulting
added tax burden must be spread over
relatively fewer middle-class and

www.phil.frb.org

UNDERSTANDING THE ECONOMICS BEHIND TABLE 1

The estimated effects of
weak city finances reported in Table
1, columns (1) and (2), are statistical
estimates of the effects of weak city
finances on city and suburban home
values for a sample of the largest 195
U.S. metropolitan areas for the
decade 1980-1990. The estimates in
Table 1, columns (3) and (4), are
estimates for the effects of changes in
the same city fiscal variables on
Philadelphia city and area suburban
home values for the decade 19801990 derived from an economic
model of the metropolitan area
economy. An Appendix to this
article provides a brief summary, and
Haughwout and Inman’s 2002 article
provides the technical details. An
important question to ask of any
statistical estimation or model
prediction is: Do the numbers make
economic sense?
These do. Here is a working
example using typical family incomes,
consumption, and interest rates for
the 1980s. Start with a suburban

family that, over the 1980s, spent
$50,000 per year, of which $10,000 is
allocated to city-produced goods and
services. Remember, these goods do not
have to be consumed in the city, but
simply made there or processed and
shipped by city businesses. If a higher
city tax share leads to the exit of city
businesses and lost agglomeration
economies, how much would prices of
city-produced goods have to increase to
justify a $2500 decrease in suburban
house values (Column 2, Table 1)?
Prices of city-produced goods would
have to rise only 1.25 percent. Suburban
residents would pay $125 more per year
for their city-produced goods, and they
would lose this $125 every year. An
annual loss of $125 is economically
equivalent to the estimated loss in house
value of $2500 when the real interest
rate is 5 percent.
What about city residents?
They suffer the same losses from high
prices for city goods, so their house
values should also fall by $2500 because
of lost agglomeration. But as seen in the

first entry of column 1 in Table 1, the
estimated average decline in city
house values is $3600. The additional
$1100 loss in value must come from
the direct adverse effects of higher
city tax shares. In our sample, on
average, the city budget is $6000 per
family and the typical middle-income
resident’s share of city taxes is about
0.50. So an increase of 0.01 in the tax
share means an increase of about $60
each year in tax payments ((0.51-0.50)
x $6000 = $60).
But again this is an annual
loss that will be economically equivalent to a one-time value loss of $1200
when real interest rates are 5 percent.
City residents lose $2500 because of
lost agglomeration and an additional
$1200 because of higher taxes. In this
example, the total loss for a typical city
house will then be $3700, again very
close to the estimate in Table 1.
Similar calculations can be made for
all the numbers in Table 1; they are all
plausible.

business taxpayers. The rise in city taxes
because of the average increase in the
rate of city poverty, which was 3
percent, led to an average loss in city
house values over the decade of
$12,345, a decline of more than 25
percent.
Rising city poverty, rising city
taxes, and a shrinking city economy
affect the suburbs, too. We estimate the
average decline in suburban house
values over the decade from the growth
in city poverty equaled $6696, or 13

percent of the original 1980 suburban
house value. When central cities
become poorer or the fiscal burden from
each poor household increases, city
residents lose, but importantly, so too do
residents in the city’s surrounding
suburbs. The path from city poverty to
suburban house values may be roundabout — from higher city poverty and
city taxes to a weaker city, then
regional, economy — but the impact is
significant nonetheless.
Finally, each of the four

sources of weak city finances is largely
outside the direct control of any city’s
mayor. To be sure, poor fiscal management by a city’s mayor or elected
council will also lead to higher city taxes
and lower city and suburban property
values, but that is not what we are
measuring in Table 1. Tax laws, the rules
of labor bargaining, the city charter,
federal and state mandates, the rate of
city poverty — these are the determinants of weak city finances as measured
here, and each is given to, not chosen

www.phil.frb.org

Business Review Q2 2003 29

by, the mayor. If a mayor is dealt weak
fiscal institutions, it should be no
surprise that the city and its region lose
economically over time.
THE PHILADELPHIA
CONNECTION
The estimated effects of weak
city finances on city and suburban
house values reported in columns (1)
and (2) of Table 1 are for a national
average city and its suburbs. While
Philadelphia is included in the
Haughwout-Inman national study, those
results cannot be applied directly to
Philadelphia. Perhaps Philadelphia was
one of the lucky cities dealt a winning
fiscal hand. Unfortunately, my estimates
from an economic model of the
Philadelphia metropolitan economy
show this is not the case.8 Philadelphia
faced many of the same fiscal difficulties during the 1980s as our national
sample. The results in columns (3) and
(4) of Table 1 show that while the losses
were not as large as those felt nationally,
Philadelphia and its suburbs suffered

The estimates in columns 3 and 4 of Table 1
for the Philadelphia economy are computed
using a general equilibrium simulation model
for the Philadelphia metropolitan area,
calibrated to match the economic and
political structure of the five Pennsylvania
counties in the Philadelphia MSA for the
decade 1980 to 1990; see the Appendix.
Ideally, I would have replicated the statistical
analysis of the national city sample for a
sample comprising only Philadelphia and its
suburbs, but unfortunately, this is not possible
because the required number of years of
suburban data are not available. In the
simulation model, the fundamental economic
relationship that determines the results in
columns 3 and 4 of Table 1 is the efficiency
advantage of city agglomeration economies;
the stronger these economies are, the larger
will be the adverse effects of weak city public
finances. For this analysis, I select a very
conservative elasticity of Philadelphia city
output with respect to city firm density of
only 0.01; the national average elasticity of
worker productivity with respect to firm
density is 0.06, as estimated by Ciccone and
Hall.
8

30 Q2 2003 Business Review

significant economic losses during the
1980s from weakened city finances.9
(See An Economic Model of Philadelphia
and Its Suburbs.)
First, as was true nationally, the
share of Philadelphia taxes borne by its
middle- and upper- income households
(Change in SHARE) rose 1 percentage
point, from 50 percent in 1979 to 51

suburban home values. This appears to
have been the case in Philadelphia as
well. Over the decade, Philadelphia’s
public-employee unions were able to
negotiate labor contracts that increased
real — that is, inflation-adjusted —
compensation for city employees 33
percent, equal to an annual real rate of
growth of 2.89 percent in city workers’

The city’s adverse fiscal changes over the
1980s are estimated to have reduced the
value of a typical city and suburban house
each by about $8000.
percent by 1989. The causes in
Philadelphia were the aging of the city’s
population and the city’s continued loss
of manufacturing jobs. The rise in
middle-class tax burdens led to a
predicted fall in city and suburban
house values (Table 1, columns 3 and 4).
Second, Philadelphia’s public
employees enjoy an exclusive right to
bargain with the city. Thus, Philadelphia
qualified as a strong union city (BARGAIN) in the Haughwout-Inman
national study reported in columns (1)
and (2) of Table 1. Each strong union
city in that study had its own bargaining
experience with unions, but on average,
those unions increased labor costs and
city taxes and depressed city and

The four New Jersey counties of the
Philadelphia MSA are not included in the
simulation analysis for two reasons. First, New
Jersey now rebates the Philadelphia commuter
tax for New Jersey suburban residents. This
different treatment of an important city tax
requires separate analyses for the Pennsylvania and New Jersey suburbs of Philadelphia.
Second, since our focus is on the economic
returns to reforming city and suburban
financing of city services, I have chosen what
seemed to be the most politically realistic
group of suburban counties to be included in
any such reforms. Those counties are in
Pennsylvania alone.
9

pay. This increase in Philadelphia’s real
costs of public employees’ labor services
led to higher city taxes and, as was true
for the national sample of strong union
cities, significantly lower city and
suburban house values. Our estimates of
the effect that strong public-employee
unions have on Philadelphia city and
suburban house values appear in
columns (3) and (4) of Table 1.
Third, there was no change in
city governance (GOVERNANCE)
over the decade; Philadelphia had, and
continues to have, the strong mayor
form of city government. This row in
Table 1, therefore, shows no changes.
Finally, while the burden of
city poverty on Philadelphia’s budget is
high and has an important negative
effect on the performance of the
regional economy, as we will see below,
the 1980s did not add significantly to
that burden. Thus, the net economic
effects of the change in the rate of
poverty (Change in POVERTY) were
small as well. The share of Philadelphians living in poverty fell slightly over
the decade from 0.206 to 0.203. There
was, however, a small offsetting increase
in the city’s cost of serving that population.10 Overall, poverty’s fiscal burden
on the city rose only slightly, and the
estimated additional damage done to

www.phil.frb.org

city and suburban house values is barely
noticeable.
All together, the city’s adverse
fiscal changes over the 1980s are
estimated to have reduced the value of
a typical city and suburban house each
by about $8000, a loss in value of 9
percent for Philadelphia home owners
and 6 percent for suburban home
owners.11 Far and away the most
important cause of these economic losses
for our region was the increase in city
taxes required to fund the significant
growth in the real compensation of the
city’s public employees.
INVESTING IN STRONGER
CITY FINANCES: A STRATEGY
FOR GROWTH IN THE
PHILADELPHIA REGION
Both in the nation and in
Philadelphia, weak city finances lead to
the exit of mobile city firms and
households, a less efficient city
economy, and lower incomes and house
values for the region as a whole. Strong
city finances protect a city’s economic
efficiency and a region’s income and
wealth. What might we do to
strengthen Philadelphia’s city finances,
and what will be the gains to city and
suburban residents from such a strategy?
Three Possibilities. The
analysis in Table 1 identifies three
possible directions in which the current
structure of Philadelphia’s finances
might be improved. First, reduce the
city’s relative tax burden on mobile
middle-class households and city firms.

10

See the article by Anita Summers and Lara
Jakubowski.
11

The decline in house values from the combined fiscal changes will not equal the sum of
the three isolated changes because the exit of
households and firms has accelerating effects
in the presence of agglomeration economies.
Large adverse fiscal changes will be proportionally more harmful than small changes.

www.phil.frb.org

Second, control the ability of the city’s
public employee unions to win favorable
compensation packages with greaterthan-inflation increases. Third, reduce
the city’s fiscal obligation for services to
lower income households.
On its own, Philadelphia has
already made significant progress on two
of these three fronts. First, from 1995 to
today, the city has lowered its wage-tax
rates 9.27 percent and its gross-receipts
tax rates 29.3 percent. Increases in the
rates of these two taxes over the past 30
years have caused significant damage to
the city’s economy; so lowering these
rates is an important step toward
restoring our fiscal competitiveness.12
Second, since 1992, and in sharp
contrast to the 1980s, Philadelphia’s

tax burden of poverty spending as a
percent of county residents’ income is
roughly four to seven times higher in
Philadelphia than in the suburban
counties. While the residents of Bucks,
Chester, Delaware, and Montgomery
counties pay only 0.17 percent to 0.38
percent of their income in taxes to fund
poverty services, Philadelphia residents
pay taxes equal to 1.4 percent of their
income to fund city-provided poverty
services.
The root causes of these
spending and tax disparities are, first, the
geographical concentration of the
region’s poor and low-income elderly
households within the city, a concentration due in large measure to the
availability of older, lower cost housing

Poverty spending in Philadelphia is greater
than comparable spending in all suburban
counties combined.
compensation per public employee has
declined, in real dollar terms, at an
annual rate of 0.60 percent. The city’s
improved labor compensation policy has
allowed balanced-budget tax reductions, significantly improving the city’s
ability to attract firms and households.
The one important fiscal weakness that
has not yet been addressed is the city’s
continuing high budgetary obligation for
support of its poor population.
The direct tax costs needed to
fund poverty-related county spending in
Philadelphia and in each of the four
surrounding Pennsylvania counties are
reported in Table 2. Poverty spending in
Philadelphia is greater than comparable
spending in all suburban counties
combined (Table 2). Further, the direct

within the city,13 and second, the state
of Pennsylvania’s decision to make
counties the primary providers and
administrators of poverty-related
services. Seventy percent of the fivecounty region’s poor live in Philadelphia,
and because the city is also legally a
county, Philadelphia must assume
primary fiscal responsibility for the
unreimbursed portion of the services
provided to those families. Philadelphia
spends more per taxpayer for poverty
services than the suburban counties, not
because poverty spending is a successful
election strategy or the city’s middle
class is particularly generous, but
because, as the region’s oldest and
largest city, it has more poor families as
residents and because it is a city-county,

12

For the most recent analysis of the effects of
city taxation on city business, see the article by
Haughwout, Inman, Craig, and Luce.

13

See the article by Edward Glaeser and
Joseph Gyourko.

Business Review Q2 2003 31

TABLE 2
Direct Tax Cost of Poverty Spending to County Governments
in the Philadelphia Region: FY2002
(Millions of Dollars)

County
(County % Poor)

Bucks
(4.92%)
(1)

Chester
(4.88%)
(2)

Delaware
(8.31%)
(3)

Montgomery
(4.55%)
(4)

Philadelphia
(19.74%)
(5)

Region
(11.93%)
(6)

PUBLIC
HEALTH

$3.816m

$12.951m

$0m

$1.387m+

$76.203m

$94.357m

HUMAN
SERVICES

$5.198m

$8.169m

$4.992m

$53.069m+

$67.993m

$139.421m

CORRECTIONS

$39.485m

$20.828m

$18.634m

$33.102m

$183.120m

$295.169m

EMERGENCY
SERVICES

$6.658m

$1.545m

*

*

$15.564m

$23.767m

TOTAL
($ per Non-Poor)
(% of Income)

$55.157m
($97.07)
(0.34%)

$43.493m
($105.47)
(0.32%)

$23.626m
($46.78)
(0.17%)

$87.558m
($122.33)
(0.38%)

$342.880m
($281.52)
(1.4%)

$552.714m
($163.03)
(0.60%)

The direct tax cost to county residents of poverty spending is
defined as county poverty spending minus state and federal grants,
departmental earnings, and fees paid to the county for poverty
services, all reported in millions of dollars. In Delaware County, for
example, all of county spending in FY 2002 for public health was
supported by nontax dollars. The total dollar tax burden per nonpoor
household ($ per Non-Poor) is calculated as the total poverty
spending divided by the population of the county not below the
poverty threshold. The percent of county income required to support

state law demands it.14
Regionalization. In this
regard, Philadelphia stands in sharp
contrast to Pittsburgh. Pittsburgh too is
an older city, and the current percent of
Pittsburgh’s residents who fall below the
poverty threshold (19 percent) is almost

14

In a recent study of poverty spending by Philadelphia, I showed that the trend in city spending is unrelated to who is mayor or to the racial
and ethnic composition of city council. The
main determinant of the city’s poverty spending is the performance of the city and national
economies. That is, when economic performance
improves and there are fewer families in poverty, poverty spending falls.
32 Q2 2003 Business Review

the county’s tax cost of poverty (% of Income) is computed as the
total direct tax cost divided by total county residential income.
+
Montgomery County classifies $37.838 million for geriatric
centers as spending within human services; other counties classify
such services as part of the public health budget.
* In Delaware and Montgomery counties emergency services
for low-income households have been classified as part of the human
services budget.

identical to Philadelphia’s. But
Pittsburgh’s taxpayers share the burden
of financing services for their city’s poor
with the suburban residents of Allegheny County. As a consequence,
Pittsburgh city residents face the same
tax burden on income as their suburban
counterparts, only 0.23 percent.15 This
rate is significantly below the 1.4 percent
burden on income now paid by Philadelphia city residents. From the perspective
of regional economic growth and
welfare, the Pittsburgh metropolitan
area has the financing right. Large
disparities in the fiscal costs of regional
poverty between local jurisdictions

discourage firms and households from
moving to high poverty locations. If
these locations are also the region’s
important centers of agglomeration
economies — as is likely the case in
Philadelphia — the firms and households that create those economies leave
and economic inefficiency results. The
whole region loses.

15

Allegheny County’s direct tax costs for the
poverty-related services totaled $64.867 million in fiscal year 2002. The tax burden on a
county resident not classified as poor equaled
$62.48 per resident. This burden as a percent of
county residential income equaled 0.23 percent of county income.
www.phil.frb.org

One solution — the Pittsburgh
solution — is to regionalize the financing of poverty. The Pittsburgh area
achieved this efficient structure for
poverty financing by the luck of history.
Pittsburgh’s historic boundaries define a
geographically small city within a
geographically large county. The
Philadelphia region has not been so
lucky. To solve its financing inefficiency, the region must fashion a clear
policy for sharing the five counties’ cost
of regional poverty. If the sharing is
done correctly, however, everyone —
city and suburban residents alike — can
benefit. The best policy will entail a de
facto transfer of approximately $191
million a year in poverty relief from the
suburbs to the city, with the transfer tied
to a required proportional reduction in
the city’s nonresident wage tax. If
implemented, this policy is estimated to
increase the combined economic wealth
of city and suburban home owners by
$2.1 billion, an average of about $2250
per family.
Here is how a policy to transfer
funds and reduce the commuter tax
rate might work. The budget data in
Table 2 allow us to calculate the
required suburb-to-city transfer needed
to ensure uniform regional financing of
regional poverty. To meet the total
regional tax burden from poverty of
$552.7 million for fiscal year 2002, we
need a uniform regional income tax rate
of 0.60 percent. Suburban residents have
already made a contribution toward
their regional share, but in all cases, it is
below the uniform regional rate. For
example, the taxpayers of Bucks County
have already contributed 0.34 percent of
county income (Table 2) toward the
target contribution of 0.60 percent; so
the reform policy would ask those
residents to pay an additional 0.26
percent, or in total about $42.622 million, toward the reform policy. Similar
calculations can be completed for each
of the other three suburban counties.

www.phil.frb.org

For fiscal year 2002, the total
reform contributions from the four
suburban counties would equal $191
million, or about $220 per suburban
family.16 This total would then be paid
to the city to lower its poverty-related
tax burden. Importantly, no money
need actually change hands among the
five counties. Since each suburban
county receives from the state povertyrelated grants and reimbursement
revenues greater than its required
contribution for uniform regional
poverty financing, the state can
implement the regional policy by
reallocating a portion of suburban grants
to Philadelphia. There is no need for
regional taxation or regional government
to implement poverty-financing reform.
Transfer of Money. How the
money is given to Philadelphia matters,
however. Table 3 estimates what might
happen to the values of typical city and
suburban houses and to total house
values regionwide when the reform
transfer is given to the city in one of
three ways. For purposes of comparison,
Table 3 also reports census year 2000
house values for the current “No
Reform: Status Quo” policy. Reform
Policy 1 gives the poverty-relief funding
to the city with “no strings attached” —
that is, the city is free to allocate the
funds any way it wishes.
Reform Policy 2 requires relief
funding to be allocated to a uniform

16

Chester County residents must contribute
an additional 0.18 percent of county income,
or $38.389 million; Delaware County an
additional 0.43 percent of county income, or
$59.313 million; Montgomery County an
additional 0.22 percent of county income, or
$50.988 million. The total additional
contributions from the four suburban
counties is $191.312 million. When these
funds are given to Philadelphia for poverty
relief, the city’s net contribution to regional
poverty spending becomes $342.880m $191.312m = $151.568m, which is 0.60
percent of city resident income. For suburban
counties as a whole, $191 million equals 0.29
percent of aggregate suburban income as
reported in the 2000 census.

percentage reduction in the city’s wage
tax rates for residents and nonresidents.
The proposed level of city poverty relief
will permit a 10 percent reduction in
each of the two wage-tax rates.
Finally, Reform Policy 3
requires all of city poverty relief to be
allocated to reducing the wage-tax rate
for nonresidents. This strategy will be
particularly valuable to city businesses,
since they bear a large share of the
burden of the nonresident wage tax as
higher labor costs. Reform Policy 3 is
likely to be suburban residents’ favorite
option too, since they benefit most from
the larger and more productive city
economy that this strategy encourages.
Under Reform Policy 3, the nonresident
wage-tax rate can be reduced 22
percent.
Reform Policy 1. Under the “no
strings attached” policy, the city is free
to spend its poverty relief funds as it
chooses. The results reported in Table 3
assume the city allocates the new
monies to additional public services.
Under this assumption, Reform Policy 1
improves estimated city house values by
$155 per house, but suburban house
values are estimated to fall by $95.
Neither effect should be considered
economically significant. The total gain
in wealth for regional home owners is a
very modest $335 million, about 0.03
percent of their initial wealth, and all
the gain goes to city residents.
Reform Policies 2 and 3. Reform
Policies 2 and 3 look more promising.
Under the second reform strategy, the
city is required to allocate its povertyrelief funding to a 10 percent reduction
in wage-tax rates for residents and
nonresidents. The average city house is
estimated to rise in value by $1087, or
about 1.82 percent of its pre-reform
value. But, again, the value of the
average suburban house remains
essentially unchanged, falling by $63.
Overall, under Reform Policy 2, total
home-owner wealth, regionwide, rises

Business Review Q2 2003 33

TABLE 3
Regional Financing for Regional Poverty

POLICY REFORMS

City Average House Value
(% Change from Status Quo)
(1)

Suburban Average House Value
(% Change from Status Quo)
(2)

Regional Total House Value
(% Change from Status Quo)
(3)

NO REFORM:
STATUS QUO

$59,700
(-)

$157,836
(-)

$111.76 billion
(-)

REFORM POLICY 1:
“NO STRINGS ATTACHED”

$59,855
(0.26%)

$157,741
(-0.06%)

$111.99 billion
(0.03%)

REFORM POLICY 2:
UNIFORM WAGE TAX CUT

$60,787
(1.82%)

$157,773
(-0.04%)

$112.21 billion
(0.50%)

REFORM POLICY 3:
NONRESIDENT WAGE
TAX CUT

$60,960
(2.11%)

$160,614
(1.76%)

$113.74. billion
(1.87%)

Because all the required data for census year 2000 are not yet available,
the estimates of the post-REFORM POLICY house values reported
above were computed from the simulation model of the Philadelphia
economy calibrated for the census year 1990 (see Technical Appendix).
For the suburban counties as a whole, the required equalizing transfer
of $191 million equals 0.29 percent of current suburban income; I
have therefore scaled the required suburban contribution to 0.29

an estimated $558 million, or about 0.5
percent, and again the benefits are
concentrated in the central city.17
Under Reform Policy 3,
however, all home owners in the region
benefit, not just those in Philadelphia.
Even though suburban residents send
money to the city, suburban house
values rise an estimated 1.76 percent.
Why? The answer lies in the more
efficient regional economy that follows

17

The estimates in Table 3 for Reform Policies 1
and 2 are reassuringly similar to the estimates
for suburban-to-city aid reported in Haughwout
and Inman (2002) for their national sample. In
their national sample, cities that share county
functions with their suburban governments,
such as Pittsburgh, have significantly higher
average house values than do cities, such as
Philadelphia, which pay for county functions
on their own.

34 Q2 2003 Business Review

percent of the 1990 suburban incomes for all policy simulations. The
percentage changes in city and suburban house values are computed
using this scaled transfer. The estimated percentage changes in regional
house values from the 1990 simulated economy (reported in
parentheses above) are then multiplied by the actual 2000 census
house values (reported here under NO REFORM: STATUS QUO)
to give estimates of the new, post-REFORM POLICY house values.

from Policy 3’s required reduction in the
wage-tax rate for nonresidents. The
economic effect of the city’s nonresident wage tax is to increase city firms’
labor costs roughly in proportion to the
tax rate. By reducing that tax rate,
Reform Policy 3 lowers labor costs in the
city, which encourages city businesses to
expand, more city jobs, and, most
important for suburban residents, more
low-cost city goods and services. In
dollars, the estimated net gain to a
typical suburban family in our simulated
regional economy will be an improvement in house values of $2780, or about
$140 a year assuming a 5 percent
interest rate. Under Reform Policy 3, a
Philadelphia area suburban family
would “invest” $220 per year in higher
county taxes but then benefit by saving
$360 per year through their consumption

of lower cost, city-produced goods and
services. The net gain is $140 a year.
Not a lot of money, perhaps, but a very
nice rate of return!
Most important, regional
financing of the cost of poverty is an
opportunity for Philadelphia and the
suburban counties to work together for
the benefit of all residents of the
Delaware Valley. If it is done correctly
— for example, city poverty relief is
exchanged for lower commuter tax rates
for suburbanites — regional fiscal reform
can be a true win-win, enhancing house
values in the city and suburbs alike.
CONCLUSION: FISCAL
COOPERATION FOR
FINANCING POVERTY
There is much to recommend
our region’s decentralized system of

www.phil.frb.org

public finance. But when there are
important economic interdependencies
across local jurisdictions, fiscal cooperation, not fiscal competition, is required.
One important economic interdependency, known as agglomeration
economies, occurs within our central
cities. This interdependency creates
significant production efficiencies and
allows valued product diversity, both of
which benefit all residents of the
economic region. Inefficient public
finances in a city, however, can undo
these economies as firms and households
leave the city. On its own, Philadelphia
has made significant progress toward
efficient city budgeting since its 1990
fiscal crisis. Growth in city workers’
compensation has been brought in line
with annual rates of inflation, and the

resulting savings in conjunction with
productivity improvements have allowed
balanced-budget reductions in the tax
rates for wages and gross receipts.
The problem that remains is
the city’s disproportionate share of the
region’s responsibility for poverty
spending, a burden it bears for historical
and legal reasons. Regional financing of
regional poverty will neutralize this
threat to Philadelphia’s productive
efficiency, and the region as a whole will
benefit. Reform can be implemented
within the existing structure of state
financing of county poverty spending;
no new metropolitan government is
necessary, nor is there any need for
regionwide taxation. What will be
required, however, is a commitment on
the part of the city and the four

suburban counties to work together.
One promising option would provide city
poverty relief in exchange for lower
wage tax rates for nonresidents. Under
this reform, city and suburban residents
both gain, perhaps by as much as an
additional $2.0 billion in regional house
18
values, or $2250 per household. The
source of the gain is a more efficient
Philadelphia economy, made possible by
tax relief for suburban commuters. With
city and suburban cooperation for regional poverty financing, we all win. BR

Bartik, Timothy. Who Benefits from State
and Local Economic Development Policies?
Upjohn Institute, 1991.

Glaeser, Edward, and Joseph Gyourko.
“Urban Decline and Durable Housing,”
NBER Working Paper 8598 (2001).

Inman, Robert P “How to Have a Fiscal
.
Crisis: Lessons from Philadelphia,”
American Economic Review, 85, 1995, pp.
378-83.

Beardsell, Mark, and Vernon Henderson.
“Spatial Evolution of the Computer
Industry in the USA,” European Economic
Review, 43 (June), 1999, pp. 431-56.

Glaeser, Edward, Hedi Kallal, José
Scheinkman, and Andrei Shleifer. “Growth
in Cities,” Journal of Political Economy, 100,
1992, pp. 1126-52.

Carlino, Gerald. “Productivity in Cities:
Does City Size Matter?” Federal Reserve
Bank of Philadelphia Business Review,
November/December, 1987.

Haughwout, Andrew, and Robert P Inman.
.
“Fiscal Policies in Open Cities with Firms
and Households,” Regional Science and
Urban Economics, 31, 2001, pp. 147-80.

Ciccone, Antonio. “Agglomeration Effects
in Europe,” European Economic Review, 46,
2002, pp. 213-27.

Haughwout, Andrew, and Robert P Inman.
.
“Should Suburbs Help Their Central City?”
Brookings-Wharton Papers on Urban Affairs,
2002, Brookings Institution, pp. 45-88.

18

The total difference in Reform Policy 3 minus
the status quo (Table 3): $113.74 billion - $111.76
billion = $1.98 billion, or approximately $2
billion.

REFERENCES

Ciccone, Antonio, and Robert Hall.
“Productivity and the Density of Economic
Activity,” American Economic Review, 86,
1996, pp. 54-70.
Freeman, Richard, and Robert Valletta.
“The Effects of Public Sector Labor Laws
on Labor Market Institutions and
Outcomes,” in Richard Freeman and Casey
Ichniowski, eds., When Public Sector
Workers Unionize. Chicago: University of
Chicago Press, 1988.

www.phil.frb.org

Haughwout, Andrew, Robert P Inman,
.
Steven Craig, and Thomas Luce. “Local
Revenue Hills: A General Equilibrium
Specification with Evidence from Four U.S.
Cities,” NBER Working Paper 7603 (2000).
Henderson, Vernon, Ari Kuncoro, and Matt
Turner. “Industrial Development in Cities,”
Journal of Political Economy, 103, 1995, pp.
1067-90.

Jacobs, Jane. The Economy of Cities. New
York: Random House, 1969.
Marshall, Alfred. Principles of Economics.
New York: Macmillan Company, 1890.
Rauch, James. “Productivity Gains from
Geographic Concentration of Human
Capital: Evidence from Cities,” Journal of
Urban Economics, 34, 1993, pp. 380-400.
Rosenthal, Stuart, and William Strange.
“Geography, Industrial Organization, and
Agglomeration,” Syracuse University,
Department of Economics, 2001.
Summers, Anita, and Lara Jakubowski.
“The Fiscal Burden of Unreimbursed
Poverty Expenditures,” Greater Philadelphia
Regional Review, 1997, pp. 10-12.
Voith, Richard. “Changing Capitalization of
CBD-Oriented Transportation Systems:
Evidence from Philadelphia: 1970-1988,”
Journal of Urban Economics, 33, 1993, pp.
361-76.
Business Review Q2 2003 35

TECHNICAL APPENDIX
An Economic Model of Philadelphia and Its Suburbs

The predicted changes in
Philadelphia and suburban house values
reported in Table 1 Columns 3 and 4 and in
Table 3 were computed using a general
equilibrium model of household and firm
location, investment, employment, and
production within the Pennsylvania portion
of the Philadelphia MSA (Bucks, Chester,
Delaware, Montgomery, and Philadelphia
counties). The model is able to predict city
and suburban population, employment, firm
production, government spending and
taxation, worker wages, and finally home
values, given an initial fiscal and
demographic position for the city and its
suburbs. The analysis also takes as given
land available for the location of
households and firms, both within the city
and within the surrounding suburbs (Bucks,
Chester, Delaware, and Montgomery
counties).
The household sector consists of
three broad classes of families — those
whose adult head of household works and
earns the region’s competitive market wage,
those whose head is a manager and earns an
exogenous (outside the model) managerial
wage set by the national market for
managerial talent, and finally, those whose
adult head is unemployed and classified as a
family in poverty or over the age of 65. The
Philadelphia region must offer all its workers
and managers the same “standard of living”
— what economists also call “utility” — as
available elsewhere in the country. Working
families can choose to live either in
Philadelphia or in the suburbs. We assign
managerial households to live in the
suburbs, even though they may work in
Philadelphia. Dependent households (those
in poverty or over 65) are assigned to live in
Philadelphia or the suburbs so as to match
the actual MSA data. Dependent households do not move in response to fiscal or
market incentives. If a household lives in
Philadelphia, it receives the common level
of city services and pays city taxes, either as
property, wage, or sales taxation. If the
family lives in the suburbs, it will receive the
average level of suburban public services
and it will pay the average suburban
property tax rate. Managers who commute

36 Q2 2003 Business Review

into Philadelphia are also assessed the city’s
nonresident wage tax, but the burden of this
tax is shifted back onto the manager’s city
firm as an added cost of hiring a mobile
managerial worker.
The production sector of the
metropolitan economy consists of city firms
that produce and sell a composite private
good and suburban firms that retail the same
private good but which must import that
good either from city firms or from firms
located outside the metropolitan area. The
composite good should be viewed in the
broadest sense to include all goods and
services a family might purchase in the
marketplace, from food to clothing to
entertainment to legal services to health
care. For suburban retailers, it is cheaper to
import this composite private good from
Philadelphia firms for two reasons. First,
because of agglomeration economies,
Philadelphia firms might be the low-cost
producer. Second, suburban retailers are
closest to Philadelphia producers so this
saves on transportation costs. If the demand
for the composite good by suburban
residents exceeds the exports available from
Philadelphia firms, the suburban retailers
must import the private good from more
expensive regional or national providers.
The government sector consists of
a single central city, specified to
approximate the finances of Philadelphia,
and a single, all-encompassing suburban
government specified to approximate the
finances of an average local government plus
school district in the four counties
surrounding Philadelphia. Each local
government produces one common public
service and pays for the service using local
taxes and intergovernmental transfers from
state and federal governments net of
payments for local debt and underfunded
pensions. In addition, both the city and the
suburb must meet required spending
obligations for their low-income and agedependent populations. City public services
benefit city firms and city residents, while
suburban public services benefit suburban
retailers and suburban residents. For
financing, Philadelphia uses a property tax,
resident and nonresident wage taxes, and a

gross receipts tax on sales by city firms
within the city. The suburban government
uses a local property tax. Both local
governments are free to choose their own
local tax rates and thus the level of local
government spending on the public service.
The regional economy will be in
equilibrium when all firms within the region
earn the national competitive rate of return
on invested capital, all households living in
the region receive the national “standard of
living” or “utility,” managers earn their
nationally competitive after-tax managerial
salary, and the city and suburban
governments choose locally balanced
budgets. If firms in the Philadelphia region
make more than the competitive rate of
return, more firms move into the metropolitan area; if they make less, firms exit.
When firms move into the region, the
demand for land and labor increase, leading
to a rise in land prices and worker wages.
The increase in worker wages raises the
standard of living for residents in the region,
leading to regional population growth.
Having more regional workers moderates
the initial increase in wages but reinforces
the initial rise in land prices. The opposite
effect on wages and land prices occurs when
firms leave the metropolitan area.
Firms will choose to enter the
region when they are more productive and
thus more profitable here than elsewhere.
An important source of production
efficiency will be city agglomeration
economies. Efficient city firms are more
profitable, attracting firms into the city.
Land prices and city wages rise. Having
more city firms also means more low-cost
city output for export to the suburbs.
Philadelphia suburbs are now more
attractive. Suburban retailing output
expands and this in turn increases the
demand for suburban labor. Both effects
raise suburban land values and suburban
wages. In the end, improving central city
agglomeration economies makes all residents
living in the Philadelphia region richer.
A complete description of the
simulation model can be found in the 2002
article by Haughwout and Inman.

www.phil.frb.org