Full text of Review (Federal Reserve Bank of Dallas) : Third Quarter 1993

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

FEDERAL RESERVE BANK OF DALLAS
Third Quarter 1993

Six Causes of the Credit Crunch
Robert T. Clair and
Paula Tucker

America's Health Care Problem:
An Economic Perspective
Beverly J. Fox, Lori L. Taylor and
Mine K. YOcel

Rethinking the IS in IS-LM:
Adapting Keynesian Tools
to Non-Keynesian Economies
Part 1
Evan F Koenig

The Long (and Short) on
Taxation and Expenditure Policies
Zsolt Becsi

This publication was digitized and made available by the Federal Reserve Bank of Dallas' Historical Library (FedHistory@dal.frb.org)

Economic Review
Federal Reserve Bank of Dallas
Robert O. McTeer. Jr.
President and Chief Executive Officer

Tony J. Salvaggio
Firsr Vice Presldenr and Chief Opera ring Ollieer

Harvey Rosenblum
Senior Vice President and Director of Research

W. Michael Cox
Vice President and EconomiC Advisor

Gerald P. O'Oriscoll. Jr.
Vice President and Economic Advisor

Stephen P. A. Brown
Assistant Vice President and Senior Economist

Robert W. Gilmer
Research Economist, Houston Branch

Economists
Zsolt Becsi
D'Ann M Petersen
Robert T Clair
Keith R Phillips
John V Duca
Fiona D Sigalla
Kenneth M Emery
Lori L Taylor
Beverly J Fox
John H Welch
David M Gould
Mark A Wynne
William C. Gruben
Kevin J Yeats
Joseph H Haslag
Mine K YOcel
Evan F Koenig
Research Associates
Professor Nathan S Balke
Southern Methodist University
Professor Thomas B Fomby
Southern Methodist University
Professor Scott Freeman
University of Texas at Austin
Professor Gregory W Huffman
Southern Methodist University
Professor Roy J Ruffin
University of Houston
Professor Ping Wang
Pennsylvania State University
Editors
Rhonda Harris
Virginia M Rogers
Design
Gene Autry
Graphics and Typography
Laura J Bell
Kurt Phippen
The Economic Review is published by the Federal Reserve Bank of Dallas The views
expressed are those of the authors and do not necessarily reflect the positions of the Federal
Reserve Bank of Dallas or the Federal Reserve System
Subscriptions are available free of charge Please send requests for single-copy and multiplecopy SUbscriptions, back issues. and address changes to the Public Affairs Department. Federal
Reserve Bank of Dallas, PD Box 655906, Dallas. TX 75265-5906. (2141922-5257
Articles may be reprinted on the condition that the source is credited and the Research
Department is provided with a copy of the publication containing the reprinted material

On the cover: an architectural rendering 01 the new Federal Reserve Bank 01 Dallas
headquarters.

Contents
Page 1

Six Causes ofthe
Credit Crunch
Robert T. Clair and
Paula Tucker

Bank lending typically moves with the business cycle.
In Texas from 1987 to 1992, however, bank loans declined
while nonagricultural employment rose. Robelt T. Clair and
Paula Tucker consider this evidence of a constrained supply
of bank loans, or credit crunch.
Clair and Tucker find that multiple factors have reduced
banks' willingness and ability to supply loans. The resolution
of failed banks and thrifts, tightening of bank examination
standards, new capital reqUirements, new regulations and
increased enforcement of old regulations, and increased
exposure to lawsuits have each had an effect. Many of these
regulatOly changes where made to address important economic
and social goals, but their side effects, often unintended and
perhaps unavoidable, have been to reduce bank lending in
the short run.

Page 21

America sHealth
Care Problem:
An Economic Perspective
Beverly J. Fox, Lori L. Taylor and
Mine K. YOcel

Soaring health care expenditures and the large number
of uninsured Americans-now estimated at 35 million-have
received much public attention in recent years. The widespread concerns have led to demands for substantial reform
of the U.S. health care system.
Beverly Fox, Lori Taylor, and Mine Yi.icel identify several
distortions in the current health care system that may be contributing to overconsumption of health care by some and
underconsumption of health care by others, and thus may be
leading to excessively high expenditures and the problems of
the uninsured. These distortions include tax subsidies for
employer-provided health insurance, regulations and industry
practices that restrict the supply of health care professionals,
and the noncompetitive nature of the health insurance industry.
Effective health care reform must address these distortions
rather than nondistortionary elements of the system, such as
producer and consumer uncertainty and the changing demographic composition of the U.S. population.

Contents
Page 33

Rethinking the IS in
IS-LM: Adapting
Kevnesian Tools to NonKeynesian Economies
Part 1
Evan F. Koenig

The IS-LM diagram was developed as a tool for analyzing Keynesian economies-economies with "sticky" prices
and myopic households. In a series of two articles, Evan Koenig
shows that a graphical apparatus similar to the traditional
IS-LM diagram can be used to analyze economies that have
optimizing, forward-looking households. In particular, an
expectations-augmented variant of IS-LM analysis is fully consistent with a popular real-business-cycle model. Thus, the
IS-LM diagram has wide applicability as a pedagogical device
and as a framework within which to discuss policy.
This article deals with an economy in which the capital
stock is fixed. A subsequent article will discuss how the
expectations-augmented IS-LM framework developed here
can be extended to an economy with capital investment.

Page 51

The Long (and Short)
on Taxation and
Expenditure Policies
Zsolt Becsi

Much of the 1992 presidential campaign focused on
which fiscal policies would best promote economic growth.
In this article, 2solt Becsi develops an analytical and graphical
framework to evaluate the long- and short-run effects of a
variety of taxation and expenditure policies.
Becsi shows that many tax schemes in their macroeconomic effects are essentially taxes on labor or capital or
both. While taxes on labor and capital both tend to depress
private consumption and output in the long run, Becsi shows
that a revenue-neutral reduction of capital taxes and increase
in labor taxes are likely to be contractionary in the short run
and expansionary in the long run.
Becsi discusses several ways of spending the peace dividend from a reduction in defense expenditures. He shows that
use of the dividend to reduce capital taxes causes consumption to rise in the long run with ambiguous effects on output.
In the short run, output and consumption will move in opposite
directions, but whether output rises or falls is uncertain. Using
the peace dividend to increase public investment will also
promote a long-run rise of consumption with ambiguous longrun output effects, but without short-run contractionary effects.

Robert T. Clair

Paula Tucker1

Senior Economist and Policy Advisor
Federal Reserve Bank of Dallas

Student
University of Texas at Austin Law School

Six Causes of the Credit Crunch
(Or, Why Is It So Hard to Get a Loan?)

M

any bankers, legislators, borrowers, and
regulators have expressed their views about
the cause of the credit crunch. Like the blind men
examining an elephant, each has an opinion that
has been formed from his perspective. Each has
characterized the problem and potential solutions
differently. None are completely correct or completely wrong. Bankers cite the lack of high quality
loan demand. Legislators blame overzealous regulators. Borrowers say banks are too conservative.
Regulators encourage bankers to lend and tell
their examination staffs to facilitate the extension
of credit but maintain the safety and soundness of
the banking system.
Many economists studying the credit crunch
explain it as a cyclical decline in credit demand.
They often suggest that the cyclical swing is
reinforced by structural changes in the demand
for credit. These economists have minimized the
numerous important factors that have reduced the
ability of banks to supply credit or, at a minimum,
have increased the cost of providing it.
In this article, we view credit crunches as
localized events that occur at different times in
different parts of the country. The Texas banking
industry provides an important case study. The
causes of the Texas credit crunch are highly
similar to the causes of credit crunches that have
developed elsewhere in the country. We focus on
the past seven years because the contraction of
bank credit began in Texas in 1986.
While demand may play an important part
in the decline in loans outstanding during some of
this period, we focus on the factors affecting the
supply of loans from banks over the past seven
years. While there are other sources of credit to
business, banks continue to be vitally important,
Economic Review — Third Quarter 1993

especially to small and mid-size businesses (Elliehausen and Wolken 1990). Many of the factors
that are limiting credit supply from banks also
affect other suppliers of credit. In some cases,
however, the factors limiting credit from banks are
unique to banks and place banks at a competitive
disadvantage. As discussed in the next section, the
definition of a credit crunch is fundamentally
related to the supply of credit, as opposed to the
demand for it. The following section presents the
complexity of the credit crunch as it developed in
Texas, where supply was reduced at both financially healthy and unhealthy banks. In the remainder of the article, we present six general factors
that caused the supply of credit to contract.
What is a credit crunch, and are we in one?
The economics profession is unclear as to
what constitutes a “credit crunch.” The crucial
differences in definition depend on the cause of
the contraction and whether credit is rationed by
means other than price.
Bernanke and Lown (1991) define a credit
crunch as a decline in the supply of credit that is
abnormally large for a given stage of the business
cycle. Credit normally contracts during a recession, but an unusually large contraction could be
seen as a credit crunch.
In their analysis, Bernanke and Lown compare the contraction in credit during the most

1

Paula Tucker is also a former economic analyst and writer
for the Federal Reserve Bank of Dallas.

1

recent recession to those in the previous five
recessions. Total loans at domestically chartered
commercial banks grew only 1.7 percent during
the 1990–91 period, compared with an average of
7.1 percent during the previous five recessions.
They conclude that there has been a credit crunch.
Bernanke and Lown attribute this reduced
lending activity to demand and supply factors.
Loan demand has been weak because borrowers’
balance sheets have been weaker than normal,
and as a result, borrowers have been less creditworthy than usual. The supply of credit has been
reduced by the decline in bank capital caused by
severe loan losses during the recession (Clair and
Yeats 1991). Bernanke and Lown’s analysis indicates that the demand factors have been far more
important, accounting for three-fourths of the
decline in lending in New England.
There is a disturbing dissonance created by
the Bernanke and Lown definition of a credit
crunch, the results of their analysis, and their conclusion that there was a credit crunch. They define
the credit crunch as an abnormally large decline
in the supply of credit. They argue that demand
factors largely caused the reduction in lending.
They then conclude that there is a credit crunch.
A second problem with the Bernanke and
Lown analysis is their use of national data to
determine if a credit crunch exists. Their crosssectional analysis using state-level data assumes
the imperfect substitutability of bank and nonbank
credit and of bank credit from banks located in
different states. Samolyk (1991) provides empirical
evidence supporting this assumption. If bank credit
cannot flow perfectly across state lines, however,
then the problems of a credit crunch would be
more likely to develop at the state level, not the
national level, unless a nationwide economic shock
caused the decline in bank capital.
The second definition of a credit crunch
relies not on the contraction in lending but on the

2

2

At the same time that Bush administration officials were
encouraging additional lending, Congress was holding
hearings on bank failures and sending a signal to examiners
that they should be conservative if they wished to avoid
testifying before Congress.

microeconomic principle of a shortage. If at the
current market price the demand for a good
exceeds the supply, then there is a shortage. The
available supply will be rationed but by some
means other than pricing. Nonprice credit rationing
may occur even in a market that might not be
described as experiencing a credit crunch (Stiglitz
and Weiss 1981). Owens and Schreft (1992) define
a credit crunch as a period of sharply increased
nonprice rationing.
Owens and Schreft review historical episodes
of nonprice rationing—that is, credit crunches that
were accompanied by binding interest rate ceilings,
credit controls, or coercive posturing by administrative officials and bank regulators to discourage
banks from lending. In the current recession,
researchers argue, administrative officials and bank
regulators have actively encouraged banks to
lend.2 Owens and Schreft do state that there was
probably nonprice rationing in loans secured by
real estate, resulting from bank examiners’ reaction
to real estate loan losses. They cite the statements
made by Robert Clarke, then comptroller of the
currency, that discouraged banks from making
real estate loans.
Owens and Schreft conclude that there is
not a general credit crunch, but there might have
been a sector–specific crunch in real estate. Since
nonbank providers of credit also contracted their
lending, Owens and Schreft attribute the decline
in lending to ebbing loan demand.
The Owens and Schreft definition of a credit
crunch has intuitive microeconomic appeal but
may not provide the insights needed for economic
policy analysis. Their definition does not consider
actual lending activity. Consequently, a “credit
crunch” can occur during a period of expanding
credit as easily as during a contraction of credit.
Furthermore, Owens and Schreft dismiss
anecdotal evidence from borrowers. They may be
correct that borrowers would complain during any
period of tight credit, but the type of complaint
could be quite different. During nonprice rationing,
borrowers complain about not being able to get a
loan at any price. During periods of simply tight
credit, borrowers complain about the cost of credit.
Despite their differences, both the Bernanke–
Lown and Owens–Schreft studies agree that a
decline in credit demand explains the major part of
the credit contraction, and both find little support
Federal Reserve Bank of Dallas

for the explanation that more stringent bank
examination practices account for the contraction
in loan supply. Owens and Schreft link the decline
in credit demand to the deterioration of real estate
asset values, similar to the Bernanke and Lown
view of weakened balance sheets. In determining
if the nation is in a credit crunch, Bernanke and
Lown cite the abnormally slower growth of credit
as a sign of the credit crunch, while Owens and
Schreft see few signs of nonprice credit rationing
and conclude that there is no general credit crunch.
In ascertaining that demand factors are a
primary cause of the decline in bank credit, both
studies cite the lack of credit supply response
from nonbank sources of credit. Nonbank sources
of credit to businesses are growing increasingly
important (Pavel and Rosenblum 1985). Approximately 25 percent of small and mid-size businesses
obtain credit from nonbank sources (Elliehausen
and Wolken 1990).
If bank credit alone were being rationed or
constrained, both studies argue, other providers
of credit should have increased their activity. Most
nonbank sources of credit to corporate businesses
have contracted during the 1990 –91 recession.
From 1989 to 1991, not only did the annual flow
of funds from bank loans contract but so did the
flow of funds from finance companies, commercial
paper, mortgages, and trade credit. At the same
time, the flow of funds needed for capital expenditures contracted sharply. These national aggregate data are consistent with the hypothesis that
demand factors have driven the credit contraction.
In a comment on the Bernanke and Lown
study, Benjamin Friedman points out that they
assumed that other nonbank credit providers did
not suffer the same constraints (Bernanke and
Lown 1991). If loan losses have caused capital to
decline at banks, might not similar losses reduce
the capital of nonbank creditors, such as insurance
companies? Michael Keran (1992), vice president
and chief economist of the Prudential Insurance
Company of America, has acknowledged that
financial intermediaries other than banks have also
suffered declines in capital resulting from real
estate and other loan losses.
Because of their analytical approaches, both
of these empirical analyses have misdated the
beginning of the credit crunch. The Bernanke and
Lown analysis uses national data that mask imporEconomic Review — Third Quarter 1993

tant differences among various regions of the
country. The Owens –Schreft analysis begins in
late 1989 and focuses on New England. Texas
suffered a severe contraction of its economy and
of bank credit during the last half of the 1980s
when the national economy was growing. By
failing to examine state-level data, both studies
misdate the start of the credit crunch by several
years and fail to establish its regional nature
(Rosenblum and Clair 1993).
Because Texas began its credit crunch earlier,
Texas is a better case study to examine long-term
effects. Texas’ banking industry was so severely
affected that even after the state’s economy began
a recovery in 1987, the banks did not increase
their lending. Even though Texas’ economy outperformed the nation’s during the 1990 –91 recession
and experienced only a modest slowdown, lending at Texas banks did not increase for six years.
The life cycle of a credit crunch:
the Texas experience
Until 1987, Texas’ loan cycle was in line
with the regional economic cycle. During the
economic expansion of the first half of the 1980s,
loans extended by Texas banks more than doubled
from $52 billion in 1980 to $119 billion in 1985.
In the midst of continued growth in the national
economy, Texas entered a recession, triggered by a
precipitous decline in oil prices in 1986. Declines
in lending during an economic downturn are
normal.
The abnormality in the Texas lending pattern
surfaced about 1987. Despite an economic recovery,
lending continued to decline. From 1987 to 1990,
lending declined another 30 percent, even though
employment increased 6.8 percent. Even the
modest increase in loans outstanding that began
in 1992 does not reflect new lending as much as it
does acquisition of failed savings and loan associations (S&Ls), their assets, or assets from other
nonbank institutions and consolidation of national
lending operations into Texas banks.
A credit crunch is not a necessary consequence of an economic downturn. Lending declines
during an economic downturn, but primarily
because of decreases in business and consumer
loan demand. In Texas, however, the economic
climate has played an important role in the credit
3

crunch. A chain reaction of huge shocks to the
Texas economy resulted in the near destruction of
several key industries the state had relied on for
growth throughout the 1970s and 1980s. To
under-stand the Texas credit crunch, we must first
understand the nature of this abnormally strong
downturn and its repercussions on the economy.
In the late 1970s and early 1980s, the Texas
economy prospered as the oil and gas industry
boomed. Growth in the oil industry fostered
employment growth in all sectors of the Texas
economy. The climate was especially hospitable
for commercial real estate.3 Low vacancy rates,

4

3

See Petersen (1992) for an excellent description of and
outlook for the Texas commercial real estate industry.

4

Petersen (1992) explains that the Economic Recovery Tax
Act of 1981 redefined the business depreciation allowance
for some real estate properties to allow for an accelerated
recovery of investments, thus making those investments
more attractive. For an extended discussion of the effects of
depreciation rates on real estate decisions, see Yeats
(1989). Also, the Depository Institutions Deregulation and
Monetary Control Act of 1980, which helped phase out
interest rate ceilings on time and savings deposits, and the
Garn–St Germain Depository Institution Act of 1982, which
created the money market deposit account, resulted in a
large source of new funds. The Garn–St Germain Act further
liberalized investments that S&Ls could make (although
Texas state-chartered S&Ls already had these powers) and
included provisions for the creation of nonexistent capital
through the issuance of capital certificates. Together, these
changes provided tremendous incentives favoring investment in commercial real estate.

5

When loans are charged off as losses, these losses are
deducted from the allowance for loan loss (a reserve
account on the balance sheet), which historically was
considered a part of regulatory capital. If the charge-offs
are large, then the allowance must be replenished because the adequacy of the allowance is judged relative
to the size of the loan portfolio and its risk. This is done by
increasing the provision for loan losses (an expense item on
the income statement). If this provision is large enough,
it can cause net income to be negative; in other words,
the bank sustains a net loss. If income is negative, then
the equity capital position is reduced by the amount of
the loss. Essentially, if the decline in the allowance for
loan loss is so large that it cannot be absorbed by current
income, then monies are diverted from equity capital to
the allowance for loan losses.

changes in tax laws, and financial deregulation in
the early 1980s motivated investment in commercial real estate.4 The state’s strong economy and a
drop in interest rates also encouraged the flow of
funds to the real estate sector (Petersen 1992). As
a result, office building permit values nearly
doubled from $1,143 million in 1980 to $2,184
million in 1985.
Even when an initial weakening of oil prices
in 1982 triggered a downturn in parts of the Texas
economy, commercial real estate activity continued.
Bankers’ and other investors’ interest in office
buildings persevered in the face of skyrocketing
vacancy rates. Office vacancy rates in major Texas
cities increased from 8 percent in 1980 to 24.3
percent in 1985, as office building permits continued to rise (Petersen 1992).
Texas was not so lucky after a second sharp
decline in oil prices in 1986. Recession struck the
state but not the nation. Texas is an oil-producing
state and an exporter of oil-field machinery, while
the nation is an oil importer. Reversals of the tax
laws that had favored commercial real estate
investments exacerbated the state’s economic
problems by accelerating the flow of funds out
of the office construction arena. The state lost
250,000 jobs and gained the burden of an extraordinary amount of vacant office space. In 1987,
vacancy rates were near 30 percent in most major
Texas cities.
Unfortunately, the shocks engendering the
collapse of petroleum and construction were only
the beginning for Texas. Like other investors,
many aggressive banks were caught holding loans
to both oil and gas producers and commercial real
estate developers (Gunther 1989). Nonperforming
loan rates at Texas banks increased steadily from
1984 to 1987, and troubled assets caused declines in
equity capital and bank failures (Robinson 1990).
During the 1980s, equity capital at Texas
banks followed the same pattern as lending. From
1980 to 1985, equity capital increased by 85 percent, or $6.2 billion. After the downturn in the
Texas economy, equity capital declined by 41 percent, or $5.6 billion. The declines in equity capital
resulted from $10.8 billion in loan losses experienced by Texas banks during the second half of
the 1980s.5 Although equity capital improved
somewhat in 1989 and 1990, it was 23 percent
below its peak.
Federal Reserve Bank of Dallas

Table 1

Texas Banking Statistics
(all figures are percentages)

Healthy Bank Index

1

1988

1989

1990

1991

1992

38.12

49.53

60.60

68.30

81.57

–1.21

–.33

.41

.65

1.07

Nonperforming Loan Ratio2

6.41

6.59

3.14

2.85

1.70

Primary Capital Ratio3

6.40

6.02

7.20

7.44

7.68

Return on Assets

Growth Rate of Securities
Growth Rate of Loans
1

2
3

10.42

10.95

18.66

14.65

3.67

–17.61

–7.43

–4.59

–2.53

6.69

This index is the percentage of assets held by healthy banks. A bank is defined as healthy if it is earning a profit, has a troubled
asset ratio below 3 percent, and has a capital ratio at least one-half percentage point above the regulatory minimum.
Nonperforming loans are all loans 90 days or more past due or nonaccruing divided by total loans.
Primary capital ratio is the sum of bank equity and loan loss reserves divided by the sum of total assets and loan loss reserves.

SOURCE: Federal Reserve Bank of Dallas.

For many banks, the decline in bank capital
was fatal. Bank failures skyrocketed to levels not
seen since the Great Depression. No Texas banks
failed in 1981, but thirty-seven did in 1986, and
the numbers kept climbing.6 Texas bank failures
peaked in 1988 at 149 and were down to 31 in
1992. The savings and loan industry suffered an
even higher failure rate.
Pathology of a credit crunch
Researchers at the Federal Reserve Bank of
Dallas have examined the connection between
financial health of banks and their lending activity
and have found that during the latter half of the
1980s, many Texas banks were too unhealthy to
lend. Financially unhealthy banks are those with
capital-asset ratios below 6 percent, with negative
income, or with a troubled-asset ratio of 3 percent
or more. In 1986, 55 percent of Texas banks
holding 72 percent of the state’s total banking
assets were unhealthy by this standard. Increased
lending by these banks would have exposed them
to unacceptable risk of failure. Lending would
have been discouraged or prohibited by bank
supervisors and, in all likelihood, by the banks’
own boards of directors.
Economic Review — Third Quarter 1993

Since the second quarter of 1988, the health
of the state’s banking industry has steadily improved. By the fourth quarter of 1992, 72 percent
of Texas banks, with 82 percent of the state’s
assets, were healthy (Table 1 ). The improvement
resulted from the failure of many unhealthy banks,
from customers’ switching their business from
unhealthy banks to healthy banks, and the financial recovery of some unhealthy banks. However,
the fact that 304 unhealthy Texas banks were
holding approximately one-fifth of the state’s
assets as of the fourth quarter of 1992 indicates
the im-provement has been slow (Clair and Sigalla
1993).
The inability of healthy Texas banks to take
market share away from unhealthy banks in a
timely manner contributed to the slow recovery of
Texas banking. When banking problems escalated

6

Banks failures had slowly increased in the four years
preceding the oil and construction bust. Although no Texas
banks failed in 1981, five banks failed in both 1982 and
1983, six banks failed in 1984, and thirteen in 1985. This
slow but steady increase signaled the increasing fragility of
the banking industry before the economic shock.

5

in 1986, healthy banks were small compared with
their unhealthy competitors. The average healthy
bank had only $67 million in assets compared
with $136 million for unhealthy banks. Healthy
banks controlled only 28 percent of Texas banking assets. Thus, for healthy banks to take over
the market share of unhealthy banks would have
required an inconceivably rapid expansion.
The rate at which healthy banks can take
over the market share of unhealthy banks is
limited by healthy banks’ capital in excess of
regulatory minimums. Raising capital in the equity
markets was and is difficult for these healthy
banks. Their small size means small equity offerings, which are costly to sell. Moreover, the chaotic
state of the Texas banking market caused investors
to shy away from Texas bank stocks. The only
alternative left open to banks was to raise capital
through the slow process of retaining earnings.
Even if the average-size healthy Texas bank
retained 75 percent of its earnings and maintained
its primary capital-to-asset ratio, individually it
could increase lending by only $2.4 million per
year. If all healthy banks followed the same strategy,
they could have only increased total lending by
$2 billion— only a 1.7-percent annual increase.
But not all healthy banks increased their
lending activity, which indicates an important
pathology. The term pathology applies in this case
because a bank in good financial condition in a
growing region would normally be expected to
increase its lending activity (Rosenblum 1991).
Those healthy banks not building their loan
portfolios represented a significant share of the
healthy banks in Texas. Of the 619 banks that
were healthy as of the first quarter of 1991 and
that had reported data for the past ten quarters,
nearly 40 percent did not increase their lending
from the first quarter of 1990 to the first quarter of
1991 (Rosenblum 1991). These banks accounted
for 40 percent of the assets and 35 percent of the
loans of healthy Texas banks at that time.
The pathology of financially healthy banks
in a growing state not increasing their lending
raises the need for an examination of the possible
causes. The extension of bank credit, especially to
small and mid-size businesses, supports new job
creation and economic expansion. The remainder
of this article discusses serious impediments
affecting the supply of credit.
6

Six causes of the credit crunch
Declines in bank capital
Business-cycle effects typically do not cause
a credit crunch. Business lending after adjusting
for inflation typically moves with the business
cycle with a lag. Both demand and supply shifts
contribute to the cyclical movement. During a
slowdown, demand for credit declines and the
supply of credit also contracts because loans
become riskier. After the recovery is established,
demand increases and banks begin lending again.
In an atypically severe cycle, the ability of
banks to begin lending after the recovery is established may be hindered. During the recession
phase of a severe cycle, the larger than normal
loan losses result in larger than normal reductions
in bank capital and numerous bank failures. Loan
losses in the recent regional and national recessions have been severe, especially when viewed
relative to bank capital. During the 1985–90 period,
banks in Texas made provisions for $14.5 billion
in loan losses, and their total capital at the end of
this period was $10.3 billion. Even among surviving banks, capital may fall below either the level
desired by bank management or the minimums
established by regulatory agencies. In either case,
the expansion of credit will be limited by the bank
capital levels (Clair and Yeats 1991, Hancock and
Wilcox 1992).
Not only did loan losses reduce bank capital,
but minimum capital standards rose. Baer and
McElravey (1993) have examined the factors
causing an increased demand for bank capital in
the two-year period beginning in June 1989. By
their estimates, meeting higher capital standards,
whether imposed by regulators or adopted by more
conservative bankers, had twice the effect of loan
losses in creating the need for new bank capital.
Bank capital standards rose substantially over
the 1980s and early 1990s (Baer and McElravey
1993). In the 1970s, bank supervisors set minimum
capital ratios for each bank, based on ratios at
similar banks. Bank capital ratios had been declining during the 1970s, and concerned regulators
established a minimum primary capital ratio of 5.5
percent in late 1981, to be phased in over time.
By the latter half of the 1980s, bank regulators, as part of an international agreement, established risk-based capital ratios. Regulators assign
Federal Reserve Bank of Dallas

risk weights to various types of assets and offbalance-sheet risks and require capital to be held
in proportion to the credit risk of the bank portfolio. For example, short-term Treasuries have a
zero credit risk weight, and business loans have a
100-percent risk weight.
Risk-based capital ratios may have raised the
relative cost of lending compared with investing
in securities. If these risk-based ratios are a binding constraint on banks, then increasing business
lending will require additional capital to be raised,
but investing in short-term Treasuries requires no
additional capital.7 Since capital is costly, the riskbased system increases the cost of business loans
relative to securities, thereby discouraging business lending.8
In addition to risk-based capital ratios, regulators removed the primary capital ratio requirement
and replaced it with a leverage ratio requirement.
Whether the leverage ratio is a higher constraint is
uncertain. The required leverage ratio is dependent
on a bank’s risk rating. Nominally, a top-rated bank
could have a leverage ratio of 3 percent, but most
banks were expected to maintain leverage ratios
in the neighborhood of 4 percent to 5 percent.9
The old primary capital ratio was 5.5 percent, but
it included loan loss reserves in the definition of
capital, which the new leverage ratio does not.
While a direct comparison of these new
capital regulations is not possible, an empirical
analysis by Baer and McElravey (1993) indicates
that banks are behaving as though their minimum
capital requirements have risen substantially over
the past few years. Based on their analysis, banks
now respond as though their required leverage
ratio has risen from 4 percent in the 1973–75
period to 7 percent in the 1989–91 period. Banks
are behaving as though they are setting internal
minimum capital standards much higher than the
regulatory minimums. The pressure on banks,
whether from regulators or internal management,
to maintain higher capital ratios has severely
limited their ability to extend new credit.
FDIC and RTC resolution of failed
depository institutions
While loan losses directly reduced capital,
the resolution of failed banks and thrifts increased
the demand for capital. After a depository institution fails, its assets are taken over by the deposit
Economic Review — Third Quarter 1993

insurer. Typically, the insurer sells the institution,
often after cleaning the portfolio of the nonperforming assets.10 The acquiring institution must
have sufficient capital in excess of regulatory
minimums to be able to increase its total asset
holdings without becoming undercapitalized.
The resolution of failed banks and thrifts was
not the only source of assets to be acquired. Many
banks that did not fail but were undercapitalized
reduced their assets to improve their leverage ratios.
They had to sell these assets to healthier institutions that had sufficient excess capital to purchase
the assets and remain sufficiently capitalized.
Baer and McElravey (1993) term this process
the recycling of assets, suggesting that assets are
recycled from undercapitalized to well-capitalized
banks and thrifts. During the two-year period
beginning June 1989, undercapitalized bank holding companies sold $82.8 billion in assets, failed
banks accounted for $58.6 billion, and failed thrifts
accounted for $177 billion, for a total $318.4 billion
in recycled assets. Recycling these assets increased
the need for capital by more than $22 billion, a

7

There is a requirement for a minimum leverage ratio that
requires a bank hold some capital regardless of the composition of its asset portfolio.

8

Risk-based capital is not a bad idea in theory. That riskier
institutions should hold greater capital is logical. If the loans
diversify the bank’s overall portfolio, however, then increased lending may decrease a bank’s risk.

9

It is erroneous to think that a bank is permitted to operate
with a leverage ratio of 3 percent. There is a catch-22. Banks
are rated from one to five on the CAMEL scale, with one
being the highest rating possible. CAMEL is an acronym for
capital, asset quality, management, earnings, and liquidity.
A bank can’t get a CAMEL-one rating with only 3 percent
capital, but if the bank has a CAMEL-one rating, it is
permitted to have only 3 percent capital.

10

In some cases, the acquiring institution also acted as a
collecting bank for the Federal Deposit Insurance Corporation (FDIC). In these cases, it was common for the bank to
carry the assets in the collection operation under a special
classification of “other assets,” and the bank was not
required to hold capital against these assets. Since the
losses incurred from these collecting bank assets would be
borne by the FDIC, the bank did not need to hold capital
against these assets.

7

28.7-percent increase in capital at the time.
Beyond increasing the demand for capital by
recycling assets of failed institutions, the failure–
resolution process destroyed valuable information—reducing the ability of many borrowers to
obtain credit (Board of Directors of the Federal
Reserve Bank of Dallas 1991). Effective lending
involves the ability of bankers to develop specialized information regarding their borrowers. This
information allows bankers to make informed
credit decisions at minimal cost. Anything that
disrupts the banker–borrower relationship can
lose or destroy the specialized information a
banker has about a specific borrower.
One type of this specialized information is
the banker’s assessment of the borrower’s character
— a signal of the borrower’s commitment to repay
a loan under adverse conditions. Bankers attempt
to assess the character of a borrower prior to
making a loan. This assessment is hard to quantify
or document and is an important judgment call
that a bank officer must make.
Many borrowers will face difficulty in repaying during an economic downturn. Some will be
unwilling to accept any personal sacrifice and will
be quick to declare bankruptcy or otherwise force
a bank into losses. Other borrowers, those with
greater character, will make every reasonable
effort to repay their obligations and will make
personal sacrifices in the process.
During an economic downturn, the loan
documentation of borrowers with radically different
characters may appear very similar. The repayment
may appear poor—that is, late or partial payments or violated loan covenants. Bankers know
which borrowers are making tremendous efforts
to meet their obligations and which borrowers
expect the bank to be the first to forgo payment.
Both loans may be classified as nonperforming.
When a failed bank is resolved, nonperforming loans are often either placed in a collecting
bank or are held by the FDIC for liquidation.
Borrowers must establish new banking relationships. But being placed in these collecting or
liquidating operations places an equal stigma on
borrowers of good and poor character. Resolving
the failed bank destroyed the information that
distinguished low-risk from high-risk borrowers.
Being placed in a collecting bank can even
tarnish the reputation of borrowers with perfect
8

repayment records. In the late 1980s, regulators
created the “nonperforming performing” loan
category. These loans were current on payments
and not in violation of any loan covenants. Because
the examiners considered the loans unlikely to be
repaid given the examiners’ current economic
outlook, they classified them as nonperforming.
As a result, another group of borrowers may have
been inappropriately placed in the collecting bank
and thereby faced substantial damage to their
reputations.
The resolution of the failed banks and thrifts
was inevitable, and it improved the health of the
financial industry. The huge demand for capital
required to recycle assets was unavoidable. Still,
the increased demand for capital to fund these
assets limited the capital available to fund new
loans. The resolution process, however, destroyed
valuable information on borrower relationships,
and a reevaluation of the process to determine if
the negative economic impacts of closing failed
banks and thrifts can be reduced is warranted.
Bank supervision overreaction
The evidence that an overreaction by bank
supervisors caused the credit crunch is mixed.
Since the potential impact of bank examiners on
credit decisions is large, the evidence needs to be
presented. There are many different ways in
which bank examiners, in the process of enforcing
safety and soundness guidelines, might constrain
bank lending.
1. Examiners could criticize existing loans—
requiring banks to increase loan loss
provisions and charge-offs, and thus
reduce their capital.
2. Examiners could become more conservative in evaluating a bank’s condition and
thereby require a higher leverage ratio.
3. The specter of examiners’ criticism alone
could discourage loans from being
extended.
4. For more troubled institutions, examiners
may be directly setting restrictions on
lending activity
5. Higher loan documentation requirements
could raise costs, but these requirements
may be more directly related to the
regulatory burden and will be discussed
elsewhere.
Federal Reserve Bank of Dallas

Each of these supervisory and regulatory impositions will have different effects on bank financial
statements.
The hypothesis that bank examiner overreaction caused the credit crunch arises from the
February 1990 advisory sent by the Office of the
Comptroller of the Currency (OCC) to all banks
warning against making imprudent real estate
loans. In November 1990, as the national economy
weakened, the Bush administration blamed the
tight credit conditions on an overreaction by bank
supervisors. Most bankers responded, however,
that it had been and was the lack of loan demand
and deteriorating economic conditions that discouraged their lending and not supervisory excess
(Owens and Schreft 1992).
The evidence indicates that bank examiners
did not overreact in criticizing existing loans and
requiring good loans to be charged off. Bernanke
and Lown (1991) examine this issue by analyzing
the trend of provisions for loan losses relative to
actual net charge-offs. Certainly, provisions for
loan losses and net charge-offs rose during the
1980s, but the ratio of provision to charge-offs
was very steady, indicating that examiners did not
raise the standard for provisioning excessively.
Accordingly, Bernanke and Lown conclude that
examiners have not suddenly imposed new tighter
examination standards that have constrained credit.
Even so, there is evidence that bank examiners are enforcing a more conservative view of what
constitutes a healthy bank. David Bizer of the
Securities and Exchange Commission has argued
that bank examiners have raised the financial
standards for any given CAMEL rating (Bizer 1993).
This change to more conservative CAMEL
ratings is related to the credit crunch because the
required leverage ratio is tied to a bank’s CAMEL
rating. The minimum leverage ratio is set at 3
percent for banks rated CAMEL one and rises as
CAMEL ratings worsen. If bank examiners raise the
standards for any given CAMEL rating, they are, in
fact, increasing the minimum capital standard.
Bank examiners could also affect credit
decisions by raising the expected cost of funding
the credit. The cost of funds is a combination of the
cost of the necessary capital and the cost of deposit
funds. If bankers perceive, even erroneously, that
examiners might criticize new credit extensions,
then they expect that a larger share of new credits
Economic Review — Third Quarter 1993

will have to be funded with relatively expensive
capital, driving up the expected funding cost and
discouraging new lending. These concerns could
drive up funding costs by 70 basis points or more.
(For a detailed example of this effect, see the box
entitled “Examiners and Funding Costs.”)
It is possible that bank supervisors are constraining lending activity beyond their power to
set higher leverage ratios. Peek and Rosengren
(1993) analyzed new lending activity of banks in
New England, adjusting for whether a bank was
operating under a formal agreement with its primary
regulator. Regulators impose formal agreements
on banks considered seriously troubled or even
recalcitrant in repairing their financial condition.
Such agreements allow the regulator to seek civil
or even criminal penalties in the case of noncompliance.
After controlling for differences in leverage
ratios, Rosengren and Peek’s results indicate that
new lending was significantly lower at banks
operating under formal agreements than at banks
with equally low capital ratios but not under such
agreements. They conclude that banks may be
slow to constrain their lending or to rebuild their
capital on their own. Once the formal agreement
is in place, however, banks respond much more
quickly.
In sum, regulators constrain credit growth at
weak institutions that are unwilling to temper
their own behavior when faced with declining
capital. But constraining the credit expansion of
weak institutions does not cause credit crunches.
In fact, it may prevent them in the future. Texas
had many institutions that lent freely despite their
weak financial condition. As a result, imprudent
loans were extended, especially in commercial
real estate development. The overbuilding that
ensued affected the value of collateral supporting
what otherwise would probably have been good
loans made by financially strong institutions.
During the worst of the Texas banking crisis,
managers of well-run banks called for bank supervisors to shut down the activity of insolvent or
nearly insolvent institutions.
In conclusion, bank supervisors do not
appear to have required excessive charge-offs of
nonperforming loans. Supervisors did constrain
lending at financially weak institutions, but that is
the proper role of supervisors. Bankers’ concerns
9

Examiners and Funding Costs
Bank examiners can change the expected cost of funding a new loan by changing
the banker perception of what loans might be
criticized, which would change the required
mix of capital and deposits needed to fund the
loan. Loans are funded by a combination of
capital and deposits.1 Baer and McElravey’s
(1993) results indicate that banks would want
a loan to be funded with 7 percent capital and
93 percent deposits. Capital is more costly to
raise than deposits. For the purposes of our
example here, it is assumed that capital requires a 15-percent return, and deposits cost
3 percent. In this simple example, the funding
cost of a loan is the weighted average of these
two costs, 3.8 percent, or
(.07 × 15%) + (.93 × 3%).
If, however, bankers believe examiners
will criticize the loan, then the funding cost of
the loan will rise sharply. Suppose bankers
believe examiners will criticize the loan and
require 30 percent of the loan to be reserved.
In this case, approximately 65 percent of the
loan would be funded with deposits and 35
percent with capital (30 percent of the loan is
100 percent funded by capital by being reserved and the remaining 70 percent of the

that new loans might be criticized, however, may
have discouraged lending by raising the expected
cost of funding these loans.

11

10

Hindsight is always 20-20. For example, most cities will not
grant building permits for land within a 100-year flood plain.
If a new record flood results in the destruction of homes, it
could be said that the previous standard was too lax. Higher
standards would mean less risk, but the cost would be more
land that could not be used for buildings.

loan that still requires a 7-percent leverage
ratio.) In this case, the funding cost would rise
to 7.2 percent, or
{[.30 + (.70 × .07)] × 15%} + (.65 × 3%).
Now, if the banker believes there is only
a 20-percent probability that the loan will be
criticized, then the cost of funding would be
the weighted average of these two funding
costs, that is, 4.5 percent, or
(.20 × 7.2%) + (.80 × 3.8%).
Therefore, just the specter of examiner overreaction could increase the expected cost of
funding 70 basis points, from 3.8 percent to
4.5 percent. Given this expectation, many
loans would never be made. Beyond a lack of
lending, there would be no direct evidence of
this effect in bank financial statements, that is,
no sharp rise in provisions for loan losses or
charge-offs.

1

To be precise, loans are funded by capital and liabilities.
Liabilities include deposits in addition to federal funds purchased, other debt instruments, etc. For the purposes of this
example, we have simplified the bank’s funding to capital and
deposits only.

New credit standards set by bankers
Atypically severe recessions alter both
bankers’ and bank supervisors’ perception of risk.
After an unusually severe recession and a sharp
increase in bank failures, bankers will likely reevaluate risk and change their risk-taking behavior,
require more capital to buffer against it, or both.
Their willingness to supply credit is likely reduced.
In hindsight, the old credit standards were
too lax.11 If loans had been properly priced, banks
would have accumulated sufficient capital during
expansions to absorb loan losses during downFederal Reserve Bank of Dallas

turns. This is not what happened in Texas. Many
banks failed because their reserves and capital
were insufficient to absorb loan losses. The
inability of banks to properly price risk is related
to the disincentives inherent in the deposit
insurance system (Short and O’Driscoll 1983).
Bankers have contracted the supply of credit
by raising credit standards and denying credit to
many borrowers. Some of the rejected applicants
have qualified for loans in the past or are even
current borrowers seeking credit extensions. This
change in status from creditworthy to uncreditworthy can be difficult to accept and can damage
borrowers’ businesses since many planned on
continued access to credit.
Probably the greatest difference between
borrowers’ and bankers’ perceptions is that
borrowers perceive creditworthiness as an individual characteristic, while bankers view creditworthiness on both an individual basis and on the
basis of the entire portfolio of loans. To illustrate,
suppose that prior to a severe recession, a banker
expects a 2-percent loss rate on loans to a given
industry but during the recession, actually sustains a
5-percent loss rate. In response, the banker raises
the credit standards for borrowers in this industry
with the intention of obtaining a 2-percent loss
rate. The higher standards, however, might result
in, say, 25 percent of the previous borrowers
being unable to qualify for credit.
The majority of rejected borrowers will have
repaid their loans on time and in full. These
borrowers are not different—in either financial
characteristics or character—from the minority
that defaulted. The bank realizes that the likelihood
that an individual borrower will default is not
easily guessed but that the default rate for a portfolio of loans is fairly predictable. The borrowers
see themselves as good bank customers—not as
the inadvertently lucky ones who did not default
—and do not understand why they have been
rejected. They complain, accordingly, that there is
a credit crunch.
Returning to the issue of real, rather than
hypothetical problems, the reevaluation of riskreturn tradeoffs during the 1980s in Texas was not
uniform across industries or types of loans. During
this regional recession, some industries proved far
riskier than bankers previously thought. In the
1980s, energy prices fell by more than 50 percent.
Economic Review — Third Quarter 1993

Real estate values plunged. Consequently, many
bankers revised their expectations for these industries more than for others.
Borrowers’ confusion over their own creditworthiness was compounded by banks that were
too weak financially to lend but pretended to
consider loans for approval. If a bank’s condition
deteriorates to the point that it is unable to extend
credit, the bank would likely want to conceal that
fact. Otherwise, depositors might demand higher
interest rates, and the bank’s best borrowers might
take their business elsewhere (Rosenblum 1991).
To create the appearance of financial health, the
bank pretends to continue its lending operations
—including marketing activities. Even high-quality
loan proposals are rejected, however, under the
pretense that they are too “risky.”
This masquerade is costly. The cost of
camouflaging is borne by the borrowers that waste
time and resources applying for loans from banks
that are incapable of lending. In addition, other
banks consider the rejection of the loan proposal
a sign that the proposal really is too risky. As a
result, with each rejection borrowers find it
increasingly difficult to locate a willing lender.
Regulatory burden
Banking has been and currently is one of
the most regulated industries in the United States.
In its report on the regulatory burden, the Federal
Financial Institutions Examination Council (1992)
stated, “Certainly federal regulation of banking is
pervasive in 1992; it affects virtually every aspect
of industry behavior.” The American Bankers
Association estimates that banks employ more than
75,000 people just to comply with regulations, an
average of more than six employees per bank.
The extent of the burden of regulation is
best summarized by the following quote on the
extension of just one type of credit (Greater
Cincinnati Business Record 1992):
Our biggest concern in the banking
industry is the overregulation. It’s unbelievable
the regulations that the bank has to live with. If
one of our people at one of our banking centers
at one of our branches wants to make a car loan
to you, here’s the legislation that they have to be
totally familiar with: the Consumer Credit Protection Act, the Truth in Lending Act, the Equal
11

Credit Opportunity Act, the Fair Credit and
Charge Card Disclosure Act, the Home Equity
Loan Consumer Protection Act of 1988, the Fair
Housing Act, the Real Estate Settlement Procedures Act, the Flood Insurance Protection Act,
the Fair Credit Billing Act, the Fair Credit Reporting Act, the Home Mortgage Disclosure Act, the
Fair Debt Collection Practice Act, the Consumer
Leasing Act, the Community Reinvestment Act,
the Bank Bribery Act, and the Securities and
Exchange Act....And this isn’t an inclusive listing.
It’s absolute insanity.
— George A. Schaefer, Jr.,
president and chief executive
officer, Fifth Third Bank.

There are costs and benefits to every regulation. Consumer protection and antidiscrimination
laws are worthwhile goals. Achieving these goals
is costly, and one cost is the effect of regulatory
burden on the availability of credit. Judging
whether the costs outweigh the benefits of any
given regulation is outside the scope of this
article, which is presenting only an explanation of
some of the more hidden costs of regulation.
If banks have always faced a heavy regulatory burden, why is this now being proposed as a
source of the credit crunch? The credit crunch did
not begin until 1986 in Texas and even later elsewhere in the nation. It is crucial then to focus on
what new regulations were enforced during this
period. In addition, the financial condition of the
banking industry should to be taken into account
when, the effect of additional regulation is assessed.
Following the failures of banks and thrifts,
Congress passed three major banking bills that
increased the regulatory micromanagement of
banks: the Competitive Equality Banking Act of
1987; the Financial Institutions Reform, Recovery,

12

12

If the regulatory burden had been imposed on a healthy,
thriving banking industry, then it still would have had a
negative effect on lending. The effect might not have
been as noticeable. It might have been the difference
between slower credit growth instead of credit contraction. In either case, the effect of the regulatory burden
might be equal, but in a healthy banking industry it would
be offset by capital growth.

and Enforcement Act of 1989 (FIRREA); and the
Federal Deposit Insurance Corporation Improvement Act of 1991 (FDICIA). These statutes funded
the closure of insolvent thrifts and constrained
banking activity.
Consumer protection laws also have increased sharply, with seven new laws since 1985
(Spong 1990). In addition, enforcement of some
previously passed legislation increased suddenly in
the early 1980s. By many accounts, the Community Reinvestment Act of 1977 (CRA) caused bankers
relatively little concern until 1989, when a major
bank’s application for an acquisition was denied
because it failed to meet its CRA responsibilities.
The increased regulatory burden further
extended the time needed for healthy banks to
take over the market share of unhealthy banks and
for unhealthy banks to recover. The additional
costs imposed on banks lowered their net income,
slowing the rebuilding of capital through retained
earnings. If the capital losses of the 1980s created
a credit crunch, then the increased regulatory
burden extended its life.12
The regulatory burden’s impact on the credit
crunch is directly related to increased compliance
costs. Four different estimates of the compliance
cost are presented in Table 2 and range from $7.5
billion to $17 billion for 1992. Based on the lowest
estimate of $7.5 billion, if these funds could have
been applied to capital rebuilding, banks could
have funded an asset expansion of $93 billion
(assuming an 8-percent capital-to-asset ratio). This
analysis, however, has not attempted to measure
the benefits of regulation.
The regulatory burden not only contributed
to the credit crunch by imposing a cost on banks,
but it also discouraged lending by imposing relatively higher costs on lending than on investing
in securities. Most discussions of the cost of the
regulatory burden treat the cost as a lump-sum tax
that must be paid by the bank. A lump-sum cost
would not affect the banks’ decisions to lend
relative to invest in securities.
In reality, many regulatory requirements
impose a greater burden on lending than on
investing in securities. For example, regulatory
requirements for frequent appraisals on real estate
loans impose a cost on a type of loan that is not
imposed on securities. If compliance costs are
directly related to lending, then the regulatory
Federal Reserve Bank of Dallas

Table 2

Cost of Regulatory Burden
Group conducting the study

Estimated annual costs for
1992 (billions of dollars)

Cost as a percent on
noninterest expenses1

7.5 to 17

6% to 14%

FFIEC2
ABA3

10.7

10%

McKinsey

10.4*

8.1%

IBAA4 Grant Thorton

11.1

8.7%

*Estimated value based on 1992 total noninterest expense of $128 billion.
1
2
3

4

Costs do not include the opportunity cost of holding required reserves that are not bearing interest.
Federal Financial Institutions Examination Council (FFIEC).
American Bankers Association (ABA) estimate does not include the cost of deposit insurance premiums, which could raise their
ratio by 5.3 percentage points based on the McKinsey & Co. study, and does not include the costs of FDICIA, which McKinsey &
Co. estimates to be at least 1.5 percentage points.
Independent Bankers Association of America (IBAA).

SOURCES: FFIEC; ABA; McKinsey & Co., Inc.; and IBAA.

burden will discourage lending even after the
banks have replenished their capital.
A third way that regulation might reduce
lending is through mandates for direct credit
allocation. Through the CRA, Congress has sent a
message to banks and bank regulators that it wants
increased lending in lower income neighborhoods.
If banks are required to lend more to lower
income borrowers, then they will reduce their
lending to other borrowers (Gruben, Neuberger,
and Schmidt 1990), and they may reduce their
total lending and increase investment in Treasury
securities (Wood 1991). Borrowers from higher
income areas could perceive this shift as a constraint on credit availability.13
This analysis is based on the following
assumptions:
1. that banks are judged for CRA compliance
by the percentage of their loan portfolio
that is lent in lower income areas;14
2. that bankers are adverse to taking risk;
and
3. that bankers believe loans in lower
income neighborhoods are riskier than
loans in higher income neighborhoods.
If the cost of failing to comply with CRA is substantial, then banks will raise the proportion of
Economic Review — Third Quarter 1993

lower income loans in their loan portfolio. As
banks hold more of the riskiest assets, bankers
balance their portfolio’s risk exposure by investing
more in risk-free Treasury securities, and total
lending declines. In an extreme case, banks may
even decrease their loan portfolios so much that
the lower income group receives fewer total loans
than previously, even though their proportion of
the loan portfolio has risen.
Consequently, the enforcement of CRA, while
achieving certain goals, is an example of how
regulatory burden can reduce total lending through
three different mechanisms. First, by imposing
compliance costs, such as those for record keeping, regulation can reduce net income and slow

13

The reduction in lending to other borrowers would be
especially severe if banks are operating at the minimum
acceptable levels of capital. If banks had excess capital,
then increased mandated lending to one group of borrowers might not affect credit availability to other borrowers.

14

Currently, banks’ CRA compliance is judged on the basis of
making a good faith effort to serve the credit needs of lowincome communities within their markets.

13

the rebuilding of capital. Second, by imposing
costs on lending, such as those for geographic
loan coding, that are not imposed on investing,
regulation can discourage lending and encourage
investment in securities. Third, the direct allocation of loan portfolio shares into loans that are
perceived to be riskier can lead banks to balance
their overall risk by reducing total lending.
Cost of increased legal exposure
Lender liability lawsuits. Lender liability is a
growing concern and an important risk exposure
for banks. Bankers’ increasing concern over these
issues can be demonstrated by the rise in the
number of citations of “lender liability” in the
American Banker over the past seven years
(Figure 1 ). The sharp rise in citations that began
in 1986 is a clear indicator of bankers’ interest
and concern. The timing of the increase is likely
related to a 1985 case on wrongful termination of
credit.
One of the biggest legal problems for banks
is that uncertainty in the law makes it difficult to
determine what actions create liability. Uncertainty
raises the risk of extending loans, because banks
are unable to estimate their exposure to lawsuits.
Increased risk discourages lending and exacerbates
problems in credit availability.

Figure 1

Number of Articles on “Lender Liability”
in the American Banker
50

40

30

20

10

0
1985

1986

1987

1988

1989

SOURCE: Federal Reserve Bank of Dallas.

14

1990

1991

1992

A major area of concern for bankers is potential liability for environmental cleanup costs of
property belonging to the banks’ borrowers. Banks’
environmental liability arises from several sources
—the Comprehensive Environmental Response,
Compensation and Liability Act of 1980 (CERCLA,
or the Superfund law), the Resources Conservation Recovery Act (RCRA), and other state and
federal environmental laws.
Banking environmental litigation often concerns CERCLA provisions that make owners or
operators of contaminated properties responsible
for environmental cleanup, even if they did not
cause the contamination. CERCLA contains an
exemption for creditors that take ownership in a
foreclosure (Scranton 1992). An interpretation by
the courts, however, left banks susceptible to
liability for conducting ordinary banking activities
(Garsson and Kleege 1991).
Bankers objected to this interpretation, which
made foreclosures especially risky. A bank implicated under CERCLA faces liability for claims
limited only by the total cost of the cleanup,
possibly billions of dollars. The size of the bank’s
loan to the owner or operator of the contaminated
property does not limit the bank’s total exposure
to claims (Kleege 1992). The Environmental Protection Agency (EPA) has settled with banks for
smaller amounts, but as American Bankers Association associate counsel Thomas J. Greco notes, “If
a bank hasn’t done anything, why should it be
paying anything at all?” (Kleege 1991a).
Some relief has been given. The EPA has
written new rules that define the terms when a
bank is liable for cleanup costs. These new rules,
however, have not been fully challenged in the
courts, and attempts to make the rules into law
have been unsuccessful in Congress.
Requiring banks to pay for property contamination cleanup expenses has contributed to the
credit crunch. The cost of screening loans for
environmental risks discourages lending. Banks shy
away from extending loans to businesses that utilize
hazardous materials. Many of these businesses are
small, local businesses such as dry cleaners, funeral
parlors, gasoline stations, and farms (Garsson and
Kleege 1991). Knowledge or fears about environmental liability can halt loans to businesses of any
kind, if bankers have reason to suspect contamination (Kleege 1990). Beyond legal liability, the
Federal Reserve Bank of Dallas

bank is not protected from the borrower reducing
the value of the collateral through pollution.
Environmental issues are not the only ones
that can land banks in court. Banks face legal
liability in providing many banking services. With
regard to their borrowers, banks can be sued if
they exercise excessive control over borrowers or
if they wrongfully terminate credit. In addition,
banks can be sued under the Racketeer-Influenced
and Corrupt Organizations Act of 1970, also known
as the RICO Act, which was originally designed to
attack organized crime.
The issue of excessive control dates to a 1984
Texas court decision that established limits as to
what direct influence a bank can exert over its
borrowers. The effect of this decision is best summarized by A. Barry Cappello, a lawyer specializing in lender liability, who said, “Whenever a
lender has anything other than an arm’s-length
relationship with its borrower, the potential for
liability exists” (Adkins 1992).
This decision has discouraged lending in at
least two ways. First, lending is now riskier because
banks are more limited in the actions they can
take to enhance the probability of repayment and
protect their collateral. Second, the cost of defending against such suits and the possible damages
that must be paid are costs to supplying credit
and must be factored in the bank’s pricing. As a
result, the amount of credit a bank is willing to
supply at any given price is reduced.
Legal exposure for the wrongful termination
of credit was a problem for banks in the mid-1980s,
but it has diminished in recent years. Court rulings
had made it no longer sufficient merely to stay
within the terms of the loan agreement; banks had
to show reasonable cause. In recent years, however, the courts have allowed banks to enforce
the terms of a loan agreement without imposing
additional requirements. The timing here is important. Even if this legal issue has diminished in
importance in recent years, it could have contributed to the Texas credit crunch that began in 1985.
Though the RICO Act was passed in 1970, it
did not become a problem for bankers until 1985,
when the Supreme Court expanded RICO to
include banks. The RICO designation permits the
plaintiff to ask for treble damages. RICO is yet
another example of an increase in lender liability
that occurred in the mid-1980s. Protecting against
Economic Review — Third Quarter 1993

such lawsuits is costly and discourages lending. In
contrast, no one has ever been sued by the federal
government for buying Treasury securities.
Regulator lawsuits. In the 1980s, financial institutions failed for a variety of reasons, only some
of which might be considered criminal. Some
thrift and bank managers were guilty of criminal
misconduct because of insider dealing or other
fraudulent acts. Many other banks and thrifts
failed because they made mistakes in their loan
decisions. Often these mistakes were only apparent
after the fact and could not necessarily have been
foreseen at the time the loan was approved.
The FDIC, however, has reacted to the bank
failures with scores of lawsuits against bank officers
and directors. These lawsuits serve two purposes
for the FDIC. First, the lawsuits seek to collect on
the director and officer insurance banks routinely
purchase, shifting a portion of the costs to the
private insurance industry. Second, these lawsuits
tend to focus attention on the industry’s responsibility for the problems.
In the 1990s, the FDIC has attempted to raise
the acceptable standard for bank officers’ and
directors’ behavior. Under the proposed standard,
a bank failure in the normal course of business
would be evidence of simple negligence, and the
FDIC would sue for damages. The courts have
failed to accept this new standard (Rehm 1993).
These lawsuits contribute to the credit crunch.
Bank officers and directors are encouraged to be
more cautious in assessing risk and return tradeoffs. Directors and officers are expected to avoid
ex post any risk that resulted in a loss to the bank.
Of course, the best way for a bank to avoid risk is
to avoid lending, an inherently risky activity.
These lawsuits also increase the difficulty of
recruiting highly competent individuals for positions on banks’ boards of directors. High-quality
directors monitoring bank management reduces
regulators’ burden monitoring the industry.
Furthermore, these lawsuits have resulted in
higher premiums for directors’ and officers’
insurance for nearly all banks. These higher
costs must be factored into loan pricing.
Conclusion
Sometimes, six observers can find six different
causes for a single problem, and they can all be
15

right. The credit crunch is an example. The credit
crunch is the result of multiple factors adversely
affecting banks’ ability to supply credit at a time
when banks’ ability to adjust to these factors was
unusually limited.
Increased lending is limited to some degree
by the necessary rebuilding of banks’ capital
positions. The drains on capital in recent years
have been substantial. Loan losses have directly
reduced capital at the banks experiencing the
loss; recycling the assets of failed banks and thrifts
also created a huge need for additional capital;
and following an atypically severe economic
contraction, both regulators and bankers appear
to have raised the acceptable minimum capital
levels and credit standards.
The costs of lending have also risen substantially over this time period. The resolution of
failed banks and thrifts destroyed valuable information. The perception of an overreaction by
bank examiners has raised the expected cost of
funding lending activity. Regulatory burden and
increased exposure to legal liability has raised the
cost of doing business for banks. By decreasing
net income, these costs compound the problem of

16

raising capital through retained earnings, and in
many cases, these costs skew the cost of extending loans relative to investing in securities.
Since there are multiple causes of the credit
crunch, the solutions to the credit crunch need to
be multifaceted. Some causes are temporary in
nature and will correct themselves over time.
Other causes, however, are structural and will not
be eliminated with economic recovery. Some
causes are the unintended side effects of policies
addressing other societal problems. Simply addressing the financial condition of the banks is unlikely
to generate a quick solution. Many healthy banks
in Texas have declined to increase their lending
in the first five years of the regional economic
expansion. While it is beyond the scope of this
article to propose solutions, solutions need to be
found. Many of the causes of the credit crunch are
the result of policies addressing other problems,
such as bank failures, community development,
credit discrimination, and access to the courts.
The positive outcome of these policies must be
carefully weighed against the economic consequences of inhibiting the flow of credit and slowing economic growth and job creation.

Federal Reserve Bank of Dallas

References
Adkins, Lynn W. (1992), “Lenders Warned of
Liability Risks,” American Banker, June 24, 6.
American Bankers Association (1992), Cut the Red
Tape: Let Banks Get Back to Business (Washington, D.C.: American Bankers Association,
November).
Baer, Herbert L., and John N. McElravey (1993),
“Capital Shocks and Bank Growth: 1973 to
1991,” Federal Reserve Bank of Chicago
Working Paper (Paper presented at the meeting of the Federal Reserve System Research
Committee on Financial Structure and Regulation, Miami, Fla., February 3–4).
Bernanke, Ben S., and Cara S. Lown (1991), “The
Credit Crunch,” Brookings Papers on Economic Activity, no. 2: 205–47.
Bizer, David S., (1993), “Examiner’s Crunch
Credit,” Wall Street Journal, March 1, A14.
——— (1993), “Regulatory Discretion and the
Credit Crunch,” unpublished manuscript, U.S.
Securities and Exchange Commission, Washington D.C., revised April 19.
Bleakley, Fred R. (1993), “Tough Guys: ‘Regulators
from Hell’ Frighten Some Banks But Also Win
Praise,” Wall Street Journal, April 27, A1.
Board of Directors of the Federal Reserve Bank of
Dallas (1991), “The Credit Crunch: Statement
of the Problem and Proposed Solutions,”
transmitted to John P. LaWare, Board of
Governors, December 30.
Board of Governors of the Federal Reserve System
(1993), “Flow of Funds Accounts, Flows and
Outstandings, Fourth Quarter 1992,” Statistical
Release Z.1, March 10.
Clair, Robert T. (1989), “Credit Shortage Slows
Texas Recovery,” Federal Reserve Bank of
Dallas Southwest Economy, January.

Economic Review — Third Quarter 1993

———, and Kevin Yeats (1991), “Bank Capital
and Its Relationship to the Credit Shortage in
Texas” (Paper presented at the Meeting of the
Federal Reserve System Research Committee
on Financial Structure and Regulation, San
Antonio, Texas, January 7– 8).
———, and Fiona Sigalla (1993), “Regional
Roundup...What Do the Coming Months Hold
for Texas?” Journal of Commercial Lending,
forthcoming.
Clark, Barkley (1986), “More Than Meets the Eye
from 9 Court Rulings,” American Banker,
August 7, 4.
Elliehausen, Gregory E., and John D. Wolken (1990),
“Banking Markets and the Use of Financial
Services by Small and Medium-Sized Businesses,”
Federal Reserve Bulletin 76 (October): 801–17.
Federal Financial Institutions Examination Council
(1992), “Study on Regulatory Burden,” Washington D.C., December 17.
Furlong, Fredrick T. (1992), “Capital Regulation and
Bank Lending,” Federal Reserve Bank of San
Francisco Economic Review, Number 3, 23–33.
Garsson, Robert M., and Stephen Kleege (1991),
“High Court Rebuffs Banks On Liability,”
American Banker , January 15, 1.
Grant Thorton (1993), “Regulatory Burden: The
Cost to Community Banks” (A study prepared
for the Independent Bankers Association of
America, January).
Greater Cincinnati Business Record (1992), “Banking Survey,” March 2, 16.
Gruben, William C., Jonathon A. Neuberger, and
Ronald H. Schmidt (1990), “Imperfect Information and the Community Reinvestment
Act,” Federal Reserve Bank of San Francisco
Economic Review, Summer, 27– 46.

17

Gunther, Jeffery W. (1989), “Texas Banking Conditions: Managerial Versus Economic Factors,”
Federal Reserve Bank of Dallas Financial
Industry Studies, October, 1–18.
Hancock, Diana, and James A. Wilcox (1992),
“The Effects on Bank Assets of Business Conditions and Capital Shortfalls,” (Published in
Credit Markets in Transition: Proceedings of
the 28 th Annual Conference on Bank Structure and Competition, Federal Reserve Bank
of Chicago, May 7, 502 –20).
Jordan, Jerry L. (1992), “The Credit Crunch: A
Monetarist’s Perspective,” (in Credit Markets
in Transition: Proceedings of the 28 th Annual
Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May 7),
470 – 82.
Kang, Grace (1992), “Courts Increasingly Are
Favoring Banks in Credit Suspension Suits,”
Wall Street Journal, June 30.
Keran Michael W. (1992), “The Regulatory Recession?” Economic Review, The Prudential
Insurance Company of America, October.
Kleege, Stephen (1992), “ ‘Clean’ Lending,” American Banker, April 27, 2A.
——— (1991a), “EPA Shows Flexibility in Seafirst
Pact,” American Banker, May 15, 22.
——— (1991b), “Curbs on Cleanup Liability
Sought,” American Banker, May 15, 22.
——— (1990), “Kansas Bankers Run from the
Blob,” American Banker, December 7, 1.
Klinkerman, Steve (1993), “Banks Accused of
Inflating Compliance Cost Estimates,” American Banker, April 6, 1.
McKinsey & Co., Inc. (1992), “Estimating Bank
Regulatory Cost Burdens” (Presentation to
Federal Reserve officials, May 1).
Owens, Raymond E., and Stacey L. Schreft (1992),
“Identifying Credit Crunches,” Federal Reserve
18

Bank of Richmond Working Paper no. 92–1,
March.
Pare, Terence P. (1993), “Why Banks Are Still
Stingy,” Fortune, January 25, 73 –5.
Pavel, Christine, and Harvey Rosenblum (1985),
“Banks and nonbanks: The horse race continues,” Federal Reserve Bank of Chicago Economic Review, May/June, 3 –17.
Peek, Joe, and Eric S. Rosengren (1993), “Bank
Regulation and the Credit Crunch,” Federal
Reserve Bank of Boston Working Paper no.
93–2, February.
———, and ——— (1992), “The Capital Crunch
in New England,” Federal Reserve Bank of
Boston New England Economic Review, May/
June, 21–31.
Petersen, D’Ann M. (1992), “Will Office Real Estate
in Texas Ever Recover?” Federal Reserve Bank
of Dallas Southwest Economy , September/
October, 6– 8.
Prestridge, Sam (1992), “Mississippi Banks Coping
with Weight of Compliance,” Mississippi
Business Journal, vol. 14, April 20.
Rehm, Barbara A. (1993), “FDIC Loses Suit Against
Directors,” American Banker, February 19, 1.
——— (1992), “ABA: Cost of Compliance Equals 59%
of Bank Profits,” American Banker, June 18, 1.
Robinson, Kenneth J. (1990), “The Performance of
Eleventh District Financial Institutions in the
1980s: A Broader Perspective,” Federal Reserve Bank of Dallas Financial Industry
Studies, May, 13–24.
Rosenblum, Harvey (1991), “Pathology of a Credit
Crunch,” Federal Reserve Bank of Dallas
Southwest Economy, July/August, 6.
———, and Robert T. Clair (1993), “Banking Conditions and the Credit Crunch: Implications for
Monetary Growth and the Economy,” Business Economics 28 (April): 42– 47.
Federal Reserve Bank of Dallas

Ryan, Timothy (1992), “Banking’s New Risk—
Litigation,” Wall Street Journal, November 13,
A14.
Samolyk, Katherine A. (1991), “A Regional Perspective on the Credit View,” Federal Reserve
Bank of Cleveland Economic Review, Second
Quarter, 27–38.
Scranton, David F. (1991), “How the Proposed EPA
Rule Affects Lender Liability,” Journal Commercial Bank Lending 74 (October): 18 –28.
——— (1992), “Understanding the New EPA
Lender Liability Rule,” Journal of Commercial
Lending 74 ( July): 6 –16.
Secura Group (1992), “The Burden of Bank Regulation: Tracing the Cost Imposed by Bank
Regulation on the American Public” (Paper
prepared for the American Bankers Association, November 16).

Spong, Kenneth (1990), Banking Regulation: Its
Purposes, Implementation, and Effects, 3d ed.,
(Kansas City: Federal Reserve Bank of Kansas
City).
Stiglitz, Joseph E., and Andrew Weiss (1981),
“Credit Rationing in Markets with Imperfect
Information,” American Economic Review 71
( June): 393 – 410.
Syron, Richard F. (1991), “Are We Experiencing
a Credit Crunch?” Federal Reserve Bank of
Boston New England Economic Review,
July/August, 3 –10.
Wood, John (1991), “An Extension of Gruben–
Neuberger–Schmidt on the Impact of CRA,”
correspondence to William C. Gruben, Federal Reserve Bank of Dallas, June 14.
Yeats, Kevin (1989), “Tax Treatment of Real Estate
Depreciation,” Real Estate Accounting and
Taxation, Fall, 48 –57.

Short, Eugenie D., and Gerald P. O’Driscoll, Jr.
(1983), “Deregulation and Deposit Insurance,”
Federal Reserve Bank of Dallas Economic
Review, September, 1–10.

Economic Review — Third Quarter 1993

19

20

Federal Reserve Bank of Dallas

Beverly J. Fox

Lori L. Taylor

Assistant Economist
Federal Reserve Bank of Dallas

Senior Economist
Federal Reserve Bank of Dallas

Mine K. Yücel
Senior Economist and Policy Advisor
Federal Reserve Bank of Dallas

America’s Health Care Problem:
An Economic Perspective

H

ealth care expenditures in the United States
are expanding rapidly. Real per capita expenditures on health care more than doubled over the
period 1970 –90.1 Real expenditures for health
care are now growing nearly 4 percent per year,
while real expenditures on other consumer goods
are growing only 2.5 percent per year.2 Furthermore, health services grew more than twice as fast
as any other major industry during the recent
recession. If expenditures continue to grow at the
current rate, health care will represent a larger
share of the United States’ gross domestic product
(GDP) than manufacturing by 2000.3
The explosive growth in health care expenditures concerns many Americans. Citizens fear
that they will be priced out of the market for health
care. Business people worry that rising health care
costs will reduce the international competitiveness
of U.S. corporations. Politicians worry that rising
bills for health care programs like Medicare and
Medicaid will force the government to raise taxes
or run increasingly large deficits.
The widespread concern has led to demands
for substantial reform of the U.S. health care
system. Some groups call for controls on health
care prices. Others want to reform the insurance
industry. There are plans that call for managed
competition and plans that eliminate competition
by making the government the sole provider of
health services. There are almost as many plans as
there are interested parties.
However, before we can fix the system, we
have to know what parts of it are broken. If the
increase in health expenditures reflects distortions
Economic Review — Third Quarter 1993

in demand, then we should focus on reforming
consumer incentives. If distortions in supply fuel the
expenditures increase, then we should respond
with policies that affect suppliers. If the increase
in health expenditures reflects shifts in market
fundamentals—for example, the increasing health
care demands of an aging population—then
economic analysis suggests that the system does
not need fixing, and we should leave it alone.
Why is everyone so concerned?
Until recently, health care costs were not a
major concern of most Americans. Surveys on the
top problems facing the United States in 1984 did
not even mention health care.4 Today, however,
reforming the health care system is one of the
primary objectives of state and federal governments.
A look at health care prices suggests one
reason for this change in perspective. As Figure 1

Our thanks to Zsolt Becsi, Steve Brown, and Mark Wynne
for their comments and suggestions.
1

Levit et al. (1991).

2

Based on data from the U.S. Department of Commerce,
Bureau of Economic Analysis.

3

Based on data from the U.S. Department of Commerce,
Bureau of Economic Analysis.

4

For example, see the reader survey in Tift (1984).

21

indicates, health care prices increased at roughly
the same rate as the general price level until the
early 1980s. After the mid-1980s, however, the
medical care component of the consumer price
index shot upward. By 1992, medical care prices
were increasing at more than twice the rate of
inflation.5
This sharp increase in health care prices has
led consumers to fear that they are being priced
out of the market for health care. Publicity on the
35 million uninsured Americans lends credibility
to those fears.6 Because many Americans view
health care as essential, the prospect of being
unable to afford it frightens them.
Rising health care prices also concern business because employers pay a large proportion of
the Medicare and Medicaid taxes and 64 percent
of private insurance premiums.7 Wage and price
controls during World War II encouraged employers
to provide fringe benefits such as health insurance
in lieu of wage increases. The tax-exempt status
of fringe benefits led many employers to continue
the practice after the controls were removed.
Therefore, much of the increase in health care
expenditures is a drag on the balance sheets of
American employers.
Furthermore, government is concerned about
increasing health care costs. Federal expenditures
for Medicare, which finances health care services
for the elderly, and Medicaid, which finances
health care services for the poor and disabled,
have been growing more than 10 percent per year
since 1985.8 The Congressional Budget Office
estimates that health spending consumed 15 percent

Figure 1

Consumer Prices
Index, 1982– 84 = 100
200
180

Medical care

160
140
120

All items

100
80
60
40
20
0
’60 ’62 ’64 ’66 ’68 ’70 ’72 ’74 ’76 ’78 ’80 ’82 ’84 ’86 ’88 ’90 ’92

SOURCE: U.S. Bureau of Labor Statistics.

of the federal budget in 1992 and will consume
28 percent of the federal budget by 2002.9
Ultimately, however, consumers bear the
burden of increases in health care spending. Much
of the increase in employer health costs is passed
along to employees in the form of lower wages
(see the box entitled “Health Care Costs and
Profitability”). The increase in government health
costs is passed along to citizens in the form of
higher taxes or fewer alternative services. Therefore, consumers would be the primary beneficiaries
of health care reform.
Sources of increasing health
care expenditures

22

5

The medical care component of the consumer price index
may mismeasure medical inflation somewhat, because it is
difficult to adjust properly for changes in medical technology and the quality of care. However, it undoubtedly
influences the public’s perceptions of health care prices.

6

Garrison (1990).

7

Levit and Cowan (1991).

8

Levit et al. (1991).

9

Burman and Rodgers (1992).

It is possible to determine the best way to
reform the health care system using the basic
principles of supply and demand. If no distortions
exist, the health care market achieves the optimal
resource allocation for a given income distribution.
The increase in health care expenditures then
reflects either an increase in the public’s desire for
health services or an increase in legitimate costs.
Under these conditions, if society is unhappy with
the allocation, the best solution is to redistribute
income without meddling in the health care market.
Federal Reserve Bank of Dallas

Health Care Costs and Profitability
Businesses pay most of the nation’s
health bills, but the effect of increasing health
care costs on profits is not straightforward.
Although increases in health costs for retirees
would have a negative effect on firm profitability, increases in health costs for current employees can have a positive effect on firm
profitability.
The health care costs of current workers
are part of a total compensation offer that is
determined by the worker’s contribution to the
firm’s output. As long as the worker’s productivity is unaffected by increases in health care
costs, the amount of total compensation the
firm is willing to offer is unaffected by increases in health costs. Therefore, increases
in health costs should be offset by decreases
in wages to keep the total compensation
package unchanged.
Furthermore, the increase in health care
costs increases the value to employees of the
tax exemption for fringe benefits. The advantages of being employed by a firm that offers
health benefits increase, so more workers are
attracted to such firms. As the supply of labor
offered to firms that provide health benefits

However, if the health care system is distorted,
reform is needed to eliminate the distortions.
We have identified several distortions in the
current system of health care. First, tax subsidies for
employer-provided health insurance lead to excess
demand for health insurance and, consequently,
to excess consumption of health care. Second,
regulations and industry practices restrict the supply
of health care professionals, leading to higher
prices for health services. Finally, the structure of
the health insurance industry promotes inefficiency.
These distortions of both supply and demand lead
to excessive expenditures on health care.
Expenditures also are increasing for several
reasons that are nondistortionary. These reasons
include uncertainty about causes and appropriate
Economic Review — Third Quarter 1993

increases, the total price those firms must pay
for it decreases, and those firms’ total compensation costs can fall. Therefore, firms that
offer health insurance as a fringe benefit to
their employees can be made better off—not
worse off—by the increase in health costs.
Unfortunately, the savings on total compensation for current employees can be more
than offset by increased costs for the health
care of retirees. After all, the increases in
health costs for retirees cannot be offset by
decreases in wages. The problem has become particularly evident recently as accounting rule changes have forced firms to indicate
their commitments to retiree benefits on their
balance sheets. For example, General Motors was forced to record a $22.2 billion charge
in 1992 for retiree and future retiree health
costs.1 Firms that respond to the increase in
health care costs by modifying or eliminating
health care coverage for the retired may face
increased wage demands by current employees who fear being treated in a similar way
when they retire.
1

Dallas Morning News (1993).

treatments for health problems, changes in population demographics, and society’s reluctance to
place limits on the value of human life.
The implicit tax subsidy for health insurance
For nearly fifty years, employer-provided
fringe benefits have been exempt from both
personal income taxes and payroll taxes such as
those for Social Security. Thus, employees avoid
taxes by taking some of their compensation in the
form of health insurance. If the combined marginal
tax rate is 28 percent, an employee can receive $1’s
worth of health care instead of 72 cents’ worth of
after-tax take-home pay (Table 1 ). The difference
represents an implied tax subsidy. As Table 1
23

Table 1

The Subsidized Price of Health Care

Wage

Income
tax
(percent)

Effective marginal
tax rate*
(percent)

Price of health care
in terms of
take-home pay

$1
$1
$1
$1

0
15
28
33

14
28
40
45

$.86
$.72
$.60
$.55

*Effective marginal tax rate equals share of the last dollar of monetary compensation paid in
federal taxes and includes both payroll and income taxes.

indicates, those in the highest tax bracket receive
the largest tax subsidy, while those in the lowest
tax bracket receive a much smaller subsidy.10
Because employees will naturally buy more
health insurance at 72 cents than at $1, excluding
health-related fringe benefits from taxable income
increases expenditures on health insurance by
those receiving the subsidy. Burman and Rodgers
(1992) estimate that the subsidy costs the federal
government $65 billion per year in foregone
revenue and increases private health insurance
spending by roughly one-third.
Excessive consumption of health insurance
has a number of disquieting consequences. First,
because health insurance leads to increased consumption of health care, excessive consumption
of health insurance produces excessive consumption of health care. (For a discussion of the ways
in which health insurance increases health care
consumption, see the box entitled “The Relationship Between Health Insurance and Health Care
Consumption.”) Second, overconsumption of health
insurance by those receiving the implicit subsidy
increases the insurance premiums of the unsub-

10

24

Residents of cities and states with income taxes receive
additional subsidies because their fringe benefits are also
exempt from local income taxes.

sidized and may cause some consumers to be
underinsured. Finally, excessive health insurance
distorts medical research in favor of technologies
that extend or improve life at any price rather than
technologies that reduce the costs of treatment.
By its nature, health insurance makes consumers less sensitive to health care prices, thereby
generating more expenditures on health care than
would otherwise occur. Given this relationship,
excessive insurance consumption necessarily leads
to excessive health care consumption. Phelps
(1992) estimates that annual health care expenditures are between 10 percent and 20 percent
higher because of the subsidy.
By leading to excessive health care expenditures, the tax subsidy also can exacerbate the
problem of the uninsured. Because the subsidy
increases demand for health care, health care prices
rise, putting upward pressure on health insurer
costs. Higher payouts result in higher insurance
premiums. Thus, the subsidy distorts the distribution of health insurance so that higher income
households overconsume health insurance, while
lower income households can be priced out of
the market for health insurance and health care.
Overconsumption of health insurance also
plays a role in directing technical progress in health
care and has reinforced the development of costly
technologies. Contrary to conventional wisdom,
technological improvements in health care generally
have not lowered costs. Rather, technological
innovations have brought about a higher quality
Federal Reserve Bank of Dallas

The Relationship Between Health Insurance and Health Care Consumption
Substantial research indicates that as
the price to consumers decreases, health
care consumption increases (Long and
Rodgers 1990, Phelps 1992, Keeler and Rolph
1988, Manning et al. 1987). According to the
Rand Health Insurance Experiment, the price
elasticity of demand for health care is –0.2
(Keeler and Rolph 1988, Manning et al. 1987).
In other words, every 1 percent decrease in
consumer prices for health care increases
health care consumption by 0.2 percent.
Insurance reduces the consumer’s effective price of health care in two ways. First,
because health insurers typically pay for health
treatments rather than for health losses, insurance lowers the marginal price of treatment. If a consumer is fully insured (and,
therefore, pays none of the billable costs of
treatment), then the marginal cost of health
care becomes the opportunity cost of the
consumer’s time. If a consumer is co-insured,
then the marginal cost of treatment becomes
a predetermined fraction of the treatment
cost, plus the consumer’s opportunity costs.
For example, with a copayment of 20 percent,
a $10 prescription antihistamine costs the
consumer only $2. In either case, health insurance effectively reduces the consumer’s
marginal cost of health care.
Second, because health insurance premiums are only loosely connected to claims,
insurance insulates people from some of the
costs of their decisions. Theoretically, insurance premiums, which reflect expected losses,
are a function of health risk and the extent of

product, which is most often more expensive than
the older product. Weisbrod (1991) finds that our
system of pricing (that is, paying the health care
provider based on costs incurred or on a fee-forservice basis) has led the research and development sector to develop new technologies that
enhance the quality of care irrespective of cost
Economic Review — Third Quarter 1993

claims. In such a case, consumers have incentives to limit their health care consumption
and submit only those claims that are worth
the resulting increase in premiums. In practice, however, an individual in a large health
insurance plan pays an average premium that
is almost independent of the individual’s risk
or health care consumption. Hence, consumers do not bear the full costs of their decisions
about the extent of claims.
Furthermore, the loose connection between premiums and claims in health insurance exacerbates problems of moral hazard.
Moral hazard arises when insurance changes
the insured’s behavior in a way that increases
claims. For example, people who are insured
and who, therefore, know that they will bear
only part of the cost of illness, may not be as
careful of their health as people who are not
insured. Individuals will have no incentive to
curb unhealthy behavior if increased claims
are not reflected in higher premiums, especially if behavior cannot be easily monitored.
Thus, health insurance increases expenditures on health care. In the Rand experiment, fully insured individuals spent 30 percent
more on outpatient services than individuals
with a 25 percent copayment. In turn, individuals with a 25 percent copayment spent 28
percent more than individuals with a 95 percent copayment (Manning et al. 1987).1

1

Both co-insurance programs had an annual cap on out-ofpocket expenses.

rather than the cost-effective technologies that
probably would develop if consumers were more
sensitive to health care prices.
One could argue that the tax subsidy is
necessary because without it, poor people would
receive less medical care and there would be
greater public health risks from communicable
25

Figure 2

Average Employer-Provided Health Benefits
Dollars
3,500

3,000

Estimated tax subsidy

2,500

2,000

1,500

1,000

500

0
$0 to
$7,600

$7,600 to
$21,600

$21,600 to
$36,300

$36,300 to
$57,100

$57,100
and higher

Income quintile
SOURCE: U.S. Department of Commerce, Bureau of the Census.

diseases such as tuberculosis. However, the progressive nature of the income tax code negates
those arguments. As Figure 2 indicates, high-income
households receive a greater health insurance
subsidy than low-income households. Households
that fall in the lowest income tax bracket receive a
small subsidy because health benefits are exempt
from Social Security and other payroll taxes. Meanwhile, some of the households in the highest income
tax bracket receive a federal subsidy of nearly 50
percent when both income and payroll taxes are
considered. In combination with an exemption
from state taxes, high-income households in hightax states receive an even larger subsidy. There is
little risk that high-income households will not be
able to afford insurance and no obvious consensus that these groups deserve public assistance.
Supply constraints
Numerous restrictions on entry to the health
care profession distort health care supply and lead
to higher consumer prices. These restrictions
include limits on access to medical training, licensing and certification requirements for doctors, and
work rules that exclude paraprofessionals from
26

performing many medical tasks. The restrictions
are ostensibly designed to protect the consumer by
increasing the quality of the health care product.
Studies have shown, however, that regulations
that limit supply do not always lead to higher
quality and tend to increase expenditures because
they increase incomes in the profession.
People who want to become doctors must first
gain entry into an accredited U.S. medical school.
Doctors who train at nonaccredited schools or in
other countries frequently are not permitted to practice medicine in the United States. The market for
medical training is monopolistic, and the number of
medical school applicants greatly exceeds the number
of openings at accredited schools. Each year since
1960, medical school applications have exceeded
classroom openings by at least 50 percent (Association of American Medical Colleges 1993, Table B–
1). In the 1992–93 school year, there were two
applicants for every opening. Restrictions on the
supply of medical training necessarily restricts the
supply of physicians. Assuming that those students
who were not accepted into medical schools were
only 50 percent as likely to complete their education
as those who were accepted, the restriction reduces
physician supply by approximately 30 percent.
Once physicians have graduated from
medical school, they face additional restrictions
imposed by state and local agencies. States have
licensing and regulatory agencies or boards that
regulate the medical profession. The agencies
establish the minimum level of education and
experience required to practice, define the functions
of the profession, and limit the performance of
certain functions to licensed professionals. Restrictions include the use of trade names, restrictions
on branch offices and location of offices, and, until
1977, a ban on advertising (Haas –Wilson 1992).
Many studies have shown that occupational
licensing leads to lower consumer welfare and
higher incomes in the licensed profession. Economic
theory suggests that self-licensing by the medical
profession leads to economic rents (Friedman
1962 and Stigler 1971). Leland (1979) finds that
although minimum quality standards may be
desirable in markets in which suppliers have more
information than consumers, the minimum quality
standards set by the medical industry may be too
high. Chan and Leland (1982) show that when both
price and quality are hard to observe, uninformed
Federal Reserve Bank of Dallas

consumers may pay a higher price and receive a
lower quality of goods. Haas –Wilson (1986) finds
that increasing the restrictiveness of optometrists’
licensing examinations increased the price of eye
exams and eyeglasses significantly but had an
insignificant effect on the quality of the eye exams.
Whenever entry into a market is artificially
constrained, either through restricted access to
medical training or through obstacles such as
licensing and certification, consumer prices are
inefficiently high. Therefore, restrictions on entry
into the health care profession, together with
work rules that prevent competition within the
profession between physicians and less-expensive
paraprofessionals, increase medical costs.
Relaxing some of the restrictions on entry
into the medical profession should make consumers
better off. Shaked and Sutton (1981) show that
granting monopolistic powers to the self-regulating profession is likely to be welfare-reducing and
that the entry of paraprofessionals would be
welfare-improving. Moreover, the size of the paraprofession that leads to the greatest improvement
in welfare is the size that leads to the greatest
income loss for members already in the profession.
Evans and Williamson (1978) estimate that in
Ontario, Canada, a dental care system that made
optimal use of paraprofessionals could reduce the
cost of care by 30 percent to 40 percent. More
recent studies on restrictions in the dental profession (Liang and Ogur 1987) estimate that state
restrictions on the number of auxiliaries a dentist
can hire and the functions they may perform cost
consumers $700 million in 1982.
Counter to the principles of supply and
demand, there are some who assert that an increase
in physician supply would, in fact, cause higher
prices. They cite the phenomenon that doctors
charge higher fees in communities with high
physician-to-patient ratios than they charge in
communities that are less well supplied, even
after adjusting for input cost differences.
However, there is no need to suspend the
laws of supply and demand to explain this phenomenon. Where there is a greater density of
physicians, there also may be a greater degree of
specialization and nonprice competition. Physicians
segment a large market and respond to a greater
variety of needs and preferences by treating fewer
patients but charging higher prices.11
Economic Review — Third Quarter 1993

Inefficiencies in the insurance
industry’s structure
Another distortion in the health care system
arises from the structure of the insurance industry.
The market for health insurance is dominated by
noncompetitive firms. Medicare and Medicaid,
which represent 57 percent of the insurance market,
are government entities.12 Furthermore, much of
the private market for health insurance is dominated by not-for-profit groups like Blue Cross and
Blue Shield. Only 30 percent of the health insurance market is served by for-profit commercial
insurers. Without the discipline of competition,
the market for health insurance is inefficient and
encourages higher health care costs.
Considerable economic research indicates
that government agencies are, in general, inefficient (Breton 1974, Downs 1967, and Tullock
1965 and 1967). According to Niskanen (1971), government agencies are more likely to try to maximize the size of their budgets than to maximize
profits because budget size is a mark of the power
and prestige of the agency. Among other bureaucratic goals are salaries, office perks, and patronage.
Weatherby (1971) cites the expansion of personnel
as a goal pursued by bureaucrats. Borcherding’s
(1977) and Spann’s (1977) findings on the growth
of government and lack of productivity growth
are consistent with Niskanen’s theory. Since agencies
have to return any unused moneys to the U.S.
Treasury, they are not residual claimants on cost
savings in the budget and have few incentives to
cut costs. There is no reason to believe that
Medicare and Medicaid administrators behave
differently than other bureaucrats.
Like government agencies, not-for-profit
firms also face incentives to behave inefficiently.
(Alchian and Demsetz 1972, Eisenstadt and Kennedy 1981, and Sindelar 1988). Nonprofit health
insurers have incentives to dissipate any potential
profits through excess payments to doctors and

11

Phelps (1992, 202).

12

U.S. Bureau of the Census (1992).

27

hospitals, unusually generous insurance coverage,
or artificially low insurance premiums. Sindelar
(1988) finds that, unlike for-profit insurers, Blue
Cross and Blue Shield plans (the Blues) do not
respond to market forces by changing the price of
health insurance (measured as the ratio of premiums to benefits). In particular, Sindelar finds
that administrative costs for the Blues increase as
the size of the typical insurance claim increases,
suggesting that the Blues do not take advantage
of economies of scale that are exploited by commercial insurers.
In most industries, the existence of a competitive fringe of efficient firms would discipline
the inefficient nonprofit firms (Baumol, Panzar and
Willig 1988; Caves and Christensen 1980). However, in the insurance industry, inefficient nonprofit insurers receive tax advantages not available
to for-profit insurers. Most states tax the insurance
premiums of for-profit insurers, while they exempt
the premiums of nonprofit insurers or tax them at
lower rates. Eisenstadt and Kennedy (1981) find
that Blue Shield plans were less efficient in states
where the plans had a tax advantage than in
states where they did not.13 According to Eisenstadt and Kennedy, “the regulatory advantages
given to the ‘blues’...allow inefficient behavior to
be maintained.” 14
One could argue that nonprofit insurers
should receive tax advantages because they generally accept customers with preexisting conditions
that other insurers consider uninsurable. However,
society could subsidize insurance for individuals
with preexisting conditions without requiring that
the insurer be a nonprofit organization. For example, the government could provide Medicare
and Medicaid recipients with the resources to
purchase private insurance rather than providing
the insurance itself. There is no need to finance
an inefficient market structure.

28

13

Inefficiency is measured by the ratio of administrative costs
to premiums. Both administrative costs and premiums are
expressed as net of premium taxes, if any.

14

Eisenstadt and Kennedy (1981, 27).

Nondistortionary sources
of increasing expenditures
In addition to the distortions, a number of
nondistortionary factors lead to higher health care
expenditures. Uncertainties on the part of both
physicians and consumers as to the nature and
causes of health problems lead to more health
care consumption than would occur if all information were freely available. However, information
is not free, and some of these expenditures are the
natural result of optimization under uncertainty.
Other nondistortionary factors that contribute to
higher expenditures include changes in the demographic composition of the U.S. population and
the nearly infinite value placed on human lives.
Uncertainty has a major influence on medical
decision-making. Doctors and patients have incomplete information about causes and cures for
many health problems. Phelps (1992) shows that
there is substantial disagreement and uncertainty
within the medical profession about the marginal
productivity of alternative medical treatments.
Uncertainty about the optimal course of action for
various health problems, together with consumers’
distaste for taking risks with their health, leads to
increased testing and treatments and, therefore,
higher health expenditures.
Further, because patients lack the information to reliably judge medical care quality, they
must rely on their doctor’s advice and judgment.
But much like an auto mechanic, the doctor has
incentives to provide (and bill for) more services
than absolutely necessary and to provide those
services with less than maximum effort. Economists
refer to these situations as principal – agent problems. The usual solution to such problems is a
contract that provides the agent (in this case the
health professional) with incentives to behave
optimally and a mechanism for monitoring the
agent’s compliance with that contract. The mechanism to monitor doctors’ behavior and provide
incentives for optimal performance is the malpractice suit.
Unfortunately, asymmetric damages make
malpractice suits more effective at inducing careful
care than cost-effective care. After all, if the doctor
orders too few tests and a patient is injured or
killed, the potential damage is huge. However, if
the doctor orders too many tests, the damage is
Federal Reserve Bank of Dallas

limited to the cost of the tests. Whenever there is
uncertainty about the appropriate number of tests,
the risk-averse doctor will prescribe more tests. Thus,
malpractice laws and asymmetric damages create
incentives for defensive medicine—procedures
designed to ward off lawsuits rather than diseases.
According to the American Medical Association,
defensive medicine and malpractice insurance add
$36 billion to the nation’s medical bills each year. 15
In addition to uncertainty, the changing
demographics of the U.S. population also contribute to increases in health care expenditures. Per
capita health care expenditures increase with both
age and income. For example, consumers 65 and
over consume more than three-and-one-half times
as much health care as consumers ages 19 to 64
(Figure 3). The aging of the population is expected
to explain one-seventh of the increase in health
care expenditures over the 40 years from 1990 to
2030.16 Furthermore, real U.S. income per capita
has grown 2.2 percent per year over the past three
decades, and as populations grow wealthier, they
consume more of all normal goods, including
health care. Simple regression analysis suggests
that one-quarter of the increase in per capita health
expenditures over the period 1960 – 90 can be
explained by these two demographic factors.
Finally, the high value we place on human
life leads to higher expenditures in the health
care system. Because most consumers would be
willing to spend huge amounts to avoid dying,
insured consumers will demand any treatment,
however costly, that will prolong a patient’s life.
The Council of Economic Advisers (1993) estimates that the 5 percent of beneficiaries who are
in the last year of their lives consume 29 percent
of the Medicare budget.
Summary and conclusions
Health care expenditures in the United States
have expanded rapidly in the past twenty years.
This growth in expenditures concerns business
people, politicians, and individual consumers of
health care, although most of the burden falls on
the consumer. Hence, health care reform has
become a primary objective of policymakers.
Increasing expenditures for health care are
not a problem when they reflect consumer demands
for health care in an undistorted market, and
Economic Review — Third Quarter 1993

Figure 3

Per Capita Health Care Expenditures by Age, 1987
Dollars
10,000
9,000
8,000
7,000
6,000
5,000
4,000
3,000
2,000
1,000
0
Under 19

19 – 64

65 – 84

Over 84

Age
SOURCE: U.S. Bureau of Labor Statistics.

some of the recent increases clearly represent the
demands of an aging and increasingly wealthy
population. However, we have identified a number
of distortions in the health care market that have
a substantial impact on health care expenditures.
The personal income tax code subsidizes health
insurance consumption, thereby fostering excessive consumption of health care. Tax exemptions
for nonprofit insurers and restrictions on the
supply of health services also lead to higher costs.
To be effective, health care reform must
address these distortions in the health care system.
Eliminating the tax subsidy for employer-provided
health insurance, reducing the tax advantages of
nonprofit insurers, and reducing the restriction
on health care providers would go a long way
toward eliminating America’s health care problem.
Only after these distortions are removed can the
economy achieve an efficient allocation of health
resources.

15

Felsenthal (1993).

16

Council of Economic Advisers (1993).

29

References
Alchian, Armen, and Harold Demsetz (1972),
“Production, Information Costs and Economic
Organizations,” American Economic Review
62 (December): 777– 95.
Association of American Medical Colleges (1993),
AAMC Databook: Tabulations of Statistical
Information Related to Medical Education
(Washington, D.C.: AAMC, January).

Downs, Anthony (1967), Inside Bureaucracy
(Boston: Little, Brown, and Co.).
Eisenstadt, David, and Thomas E. Kennedy (1981),
“Control and Behavior of Nonprofit Firms:
The Case of Blue Shield,” Southern Economic
Journal 48 ( July): 26 –36.

Baumol, William J., John C. Panzar, and Robert D.
Willig (1988), Contestable Markets and the
Theory of Industry Structure (Fort Worth,
Texas: Harcourt Brace).

Evans, Robert G., and Malcolm F. Williamson
(1978), Extending Canadian Health Insurance: Policy Options for Pharmacare and
Denticare, Ontario Economic Council Research Study no. 13 (Toronto: University of
Toronto Press).

Borcherding, T.E. (1977), “One Hundred Years of
Public Spending,” in Budgets and Bureaucrats, T.E. Borcherding, ed. (Durham, N.C.:
Duke University Press).

Felsenthal, Edward (1993), “Curing Health Care:
Doctors are Spurring Effort to Remedy the
Nation’s Ailing Malpractice System,” Wall
Street Journal , March 1, B1.

Breton, Albert (1974), The Economic Theory of
Representative Government (Chicago: Aldine
Publishing Co.).

Friedman, Milton (1962), Capitalism and Freedom
(Chicago: University of Chicago Press).

Burman, Leonard E., and Jack Rodgers (1992),
“Tax Preferences and Employment-Based
Health Insurance,” National Tax Journal 65
(September): 331– 56.
Caves, Douglas W., and Laurits R. Christensen
(1980), “The Relative Efficiency of Public and
Private Firms in a Competitive Environment:
The Case of Canadian Railroads,” Journal of
Political Economy 88 (October): 958 –76.
Chan, Yuk–Shee, and Hayne Leland (1982),
“Prices and Qualities in Markets with Costly
Information,” Review of Economic Statistics 49
(October): 499 – 516.
Council of Economic Advisers (1993), Economic
Report of the President (Washington, D.C.:
U.S. Government Printing Office).
Dallas Morning News (1993), “GM To Take $22.2
Billion Health Charge,” February 2: 1D.

30

Garrison, Louis P., Jr. (1990), “Medicaid, the
Uninsured, and National Health Spending:
Federal Policy Implications,” Health Care
Financing Review Annual Supplement, 167.
Haas–Wilson, Deborah (1992), “The Regulation of
Health Care Professionals Other than Physicians,” Regulation: The Cato Review of Business and Government, Fall, 40 – 46.
——— (1986), “The Effect of Commercial Practice
Restrictions: The Case of Optometry,” Journal
of Law and Economics 29 (April): 165 – 86.
Keeler, Emmett B., and John E. Rolph (1988),
“The Demand for Episodes of Treatment in
the Health Insurance Experiment, Journal of
Health Economics 7 (December): 337– 67.
Leland, Hayne E. (1979), “Quacks, Lemons, and
Licensing: A Theory of Minimum Quality
Standards,” Journal of Political Economy 87
(December): 1328 – 46.

Federal Reserve Bank of Dallas

Levit, Katharine R., and Cathy A. Cowan (1991),
“Business, Households, and Governments:
Health Care Costs, 1990,” Health Care Financing Review, Winter, 83 – 93.
———, Helen C. Lazenby, Cathy A. Cowan, and
Suzanne W. Letsch (1991), “National Health
Expenditures, 1990,” Health Care Financing
Review, Fall, 29 – 54.
Liang, J. Nellie, and Jonathan D. Ogur (1987), Restrictions on Dental Auxiliaries: An Economic
Policy Analysis (Washington, D.C.: Federal
Trade Commission, Bureau of Economics, 89).
Long, Stephen H., and Jack Rodgers (1990), “The
Effects of Being Uninsured on Health Care
Service Use: Estimates From the Survey of
Income and Program Participation,” Congressional Budget Office Working Paper no. 9012.
Manning, Willard G., Joseph P. Newhouse, Naihua
Duan, Emmett B. Keller, Arleen Leibowitz,
and Susan Marquis, (1987), “Health Insurance
and the Demand for Medical Care,” American
Economic Review 77 ( June): 251–77.
Niskanen, William A., Jr. (1971), Bureaucracy and
Representative Government (Chicago: Aldine
Publishing Company).
Phelps, Charles E. (1992), Health Economics (New
York: HarperCollins Publishers).
Shaked, Avner, and John Sutton (1981), “The SelfRegulating Profession,” Review of Economic
Studies 48 (April): 217– 34.

H.E. Frech III, ed. (San Francisco: Pacific
Research Center for Public Policy).
Spann, R.M. (1977), “Rates of Productivity Change
and the Growth of State and Local Government Expenditures,” in Budgets and Bureaucrats, T.E. Borcherding, ed. (Durham, N.C.:
Duke University Press).
Stigler, George (1971), “The Theory of Economic
Regulation,” Bell Journal of Economics , Autumn,
2 – 31.
Tift, Susan (1984), “Basking in the Sunshine:
During the Final Heat, a Time Poll Shows a
GOP Romp,” Time, November 5, 21.
Tullock, Gordon (1967), Toward a Mathematics
of Politics (Ann Arbor, Mich.: University of
Michigan Press).
——— (1965), The Politics of Bureaucracy,
(Washington, D.C.: Public Affairs Press).
U.S. Bureau of the Census (1992), Statistical
Abstract of the United States: 1992, 112th ed.
(Washington, D.C.: U.S. Goverment Printing
Office).
Weatherby, J.L. (1971), “A Note on Administrative
Behavior and Public Policy,” Public Choice,
Fall, 107–10.
Weisbrod, Burton A. (1991), “The Health Care
Quadrillemma: An Essay on Technological
Change, Insurance, Quality of Care, and Cost
Containment,” Journal of Economic Literature
29 ( June): 523 –52.

Sindelar, Jody L. (1988), “The Declining Price of
Health Insurance,” in Health Care in America,

Economic Review — Third Quarter 1993

31

32

Federal Reserve Bank of Dallas

Evan F. Koenig
Senior Economist and Policy Advisor
Federal Reserve Bank of Dallas

Rethinking the IS in IS–LM:
Adapting Keynesian Tools to Non-Keynesian Economies
Part 1

T

his article attempts to narrow the gap between
two macroeconomic paradigms by showing that,
in modified form, a graphical tool taken from one
of these paradigms can be used to analyze models
drawn from the other. The two paradigms are the
Keynesian and real-business-cycle approaches to
macroeconomics. The graphical tool is the IS –LM
diagram.
The IS–LM diagram was originally developed
by Hicks (1937) as a graphical representation of
ideas put forth by Keynes in his General Theory.
Not surprisingly, given its origins, the IS –LM
diagram has come to be associated with traditional
Keynesian macroeconomic analysis—analysis that
treats household expectations as either irrelevant
or exogenously determined and in which prices
fail, in the near term, to clear the markets for goods
and for labor. Not all Keynesians are comfortable
with the assumption that households are myopic.
Nevertheless, the IS –LM diagram remains the
graphical framework of choice among those who
treat sluggish price adjustment seriously.1 Expectations have usually been incorporated into textbook
IS –LM analysis in only the most rudimentary way.
Given its pedigree, the IS –LM diagram
would seem ill-suited to analyzing an economy
like that described by Barro (1990), in which prices
adjust instantaneously to clear all markets, households are forward-looking, and macroeconomic
fluctuations are due solely to shocks to tastes,
technology, and government purchases. The point
of this article, however, is that household myopia
is not an essential component of the IS –LM framework. Once this myopia is eliminated, the IS –LM
framework becomes flexible enough to encompass
a simple Barro-style real-business-cycle model as
a special case. Furthermore, in working through
Economic Review — Third Quarter 1993

the modified IS –LM model, one gains an appreciation for which of the traditional Keynesian
results flow from the assumed myopia of households, as opposed to sluggish price adjustment.
No attempt is made here to pass judgment
on the relative merits of alternative models; nor
does this article attempt to develop new theoretical insights. The models examined are simple,
comparable to those typically included in popular
undergraduate textbooks. A more intellectually
satisfying reconciliation of the Keynesian and realbusiness-cycle paradigms would move away from
the assumption—maintained throughout this
article—that households and firms are price takers.
Until research in this direction makes further
progress, any device that provides common ground
for macroeconomists and policymakers with differing perspectives provides a valuable service.
Overview
The article begins with a review of how a
simple market-clearing economy responds to policy
and technology shocks. Then, under the assumption that people comprehend the long-run implications of such shocks, the short-run responses of

Stephen P. A. Brown, Zsolt Becsi, and Mark A. Wynne
offered valuable comments and suggestions. The views
expressed are not necessarily those of the Federal Reserve
Bank of Dallas or the Federal Reserve System.
1

Examples of intermediate-level texts that rely heavily on the
IS–LM model are Mankiw (1992), Hall and Taylor (1988),
Dornbusch and Fischer (1987), and Gordon (1987).

33

the economy to current and anticipated future
changes in the money supply, government purchases, and technology are determined. The shortrun analysis has two parts. The first part is a
thought experiment in which the dollar price of
output is held fixed at an arbitrary level and output and employment are sales-determined. It is in
this thought experiment that variants of the traditional IS and LM curves play an important role in
determining the level of output. While the LM
curve is fairly standard, the IS curve, because it
reflects the savings decisions of households,
depends heavily on expectations about the future.
The final step in the analysis is to determine what,
in fact, the short-run equilibrium price level will
be. Depending on the speed of price adjustment,
results either are identical to those obtained from
Barro’s real-business-cycle model or are reminiscent of those obtained from Keynesian models.
The essential features of the expectationsaugmented IS –LM approach can be presented in
a setting that abstracts from capital investment. In
models without investment, causality runs entirely
from the future to the present: the current actions
of private decisionmakers depend on expected
future economic conditions, while future economic
conditions are independent of people’s current
actions. This one-way causality considerably simplifies the analysis of policy and technology shocks.
Accordingly, I defer discussion of macroeconomic
models with investment to Part 2 of the article,
which will be published in a subsequent issue of
this Review.
Long-run equilibrium
In real-business-cycle models, the economy is
assumed always to be in a full-information, marketclearing equilibrium. This section reviews how a
typical real-business-cycle economy responds to
technology and policy shocks. Similar, but more
detailed, analyses have been presented by Barro
(1990) and Barro and King (1984). Later in the
article, we allow for the possibility that sluggish
price adjustment or imperfect information may
delay the economy’s response to current shocks.
Accordingly, for our purposes, it is both convenient
and accurate to call the full-information, marketclearing equilibrium “long-run equilibrium.”
Briefly, the analysis of this section shows
34

that an increase in government purchases raises
long-run equilibrium output and employment
while reducing long-run equilibrium consumption
and the real wage. An adverse technology shock
reduces long-run equilibrium output, the long-run
equilibrium real wage, and long-run equilibrium
consumption. Its effect on long-run equilibrium
employment is ambiguous. The real interest rate
is determined by the requirement that aggregate
saving equal zero. Monetary policy determines the
long-run equilibrium price level.
The representative household. The representative household is endowed with a certain quantity
of time, L. The household divides this time between market and nonmarket activities (between
“labor” and “leisure”). Earnings from market activities are used to purchase output from firms (“consumption”). To maximize its total satisfaction, the
household will allocate its time so as to equate the
rate at which it is willing to trade leisure for consumption to the rate at which leisure and consumption trade for one another in the marketplace. Thus,
(1)

MRSl c = w,

where MRSl c denotes the household’s willingness
to exchange leisure for consumption (its marginal
rate of substitution between leisure and consumption) and where w denotes the real wage rate. It is
usual to assume that both leisure and consumption are “normal” goods, meaning that as the household’s wealth increases (holding the real wage
constant), the household chooses more of both.
In equation 1, assuming normality is equivalent to
assuming that the marginal rate of substitution is
a decreasing function of leisure and an increasing
function of consumption.
Graphically, in a plot with leisure on the
horizontal axis and consumption on the vertical
axis, the real wage is the negative of the slope of
the household’s budget line, while the marginal
rate of substitution is the negative of the slope of
the household’s indifference curve map. Equation
1 says that the household selects the point on its
budget line that is tangent to one of its indifference curves. See Figure 1.
The representative firm. The representative firm
hires labor from the representative household and
produces output, which is sold either to households
or to the government. The firm finds it profitable
Federal Reserve Bank of Dallas

Figure 1

Figure 2

The Representative Household

The Representative Firm

The representative household chooses l units
of leisure and c units of consumption.

The representative firm hires n units of labor
and produces y units of output.

Consumption

Output
Slope = w
Slope = –w

y
c

L
l

Leisure

to hire labor up to the point where the output
produced by an additional unit of labor equals the
real wage. Thus,
(2)

n

Labor

against leisure. Figure 2 depicts the optimum of
the representative firm in a plot of output against
labor. Leisure and labor are related to one another
by the equation

MPn = w,
(3)

where MPn denotes the marginal product of labor.
It is usual to assume that labor is subject to the law
of diminishing marginal returns: in equation 2, the
marginal product of labor is a decreasing function
of the hours of work purchased by the firm.
Graphically, equation 2 says that the firm will
operate at that point on its production function
where the slope of its production function equals
the real wage. See Figure 2.2
The government. The government purchases output from firms, financing its spending with lumpsum (that is, nondistortionary) taxes. For simplicity,
changes in government purchases will be assumed
to have no effect on household preferences for
private consumption and leisure and to have no
effect on the production technology. As a practical
matter, these simplifying assumptions mean that
fluctuations in government purchases are probably best interpreted as the counterpart to threatoffsetting changes in real-world military spending.
Equilibrium. Figure 1 depicts the optimum of the
representative household in a plot of consumption
Economic Review — Third Quarter 1993

n = L – l,

where n and l denote hours of work and of leisure,
respectively. Output and consumption are related
to one another by the equation
(4)

y = c + g,

where y, c, and g denote output, consumption,
and government purchases, respectively. Equations
3 and 4 allow one to transfer Figure 2 into the
same space as Figure 1. This transfer is accomplished by taking the mirror image of Figure 2
and shifting the resultant graph downward by the
amount g. See Figure 3. Finally, Figure 4 combines
Figures 1 and 3 to depict the overall long-run
equilibrium of the economy. In the figure, the

2

This and all subsequent figures assume that labor is essential to production.

35

Figure 3

The Leisure–Consumption Opportunity Set
of the Representative Household
Consumption

Slope = –w

c
y

g

x

l

n

Leisure

L

equilibrium levels of leisure and consumption are
l e and c e, respectively, and the equilibrium real
wage is w e.
Comparative statics. An increase in government
purchases reduces the amount of output available
for consumption at any given quantity of leisure.
In Figure 4, the effect of an increase in government
purchases is to shift the leisure–consumption
opportunity locus downward by the amount of
the increase in g. Because leisure and consumption
are normal goods, the representative household
will choose to absorb the impact of the downward
shift in its opportunity locus by cutting back on
both leisure and consumption, rather than on consumption alone. As shown in Figure 5, the new
equilibrium of the economy is below and to the
left of the original equilibrium. Because the new
equilibrium lies to the left of the original equilibrium, it corresponds to a point that is farther out
along the production function. Thus, output is
higher—and the real wage is lower—than before.
The intuition underlying these results is that households, feeling poorer, are more willing to work
than before. The resultant rightward shift in the
labor supply schedule drives down the equilibrium
real wage, making it profitable for firms to increase
hours of work and expand production.
An adverse technology shock, which might
36

be due, for example, to a deterioration in the
weather, can be modeled as a constant-percentage
reduction in the amount of output produced at any
given quantity of labor. In Figure 4, the leisure–
consumption opportunity locus rotates downward,
falling more (in absolute terms) at low levels of
leisure than at high levels of leisure. The representative household is unambiguously worse off than
before, and the reduction in its wealth tends to
induce declines in both leisure and consumption
(much as in Figure 5). On the other hand, the new
opportunity locus is flatter than the old. The flattening of the opportunity locus reduces the marginal
reward for working, so it provides households with
an incentive to substitute leisure for consumption.
The net effect of these wealth and substitution
effects is negative for consumption but ambiguous
for leisure. The decline in consumption is accompanied by an equal decline in output. The real wage
falls, reflecting both households’ increased supply
of labor (due to the wealth effect) and firms’
increased hesitancy to demand labor (due to labor’s
lower marginal productivity). See Figure 6.
The interest rate. In a full-information, marketclearing economy without capital investment, the
real interest rate plays a largely passive role. As we
have seen, the equilibrium levels of consumption,
output, and hours can be found, in each period,
without bringing the interest rate into the analysis

Figure 4

Long-Run Equilibrium
Consumption

Slope = –we

ce

L
g

e

le

Leisure

Federal Reserve Bank of Dallas

Figure 5

Figure 6

Effects of Increased Government Purchases

Effects of an Adverse Technology Shock

An increase in government purchases lowers
equilibrium leisure and consumption.

An adverse technology shock lowers equilibrium
consumption and has an ambiguous effect on
equilibrium leisure.

Consumption
Consumption

ce
ce

ce′

c e′

L
le′ le

Leisure

L
le ′ ⭵ l e

Leisure

∆ge

at all. It will, nevertheless, be useful later to have
an expression for the equilibrium real return on
bonds.3 We can obtain such an expression from the
representative household’s optimality conditions.
Although aggregate saving must be zero in
equilibrium, each individual household feels free
to borrow and lend. The representative household
will want to adjust its borrowing and lending until
the rate at which it is willing to trade current
consumption for future consumption matches the
rate at which current consumption trades for
future consumption in the marketplace. Thus,
(5)

MRSc c僓 = r,

where MRSc c僓 denotes the amount of future consumption (c僓 ) that the household requires as compensation for a one-unit reduction in current consumption
and where r denotes the (gross) rate of return
on bonds (equal to 1 plus the real interest rate).
Turning equation 5 around and substituting into it
equilibrium levels of current consumption and
future consumption, each determined as in Figure 4,
yields the real return on bonds that is consistent
with zero desired aggregate saving.
Economic Review — Third Quarter 1993

Because current consumption and future
consumption are normal goods, the marginal rate
of substitution between current consumption and
future consumption is negatively related to current
consumption and positively related to expected
future consumption. Hence, the equilibrium real
rate of return rises in response to shocks that
increase expected future consumption relative to
current consumption. Intuitively, when they expect
the future to be bright in comparison to the present,
households are tempted to borrow against their
future prosperity. In an economy without investment opportunities, the real interest rate must rise
to choke off this incipient borrowing. When they
expect the future to be dark in comparison to the
present, households are tempted to save for the
coming “rainy day.” The real interest rate must fall
until the desire to save is eliminated.

3

Note that in an economy with identical households and no
capital investment, government bonds are the only securities traded in equilibrium.

37

Money and prices. There is no single, generally
accepted way of modeling the demand for money.4
Here, we assume that the representative household
makes trade-offs between real money balances
and consumption in much the same way it makes
trade-offs between leisure and consumption. We
assume, in particular, that the demand for real
money balances is determined by the equation

Assuming that consumption and money
balances are both normal goods, equation 6 implies
that the demand for money is an increasing function of consumption and a decreasing function of
the nominal return on bonds.6
The nominal rate of return on bonds is
related to the real rate of return and inflation by
the identity

(6)

(7)

MRSmc = (R – 1)/R,

where MRSmc denotes the additional current consumption that the household demands in compensation for a one-unit reduction in end-of-currentperiod real money balances and where R denotes
the gross nominal return on bonds. (Thus, R
equals 1 plus the nominal interest rate.) Intuitively,
MRSmc is the rate at which the household is willing to trade money (m) for consumption, while
(R – 1)/R is the opportunity cost of holding money,
measured in units of current consumption.5

4

5

38

Most undergraduate macro textbooks assume that the
demand for money is an ad hoc function of gross income.
Real money balances are also sometimes modeled as an
argument of the production function, as an argument of the
household utility function, or as a constraint on current
household spending (the “cash in advance” approach).
By transferring $1 from cash into bonds, the household
raises its purchasing power in the next period by $(R – 1),
which has a current purchasing power of $(R – 1)/R. See
Barro (1990, 96–98). The Baumol–Tobin money demand
model is obtained in the special case in which the household utility function takes the form u(c,m) = ln(c) + ln[2m/
(2m + γ /P)] = ln(C), where γ /P is the real transaction cost
associated with each exchange of interest-bearing assets
for money and where C ≡ c[ 2m/(2m + γ /P)] is consumption
net of transaction costs.

6

Both consumption and money balances are assumed to be
additively separable from leisure in the household utility
function, so that the marginal rate of substitution between
real balances and consumption is independent of leisure.

7

By assuming that production adjusts to match changes in
sales, I avoid having to deal with the possibility that households might be rationed in the output market. Whether such
rationing is of practical significance is controversial. For an
attempt to analyze an economy in which such rationing
occurs, see Neary and Stiglitz (1983).

R = r π,

where π denotes the ratio of the future price level
to the current price level. Hence, equation 6 can
be rewritten as
(6′)

MRSmc = 1 – P/(rP僓 ),

where P and P僓 denote the current price level and
next period’s price level, respectively.
Suppose that equation 6′ is satisfied in the
current and all future periods. Then a simultaneous
doubling of the current and all future nominal
money supplies and of the current and all future
price levels leaves equation 6′ unaltered. In general,
the only effect of a once-and-for-all surprise change
in the level of the nominal money supply is to
cause a proportionate change in the price level.
The equilibrium values of real variables are entirely
unaffected.
A short-run thought experiment
There is considerable debate among economists about whether prices actually adjust sufficiently, in the short run, to keep the economy in
full-information, market-clearing equilibrium. It is
worthwhile, therefore, to adopt an analytical framework that allows price adjustment to be less than
immediate. Even if one is personally convinced that
market imperfections are of negligible importance,
a framework that does not impose market clearing
has the advantage of keeping channels of communication open to those holding contrary views.
Accordingly, this section attempts to answer
a “What if…” question: What would happen to output and interest rates, in response to policy and
technology shocks, if the price level were to remain
fixed in the short run, with output and employment adjusting to match the level of sales? 7 The
twist on traditional IS –LM analysis here is that
Federal Reserve Bank of Dallas

people’s expectations of future economic conditions
are acknowledged to be an important determinant of their current behavior, and people are
assumed to comprehend fully the implications of
each shock for the future course of the economy.
For example, people recognize that a sustained
increase in government purchases will eventually
reduce the amount of output available for private
consumption. Consequently, the announcement
of a defense buildup may have an adverse impact
on current household demand and, hence, on
current output and employment.
We will assume that the short run in our
thought experiment—the interval over which the
price level is held fixed and output and employment are sales-determined—lasts only one period.8
Next period, all markets are expected to clear,
with equilibrium determined as in Figure 4.
The IS and LM curves. We adopt the standard
Keynesian assumption that the markets for money
and bonds must continue to clear, even if the
markets for output and employment do not. The
requirement that the demand for money equal the
supply of money yields the LM schedule. The
requirement that the supply of bonds equal the
demand for bonds — or, equivalently, that investment equal savings —yields the IS schedule.
Here, the demand for money is determined
by equation 6′. Because the current price level is
held fixed in our thought experiment, the monetary
authority controls the short-run real money supply
through its choice of the short-run nominal money
supply. Thus, for a given short-run nominal money
supply and long-run price level target, equation 6′
defines an upward-sloping LM schedule.9 The LM
schedule shifts to the right in response to increases
in the short-run nominal money supply and in
response to increases in the monetary authority’s
perceived long-run price level target. The only
unconventional feature of the LM schedule is that
it is a relationship between consumption and the
real rate of return on bonds rather than between
income and the real rate of return on bonds.
The condition that investment equals savings,
in the current model, reduces to the requirement
that equation 5 be satisfied. Recall that the marginal
rate of substitution between current consumption
and future consumption, which appears on the lefthand side of equation 5, is a decreasing function
of current consumption and an increasing function
Economic Review — Third Quarter 1993

of expected future consumption. Consequently,
for any given level of expected future consumption, equation 5 defines a negative relationship
between current consumption and the real return
on bonds. It is this negative relationship that takes
the place of the traditional IS curve.10
Unlike the traditional IS curve, the IS curve
defined by equation 5 depends explicitly on expectations of the future. Intuitively, households want
to smooth consumption through time. If expected
future consumption rises, households desire more
consumption today as well. Thus, with current consumption plotted on the horizontal axis and the
real return on bonds on the vertical axis, increases
in expected future consumption shift the expectations-augmented IS schedule to the right. Indeed,
when MRSc c僓 depends only on the ratio of future
consumption to current consumption (that is, when
household preferences are homothetic), the rightward shift in the augmented IS schedule is exactly
proportionate to the increase in expected future
consumption. The stronger is the desire to smooth
consumption, the steeper is the IS curve.
It is important to note that the optimality
conditions from which the expectations-augmented
IS and LM curves are derived are not ad hoc additions to the full-information, market-clearing

8

See Koenig (1987) for an analysis of the case in which the
short run lasts several periods.

9

I have chosen to assume that the monetary authority targets
the long-run price level because critics of Keynesian economic models have often advocated just such a policy.
(See, for example, Barro 1986.) One could assume equally
well that the monetary authority holds the long-run money
supply fixed, that the monetary authority fixes money growth
between the two periods, or that the monetary authority
targets the rate of inflation between the two periods. Results
differ, somewhat, depending on the long-run policy rule
adopted (though the basic methodology outlined here
carries through).

10

As defined here, the expectations-augmented IS curve is a
relationship between consumption and the interest rate,
rather than between income and the interest rate. Given the
manner in which I have chosen to model the demand for
money, plotting the expectations-augmented IS and LM
curves in consumption × interest rate space seems natural.
For further discussion, see the Appendix.

39

economy we analyzed earlier. In that economy,
equations 5 and 6′ played critical roles in determining the real return on bonds and the equilibrium price path. Now, with the current-period
price level held exogenously fixed, these same
equations determine the current-period levels of
consumption, output, and employment.
Comparative statics. Figure 7 plots the expectations-augmented IS and LM curves defined by
equations 5 and 6′ and illustrates the effects of an
increase in the current-period money supply. As
in traditional IS –LM analysis, consumption (and,
so, output) rises and the real interest rate falls,
eliminating what would otherwise be an excess
supply of money.11 Similar results are obtained if
it becomes known that the monetary authority has
adopted a higher long-run price level target.
An increase in current-period government
purchases has no impact on either the IS or the LM
curve and, hence, has no impact on consumption
or interest rates. With consumption unchanged,
output must rise by the full amount of the increase
in government purchases. These results will seem
strange to those used to textbook Keynesian
analysis, and they merit explanation.
First, given that interest rates fail to rise, why
is there no multiplier effect in the model developed
here? That is, why does consumption remain constant, rather than increase, as aggregate income
expands? In the standard textbook IS –LM model,
households are assumed to ignore the future tax

40

11

Figure 7 assumes that household utility is additively separable between consumption and money balances. Though
not essential to the analysis, this assumption is convenient
and will be retained throughout the remainder of the article.
For empirical evidence on the separability question, see
Koenig (1990). For an analysis of the nonseparable case,
see Koenig (1989).

12

That the representative household cares only about the
present discounted value of tax payments follows from
equation 5.

13

See Mankiw and Summers (1986) for an analysis of fiscal
policy in a model in which the demand for money is a
function of consumption, rather than income, but households are myopic.

Figure 7

Impact of an Increased Money Supply

In the short run, for a given price level, an
increase in the money supply raises consumption and lowers the real return on bonds.
Real return
LM

LM′
r

r′

IS

c

c′

Consumption

liabilities implied by an increase in government purchases. Consistent with most of the real-businesscycle literature, here we have implicitly gone to the
opposite extreme and assumed that people are
fully cognizant of the tax implications of changes
in government spending. The absence of a multiplier effect is exactly what one would expect in a
model in which the timing of (lump-sum) taxes is
irrelevant, so that any change in government purchases might just as well be financed through an
increase in current taxes.12 After all, when a balancedbudget constraint is imposed on the textbook
model, multiplier effects disappear from it too.
Second, interest rates fail to rise in response
to increased government purchases because the
demand for money is a function of consumption
rather than income.13 If the more conventional
textbook specification of money demand were
adopted, the current model would yield the standard result that increases in government purchases
tend to raise interest rates. (See the Appendix.)
Consumption would then tend to fall somewhat
(though by less than the increase in government
purchases) rather than remain constant.
Summarizing, the response of the current
model to monetary policy is entirely conventional.
The response of the current model to near-term
Federal Reserve Bank of Dallas

changes in fiscal policy would be much like that
of a conventional IS –LM model with a balancedbudget constraint, were it not for our assumption
that the demand for money depends on consumption rather than on gross income. This assumption
is not essential to the expectations-augmented
approach to IS –LM analysis.
Expectations-augmented IS –LM analysis,
unlike conventional IS –LM analysis, explicitly
recognizes that prospective fiscal and technology
shocks can have every bit as much near-term
impact on the economy as realized shocks. The
impact of prospective shocks is transmitted to
today’s economy through changes in expected
future consumption, which proxy for changes in
permanent income. For example, we saw (in
Figure 5) that an increase in long-run government
purchases tends to lower long-run consumption.
Thus, the prospect of a defense buildup will lower
expectations of future consumption and, by equation 5, shift today’s IS curve to the left. Today’s
consumption and today’s interest rates, consequently,
fall at any given current price level (Figure 8 ).
Today’s output falls, too, if current government
purchases are unchanged. (See the box titled “The
Short-Run Impact of a Permanent Defense Cut” for
a discussion of what happens if current purchases
and expectations of future purchases change simultaneously.) Effects qualitatively similar to those
displayed in Figure 8 are also observed in response
to a prospective adverse technology shock.
Closing the short-run model
Thus far, our analysis has taken the short-run
price level as given. This assumption is unnecessarily restrictive, and we now take steps to relax
it. Several alternative models of short-run price
determination are considered. At one extreme we
have the real-business-cycle model, which assumes
that the wage rate and price level adjust instantaneously to clear the labor and output markets. At
the other extreme is a model in which output prices
are set in contracts before complete information
on technology and government policies is available. Between these extremes are models in which
the price of output is flexible, but labor contracts
prespecify the wage, and models in which firms
adjust their output in partial ignorance of the prices
prevailing in other markets.
Economic Review — Third Quarter 1993

Figure 8

Impact of an Anticipated Increase
in Future Government Purchases

In the short run, for a given price level, an anticipated increase in future government purchases
lowers consumption and the real return on bonds.
Real return
LM

r

r′

IS
IS′
c′

c

Consumption

As in the preceding section, we find that
the tools of traditional Keynesian analysis can be
adapted to analyze models in which people’s
current behavior depends nontrivially on their
expectations of the future course of the economy.
In particular, we can derive well-defined counterparts to the traditional Keynesian “aggregate
demand” and “aggregate supply” curves. The intersection of these two curves determines the shortrun equilibrium price and quantity of output.
The aggregate demand schedule. As a first step
in the direction of relaxing the fixed-price assumption, consider what happens to the IS –LM intersection as the current price of output declines.
The position of the IS curve depends only on
future consumption, which is independent of the
current price level. For any given current nominal
supply of money, however, a decline in the current
price level raises the real money supply, shifting
the LM curve to the right, as in Figure 7. Given
our assumption that the monetary authority targets
the long-run price level, this rightward LM shift is
not quite the end of the story. A lower current
price level raises expected inflation (or lowers
expected deflation) and, hence, lowers the demand
for money at any given real rate of return on
41

bonds, shifting the LM curve a bit further to the
right (or, more accurately, down).14 Thus, the level
of current consumption determined by the intersection of the IS and LM curves rises as the
current price level falls.
The negative short-run relationship between
consumption and the price level is plotted in
Figure 9 and labeled “AD.” Like the so-called
aggregate demand curve of traditional Keynesian
analysis, the AD schedule represents output–price
combinations (or, in the present model, consumption–price combinations) in which the demand
for money equals the supply of money and, simultaneously, the representative household is content
with the intertemporal allocation of output.
Obviously, any disturbance that shifts the
IS –LM intersection to the right for a given price
level will shift the AD schedule to the right by
exactly the same amount. Thus, an increase in the
current-period nominal money supply, an increase
in the monetary authority’s long-run price target,
a cut in long-run government purchases, and
positive long-run technology shocks will tend to
move the AD schedule to the right. Changes in
current technology and current government purchases, on the other hand, have no effect on the
AD schedule.15
The aggregate supply schedule:
alternative models
The real-business-cycle model. In the real-businesscycle model, prices adjust instantaneously to clear
all markets. In the present context, equations 1
and 2, which define the supply of labor and the
demand for labor, must both be satisfied—even

42

14

Note that this endogenous response of expected inflation to
changes in P implies a more elastic aggregate demand
curve than does an inflation target.

15

If the demand for money were assumed to be a function of
income rather than consumption, the IS, LM, and AD schedules would be more appropriately plotted with income on
the horizontal axis. Increases in current government purchases would shift the IS schedule—and, hence, the AD
schedule also—to the right, much as in a traditional
Keynesian analysis.

Figure 9

Aggregate Demand Curve with Alternative
Aggregate Supply Schedules
Output price

ASRBC
ASII

ASSP

AD

Consumption

in the short run. Thus, short-run equilibrium levels
of consumption and leisure are determined as in
Figure 4.
In Figure 9, the combinations of price and
consumption consistent with the clearing of the
labor market are labeled “ASRBC .” The curve ASRBC
is very much the counterpart of the traditional
Keynesian “aggregate supply schedule,” except
that the curve ASRBC represents the total amount of
output available to the private sector rather than
the total amount of output available to the public
and private sectors combined. That the ASRBC
schedule is vertical reflects the fact that the indifference curves and production function plotted in
Figure 4 are independent of the price of output.
The sticky-price model. The sticky-price model
assumes that the price of output is fixed in advance.
Usually, this approach also assumes that output
adjusts one for one in response to unanticipated
changes in sales. (Presumably, either labor contracts give employers discretion in setting hours of
work or the wage rate adjusts so that employees
are content with whatever hours are required of
them.) Equation 2 may be satisfied ex ante but is
not, in general, satisfied ex post.
In Figure 9, the assumption that output
adjusts one for one in response to changes in
sales at a preset price is reflected in a horizontal
aggregate supply schedule, ASSP .
Federal Reserve Bank of Dallas

The Short-Run Impact of a Permanent Defense Cut
In traditional textbook Keynesian analysis, no distinction is made between permanent changes and temporary changes in
government purchases. Implicitly, the traditional analysis assumes that it is only the
contemporaneous change in government
purchases that affects the short-run equilibrium of the economy. In expectations-augmented IS–LM analysis, in contrast, whether
a change in government purchases is thought
to be temporary or thought to be permanent is
of considerable importance. An analysis of
the impact of a permanent cut in defense
spending illustrates the point.
Recall that changes in government purchases that are expected to be transitory
have no impact whatsoever on the expectations-augmented IS and LM schedules. In our
short-run thought experiment, therefore, consumption and the real return on bonds are not
affected by short-term defense cuts. Output
falls one for one with government purchases:
a $20 billion cut in defense spending results in
a $20 billion decline in gross domestic product. A prospective change in government
purchases, however, shifts the expectationsaugmented IS schedule in the same direction
as the resultant prospective change in longrun consumption. A prospective cut in the
defense budget would, therefore, shift the IS
schedule to the right. Assuming that preferences are homothetic in current and future

The sticky-wage and imperfect-information models.
Sticky-wage and imperfect-information models
yield an aggregate supply schedule with an elasticity
that lies between the elasticities of the real-businesscycle and sticky-price supply schedules. The stickywage model assumes that money wages are set in
advance. If the price of output rises, unexpectedly,
relative to the preset wage, firms find it profitable
to expand their production and hiring (Fischer
Economic Review — Third Quarter 1993

consumption (so that the marginal rate of
substitution between current and future consumption depends only on the ratio of current
to future consumption) and assuming that
current consumption and future consumption
are initially equal, the rightward shift in the IS
curve will exactly match the increase in future
consumption. Because the LM schedule
slopes upward, the actual short-run equilibrium level of consumption in our thought experiment rises by less than the increase in
future consumption, which, in turn, rises by
less than the future cut in government spending. If a $20 billion cut in the defense budget
raises long-run consumption by $15 billion,
then short-run consumption might rise by only
$5 billion.
What, then, is the impact of an immediate, permanent cut in the defense budget?
Consumption rises in the short run but not as
much as it will rise in the long run. The
prospect of a rising consumption path puts
upward near-term pressure on interest rates.
Output falls in the short run, by more than it will
fall in the long run. In our numerical example,
consumption rises by $5 billion in the short run
and by $15 billion in the long run. Output falls
by $15 billion in the short run and by $5 billion
in the long run.
Of course, these results assume that the
monetary authority holds both the long-run
price level and the current money supply fixed.

1977; Taylor 1980). Equation 1 may be satisfied
ex ante but is not, in general, satisfied ex post.
In the imperfect-information model, when a
firm sees the price of its product rise, the firm is
not certain whether this rise reflects an increase in
the price of its product relative to the prices of
other goods or, instead, an increase in the general
level of prices. Because of this confusion, an unexpected increase in the general price level is
43

Figure 10

Impact of Monetary Stimulus

An increase in the current-period money supply
or in the monetary authority’s long-run price level
target will shift the aggregate demand schedule
to the right.
Output price
ASRBC

ASII

E1
E3
E2

ASSP

E

AD′
AD
Consumption

usually accompanied by some increase in each
firm’s output level. Each firm believes that its
behavior is consistent with profit maximization
but discovers, after the fact, that it was mistaken.16
In Figure 9, the sticky-wage and imperfectinformation models yield aggregate supply curves
like that labeled “ASII .”
Comparative statics. Figure 10 illustrates the
price and output (consumption) effects of a variety
of economic shocks. Much as in the traditional
Keynesian model, expansionary monetary policy—
as reflected in either an unexpected increase in
the current money supply or an upward revision
in the monetary authority’s perceived long-run
price level target—shifts the aggregate demand
schedule to the right. In the real-business-cycle

model, the only effect of this shift is to cause an
increase in the current price level.17 (The economy
moves from point E to point E1 .) In the stickyprice model, it is output, rather than the price
level, that increases. (The economy moves from
point E to point E2.) The sticky-wage and imperfect-information models yield increases in both
consumption and the price level. (The economy
moves from point E to a point like E 3.)
As shown in Figure 11, an unexpected increase in current-period government purchases
shifts each aggregate supply curve to the left. (Recall
that “aggregate supply” in the current model refers
to the amount of output available to the private
sector, rather than the amount of output available
to the economy as a whole.) In the real-businesscycle, sticky-wage, and imperfect-information
versions of the model, consumption falls (but by
less than the increase in government spending)
and the price level is driven up. The economy
moves from E to E1 in the real-business-cycle
model and from E to a point like E3 in the stickywage and imperfect-information models. In the
sticky-price model, output rises by the full amount
of the increase in government spending, leaving
consumption and the price level unchanged.
(The economy stays at point E.)

Figure 11

Impact of Increased Government Purchases

An increase in current-period government
purchases shifts the aggregate supply curve
to the left.
Output price
AS′RBC

ASRBC
AS′II

ASII

E1

E3
E
16

17

44

For additional explanation of the imperfect-information
model, see Lucas (1972), Barro (1990, chap. 19), or Mankiw
(1992, chap. 11).
Recall that, for simplicity, we are assuming that real money
balances are additively separable from both consumption
and leisure in the household utility function.

AS′SP = ASSP

AD

Consumption

Federal Reserve Bank of Dallas

An adverse current-period technology shock
has consumption and price effects very like those
associated with an increase in current-period
government purchases.
Prospective changes in government purchases
and technology affect the current-period equilibrium of the economy by altering households’
long-run consumption prospects. The announcement of future defense cuts or the future implementation of improved technology will shift the
current-period aggregate demand schedule to the
right in much the same way as expansionary
monetary policy. In Figure 10, the economy will
move from E to E1, E2, or a point like E3, depending on whether markets clear instantaneously, the
short-run price level is fixed, or firms have difficulty
distinguishing general price level movements from
relative price level movements.
Concluding remarks
The basic idea underlying IS –LM analysis is
that supply and demand in the financial markets
determine the economy’s short-run equilibrium
quantities of labor and output in the event that
the wage rate and price level fail to achieve their
full-information, market-clearing levels. Traditional
Keynesian analysis, in addition, treats household
expectations as either exogenous or irrelevant. In
this article, we have seen that it is possible to
abandon traditional Keynesian myopia without
abandoning the basic IS –LM framework.
Admittedly, the thought experiment that
underlies IS –LM analysis seems artificial in realbusiness-cycle models, where prices adjust instantaneously to clear all markets. Even in real-businesscycle models, however, the equilibrium conditions
used to derive the expectations-augmented IS and
LM curves are indispensable. Thus, the “money
demand equals money supply” condition that
defines the LM curve determines the equilibrium
price path in a real-business-cycle world, while
the intertemporal optimality condition that defines
the expectations-augmented IS curve determines
the real interest rate. In brief, real-business-cycle
models impose instantaneous market clearing.
Expectations-augmented IS –LM analysis is consistent with instantaneous market clearing but allows
for the possibility that price adjustment in the labor
and output markets is less than immediate.
Economic Review — Third Quarter 1993

By analyzing a variety of macroeconomic
models within a common framework, one obtains
insights into how the models relate to one another,
facilitating discussion. A particular advantage of
the IS –LM approach developed here is that, in
using it, one gains some appreciation for which
of the traditional Keynesian results flow from the
assumed myopia of households and firms, which
flow from sluggish wage and price adjustment,
and which flow from special assumptions about
the determinants of the demand for money.
For example, the impact of monetary policy
in the current model is quite traditional, despite
forward-looking expectations and despite our use
of consumption rather than income as the scale
variable in the money demand function. On the
other hand, we found that forward-looking expectations eliminate the short-run multiplier effect
usually associated with an increase in current
government purchases. And whether an increase
in current government purchases puts near-term
upward pressure on interest rates depends critically
on how one models the demand for money.
Finally, the traditional distinction between
demand shocks and supply shocks is blurred when
household consumption demand is forwardlooking, rather than myopic. Thus, the expectation
of a future shift in aggregate supply—the result,
perhaps, of an anticipated change in technology—
affects current aggregate demand.

Postscript. The analysis presented here is incomplete
in that it fails to allow for endogenous changes in
capital investment. This omission is potentially
serious. Fluctuations in investment were given a
prominent place in Keynes’ own account of the
business cycle. Recently, a study by Fama (1992)
has confirmed that fluctuations in investment are
an important source of transitory movements in
real-world aggregate output. Accordingly, Part 2
of this article, to be published in a future issue of
the Economic Review, extends the expectationsaugmented IS –LM framework developed here to an
economy in which investment is endogenous.
45

Appendix
Derivation of the Comparative Statics Results
This Appendix formally derives many of
the comparative statics results presented in
the main text, and it clarifies the relationship
between the model developed in this article
and standard textbook Keynesian models.
The basic model
Suppose, for analytical convenience,
that the representative household’s willingness to trade current consumption for future
consumption and its willingness to trade consumption for money balances depend only on
the ratios of the quantities of the goods in
question, so that, for example, the marginal
rate of substitution between current consumption and future consumption depends only on
the ratio of current consumption to future
consumption.1 Equations 5 and 6′ then imply
(A.1)

c = c φ(r )

and
(A.2)

m = cκ (rP /P ),

respectively, where both φ (•) and κ (•) are
strictly decreasing.
Equation A.1 defines an IS schedule,
and equation A.2 defines an LM schedule.
Differentiating logarithmically,
(A.1′ )

d ln(c ) = d ln(c ) − ⑀φd ln(r )

where ⑀φ ≡ –rφ′/φ > 0 and ⑀κ ≡ –Rκ ′/κ > 0 equal,
in absolute value, the rate-of-return elasticities of consumption demand and money demand, respectively.
Solving for the percentage change in
consumption and the percentage change in
the real return on bonds,
(A.3)

d ln(c ) = {⑀κ d ln(c )
+ ⑀φ [d ln(M ) − d ln(P )]
+ ⑀κ ⑀φ [d ln(P ) − d ln(P )]}
/(⑀κ + ⑀φ )

and
(A.4)

d ln(r ) = {d ln(c ) − [d ln(M ) − d ln(P )]
− ⑀κ [d ln(P ) − d ln(P )]}
/(⑀κ + ⑀φ ).

Note that consumption and the real rate of
return are increasing in expected future consumption. If households are forward-looking,
we know that expected future consumption
will be increasing in expected future productivity and decreasing in expected future government purchases: c僓 = ␺(θ僓,g僓 ), where θ僓 is a
positive technology-shock variable and where
␺1 > 0 and –1 < ␺2 < 0. Increases in both
current real money balances and expected
inflation have a positive effect on current
consumption and a negative effect on the real
rate of return. Both consumption and the real
rate of return are completely independent of

and
1

(A.2′)

d ln(c ) = d ln(M ) − d ln(P )
+ ⑀κ [d ln(r ) + d ln(P )
− d ln(P )],

46

This condition will be satisfied if the household utility function is
additively separable in its arguments and preferences are
homothetic. Additional realism can be obtained—at the expense of some additional complexity—by relaxing the separability conditions.

(Continued on the next page)

Federal Reserve Bank of Dallas

Appendix
Derivation of the Comparative Statics Results — Continued
current government purchases.
Thus far, we have treated the current
price level, P, as fixed. If, at the opposite
extreme, the wage rate and price level adjust
instantaneously to clear the labor and output
markets, current consumption is determined
as in Figure 4—that is, c = ␺ (θ,g). Given c and
c僓, equation A.1′ determines the real rate of
return on bonds. Equation A.2′ determines the
current price level. In general, one might expect
the aggregate supply schedule to be neither
horizontal nor vertical, so that the changes in
consumption predicted by equation A.3 will
be only partially offset by changes in P.
Encompassing traditional IS–LM analysis
By generalizing equations A.1 and A.2,
we can formulate a model that includes traditional Keynesian IS–LM analysis as a special
case. Suppose, in particular, that

(A.5′)

d ln(c ) = [(1 − λ )d ln(c )
− (1 − λ )⑀φd ln(r )
+ λα gd ln(g )] /(1 − λα c ).

(A.6′) d ln(c ) = {d ln(M ) − d ln(P ) − γα gd ln(g )
+ ⑀κ [d ln(r ) + d ln(P ) − d ln(P )]}
/(1 − γ + γα c ).

In deriving these expressions, use has been
made of the fact that d ln( y ) = αcd ln(c ) +
αgd ln(g), where αc and αg are the respective
shares of consumption and government purchases in national income. Alternatively, one
can write

(A.5′′)

d ln(y ) = [α c (1 − λ )d ln(c )
− α c (1 − λ )⑀φd ln(r )
+ α gd ln(g )] /(1 − λα c )

and
(A.5)

c = [c φ(r )]1−λ y λ

(A.6′′)

+ (1 − γ )α gd ln(g )

and
(A.6)

d ln(y ) = {α c [d ln(M ) − d ln(P )]
+ α c⑀κ [d ln(r ) + d ln(P )

m = c 1−γ y γ κ (rP /P ),

where 0 λ < 1 and 0 γ 1. The parameter
λ measures the “excess sensitivity” of consumption to current income.2 The parameter γ
will be positive to the extent that the demand
for money depends on components of income
other than consumption (Mankiw and Summers 1986). Standard textbook Keynesian
analysis assumes that λ is close to 1 and that
γ is equal to 1. Furthermore, expected future
consumption (c僓 ) is held fixed.3
Logarithmic differentiation of A.5 and
A.6 yields IS and LM curves:

Economic Review — Third Quarter 1993

− d ln(P )]}/(1 − γ + γα c ).

2

Campbell and Mankiw (1989) put λ at 0.5, but most empirical
studies suggest that a value like 0.1 is closer to the mark.
Koenig (1990) tests the Campbell–Mankiw specification and
finds it inferior to an alternative model in which all households
are forward-looking but utility is not separable between consumption and money balances.

3

An alternative interpretation of the standard Keynesian model
is that λ is equal to zero, but households base their expectations of future consumption solely on their current incomes.

(Continued on the next page)

47

Appendix
Derivation of the Comparative Statics Results — Continued
In the special case in which γ equals 1, so that
the demand for money depends on income
rather than consumption, the LM equation
simplifies to

and
(A.9) d ln(r ) = {− (1 − λα c )[d ln(M ) − d ln(P )]
− (1 − λα c )⑀κ [d ln(P ) − d ln(P )]
+ α g [1 − (1 − γ )(1 − λ )]d ln(g )

d ln(y ) = d ln(M ) − d ln(P )
+ ⑀κ [d ln(r ) + d ln(P ) − d ln(P )].
The importance of excess sensitivity in
determining the strength of the “multiplier
effect” is clear from equations A.5′ and A.5″.
In equation A.5′, increases in government
purchases have a positive impact on consumption demand (for a given rate of return
on bonds) only insofar as λ, which measures
the excess sensitivity of consumption to current income, is greater than zero. Similarly, it
follows from equation A.5″ that dy /dg = 1/(1 –
λαc ). Thus, changes in government purchases
have a larger than one-for-one impact on the
demand for output only to the extent that λ is
greater than zero.
The IS and LM equations can be solved
for percentage changes in income, consumption, and the real rate of return. The general
solutions follow:

(A.7) d ln(y ) = {α c (1 − λ )⑀φ [d ln(M ) − d ln(P )]
+ α c (1 − λ )⑀φ⑀κ [d ln(P ) − d ln(P )]
+ α g [(1 − γ )(1 − λ )⑀φ + ⑀κ ]d ln(g )

+ (1− γ + γα c )(1 − λ )d ln(c )} /∆,
where

∆ ≡ (1 − λα c )⑀κ + (1 − γ + γα c )(1 − λ )⑀φ > 0.
Standard textbook results are obtained in the
special case where γ = 1, λ > 0, and c僓 is held
fixed (so that d ln(c僓 ) = 0).
Equations A.7 and A.8 imply that dy /dg
= [(1 – γ )(1 – λ )⑀φ + ⑀κ ]/∆ and dc /dg = –αc [γ (1
– λ )⑀φ – λ⑀κ ]/∆, respectively. The importance
of γ in determining the extent to which crowding out reduces the stimulatory effects of
increased government purchases can be seen
by differentiating these multipliers with respect to γ . One obtains
(A.10) ∂ (dy /dg ) /∂γ = ∂ (dc /dg ) /∂γ
= −α c⑀φ (⑀κ + ⑀φ )(1 − λ )2 /∆ 2 < 0.

Not surprisingly, the more sensitive is the
demand for money to changes in government
purchases (the larger is γ ), the more increases
in such purchases tend to crowd out private
spending.

+ α c (1 − λ )⑀κ d ln(c )} /∆,
(A.8) d ln(c ) = {(1 − λ )⑀φ [d ln(M ) − d ln(P )]
+ (1 − λ )⑀φ⑀κ [d ln(P ) − d ln(P )]
− α g [γ (1 − λ )⑀φ − λ⑀κ ]d ln(g )
+ (1 − λ )⑀κ d ln(c )} /∆,

48

Federal Reserve Bank of Dallas

References
Barro, Robert J. (1990), Macroeconomics, 3d ed.
(New York: John Wiley and Sons).
——— (1986), “Recent Developments in the
Theory of Rules Versus Discretion,” Economic
Journal 96 (Supplement): 23–37.
———, and Robert G. King (1984), “Time-Separable
Preferences and Intertemporal-Substitution
Models of Business Cycles,” Quarterly Journal
of Economics 99 (November): 817–39.
Campbell, John Y., and N. Gregory Mankiw (1989),
“Consumption, Income, and Interest Rates:
Reinterpreting the Time Series Evidence,” in
NBER Macroeconomics Annual 1989, ed.
Olivier Jean Blanchard and Stanley Fischer
(Cambridge, Mass.: MIT Press), 185 –216.
Dornbusch, Rudiger, and Stanley Fischer (1987),
Macroeconomics, 4th ed. (New York: McGraw–
Hill Book Company).
Fama, Eugene F. (1992), “Transitory Variation in
Investment and Output,” Journal of Monetary
Economics 30 (December): 467– 80.
Fischer, Stanley (1977), “Long-Term Contracts,
Rational Expectations, and the Optimal Money
Supply Rule,” Journal of Political Economy 85
(February): 191–205.
Gordon, Robert J. (1987), Macroeconomics, 4th
ed. (Boston: Little, Brown and Company).
Hall, Robert E., and John B. Taylor (1988), Macroeconomics: Theory, Performance, and Policy,
2d ed. (New York: W.W. Norton and Company).

Koenig, Evan F. (1990), “Real Money Balances
and the Timing of Consumption: An Empirical
Investigation,” Quarterly Journal of Economics
105 (May): 399 – 425.
——— (1989), “A Simple Optimizing Alternative
to Traditional IS –LM Analysis” (Federal Reserve Bank of Dallas, May, Typescript).
——— (1987), “A Dynamic Optimizing Alternative
to Traditional IS –LM Analysis,” University of
Washington, Institute for Economic Research,
Discussion Paper Series, no. 87-07 (Seattle,
May).
Lucas, Robert E., Jr. (1972), “Expectations and the
Neutrality of Money,” Journal of Economic
Theory 4 (April): 103 –24.
Mankiw, N. Gregory (1992), Macroeconomics
(New York: Worth Publishers).
———, and Lawrence H. Summers (1986), “Money
Demand and the Effects of Fiscal Policies,”
Journal of Money, Credit, and Banking 18
(November): 415 –29.
Neary, J. Peter, and Joseph E. Stiglitz (1983),
“Toward a Reconstruction of Keynesian Economics: Expectations and Constrained Equilibria,” Quarterly Journal of Economics 98
(Supplement): 199 –228.
Taylor, John B. (1980), “Aggregate Dynamics and
Staggered Contracts,” Journal of Political
Economy 88 (February): 1–23.

Hicks, J.R. (1937), “Mr. Keynes and the ‘Classics’;
a Suggested Interpretation,” Econometrica 5
(April): 147–59.

Economic Review — Third Quarter 1993

49

50

Federal Reserve Bank of Dallas

Zsolt Becsi
Economist
Federal Reserve Bank of Dallas

The Long (and Short) on Taxation
and Expenditure Policies

O

ne of the central issues in the 1992 presidential campaign was how best to promote
economic growth. Because much of the growth
debate concerned fiscal policy, taxation and
expenditure plans came under intense public
scrutiny. At issue were both the level of taxation
and the proper mix of taxes. Similarly, voters
were concerned with the composition as well as
the level of government expenditures. While voter
interest was high, the various programs put forth
grew so detailed that their long- and short-run
effects became difficult to evaluate and compare.
What distinguished the 1992 campaign was a
fiscally sober post–Cold War reassessment of the
government’s economic priorities. All major candidates argued for cuts in defense spending and
agreed that the resources saved—the peace dividend—should be spent on enhancing the nation’s
productivity. This productivity enhancement was
to come from some combination of public investment and a more investment-friendly business tax
structure. While the candidates’ broad visions
were similar, they disagreed on how much public
investment to allocate to human capital (such as
education and training) versus physical capital
(such as roads, bridges, mass transit, and so on).
Proposals also differed on how to change business taxation to promote investment, although all
called for lower costs of private capital. Most
candidates did not openly acknowledge that promoting growth usually entails a current sacrifice
for future public and private consumption.
This article presents an analytical and graphical
framework for evaluating the long- and short-run
effects of a broad range of fiscal policies. Except
for two simplifying assumptions on the structure
of preferences and the production process, the
Economic Review — Third Quarter 1993

model is fairly general. The model is well-suited
for insights into the dynamic effects of some of
the 1992 fiscal policy proposals, and it can easily be
expanded to analyze distributional, educational,
and industrial policy questions.1 To set the stage,
I focus first on the effects of changes in factor
income taxation. Factor income taxes are the main
components of an income tax. Factor income
taxes also have a simple connection to most tax
proposals, and this article shows how they relate to
consumption and corporate taxes. Lastly, to frame
the debate on what to do with the peace dividend, I analyze the effects of changes in government defense and investment expenditures.
Description of the model
This section presents a simple growth model
that can be used to analyze the macroeconomic
effects of alternative fiscal policies. The model
consists of three sectors. The household sector
determines current and planned future levels of

I wish to thank my reviewer, Evan Koenig, for helpful
suggestions and my readers, Steve Brown, Ping Wang, and
Mark Wynne, for their comments. Of course, any remaining
errors are my own.
1

The analytical model used is a variant on Aschauer (1988,
1989) and Barro (1989). The graphical exposition is based
on and complements that of Wynne (1991). This framework
is extended by Becsi (1991) to deal with heterogeneity and
distributional concerns. For extensions to an endogenous
growth framework with education and industrial policies,
see, for example, Barro and Sala-I-Martin (1992).

51

consumption, labor, and savings. These plans are
optimal in the sense that households’ choices
maximize lifetime utility subject to after-tax budget
constraints. In the production sector, firms maximize after-tax profits. This is accomplished by
choosing optimal paths for output and for capital
and labor inputs. In the government sector, tax
receipts from various sources are used to finance
government consumption and investment. Equilibrium occurs when factor and goods markets clear
in every period.2
The household sector is represented by an
average household that values the amount of consumption, c, and leisure it obtains in each period
of its life.3 Individuals have a certain number of
hours, H, per year to allocate to leisure and labor.
Let h, where 0 h H, be the number of hours
devoted to market labor. Thus, H – h is the amount
of time devoted to leisure. For simplicity, preferences between any two time periods are described
by U (c, H – h) + (1 + ρ)–1U (c+1, H – h+1 ).4,5 The

52

2

For simplicity, the model abstracts from money and uncertainty. It also does not consider market imperfections and
intergenerational issues.

3

The representative household is assumed to live infinitely
long. An infinite lifetime can be viewed as dynastic families
that care about the welfare of future generations. Or, it can
be viewed as a useful abstraction of long lives. Time begins
in period one, at which point the individual is endowed with
k1 units of capital.

4

From time to time, it will be convenient to assume, additionally, that utility is separable between consumption and
labor, so that U(c, H – h) = u(c) + v( H – h).

5

Alternatively, one could easily expand utility to include
composite consumption, where composite consumption is
defined as private consumption plus the consumption services derived from public spending. The services of such
spending as health care, education, food stamps, and transportation enter individual utility as substitutes for private
consumption. Similarly, services from some government
expenditures may substitute or complement private inputs
into production. Thus, the only difference between public
consumption and public capital is that the latter takes time
to be productive and depreciates over time. I abstract from
these considerations by assuming that the consumption
services from public spending enter utility separably, and
that only public investment has productive services. However, in several footnotes below, extensions are considered.

pure rate of time preference that discounts future
utility is given by ρ > 0. An increase in this parameter reflects an increase in the individual’s desire
for early gratification.
The representative individual chooses those
feasible time streams of consumption and labor
that maximize lifetime utility. Feasibility is determined by the individual’s period-by-period budget
constraint. The budget constraint requires that
purchases of consumption goods and purchases
of assets (which are held until the next period)
not exceed current period after-tax income. Aftertax income is defined as the sum of after-tax labor
income, after-tax income from assets, and lumpsum transfers. Savings are put into interest-earning
productive capital. The budget constraint for each
period is summarized by
(1)

c + (k +1 − k ) = (1 − tw )wh + (1 − t r )rk + l ,

where l僓 is the lump-sum transfer (or tax), k is the
physical capital accumulated up to the current
period, and k+1 – k is the net purchase of capital.
The pretax real wage and real interest rate are given
by w and r, while the ti (i = w, r) are the tax rates
on wage and interest income. For example, the
wage tax encompasses payroll taxes for social
security and the salary component of personal
income taxes. The interest income tax is a tax on
the real returns to capital including dividends,
capital gains, and so on.
Households choose feasible streams of consumption, labor, and savings that maximize utility.
This leads to well-known optimality conditions for
constrained utility maximization: the marginal rate
of substitution (MRS )—which equals the rate at
which the household is just willing to trade one
good for another—is equated to the price ratio of
the two goods. The price ratio is the rate at which
the two goods can be substituted and still satisfy
the budget constraint.
Within a time period, households adjust private consumption and labor until the MRS between
consumption and leisure is equal to the ratio of
the after-tax wage to the price of consumption
goods, that is, is equal to the after-tax real wage:
(2)

MRS (c, H − h ) = (1 − tw )w .

The MRS between consumption and leisure tells
Federal Reserve Bank of Dallas

how much additional consumption is required to
compensate for a reduction in leisure (an increase
in labor). Since a reduction in leisure lowers utility,
consumption must rise to increase utility to its
original level. However, as leisure falls, a unit of
leisure becomes more valuable to individuals, so
that progressively more consumption is required
to compensate for a unit loss of leisure. In other
words, the additional consumption required as
compensation for lost leisure rises as leisure falls.
Thus, the MRS is negatively related to leisure
(positively related to labor) and, by the same logic,
positively related to consumption.
From the budget constraint, increasing labor
by one hour of work increases the household’s
take-home pay by (1 – tw )w units. This allows
consumption purchases to rise by (1 – tw )w , which
is the ratio of the price of leisure to the price of
consumption. If the MRS is smaller than this price
ratio, consumers require less consumption to make
up for the disutility of working than they actually
can get. Thus, households find it desirable to work
more, because utility rises when work effort (and
consumption) are increased. Since the MRS is
positively related to consumption and labor, as
households increase their labor and consumption
the MRS rises until condition 2 is satisfied.
When hours of labor are plotted on the horizontal axis and private consumption on the vertical
axis, one can trace the trade-offs between consumption and labor for a given level of utility
(Figure 1). These indifference curves are convex to
the origin and curve upward because an increase
in labor requires an increase in consumption to
keep utility constant. The slope of the indifference
curve is the MRS and increases with labor and
consumption. Higher indifference curves represent
higher levels of utility. The budget constraint also
slopes upward and has as its slope the after-tax
real wage rate. The vertical intercept of the budget
line is nonlabor household income such as capital
income and transfers.
Household plans for consumption and labor
are determined by the tangency of household
indifference curves and budget constraints at
point I in Figure 1. At point A the MRS is below
the after-tax real wage rate. Since the slope of the
indifference curve is less than the slope of the
budget line, the household can increase its utility
while staying on its budget line. The household
Economic Review — Third Quarter 1993

Figure 1

Optimal Consumption–Labor Combination
c

I
(l –tw)w
I′
A

h

moves to a higher indifference curve by substituting leisure for consumption or increasing labor
and consumption. Consumption and leisure are
assumed to be normal goods: a good is said to
be “normal” when consumption of the good
increases for a parallel upward shift of the budget
line. Reducing the after-tax wage rate is equivalent to a flattening of the budget line. When consumption and leisure are normal goods, this will
move the individual to point I′ where consumption and labor are lower.
Households adjust consumption and savings
across time until the MRS between consumption
in adjacent periods equals the price of current
consumption in terms of future consumption:
(3 )

MRS (c +1, c ) = 1 + r+1 (1 − t r ).

The MRS tells how much next period’s consumption must rise to compensate for the fall in lifetime
utility that occurs when current consumption is
reduced. When current consumption is low relative
to future consumption, its value is relatively high
for the individual. Thus, the compensation required
for a fall in current consumption rises as current
consumption is reduced. In turn, progressively
more future consumption is required to compensate for a unit loss of current consumption. Thus,
the MRS is negatively related to current consumption. Similarly, it is positively related to future
53

consumption.6 An impatient household has a high
rate of time preference, ρ. This means that an
impatient household requires a higher return of
future consumption for a sacrifice of current consumption than a patient household. Thus, the MRS
will tend to be higher the more impatient the individual is.
From the budget constraint, decreasing current consumption by one unit allows the household to increase savings by one unit. In turn, this
increased saving allows future consumption to
rise by [1 + r+1 (1 – tr )]. Thus, the after-tax interest
rate affects how much additional future consumption one can have for a unit reduction of current
consumption. As long as the MRS exceeds this
relative price, the individual requires more future
consumption to keep utility constant for a sacrifice
of current consumption than the budget constraints
allow. Thus, households will have an incentive to
raise current consumption relative to future consumption. As current consumption rises relative to
future consumption, the MRS falls until equality in
equation 3 is reestablished.

6

When utility is separable in consumption and leisure, the
intertemporal MRS has the following form:
MRS (c +1,c) =

u ′(c)
1
1+ ρ

.

u ′(c +1)

In steady state, consumption is constant across time, so
that the MRS equals 1 + ρ.

54

7

Constant returns to scale means that if all inputs are scaled
up by the same proportion, output will rise by the same
scaling factor.

8

To simplify the analysis, public capital is assumed to enter
production separably. Thus, public capital does not raise
(or lower) the marginal product of a private input. Empirical
evidence suggests that this is an oversimplification. For
instance, Lynde and Richmond (1992) estimate that a
constant-returns-to-scale production function with a positive marginal product of capital is plausible. However, they
find that public capital raises the marginal product of private
capital and lowers the marginal product of labor. Their
evidence on the complementarity in production of private
and public capital is consistent with previous findings.

9

Equation 6 is inoperative in the short run when the capital
stock is fixed.

Given the stock of public capital, k僓, the
representative firm chooses two inputs, labor and
private capital, to maximize its after-tax profit
from selling its final output, y. The firm’s profits
are given by
(4 )

y − wh − (r + δ )k − (k +1 − k ),

where δ is the physical rate of depreciation of
capital, and (r + δ ) is the cost of capital. I assume
that the output production function is constant
returns to scale in all inputs and given by y = f (k,h)
+ g (k僓 ).7 In other words, total final output is the
sum of output produced by private inputs and
output produced by public inputs.8
Profit maximization by the firm implies that
the firm adjusts private inputs until their marginal
products equal their factor costs:9
(5)

fh (k , h ) = w , and

(6 )

fk (k , h ) = r + δ .

If the marginal product of labor is greater than the
cost of labor, an additional hour of labor will add
more to revenues than to costs. Thus, firms can
increase profits by hiring more labor. As labor is
increased, each additional unit of labor becomes
less productive. The marginal product of labor
falls until equality in equation 5 is reestablished.
Similarly, if the cost of capital is greater than the
marginal product, firms will cut back on capital to
raise the marginal product of capital.
For a given stock of private capital, the production sector’s plans for output and labor are
determined by the point on the firm’s production
function where the slope—the marginal product
of labor—equals the ratio of the after-tax wage
cost to the after-tax output price. Increasing labor
increases output at a decreasing rate so that the
production function is concave to the origin. In
other words, the slope decreases as labor is
increased. Point F in Figure 2 gives the profit
maximizing labor–output combination for a given
stock of capital.
Increasing public capital causes a parallel
upward shift in the production function, and the
firm’s optimal combination of labor and output
moves from F to F′. An increase in private capital
causes the production function to twist upward.
Federal Reserve Bank of Dallas

This causes the firm’s labor and output to move
from F to F″. Also, an increase in wage costs
increases the slope of the tangency line and causes
point F to move down the production function.
The public sector purchases consumption
and investment goods. It finances its expenditures
and lump-sum transfers with tax revenues. For
simplicity, in the model the government’s budget is
balanced in each period. In this case, the revenue
constraint of the government is described by
(7 )

Optimal Output –Labor Combination
y

F″
F′

f(k,h)+g(k僓 )

F

d + [(k +1 − k ) + δk ] + l = twwh + t r rk ,

where d僓 denotes defense expenditures. This is a
comprehensive revenue constraint that aggregates
federal, state, and local levels of the government.10
Finally, the goods, factor, and asset markets
are assumed to clear in all periods. In particular,
equilibrium in the goods markets is
(8 )

Figure 2

c + (k +1 − k ) + δk + d + (k +1 − k )
+ δk = f (k , h ) + g (k ).

A dynamic equilibrium occurs when all markets
clear. Also, households and firms must be behaving optimally subject to their feasibility constraints
and the government’s actions.11
Graphing the model
In this model, the short run is defined as the
amount of time it takes to adjust the capital stock.
The short-run equilibrium can be described by
equations 2, 5, and 8 and by the fact that the
capital stock is constant. To study the long-run
effects of fiscal policies, one needs the steadystate version of equations 2, 3, 5, 6, and 8. In a
steady state, all variables are constant through
time. Thus, time subscripts may be dropped. In
particular, this means that the net increments to
capital are zero. The only investment is replacement investment to offset physical depreciation.
The optimality conditions can be jointly
analyzed by combining Figures 1 and 2. Subtracting private investment and public spending from
output gives the amount of output available for
private consumption. This is equivalent to a
parallel downward shift of the production function and causes points F and I to coincide. Where
the two points coincide is depicted as point O in
Economic Review — Third Quarter 1993

g(k僓 )
h

Figure 3. At point O, the (downward-shifted)
production function and the indifference curve
intersect at their points of tangency with their
respective budget lines. Thus, point O determines
the profit and utility maximizing aggregate consumption and labor levels.
Point O is optimal for individual households
and firms. However, it is suboptimal for the
economy as a whole as long as the slope of the
indifference curve does not equal the slope of the
production function. If firms increased labor by
one unit, the additional output produced would
increase utility for the household sector. However,
the tax structure makes this move unprofitable for

10

As a point of reference, defense expenditures averaged
18.4 percent of total government expenditures, transfers to
the private sector were 35.8 percent, and gross public
investment averaged around 6.8 percent for the period
1986–90. (See Akhtar and Harris [1992] and Council of
Economic Advisers [1992]. Also, see footnote 15.)

11

A perfect foresight equilibrium is defined as sequences of
optimal household consumption, labor, and savings plans
and sequences of optimal firm plans of output and inputs
that perfectly forecast the time path of all prices and
government variables. These optimal plans also clear product and factor markets.

55

the firms. In fact, the difference between the intercepts of the two tangency lines is a measure of
the aggregate distortion from the tax system. This
distortion is termed the aggregate tax wedge.
Figure 3 can be augmented to show the long
and short-run equilibrium levels of consumption
and labor. First, market equilibrium is given by
equation 8. In steady state, this equation reduces to
(9 )

⎡
c + d + δk = h ⎢ f
⎢⎣

Figure 3

Combining Intratemporal Optima
c,y

f(k,h)

⎛k ⎞
k⎤
⎜ , 1⎟ − δ ⎥ + g (k ).
h ⎥⎦
⎝h ⎠

O

Aggregate
tax wedge

Also, combining household and firm optimality
conditions and imposing steady state yields

⎛k ⎞
MRS (c, H − h ) = (1 − tw ) fh ⎜ , 1⎟ .
⎝h ⎠

(10 )

Since consumption in steady state is constant across time, the MRS between two adjacent
consumptions only depends on the individual’s
impatience for early consumption:
⎤
⎡ ⎛k ⎞
ρ = (1 − t r )⎢ fk ⎜ , 1⎟ − δ ⎥ .
⎥⎦
⎢⎣ ⎝ h ⎠

(11)

Since the rate of time preference is the required
rate of return to compensate for the individual’s
impatience, the MRS in steady state equals the
constant ρ. Equation 11 determines the marginal
product of private capital, and it also pegs the private capital–labor ratio to the rate of time preference.12 Raising the tax rate on interest income
reduces the after-tax marginal product of private
capital below its long-run equilibrium level. To
restore it to its long-run level, the steady-state
marginal product of capital must rise, and the
capital–labor ratio must, in turn, fall.
The market equilibrium condition, equation
9, determines the long-run market equilibrium

g(k僓 )–(kt 1–k)–δk
–d僓–(k僓t 1–k僓)–δk僓

h

relationship of consumption and labor for a given
capital–labor ratio and for a given level of government expenditures. In the long run, equilibrium
consumption is positively related to equilibrium
employment. This is graphed as the line MML in
Figure 4. Consumption and labor are linearly
related because the capital–labor ratio is fixed.
When the capital–labor ratio increases, labor is
more productive at all levels of employment. This
causes line MML to twist upward from its intercept.
As will be discussed below, changing government
expenditures causes a parallel shift of line MML.

Figure 4

Short- and Long-Run Equilibria
c
OOS
MML

OOL
MMS

c*
12

56

This is because of the homogeneity properties of the production function and because public capital enters separably. Thus, the capital–labor ratio is not affected by wage
taxation and government consumption and investment.
See footnote 20 on how the analysis changes when public
capital is not separable in production.

h*

H

h

Federal Reserve Bank of Dallas

In the short run, the capital stock is fixed
and the market equilibrium condition is given by
equation 8. The short-run equilibrium relationship
between consumption and labor is represented by
line MMS. This line is just the parallel-shifted
production function from Figure 3. Note that as
labor increases beyond h*, output will increase
more in the long run than in the short run. This is
because an increase in labor lowers the capital–
labor ratio in the short run, which has a partially
offsetting effect on output. In the long run, this
partial offset does not occur because capital and
labor move together.
Equation 10 determines the aggregate tradeoff between private consumption and labor. One
can use this equation to trace all intersections of the
indifference and production functions in Figure 3
that are compatible with utility and profit maximization. In other words, one can trace all possible
points O in Figure 3 for parallel shifts in the tangency lines for the production function and the
indifference curves. Given the private capital–
labor ratio, these points constitute the line OOL in
Figure 4. Line OOL gives all the desired steadystate combinations of consumption and labor for
a given wage rate. In essence, line OOL traces
how consumption and labor respond to changes
in wealth, holding relative after-tax prices constant.13 For a given tax system, there are an infinite
number of similar lines associated with different
capital–labor ratios (or wage rates). The paths
lying above and to the right of OOL are associated
with higher after-tax wages or a higher capital–
labor ratio. For a constant capital–labor ratio, a
lower output tax or a lower consumption tax also
causes OOL to shift up and to the right. Thus,
shifts of line OOL represent substitution effects
on labor and consumption.
When capital rather than the capital–labor
ratio is held fixed, equation 10 gives the desired
short-run combinations of labor and consumption.
This is graphed as OOS in Figure 4. OOS is steeper
than OOL, because as labor is reduced from h*
the capital–labor ratio and, hence, the wage rate,
rises in the short-run. Households, therefore,
require a larger compensation in terms of current
consumption than in the long run when the wage
rate is fixed.
In sum, the OO lines give desired combinations of labor and consumption, while the MM
Economic Review — Third Quarter 1993

Figure 5

Effects of a Higher Wage Tax
c
OOS
MML
OOL

MMS

E′

E

E″
H

h

lines represents technologically feasible combinations. The intersection of the curves yields the
overall equilibrium for the economy (in the short
and long run).

The effects of tax policies14
What happens when the government raises
wage taxes? A permanent increase in wage taxes
is the analytical counterpart to increasing payroll
taxes. Wage taxes do not affect the long-run market
equilibrium relationship. Thus, line MML is unchanged in Figure 5. Since the after-tax interest
rate is pegged to the constant rate of time preference in the long-run, a wage tax does not alter
the steady-state capital–labor ratio.

13

When preferences are homothetic, scaling consumption
and leisure by the same scaling factor will leave the MRS
unchanged. This implies that the MRS is constant along
OOL and that OOL is a straight line.

14

To isolate the effect of each fiscal policy instrument, I
assume that the government uses lump-sum transfers to
balance its budget when tax rates are increased. For the
same reason, lump-sum taxes are used to finance increases in government expenditures.

57

Since the capital–labor ratio does not change,
the wage rate before taxes is unaffected. But since
the after-tax wage rate received by households
falls, households substitute away from work and
consumption towards leisure. This is equivalent to
a downward shift of OOL in Figure 5. Since OOL
shifts down and the intersection of OOL and MML
determines the long-run effect of the wage tax on
consumption and labor, the long-run equilibrium
moves from E to E″. Thus, consumption and labor
fall in the long run. Because the capital–labor ratio
is unchanged in the long run, raising the wage tax
causes the capital stock to decline proportionately
to the fall in labor. In turn, output will fall in the
long run.
In the short run, capital is fixed, and line
OOS shifts to the left. As households substitute
away from labor, the short-run capital–labor ratio
rises. This causes the wage rate before taxes to rise.
Thus, the short-run fall in the after-tax wage rate
is less than the long-run fall. The increase in the
short-run capital–labor ratio also affects the market
equilibrium line MMS given by equation 8. Since
the capital–labor ratio is unchanged in the long
run, investment must fall over time to return the
capital–labor ratio to its original level. A reduction
in investment tends to offset the necessary reduction in consumption, given that labor and output
fall. For any level of labor (and output), a reduction of investment means that there is more output
available for consumption. Thus, consumption
increases according to equation 8, and MMS shifts
upward in the short run. Assuming the effect on
OOS dominates, the economy jumps from E to E′,
and labor and consumption fall in the short run.
What happens when the government raises
taxes on interest income? This tends to reduce the
after-tax interest rate received by households for
any given pretax interest rate. But since the aftertax interest rate is pegged in the long run, the
pretax interest rate must rise. To accomplish, this
the long-run capital–labor ratio falls in order to
increase the marginal product of capital. In turn, a
fall in the steady-state capital–labor ratio will affect
lines MML and OOL. For any given level of labor,
reducing the capital–labor ratio means that labor
is less productive. This reduces long-run output
and consumption. Thus, MML rotates down and to
the right. At the same time, the after-tax wage rate
falls with a reduction in the capital–labor ratio.
58

Thus, households substitute away from work and
consumption. This is equivalent to a leftward shift
of OOL. Figure 6 shows how increasing the interest
rate tax twists MML and OOL downward. This
causes the equilibrium to move from E to E″. Consumption falls in the long run. Whether labor falls
or rises is unclear and depends on how much
OOL falls relative to MML. Nonetheless, for all
reasonable parametrizations, capital and output
will fall in the long run. Since the capital–labor
ratio falls in the long run, investment will fall in the
short run. Thus, MMS shifts up and the economy
moves from E to E′ in the short run.
To summarize, increasing either factor tax
will lower labor and output in the short run and
increase the capital–labor ratio. But a wage tax
will lower consumption in the short run, while an
interest rate tax will raise consumption. In the long
run, both taxes depress consumption and output.
However, they affect labor and the capital–labor
ratio differently. A wage tax leaves the capital–labor
ratio unchanged and depresses labor. On the
other hand, an interest rate tax lowers the capital–
labor ratio and has an uncertain effect on labor.
A brief glance at the box titled “Equivalence
of Permanent Tax Policies,” shows how taxes on
consumption and corporations are equivalent to
the factor income taxes introduced in this article.
In short, most taxes correspond to taxes on capital
or labor. Ostensibly, having a personal and cor-

Figure 6

Effects of a Higher Interest Rate Tax
c

OOS

MML

OOL

MMS

E′
E
E″

H

h

Federal Reserve Bank of Dallas

Equivalence of Permanent Tax Policies
One can easily expand the model by
including household consumption taxes, tc ,
that comprise excise and sales taxes. In this
case, equation 1 expands to

(A)

(1 + t c )c + (k +1 − k ) = (1 − t w )wh
+ (1 − t r )rk + l .

Additionally, various taxes can be levied on
the firm so that after-tax profits are given by

(B)

(1 + t o )y − (1 + t h )wh − (1 + t k )(r + δ )k
− (k +1 − k ),

where to is a tax rate on the output of the firm.
Also, th is a tax surcharge on firms’ labor costs,
such as contributions for social insurance.
The term tk is the tax surcharge on the rental
payments of capital and adds to (or subtracts
from) the cost of capital through alternative
tax depreciation schedules, capital consumption allowances, and taxation and deductibility of dividends, debt, and capital gains.
In this case, the combined household
and firm steady-state optimality conditions
generalize to
⎡ (1 − t w )(1 − t o ) ⎤ ⎛ k ⎞
(C) MRS (c ,H − h ) = ⎢
⎥fh ,1 ,
⎣ (1 + t c )(1 + t h ) ⎦ ⎝ h ⎠

(D)

⎡1+ t o ⎛ k ⎞
⎤
ρ = (1 − t r )⎢
fk ,1 − δ ⎥ .
⎝
⎠
h
⎣1 − t k
⎦

porate income tax implies that tax rates on labor
and capital are equal or that a “simple income
tax” exists. This is misleading because of special
tax considerations for capital, such as depreciation
schedules, capital gains taxes and so on. In fact,
there is evidence that tax rates on capital far exceed
tax rates on labor.15 Thus, it is natural to ask what
are the effects of a reduction in interest rate taxes
Economic Review — Third Quarter 1993

From the optimality conditions, one can
show that the following taxes are equivalent,
in the sense that their qualitative effects on
aggregate consumption, investment, labor,
and output (in the short run and long run) are
the same. The equivalence relationships show
that taxes on corporations imitate taxes on
households by ultimately taxing labor and
capital. It also can be shown that a consumption tax is equivalent to a tax on labor:
1. A tax on the wage income of households, tw , is equivalent to a surcharge
on the labor costs of firms, th .
2. A tax on the interest income of households, tr , is equivalent to a surcharge
on the capital costs of firms, tk .
3. A (simple) income tax, ty , is equivalent to taxing households’ wage and
interest incomes at the same rate—
that is, tw = tr .
4. An output tax, to , is equivalent to
taxing firms’ wage and interest costs
at the same rate—that is, th = tk .
5. An output tax, to , is equivalent to a
(simple) income tax, ty .
6. An output tax, to , is equivalent to a
consumption tax plus a tax on interest income —that is, tc = tr . Or, a
consumption tax is equivalent to a
sales tax with capital costs exempt.
7. A consumption tax, tc , is equivalent to
a tax on wage income, tw .

15

Marginal tax rates have been estimated by a number of
authors. For instance, Hansson’s (1985) survey concludes
that the labor tax rate lies between 0.2 and 0.3, while the
capital tax rate is bounded above by 0.5. McGrattan (1991)
estimates that the labor tax rate fell in the interval between
0.1 and 0.35, and the capital tax rate ranged between 0.3
and 0.6.

59

and an increase in wage taxes. Such a scheme
may be considered a variant of the investmentfriendly restructuring of business taxes proposed
by the presidential candidates.16 From the analysis
above, one sees that the short-run effect is to shift
OOS leftward. However, the individual taxes
affect short-run investment in opposite directions.
Since the capital–labor ratio rises in the long run,
it is likely that investment will increase in the
short run and that the MMS will shift down. However, even if the effects on investment approximately cancel, consumption and labor will fall in
the short run.
Line MML will shift up in the long run,
because the long-run capital–labor ratio rises.
Assuming that the different effects on line OOL
approximately cancel, consumption will increase
and labor will fall. Also, output will rise in the
long run along with the capital–labor ratio. This
exercise is intriguing, because it is possible to get
a long-run expansionary effect simply by changing the tax mix from capital to labor taxation.
However, the short-run economic costs of such
policies may outweigh the long-run benefits.

60

16

Reducing the cost of capital can be accomplished by an
investment tax credit or by reducing the capital gains tax.
The cost of labor would rise if a tax for worker training were
instituted, or employer health care costs were raised. Since
consumption taxes and labor taxes are equivalent, one
would get the same result by increasing sales taxes.

17

This policy exercise was analyzed by Wynne (1991). He
also considers the aggregate effects of military employment policies.

18

What if government consumption enters private utility, as in
footnote 5? In this case, the MRS is a function of composite,
not private, consumption. Also, public consumption enters
the market equilibrium condition just like defense expenditures. These two facts can be attached to the graphical
analysis for defense spending.
If government consumption falls, lines MMS and MML
shift up just as they do with a reduction in defense spending.
However, lines OOS and OOL will also shift down, because
private consumption must rise to offset the fall of the MRS
induced by public consumption. If public and private consumption are less than perfect substitutes, households will
work more to raise output and to mitigate the negative effect
on private consumption. In the short and long run, private
consumption will rise, while the effect on labor is uncertain.

The effects of spending policies
Suppose that defense spending falls permanently.17 Since capital tax rates do not change, the
capital-labor ratio is unaffected in the long run.
Thus, the long-run market equilibrium relationship
MML depends solely on the demand and supply
effects of the change. Since defense spending
does not enter the production function, there is
only a demand effect. This means that more output
is left for consumption than before the shock. For
all levels of equilibrium labor, consumption rises.
Thus, line MML shifts up in Figure 7, while line
OOL is unaffected. Consequently, the long-run
equilibrium moves from E to E″, with private consumption rising and labor falling. Also, since the
long-run capital–labor ratio does not change,
capital must fall proportionately to labor.
The short-run effects are qualitatively similar;
only MMS shifts up. Private consumption will be
crowded in and labor will fall because households
feel wealthier. Since the capital–labor ratio remains
unchanged in the long run, investment will fall.
This reinforces the positive effect on private consumption. Since labor falls, output will fall, too.18
The effects of increased public investment
differ from those of increased defense spending.
Because of the separability of the production
function, public investment does not affect the
private capital–labor ratio. While higher public
investment does not affect OOL, it has two effects
on MML. Not only does public investment have a
demand effect, it also has a supply effect on the
market equilibrium condition. If the marginal product of public capital is greater than the depreciation rate, then the supply effect will dominate.
Since this is likely, output increases relatively more
than demand does, and consumption rises for all
levels of labor. Thus, line MML shifts up in Figure
8. Therefore, labor, capital, and private output fall
in the long run, while consumption and total
output increase.
The short-run effect of public investment
does not include a supply effect. This is because
investment takes time to be productive. Thus, the
demand effect governs the short-run market equilibrium relationship. For all levels of labor, higher
investment means lower consumption. Thus, line
MMS shifts down in the short run in Figure 8. This
means that at the short-run equilibrium point E′,
Federal Reserve Bank of Dallas

Figure 7

Figure 8

Effects of Lower Defense Spending

Effects of Higher Public Investment

c

c
MML
OOS

MML
OOL

OOS
OOL
MMS

MMS

E′
E″
E″

E

E
E′

H

labor is higher and consumption lower than at the
point of departure, E. Over time, as the economy
moves from E′ to E″, consumption will rise and
labor fall.19
In sum, in the short run, more public investment tends to lower consumption and increase
labor and output. In the long run, public investment raises consumption and output and lowers
labor. A reduction of defense expenditures will,
on the other hand, raise consumption and reduce
labor and output in the long run and in the short
run. Spending the peace dividend from reduced
defense outlays on public investment is equivalent
to increasing public investment and reducing
defense spending by an equal amount. In this
case, demand effects will cancel in the market
equilibrium condition MMS. While there are no
supply effects in the short run, in the long run
there will be positive demand and supply effects
on MML. Since the supply effect dominates, MML
shifts leftward by more than if public investment
were increased by itself. Thus, in the short run
there is no effect on the aggregate variables. However, in the long run private consumption will rise
and labor will fall. Since public output rises in the
long run and private output falls, the effect on
total output is indeterminate.20
Finally, what if the government reduces defense
spending and legislates an investment-enhancing
reduction in capital taxes? Briefly, in the short run
there are no effects on OOS. There are offsetting
Economic Review — Third Quarter 1993

H

h

h

19

If public capital is not separable in production and production is constant-returns-to-scale in all inputs, then the public
capital to private labor ratio enters equations 9 through 11.
Thus, the private capital–labor ratio is not pegged by the
constant rate of time preference in equation 11 and will
adjust with changes of the public capital–labor ratio. The
government can peg the private capital–labor ratio to the
discount rate by varying the public capital–labor ratio
(using lump-sum taxes to balance its budget). It then is free
to pursue the policies discussed above with the same
aggregate effects.
If the government targets a higher public capital–labor
ratio, the marginal product of private capital rises in the long
run, raising the private capital–labor ratio. In the short run,
public and private investment increase, causing MMS to
shift down. Assuming that supply effects dominate demand
effects, line MML will shift up. Line OOL shifts rightward
because wages rise, while OOS remains unaffected. Thus,
the graphical analysis resembles the case of increasing
public investment and simultaneously reducing capital
taxes.

20

What if (separable) public capital is raised and public
consumption is reduced dollar for dollar? In this case, line
OOS shifts up. Since labor rises in the short run and the
capital–labor ratio falls, private investment will increase.
Thus, MMS shifts down. Therefore, labor, investment, and
output rise in the short run, while private consumption may
rise or fall. Also, line MML will shift upwards because of the
supply effect of public investment. At the same time, line
OOL will shift down when public consumption is not separable. Thus, it follows that labor and output fall in the long
run, but the effect on consumption is uncertain.

61

Table 1

Summary of Policy Effects
Short-Run Effects

Policy
Higher labor tax
Higher capital tax
Higher labor tax
and lower
capital tax
Lower defense
spending
Higher public
investment
Lower defense
spending and
higher public
investment
Lower defense
spending and
lower capital tax

Consumption

Labor

Capital–
labor
ratio

–
+

–
–

+
+

–

–

+

Long-Run Effects

Labor

Capital–
labor
ratio

Capital

Output

–
–

–
?

0
–

–
–

–
–

–

+

–

+

+

+

–

–

+

–

0

–

–

–

+

+

+

–

0

–

+

0

0

0

0

+

–

0

–

?

?

?

?

?

+

–

+

+

?

Investment

Output

Consumption

–
–

–
–

+

–

–

+

–

+

0

?

effects on investment, so that the effects on MMS
are unclear. In the long run, both MML and OOL
will shift upward, except that the shift of MML will
be magnified. Since the short-run effects depend
on how MMS shifts, consumption will move in the
opposite direction of labor (and output). In the
long run, consumption rises, labor falls, and the
effect on output is ambiguous.
Conclusion
In this article, I have developed a simple
framework to analyze the effects of various fiscal
policies. Abstracting from distributional considerations, this model is useful for looking at the shortrun and long-run effects of various taxation and
expenditure schemes. In particular, I contrast wage
income taxation (or taxes on labor) with interest
income taxation (or taxes on capital), and I contrast
defense spending with government investment.
The effects of the policy experiments are summarized in Table 1. These particular instruments are
chosen because they figured prominently in the
62

fiscal policy debate of 1992. Also, many fiscal
policies can be described as a combination of these
four instruments. For instance, it is shown that
most corporate taxes are equivalent to personal
income taxes in their effects on macroeconomic
aggregates. This is because corporate and personal
income taxes ultimately tax the inputs to production. This equivalence lies at the heart of economists’ observation that the current tax system
heavily taxes capital.
The model suggests that increases in taxes
on inputs will depress output and consumption in
the long run. While labor taxes tend to lower labor
and capital in the long run, capital taxes lower
capital but may raise or lower labor in the long
run. The model also shows that a consumption
tax is equivalent to a tax on labor. Thus, a differential change in factor taxes may have been implicit
in some of the 1992 campaign proposals for a
pro-investment restructuring of business taxes.
Suppose that capital taxes are lowered and labor
(or consumption) taxes raised such that the effect
is neutral on government revenues. In this case, it
Federal Reserve Bank of Dallas

is likely that consumption, labor, and output will
fall in the short run. In the long run, labor still
may fall, but consumption and output will rise.
Thus, it is possible that changing the tax mix will
have expansionary long-run effects on the economy
and still be revenue neutral. However, these longrun benefits must be weighed against their shortrun costs.21
On the other hand, the model proposes that
spending the peace dividend from reduced defense
spending on public investment will yield long-run
benefits and no short-run costs. A reduction in
defense expenditures tends to increase consumption and reduce capital, labor, and output in the
short and long run. By contrast, public investment
will raise labor and output and lower consumption in the short run; it will reduce labor and raise
consumption and output in the long run. Thus, if
government investment increases and defense
spending falls dollar for dollar, consumption, labor,
and output are not affected in the short run, but
in the long run, consumption rises and labor and
private capital fall. Whether output rises depends
on the output effects of shifting from private capital
to public capital. The model also has implications
for when the peace dividend is used to create a
more investment-friendly business tax structure by
reducing tax distortions on capital. In the short
run, output may rise or fall; however, consumption and output (and labor) will move in opposite

directions in the short run. In the long run, consumption and capital will rise while labor falls.
Whether output rises depends on the output effects
of shifting from labor to private capital.
Whether this last option is preferable to
increasing public investment was a principle difference between the major contending fiscal policy
platforms. However, it turns out that both options
would be likely to have very similar qualitative
outcomes in the long run. And they also appear to
be similar to a shift from capital to labor (or consumption) taxes in their long-run effects. While
these three policies have qualitatively similar longrun effects, their short-run effects are dissimilar.
Increasing public investment by reducing defense
spending dollar-for-dollar clearly dominates a
differential tax change (from a labor tax to a capital
tax) in the short run. Whether this policy also
dominates a reduction of defense spending and
capital taxes depends on whether output rises or
falls. And since consumption will move opposite
to output, the ranking of the short-run effects of
the last two policies depends on whether the public
puts a higher value on movements in consumption or output. Currently, the empirical testing of
these models is an active area of research. This
research will provide estimates of the short-run
and long-run policy effects and help in deciding
which policies are implemented.

21

Economic Review — Third Quarter 1993

Note that increasing the progressivity of the personal
income tax by increasing taxes on the rich (and maybe
lowering taxes on the middle class) is a capital tax in
disguise. Because the share of capital income increases
with income, taxing the rich taxes capital income (and
reducing middle class taxes lowers labor taxes). Thus,
increasing the progressivity of the income tax might
offset the pro-investment business tax restructuring discussed above.

63

References
Akhtar, M.A., and Ethan S. Harris (1992), “The
Supply-Side Consequences of U.S. Fiscal
Policy in the 1980s,” Federal Reserve Bank of
New York Quarterly Review, Spring, 1–20.
Aschauer, David A. (1989), “Does Public Capital
Crowd Out Private Capital?” Journal of Monetary Economics 24 (September): 171– 88.
——— (1988), “The Equilibrium Approach to
Fiscal Policy,” Journal of Money, Credit and
Banking 20 (February): 41– 62.
Barro, Robert J., and Xavier Sala-I-Martin (1992),
“Public Finance in Models of Economic
Growth,” Review of Economic Studies 59
(October): 645 – 61.
——— (1989), “The Neoclassical Approach to
Fiscal Policy,” in Modern Business Cycle
Theory, Robert J. Barro, ed. (Cambridge,
Mass.: Harvard University Press).
Becsi, Zsolt (1991), “Fiscal Policy in a Perfect
Foresight Model with Heterogeneous Consumer/Workers,” (Ph.D. dissertation, University of Wisconsin—Madison).

64

Council of Economic Advisers (1992), Economic
Report of the President (Washington, D.C.: U.S.
Government Printing Office).
Hansson, Ingemar (1985), “An Evaluation of the
Evidence of the Impact of Taxation on Capital
Formation,” (Paper presented at the Forty-First
Conference of the International Institute of
Public Finance, Madrid, Spain, Summer).
Lynde, Catherine, and James Richmond (1992),
“The Role of Public Capital in Production,”
Review of Economic Statistics 74 (February):
37– 44.
McGrattan, Ellen R. (1991), “The Macroeconomic
Effects of Distortionary Taxation,” Institute for
Empirical Macroeconomics, Discussion Paper
no. 37, Federal Reserve Bank of Minneapolis.
Wynne, Mark A. (1991), “The Long-Run Effects of
a Change in Defense Purchases,” Federal
Reserve Bank of Dallas Economic Review,
January, 1–16.

Federal Reserve Bank of Dallas