View original document

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

What Can Price Theory Say
about the Community
Reinvestment Act?
Robert L. Lacy and John R. Walter


he Community Reinvestment Act (CRA or Act) is credited with funneling billions of dollars in loans and investments to distressed U.S.
communities over the last two decades. Passed in 1977 in response
to concerns that banks were failing to make loans in declining communities,
the Act in essence requires federal bank regulators to assess banks’ lending
and investment activities and encourage expanded lending and investment in
lower-income communities.1 Its passage was viewed as an important step in
curbing lending practices that might discriminate against borrowers in lowincome communities and assuring that banks would provide much-needed
funding for urban and rural development. And CRA has indeed changed the
way bankers think about lending in low-income communities in the United
States. With federal regulators looking over their shoulders, bankers today
are more careful about their lending and investment practices, particularly to
the extent that such practices differ across the often-diverse communities they
serve. While it is difficult to quantify the contribution of CRA to community
development, some lower-income neighborhoods undoubtedly have benefited
from the Act. The percentage of low-income loans in the typical bank’s loan
portfolio has grown, albeit modestly, in recent years, and many cities have received sizeable grants from banks in support of their community development
The authors benefited greatly from discussions with Huberto Ennis, Marvin Goodfriend,
Thomas Humphrey, Edward S. Prescott, Daniel Tatar, and John Weinberg. The views expressed herein are not necessarily those of the Federal Reserve Bank of Richmond or the
Federal Reserve System.
1 All depository institutions except credit unions are subject to CRA regulations—i.e., CRA
is applicable to commercial banks, savings banks, and savings and loans. Throughout this article
the simpler term banks will be substituted for depository institutions.

Federal Reserve Bank of Richmond Economic Quarterly Volume 88/2 Spring 2002



Federal Reserve Bank of Richmond Economic Quarterly

Yet, 25 years after its passage, CRA continues to draw criticism and generate controversy. The controversy stems in part from differences of opinion
regarding the need for CRA as a tool to counter discriminatory lending practices given that comprehensive fair-lending laws are also in place. Critics of
the Act maintain that there is little evidence that banks discriminate by neighborhood or community when lending and that therefore there is little need
for legislation that goes beyond the protection offered by laws preventing discrimination based on minority status. Proponents of CRA, however, argue
that the Act is needed because in its absence some banks would indeed discriminate based on a borrower’s location, and perhaps even completely refuse
to serve some neighborhoods, essentially redlining those communities. Many
proponents believe that CRA offers minorities additional protection against
discrimination even if banks don’t redline.
Some proponents of CRA also argue that the Act improves lending-market
performance by increasing the information available about lending in lowincome areas. If so, then CRA can be beneficial for both borrowers and
lenders. Canner and Passmore (1995, 77–79) discuss this view of CRA as a
mechanism for correcting for market failure due to externalities and compare
this perspective to other views of how CRA affects bank lending. If CRA
partially corrects for market failure, then lending markets can become more
But critics of CRA have suggested that such government intervention in
the banking industry reduces the efficiency of credit markets. They argue that
this intervention, which in the case of lending allows some low-income borrowers to secure loans at below-market interest rates or on better-than-market
terms, can distort credit markets and cause resources to be wasted. There
are those who maintain that CRA lending represents a substantial subsidy to
low-income borrowers and suggest that direct government aid or government
loan programs provide a more efficient means of assisting low-income communities. Lacker (1995, 15–18), for example, is skeptical of the externalities
argument, wondering why lending in low-income neighborhoods should be
any more susceptible to market failure than lending in more affluent neighborhoods. He argues that funding community development directly out of general
tax revenues seems more promising than employing CRA to redistribute income to alleviate the problems of the nation’s low-income neighborhoods.
Without denying the benefits of CRA, we identify unique lending costs the
Act imposes and analyze how these costs affect credit markets. We argue that
CRA enforcement requires some banks to expand their low-income lending to
such an extent that unprofitable loans are likely to be made, creating inefficiencies in credit markets. These inefficiencies are created as banks modify lending
in response to CRA regulations and examinations. We also explore why CRA
continues to survive in today’s banking environment, despite the above costs
that it imposes on banks but not on banks’ competitors. Requirements that

R. L. Lacy and J. R. Walter: Price Theory and the CRA


impose higher costs on only one segment of an industry are notoriously difficult to maintain in a competitive environment, and since the passage of CRA
in 1977, the financial services industry has clearly become more competitive.
Banks must charge customers in some markets higher prices in order to sustain lower, unprofitable prices in other markets. As competitors successfully
target these higher-priced markets, however, one would expect the source of
funding to eventually dry up. A number of industries—telecommunications,
electric utilities, and airlines—that have also been subject to deregulation and
increased competition over the last 20 years have seen similar requirements
erode as markets became more competitive.
We believe that CRA has survived for 25 years for two reasons: (1) a
portion of CRA’s costs can be shifted to banking products for which there
are fewer alternatives offered by banks’ competitors, and (2) CRA’s costs are
relatively small. Looking ahead, while many community activists support
CRA and would like to see it play an even larger role in funding community
development, we consider it unlikely that CRA will take on an expanded role
because competition imposes very real limits on banks’ ability to draw funds
from customers to support CRA lending.



In evaluating the effects of CRA on bank lending, we model CRA as a requirement to expand output of one product of a multiproduct firm. More specifically,
this requirement is that a bank’s low-income lending equal a fixed proportion
of total lending.2 We also assume that CRA requires banks to expand lowincome lending beyond the level that would prevail in the absence of the Act.
We begin by discussing a proportion-based requirement in general terms, and
then illustrate some of the consequences for costs and output using a hypothetical example of a lawn mower manufacturer. A detailed presentation of
the consequences of a proportional CRA lending requirement concludes the

A Requirement to Expand Output
Requirements meant to encourage expanded output of one product sold by a
multiproduct firm are not unique to banking. For example, regulations adopted
2 The CRA regulations define four income categories (see, for example, Board of Governors
Regulation BB). A low-income individual or family is one with income less than 50 percent of
area median income. Moderate income is income that is at least 50 percent and less than 80
percent of area median income. Middle income is at least 80 percent and less than 120 percent
of median income. Upper income is 120 percent or more of median income. In our model the
phrase low-income corresponds to the low- and moderate-income categories in the regulations, while
middle- to high-income (MHI) corresponds to the two higher-income categories.


Federal Reserve Bank of Richmond Economic Quarterly

in California in 1990 encourage expanded production of environmentally benign motor vehicles. These regulations require that specific percentages of
passenger cars and certain light-duty trucks produced for sale in the state by
each of the seven largest auto manufacturers be zero-emission vehicles (ZEV).
The objective was that ZEVs compose 2 percent of production by model year
1998 and 10 percent by 2003.3
In a competitive industry, a requirement to expand output beyond the level
chosen by a firm will generally mean that the additional output is produced at a
loss. A producer in such an industry normally chooses to produce the quantity
at which its marginal costs just equal the price buyers are willing to pay.
Since marginal costs of production generally rise with increasing output, and
because in a perfectly competitive market a single price exists for the firm’s
entire output, profits are greatest at a level of output where marginal costs
equal the market price. If production is below this level, the cost of producing
an additional unit is below the price buyers are willing to pay, so the firm will
continue to expand output in order to increase profits. The producer will not
choose to expand output beyond the point at which marginal costs equal price
because it would suffer losses on this additional output.
If the firm produces two goods and a regulation requires that the production
of one good be expanded to be at least a certain percentage of total production
of both goods, then a penalty cost is incurred for the production of the second
good. The penalty cost consists of the losses from additional sales of the
first good, for these losses must be incurred when additional quantities of the
second good are sold. The imposition of a penalty cost matters little if all firms
are subject to the same regulation. But if some firms are free of regulation,
such firms will gain market share in the second good. Unregulated firms do
not sustain a penalty cost, so they can charge a lower price and acquire some
of the regulated firms’ customers.
An Example

To illustrate how a market might respond to a regulation requiring expansion of
output, consider a simple hypothetical example of a lawn mower manufacturer
that sells both electric- and gasoline-powered lawn mowers in a competitive
market. Legislators in this example impose a requirement that all manufacturers of lawn mowers produce relatively more electric-powered models in order
to protect the environment.
The supply curve for electric lawn mowers identifies the quantity of such
mowers the firm will produce at various prices. As the market price increases,
3 California regulations have since been modified to allow manufacturers to meet the percentage requirements with a combination of ZEVs and very low emission vehicles. ZEV mandates
were adjusted to allow manufacturers to receive partial credit for extremely low emission vehicles
that were not pure ZEVs. See California Environmental Protection Agency (2001).

R. L. Lacy and J. R. Walter: Price Theory and the CRA


the manufacturer is able to recover higher manufacturing costs, and the number
of electric-powered mowers it is willing to produce increases. The number of
electric mowers produced at a given price depends on the firm’s marginal cost
of production. This cost rises as the firm produces more. The demand curve is
horizontal because the firm in our example operates in a perfectly competitive
market—it is only a small manufacturer in a large market for electric mowers.
It cannot affect the market price for electric mowers because its portion of the
market is almost insignificant. Given its demand and supply curves, the firm
will produce a quantity of mowers identified by the intersection of supply and
demand curves, or for our example, 100 electric mowers. The firm earns its
highest profit by manufacturing 100 electric lawn mowers.
The manufacturer also chooses to produce the quantity of gas mowers
determined by the intersection of supply and demand curves for this product.
The equilibrium price and output quantities differ, however, in the gas and
electric mower markets. In equilibrium, we will assume the manufacturer
produces 100 electric mowers and 900 gas mowers, for a total output of 1,000
lawn mowers.
Suppose legislators decide that electric lawn mowers are more environmentally friendly and that their use should be encouraged. Legislation is
therefore passed requiring that electric mowers account for 15 percent of total
mower production rather than the current 10 percent. The legislative requirement leads the manufacturer to increase its output of electric mowers beyond
100. The manufacturer’s marginal cost exceeds marginal revenue when it produces more than 100 mowers. With a regulation requiring the manufacturer to
expand production of electric mowers to 150, the manufacturer suffers losses
on each of the last 50 electric mowers produced.
Because of this requirement, the firm’s supply curve for gas mowers will
shift leftward to a position above what it would be without the requirement.
Why does the requirement raise the manufacturer’s cost of producing gas
mowers? Because the number of electric mowers produced is determined by
the manufacturer’s production of all lawn mowers, both electric and gas. As a
result, when the manufacturer wishes to expand its production of gas mowers,
it must also expand its production of electric mowers in order to comply with
the electric mower output requirement. So when a firm increases gas mower
production, it is required to produce electric mowers at a loss.
The output requirement will affect manufacturers to different degrees because their cost curves vary. Some firms are especially skillful or well equipped
for producing electric mowers; others are more skillful at gas mower production. For example, a manufacturer located in a city with battery and electricmotor suppliers can be expected to have relatively low costs for electric mower
manufacture because shipping costs for these mower components will be low.
For manufacturers especially equipped to make electric mowers, one would
expect that production of electric mowers as a percentage of all mowers would


Federal Reserve Bank of Richmond Economic Quarterly

be well above average in the absence of an output requirement, or above 10
percent in our example. For such manufacturers, electric mowers might represent 25 percent of their total mower output. The legislative requirement that
electric mowers compose 15 percent of all mowers would not result in a cost
increase for these manufacturers—the requirement is nonbinding for them.

CRA in a Competitive Banking Environment
CRA requirements impose a mandate to expand lower-income community
lending analogous to an automobile industry requirement promoting electric
vehicles and our hypothetical example of a requirement to expand electric
mower production. While CRA also promotes community development investments and service, banks focus their CRA compliance efforts on lending;
therefore, our analysis of CRA does so as well. We begin with a look at the
effect of CRA on an individual bank assumed to be operating in a competitive banking environment. A requirement to expand lower-income lending
will produce an increase in the effective cost of other types of lending. This
increase in banks’ effective lending costs means that nonbank lenders gain
a competitive advantage. We then broaden our analysis to cover the financial services industry as a whole, including nonbank competitors. As will
be shown, federal bank regulators lose much of their ability to influence the
number of loans made to low-income communities when financial institutions
not subject to CRA lending requirements emerge.
CRA and One Bank

When a bank operates in a perfectly competitive market, its supply and demand
curves for loans made in low-income communities appear as shown in Figure
1a. The bank’s supply curve identifies the number of loans it is willing to
make at various interest rates. As the market rate of interest increases, the
bank is able to recover higher costs of lending, and the number of loans the
bank is willing to make increases. The number of loans it will make at a
given interest rate, say RL , depends on the bank’s marginal cost of lending.4
The bank’s marginal cost of lending rises because to increase its lending it
must, for example, spend more to attract qualified loan officers from other
fields or make increasingly higher outlays for marketing to attract additional
borrowers. The demand curve D in Figure 1a is horizontal since the bank
is only a small part of a huge lending market. Given demand curve D and
4 While in Figure 1a, and throughout the article, the supply curve is represented as equivalent
to the marginal cost curve, this is a simplification. A firm’s output also depends on its average
variable cost curve. The supply curve would be equivalent to the marginal cost curve only at
market prices above the intersection of the marginal and average variable cost curves.

R. L. Lacy and J. R. Walter: Price Theory and the CRA


Figure 1 Single Bank, without CRA

supply curve S, the bank will make 100 low-income loans—where the supply
and demand curves intersect at point E in Figure 1a.
Supply and demand curves for lending in other markets will be similar and
are represented in Figure 1b. We define other lending as loans to all borrowers
except those in low-income communities. We will call them middle- to highincome (MHI) borrowers.5 Note that in this market too the bank chooses to
5 For expository purposes, we represent the supply and demand curves of all other loan
markets by the supply and demand curves of Figure 1b, even though each type of loan has its
own supply and demand curves.


Federal Reserve Bank of Richmond Economic Quarterly

make the number of loans determined by the intersection of supply and demand
curves. The equilibrium interest rates and loan quantities differ, however, in
the two markets. In our example, in equilibrium the bank makes 100 lowincome loans and 900 other loans for a total of 1,000 loans.
What effect do CRA requirements to expand lending have on the bank
in our model? If binding, CRA requirements oblige the bank to make more
low-income loans than it would if there were no requirements. Without CRA
requirements, the bank chooses to make 100 low-income loans representing
10 percent of its total loans. Binding regulations, however, require the bank
to extend more low-income loans in order to merit a satisfactory CRA rating.
While its profit-maximizing output is 100 low-income and 900 other loans,
binding CRA regulations require the bank to make low-income loans equal
to, say, 15 percent of all loans. Although CRA examiners do not have explicit minimum percentage guidelines that banks must meet, they do expect
bank lending to low-income individuals, in the absence of extenuating circumstances, to be roughly proportional to the low-income population. More
will be said about examiner expectations later in Section 2.
Figures 2a and 2b illustrate the case in which CRA requirements are
binding; in our example, the bank in Figure 1a must make more than 100
low-income loans. In Figure 2a the bank’s marginal cost exceeds marginal
interest earnings when it extends more than 100 low-income loans. With a
regulation requiring low-income loans equal to 15 percent of total loans, the
bank makes 150 loans, but it suffers losses equivalent to area BCE on lowincome loans between 100 and 150. The bank would like to charge borrowers
an interest rate sufficient to cover its higher costs of making each additional
loan. The bank’s low-income customers, however, are only willing to borrow
at the interest rate of RL .
Supply and demand curves for the bank’s other lending market are shown
in Figure 2b. The S curve in Figure 2b is the bank’s cost of making MHI loans,
without CRA requirements. The S curve is the bank’s cost of making MHI
loans with a binding CRA requirement. The S curve lies above the S curve
for most of its range, reflecting the additional cost of a requirement to make
at least some unprofitable low-income loans. We define unprofitable loans as
those for which marginal cost exceeds marginal revenue. The S curve begins
to diverge from S at the level of MHI lending for which CRA requirements
become binding.
In our example, with a CRA requirement that low-income loans account
for 15 percent of all loans, the point of divergence is 567 MHI loans. If the
rate of interest is such that the bank wishes to make only 567 loans—RH in
Figure 2b—it will not be required to make any low-income loans for which
costs exceed interest earnings. At 567 MHI loans the bank can make 100
low-income loans [100/(567 + 100) = 15 percent], exactly the number of
low-income loans it would make without a CRA requirement.

R. L. Lacy and J. R. Walter: Price Theory and the CRA


Figure 2 Single Bank, with CRA

If the market rate of interest for MHI loans is above RH , the bank will find
its most profitable combination of MHI and low-income lending to include
more than 567 MHI loans and, because of CRA, more than 100 low-income
loans. Since the S curve represents not only the costs of making MHI loans,
but also the losses generated by whatever quantity of low-income loans are
called forth by the profit-maximizing level of MHI loans, it represents the
bank’s most profitable levels of high-income lending, given CRA.6
6 In Figures 2a and 2b we have assumed that the bank will continue to make a total of
1,000 loans (150 low-income and 850 MHI) after CRA is imposed. By imposing an additional


Federal Reserve Bank of Richmond Economic Quarterly

Figure 3 Middle- to High-Income Loans: Banking Industry with CRA

Banking Industry and Nonbank Competition

Banks today compete with nonbanks in most banking-product markets (see
Increased Competition from Nonbanks in Table 1). For example, securities
brokers offer money market mutual funds, for some customers an attractive
alternative to holding a deposit account at a bank. Likewise, finance companies offer consumer and business loans, and nonbank mortgage lenders offer
residential mortgage loans, all in competition with similar loan products offered by banks. These nonbank competitors are free from CRA requirements.
With nonbank competitors present, regulators face limits on their ability to
employ CRA requirements to expand low-income lending; limits are present
but less severe in the absence of nonbanks.
Figure 3 depicts the banking industry’s cost curves for MHI loans—SBI
without CRA and SBI when subject to CRA, with no nonbanks present. These
curves represent the horizontal summation of individual bank MHI cost curves
like those shown in Figure 2b. Bank cost curves vary and are a function of
many factors, including bank size.7 The demand curve for MHI loans is given
by DBI in Figure 3. Unlike the single bank’s demand curves in Figures 1 and
2, the industry demand curve is downward sloping. The market interest rate,
cost on banks, CRA may instead result in a decline in total bank lending below 1,000 loans. Such
a decline in total lending does not change the conclusion that a low-income requirement raises
the effective cost of high-income lending.
7 Only banks with minimum average cost at or below the market price can earn a profit, so
only such banks will be present.

R. L. Lacy and J. R. Walter: Price Theory and the CRA


given CRA, is RBI , determined where the SBI and DBI curves intersect. The
quantity of MHI loans made is QBI .
Nonbank competitors are introduced in our analysis in Figures 4a through
4c. Figure 4a depicts the banking industry’s supply curves, which are like
those in Figure 3. Figure 4b shows the supply curve, SN , for nonbanks. We
assume that nonbanks have the same cost structure as banks but are not subject
to CRA, so the SN curve is an exact replica of the SBI curve in Figure 4a.8 The
financial industry supply curve SF I in Figure 4c is the horizontal summation of
the nonbanks’supply curve SN and the banks’supply curve SBI reflecting CRA
costs. The second supply curve, SF I , in Figure 4c, represents a summation of
bank and nonbank supply curves in a theoretical case in which nonbanks are
subject to CRA and have supply curves like SBI in Figure 4a. In this scenario,
all firms incur higher costs because of CRA and thus SF I lies above SF I . By
comparing interest rates and loan quantities when the supply curve is SF I with
rates and quantities when the supply curve is SF I , we can illustrate changes
that result when firms not subject to CRA are introduced. The industry price
and quantity are found at the intersection of the DF I and SF I curves in Figure
4c—interest rate RF I and quantity QF I . Because SF I lies below and to the
right of SF I , RF I is lower than RF I .
At the market interest rate of RF I in Figure 4a, banks will make QBI
loans, as determined by the intersection of the SBI curve and the market
interest rate of RF I that was determined in Figure 4c. In contrast, if banks
face only competitors that are subject to CRA, the market interest rate is RF I
(as determined in Figure 4c), and the quantity of bank loans is higher at QBI
in Figure 4a. The lower market interest rate RF I reflects the presence of
nonbank competitors, not subject to CRA, and causes banks to make fewer
loans. In addition, the size of the CRA subsidy is lower in the presence
of nonbank competitors. Without these nonbanks, the subsidy is the ABC
triangle in Figure 4a; with nonbank competitors, the subsidy is the smaller
triangle ADE.
This CRA subsidy represents the transfer of costs from low-income to
high-income borrowers. The subsidy is due to CRA’s lending requirement,
which results in low-income borrowers receiving loans at prices below banks’
8 For expository purposes we have assumed that the cost structures of banks and nonbanks
are the same. Such an assumption means, however, other things being equal, that banks would
immediately convert to nonbank status if they were able to do so costlessly and could avoid CRA
by converting. In practice, banks may not want to convert because they hold certain advantages
as banks. Some observers maintain that transactions accounts offer banks a low-cost source of
funds. While some nonbanks offer accounts with certain transaction features, accounts at nonbanks
typically do not provide all of the payments features of a bank checking account. Further, banks
may have cost advantages because of special lending skills, access to information not available
to nonbanks, or access to what some argue is underpriced deposit insurance. A lower cost curve
may allow banks to continue to compete successfully against nonbanks despite the costs that CRA
imposes on high-income lending, so banks would not choose to convert to nonbank status. In
Section 4 we provide further discussion of possible bank cost advantages.


Federal Reserve Bank of Richmond Economic Quarterly

Figure 4 Financial Services Industry

R. L. Lacy and J. R. Walter: Price Theory and the CRA


marginal cost of making the loans. While low-income borrowers benefit, highincome borrowers that receive a bank loan pay more than they would have paid
in the absence of the CRA requirement.
Implications for Expansion of CRA

In addition to reducing a bank’s MHI lending, the presence of nonbank competitors not subject to CRA is likely to diminish the effectiveness of CRA as a
tool for expanding lending to low-income borrowers. Specifically, an increase
in the stringency of CRA requirements in the presence of nonbank competitors will result in a smaller dollar amount of funds available for low-income
The following example illustrates this constraint on regulators’ ability to
expand low-income lending. Assume first that banks are the only type of
financial institution in the industry. With an SBI only slightly above SBI in
Figure 4a, meaning a CRA requirement causes banks to make only a small
number of low-income loans above the quantity they would make without
the requirement, the subsidy is small. The size of the subsidy will initially
increase, but as the CRA requirement becomes more stringent, the subsidy
will decline. In graphical terms, as the SBI curve swings counterclockwise
and away from SBI —that is, CRA requirements become more stringent—the
area of the ABC triangle increases at first, then later decreases.9 The area will
shrink as the SBI curve becomes ever more vertical because the BC side of
the triangle draws near point A of the triangle.
But with nonbanks competing with banks and attracting their customers,
the market rate of interest will rise less than it would in the absence of nonbanks
for any given increase in CRA stringency. The smaller increase occurs because
the nonbanks’ supply curve is unaffected by the increased stringency. As a
result, SF I in Figure 4c, derived by summing banks’ SBI curve from Figure 4a
and nonbanks’ SN curve from Figure 4b, does not swing as far as the SBI curve
because its swing is damped by the stable nonbank curve. Thus, the interest
rate changes little. Because the market interest rate rises less, the height of
the BC side of the triangle in Figure 4a is also smaller in the presence of
nonbanks. The ability of regulators to enlarge the size of the subsidy—i.e.,
the size of the triangle—is diminished by the existence of nonbanks.
Policymakers interested in assisting communities by expanding lending
therefore face very real constraints. They could impose stricter requirements
on banks in order to generate greater CRA lending benefits. For example,
they might require that low-income loans account for 20 percent of total loans,
9 In addition, as CRA requirements become more stringent, point A in Figure 4a shifts along
the SBI curve toward the origin. Point A represents the level of lending at which CRA requirements become binding. In this graph this point shifts toward the origin as the required minimum
ratio of low-income loans to total loans is increased.


Federal Reserve Bank of Richmond Economic Quarterly

rather than the 15 percent in our earlier example. However, stricter requirements eventually produce a declining subsidy as the quantity of MHI loans
made by banks declines and nonbank competitors acquire more of the banks’



The graphical analysis in Section 1 captures the essential elements of CRA’s
low-income lending requirement.10 However, there are two critical assumptions that need further elaboration. The analysis assumes that regulators expect
banks to match or exceed a specific ratio of low-income lending to total lending. While this ratio is a simple statistic and only one of several indicators
considered in a fairly complex CRA lending review, it provides regulators
with an important benchmark of acceptable lending practice. The graphical
analysis also assumes that absent CRA, banks would extend all the profitable
low-income loans possible, and it thus implies that CRA causes banks to make
unprofitable loans. Since the banking industry is currently quite competitive,
and there is little evidence of discrimination based on geographical location,
we believe our assumptions are reasonable in today’s banking environment.
These assumptions are explained in more detail below.

CRA Enforcement and Lending Ratios
The Community Reinvestment Act requires regulators to “assess the [bank’s]
record of meeting the credit needs of its entire community, including lowand moderate-income neighborhoods,” and to “take such record into account”
when evaluating a bank’s merger application (12 U.S.C. 2901, sec. 804). Federal bank regulators make their assessment, assigning a bank one of four
ratings: (1) substantial noncompliance, (2) needs to improve, (3) satisfactory,
and (4) outstanding. Banks that receive low ratings are likely to have difficulty
convincing regulators to approve merger applications. Thus, banks anticipating future mergers will tend to make at least enough low- and moderate-income
loans to ensure that a merger will not be denied, which for some banks will
likely involve an expansion in lower-income lending beyond the level they
would choose in the absence of the Act. While the CRA regulations require
examination of a bank’s record of meeting the credit needs of its “entire community,” the intent is clearly to promote more lending in low- to moderateincome communities. For example, the Federal Reserve’s Regulation BB
10 While our article focuses on CRA’s low-income lending requirement, CRA enforcement
also seeks to encourage expanded lending within banks’ local markets as well as expanded smallbusiness and small-farm lending. Measuring such lending by each bank and basing the bank’s
CRA rating, in part, on these types of lending provide the encouragement.

R. L. Lacy and J. R. Walter: Price Theory and the CRA


specifies “a very poor geographic distribution of loans, particularly to low- or
moderate-income geographies” (emphasis added) as one of the factors that can
cause a bank to receive a rating of substantial noncompliance. The Act also
requires examiners to evaluate banks’ efforts at meeting the “credit needs” of
the community, which includes lending and investments, but the CRA rating
is most dependent on lending.11
The discussion in Section 1 suggests that CRA examiners require banks
to achieve a certain minimum ratio of low-income lending to total lending.
And in practice, CRA examiners do calculate the ratio of a bank’s low-income
lending to total lending and compare that figure to the proportion of lowincome population to total population in the bank’s assessment area as a rough
benchmark of appropriate lending procedures.12 Examiners recognize, however, that there might be legitimate reasons why a bank’s low-income lending
would not be proportional to the low-income population. For example, examiners will take into consideration the prevalence of rental housing in an
area when examining a bank’s mortgage lending patterns, since the demand
for mortgage lending would likely be lower in areas with a high proportion
of rental properties. Likewise, adjustments are made for areas where income
levels are extremely low or unemployment is high, since individuals with very
low incomes or those that are unemployed are unlikely to be willing or able
to borrow.
Still, in general, given these adjustments, if low-income households compose 20 percent of the population in the bank’s assessment area, then examiners
expect approximately 20 percent of the bank’s total loans to be made to lowincome households. Barring unusual circumstances, if the bank increases its
total loans, it will be expected to increase its low-income loans proportionally.
Our assumption of a fixed proportion of low-income loans to other loans in
the graphical analysis reflects this regulatory approach to CRA enforcement.13
Whenever the bank considers making a loan that does not qualify for CRA
credit—say, a loan to an individual in a high-income community—it will take
into account that it will also be required to add to its low-income lending.
11 Large banks are rated based on three tests: a lending test, an investment test, and a
service test. On each of the three tests, the bank receives a rating. The lending test, however, is
predominant. A bank’s numerical score on the lending test receives twice the weight of scores on
either the investment or service tests before these three scores are summed to obtain the composite
CRA score (FFIEC 1997, 15–16). Small bank CRA ratings generally depend only on lending,
though investment and service activities can earn the small bank extra credit toward its final rating.
12 The shorter term low-income will be used in place of the term low- and moderate-income
throughout the remainder of the article.
13 CRA performance evaluations for individual banks are available from websites maintained
by the federal agencies that regulate banks. These evaluations provide a fairly detailed explanation
of the factors examiners consider important in the determination of a CRA rating. An evaluation
typically provides a calculation of the bank’s proportion of low-income lending to total lending,
which is compared to the proportion of low-income households in the assessment area. These
evaluations often note when the two proportions are very different.


Federal Reserve Bank of Richmond Economic Quarterly

Does CRA Encourage Unprofitable Lending?
Without a doubt CRA has increased low-income lending (Litan et al. 2001).
If in the absence of CRA, however, banks would have extended all profitable
loans possible, then it is likely that the Act has resulted in at least some unprofitable loans.14 Given the competitive nature of today’s banking industry, it is
hard to imagine that banks would overlook many opportunities to make additional profitable loans. When CRA was enacted in 1977, substantial regulatory
restrictions on entry and pricing allowed some banks to exercise monopoly
power in local banking markets. In such cases bankers would be expected to
exploit monopoly power by restricting lending. If banks did restrict loans, it
seems plausible that they would tend to prefer higher-income lending at the
expense of low-income lending. Today, however, there are fewer entry and
pricing restrictions in the banking industry, so monopoly power is likely to be
quite limited. (See Table 1 for a review of competition in banking.)
Of course, one could argue that even in competitive markets, some banks
might nevertheless prefer to avoid lending in certain low-income neighborhoods because of a bias against minorities predominant in a neighborhood
or because of a lack of experience in lending to low-income borrowers, for
example. Although not conclusive, a number of studies suggest there is little
evidence of such failure to lend to individuals in low-income communities.15
So, once again, in the absence of CRA we would expect banks to make all
possible profitable loans, and if our expectation were met, CRA pressures to
extend lending would produce unprofitable loans.
CRA may also be causing more low-income lending than is profitable because regulations generally do not make exceptions for differences in business
strategies, market niches, or capabilities. Some banks are simply less adept at
lending in low-income communities than others. For example, certain banks
are especially skilled at making loans and gathering deposits from individuals in high-income communities or providing personal banking services to
high-income individuals and make these services a large part of their deposits
and loans. For such banks, CRA requires an expansion of low-income lending
beyond the profit-maximizing equilibrium, acting as a tax on such banks’highincome lending. While one might imagine that there are few niche banks that
specialize in personal banking, all banks will be on a continuum from those
most capable at low-income lending to those most capable at high-income
14 See, for example, Gramlich (2001).
15 For reviews of the literature on geographic discrimination, see Lacker (1994, 6–9) and

Evanoff and Segal (1996, 24–25).
16 In some cases, a bank can improve its CRA rating not only by making low-income loans
but also by purchasing such loans made by other lenders. If a niche bank, for example, is able to
purchase low-income loans from a lender specializing in making such loans, then the cost imposed
by CRA on the niche bank might be lowered.

R. L. Lacy and J. R. Walter: Price Theory and the CRA


Recently the Board of Governors of the Federal Reserve System surveyed
banks in order to quantify the profitability of CRA lending. Profitability in
the Board study was measured in terms of accounting profits, that is, return
on equity. The Board asked respondents to measure profits based on revenues
and costs associated with such items as overhead and the servicing, pricing,
delinquency, and prepayment of CRA loans (Board of Governors 2000b). This
definition of profitability differs from that used in Section 1, where profitability
is defined as the difference between marginal revenue and marginal cost. The
Board survey found that bank loans that qualify for CRA credit are profitable
for most banks, with many institutions reporting that they are as profitable
as comparable non-CRA loans. However, a high proportion of institutions
reported that CRA loans are less profitable (44 percent of institutions in the
case of home mortgage lending), and almost none reported them as more
Yet, to answer our question regarding unprofitable lending, we must move
beyond an examination of the average profitability of all CRA loans and examine only those low-income loans that are made simply because of the presence
of the Act. In other words, in the absence of CRA, banks would make a
certain number of low-income loans, but only if these loans were expected
to be profitable. The Act may require banks to make additional loans that
are unprofitable. The average profitability of all low-income loans might be
positive even if banks make unprofitable loans in response to CRA. So while
the Board study does not answer our question, the fact that CRA loans are
often less profitable suggests that unprofitable loans may be lowering the average. Given the competitive nature of the banking industry, we believe it is
reasonable to assume that CRA encourages banks to make unprofitable loans.
Still, no industry is perfectly competitive. And the banking industry has
some characteristics suggesting that it is no exception, despite the fact that its
competition is quite strong. For example, observers have noted that switching
costs may be significant in banking, meaning that it is costly for a consumer
to change banks to take advantage of a superior interest rate or lower fee
(Rhoades 2000). Switching costs might be significant for a consumer because
shifting transaction accounts to a different bank would require the consumer to
contact his or her employer, utilities, and mortgage company to modify direct
deposit and direct withdrawal arrangements. As a result of switching costs,
banks might exercise some pricing power over existing customers, providing
a source of supracompetitive profits. If banks enjoy a measure of monopoly
power in pricing, then some of the additional low-income loans CRA will
induce are not necessarily unprofitable. Nevertheless, even if banks do not
lose money on these additional loans, the requirement to make additional
loans will lower profits below the bank’s profit-maximizing level of lending,
and the economic analysis found in Section 1 of this article will be unaffected.


Federal Reserve Bank of Richmond Economic Quarterly

Because profits are diminished for each additional CRA loan beyond the profitmaximizing quantity, banks will take account of the lost profits when deciding
to make an MHI loan. Accordingly, a CRA requirement will mean an increase
in the marginal cost of making MHI loans, or equivalently a leftward shift in
the supply curves for bank MHI lending in the graphs.



While CRA has expanded lending in low-income neighborhoods, any benefit
derived from the expansion may be offset by costs resulting from less efficient credit markets. Furthermore, some costs CRA imposes may be borne
disproportionately by low-income individuals. The expansion of lending to
low-income communities distorts credit decisions on loans made to individuals and businesses in both low- and high-income communities. At the same
time, CRA can place banks at a cost disadvantage in relation to nonbanks,
which are not subject to CRA, further distorting the market. Banks will attempt to shift the costs of making unprofitable loans to those customers who
have the fewest alternatives to bank products.17 Since low-income individuals
frequently have few alternatives, they will tend to bear more than their share
of such costs.

Efficiency Losses
Efficiency losses occur for several reasons. Imagine that one of the 50 borrowers discussed in Section 1 is a small-business owner in a low-income
community who plans to undertake a project that will produce a rate of return
equal to RL in Figure 2a. Consequently, this borrower is willing and able to
pay an interest rate as high as RL to borrow to fund the project. But the bank,
having already made 100 loans, finds that its cost of making this 101st loan
is greater than RL , as indicated by the S curve above RL . The 101st loan
is inefficient and results in a wasteful misallocation of resources inasmuch
as costs exceed benefits as measured by the loan’s rate of return. Because
of CRA, the project is undertaken by the business owner, even though the
project’s economic benefits—as measured by the income the owner earns on
the project—are smaller than the costs of the resources employed to fund it.
This project is funded, but the resources could be employed elsewhere in a
project that delivers benefits exceeding resource costs.
To illustrate the second source of inefficiency, imagine an owner of a
business in a high-income community. This business owner, with a project
capable of earning more than RH , would have received a loan if the bank’s
17 For an example of cost shifting in banking see Fama (1985), who discusses banks shifting
their reserve requirement costs to customers with the fewest alternatives.

R. L. Lacy and J. R. Walter: Price Theory and the CRA


cost curve had been S in Figure 2b. Instead, however, the loan is denied. The
bank will make 900 loans if its cost curve is S, but only 850 loans if S . As a
result, the economic benefits from the project in the high-income community
are unrealized.
A third type of inefficiency arises because CRA requirements result in less
business for banks and more business for other, less efficient providers. While
in Section 1 we assumed that costs of making unprofitable loans are borne by
high-income borrowers, banks may shift some of these costs to depositors as
well. The costs imposed by CRA, shifted to depositors in the form of lower
interest rates paid by banks on deposits, means some individuals will elect to
hold their transactions accounts in other places, such as money market mutual
funds (MMMFs). Yet using a MMMF for payments purposes can be far less
convenient than using a bank account and may result in an inefficient use of the
individual’s resources. Firms offering MMMFs typically do not have as many
branches as banks, nor do they offer widespread ATM networks. Further,
such firms often impose a minimum check size requirement on MMMFs,
making such accounts more difficult to use for day-to-day purchases. While
the consumer is willing to tolerate these inconveniences to earn the higher rate
paid by the MMMF, the inconveniences would have been avoided if CRA had
not led the bank to pay lower rates.
Furthermore, observers have often argued that banks can be especially
effective lenders because they have access to information about borrowers’
finances that allows them to better assess creditworthiness. For example, because bank borrowers often hold transactions accounts with the same bank,
banks have unique access to information about the financial health of borrowers. But if such transaction deposit relationships are severed because banks
lose customers by paying lower interest rates on deposits as a result of CRA,
then this information will be lost. Such information is valuable because it
lowers the cost of making loan decisions. If lost, the economy’s resources
are wasted, either because more resources are consumed by making lending
decisions or because less creditworthy projects are funded.

While CRA is intended to benefit low-income communities by ensuring that
banks do not overlook them, the law may at the same time impose additional
costs on low-income individuals. Some of the costs of such lending will
tend to be shifted to depositors in the form of lower interest rates or perhaps
higher fees on transactions deposits. In effect, individuals holding transactions
accounts are taxed so that more CRA loans can be made.
Such a tax may fall more heavily on low-income and minority individuals.
High-income individuals hold a smaller percentage of their wealth in the form
of checking account deposits than do low-income, and whites hold a smaller


Federal Reserve Bank of Richmond Economic Quarterly

percentage than do nonwhites (Davern and Fisher 2001, Tables 1 and H).
While CRA may produce additional loans for such individuals, it also tends
to tax them.

Why has CRA survived despite the presence of nonbank competitors who can
offer lower prices since they avoid cost shifts inherent in CRA’s low-income
lending requirements? After all, in a number of other regulated industries,
most notably telecommunications and airlines, the entry of unregulated competitors made it difficult to pursue social objectives that require firms to shift
costs from one customer group to another.
In the telecommunications industry, regulators were forced to slash telephone-rate subsidies as competition became more intense and deregulatory
policies were implemented. Kahn (1990, 343) argues that during the 1980s
the prices of long-distance service calling and basic residential service were
brought closer to their respective costs. He provides as evidence an increase
in the local telephone charges component of the Consumer Price Index (CPI)
and a decline in the average price of long-distance calling from December
1983 to December 1989. Local telephone charges rose 19.3 percent in real
terms, while average long-distance charges fell 44.5 percent interstate, and
24.1 percent intrastate, respectively, during the period. Temin (1990, 350)
cites telephone price data from the CPI for 1977–1987 that show a sharp rise in
the ratio of local to interstate telephone rates during the post-AT&T divestiture
period of 1983–1987 and concludes that cross subsidies from long-distance
to local calls were reduced but not eliminated.
In the airline industry, average airfares fell and fare subsidies diminished
when the industry was deregulated in 1978 and prices began to be set in
competitive markets. Prior to deregulation, fares on longer and more heavily traveled routes had been too high relative to costs, while fares on shorter
and less heavily traveled routes had been too low; in effect, heavily traveled
routes subsidized less-traveled routes (Joskow and Rose 1989, 1469). Airfares
were declining prior to deregulation and the entry of new competitors, but the
decline in real airfares was quicker and larger once the industry was deregulated (Winston 1998, 100). Morrison and Winston (1998, 484) estimate that
average airline fares are approximately 33 percent lower in real terms since
deregulation. But declines in airfares at airports in smaller communities—
those designated as small hub or nonhub airports by the Federal Aviation
Administration—were consistently smaller than declines in fares at airports in
larger communities during the post-deregulation (1978–1996) period (Morrison and Winston 1997, 43). Despite continued interest in maintaining low rates

R. L. Lacy and J. R. Walter: Price Theory and the CRA


at smaller community airports, competition today will not allow the subsidies
that could make this possible.
If banking followed the trend in the telecommunications and airline industries, one would expect CRA’s influence on bank lending to have declined, but
it has not. Two factors may help to explain CRA’s resilience. First, banks maintain some competitive advantages compared to their unregulated competitors
despite aggressive competition. These advantages allow banks to shift costs
to certain bank customers and hold at bay nonbanks, which would otherwise
lure bank customers with lower prices and cause CRA’s low-income lending
requirements to collapse. Second, as a practical matter, CRA low-income
lending requirements may not impose large costs. Both factors are further
discussed below.

CRA Costs Can Be Shifted If Banks Have
Competitive Advantages
Among financial institutions, banks are unique in offering transaction deposits
and widespread branch facilities to provide convenient deposit and withdrawal.
While there are nonbank alternatives to transaction deposits—money market
mutual funds for example—for most individuals the alternatives cannot completely substitute for a bank deposit account. According to the 1998 Board
of Governors Survey of Consumer Finances, 90 percent of households have
a bank transactions account, while only 16 percent have any kind of mutual
fund (Kennickell, Starr-McCluer, and Surette 2000, 11). Because nonbank
alternatives offer only imperfect substitutes for bank deposits, banks have
greater leeway to charge higher prices for these accounts without incurring a
loss of customers to nonbank competitors. As a result, some of CRA’s cost of
making unprofitable low-income loans can be shifted to holders of transaction
accounts with little loss of these customers to nonbank competitors. As long
as such shifts are possible, banks have less incentive to lobby for repeal of
CRA’s costs might also be shifted to small-business borrowers. Observers
argue that banks may hold a competitive edge in lending to small businesses
because of long-standing relationships with these borrowers. The special
lending skills that bank loan officers have developed over the years and the
credit information that is obtained through long-standing relationships are
difficult for nonbanks to acquire. While this competitive advantage may erode,
it could be some time before nonbanks are on an equal footing with banks in
the quality of lending services they offer.


Federal Reserve Bank of Richmond Economic Quarterly

Table 1 The Growth of Competition in Banking
Banking industry competition became more intense starting in the late 1970s for two
reasons. First, the banking industry itself became more competitive as entry barriers
were dropped, branching restrictions were removed, and interest rate restrictions fell.
Second, banks faced mounting competition from nonbank competitors.
Barriers to Entry Fall
• After the massive bank failures of the Great Depression, fairly strict requirements
were imposed on the granting of bank charters.
• The Comptroller of the Currency (Comptroller) denied applications for national bank
charters when it determined that existing banks already adequately served markets.
State banking authorities operated in a similar manner.
• In October 1980 the Comptroller ended its policy of assessing the competitiveness
of markets when making charter decisions.
Branching Restrictions Fall
• Federal and state restrictions on banks’ ability to branch were an important feature
of the U.S. banking environment throughout the 20th century. Restrictions protected
existing banks from competition.
• In addition to restrictions on bank branching, the ability of bank holding companies
(BHCs) to operate across state lines was also restricted. The Bank Holding Company
Act of 1956 largely prohibited bank holding companies from owning banks outside
the BHC’s headquarters state, but included a provision allowing ownership of banks
across state lines if legislation in the non-headquarters state specifically provided for
such rights. No states had such legislation.
• Interstate banking restrictions began to fall when, in 1978, Maine became the first
state to pass legislation to allow BHCs headquartered in other states to purchase banks
in its state. Other states followed suit, so that by 1990 all but four states allowed crossborder purchases—though interstate branching remained largely prohibited.
• In the early 1980s states began to remove restrictions on in-state bank branching.
• The Riegle-Neal Interstate Banking Act of 1994 largely eliminated restrictions on
interstate branching.
Interest Rate Ceilings Fall
• The Banking Act of 1933 prohibited the payment of interest on checking accounts and
authorized the Federal Reserve to regulate interest rates on time and savings deposits.
• Interest rate ceilings were initially set well above the interest rates banks were paying.
• Beginning in the mid-1960s the situation changed. Ceilings were set below market
rates beginning in mid-1966 and generally remained below market rates until ceilings
were eliminated in the mid-1980s. Ceilings were viewed at the time as a means of
enhancing the flow of loans to mortgage borrowers. In effect, when market interest
rates rose above the ceilings, depositors cross-subsidized mortgage borrowers.
• Bank depositors responded by moving their funds to money market mutual funds
(MMMFs), and these funds grew rapidly in the 1970s.
• In March 1980 Congress responded by passing the Depository Institutions Deregulation and Monetary Control Act, which phased out all interest ceilings on savings and
time deposits and authorized banks nationwide to pay interest on a new type of checking account, the Negotiable Order of Withdrawal (NOW) account. NOW accounts
had previously only been available in certain states.

R. L. Lacy and J. R. Walter: Price Theory and the CRA


Table 1 The Growth of Competition in Banking (continued)
Increased Competition from Nonbanks
• Companies offering MMMFs compete with banks for consumer and business savings
and checkable deposits. MMMFs gained prominence in the late 1970s. By the end
of 1999, MMMF balances amounted to $1.46 trillion. This compares to $4.54 trillion
in deposits at banks and savings institutions as of the end of 1999.
• Competition for loans increased as well. Large businesses can borrow by issuing
commercial paper, and commercial paper as a source of funds has grown rapidly.
Commercial paper outstanding in 1975 was $48 billion; as of the end of 1999 it
totaled $1.4 trillion. In comparison, all business loans made by banks and savings
institutions summed to $1.0 trillion.
• Nonbank lenders play an important and growing role in serving businesses that are
too small to effectively borrow in the commercial paper market. As of 1993, such
lending accounted for 35 percent of credit extended to small businesses.
• Nonbanks have made important inroads in consumer lending, accounting for 58
percent of such loans in 1998.
Sources: Board of Governors, Federal Reserve Bulletin (2000a), Tables 1.21 and 1.32;
Cole and Wolken (1996); Federal Deposit Insurance Corporation (1999); Gilbert (1986);
Hahn (1983); Kennickell, Starr-McCluer, and Surette (2000); Robertson (1995).

CRA May Impose Relatively Small Costs
Although CRA has undoubtedly resulted in an overall increase in low-income
lending, at many banks the increase may have been relatively small. If CRA
has caused only minor changes in bank behavior, banks are at little disadvantage to nonbank competitors because of the Act.
There is some empirical evidence to support the notion that the CRAinduced increase in low-income lending by banks has been fairly small. Evanoff and Segal (1996, 32) find that during the 1980s, growth of low-income
mortgage lending by banks lagged growth of middle- and high-income lending. According to Litan et al. (2001, 26), from 1993 through 1999, low-income
home purchase loans at institutions subject to CRA rose from 31.5 percent to
35.0 percent of total mortgage loans. Mortgage lending, the focus of CRA
lending, accounts for only 15 percent of the average bank’s assets, so as a
percentage of bank assets the change in the 1990s was fairly small.18 Moreover, a good bit of this increase seems to have been accounted for by factors
outside of pressure brought by CRA regulations, for example by declining
costs of lending to low-income borrowers. Such declines may be the result
of improvements in the quality of information available to lenders on such
18 Since banks often sell a large proportion of their mortgage loans in the secondary market,

the proportion of mortgage loans to total assets tends to understate the importance of earnings
from mortgage lending for bank profitability.


Federal Reserve Bank of Richmond Economic Quarterly

borrowers. Low-income lending by firms not subject to CRA grew even more
rapidly than such lending by banks, suggesting that factors other than CRA
have increased the attractiveness of low-income lending. For these nonbank
institutions, low-income loans grew by 30 percent from 1993 to 1997 (Gunther
2000, 58).
So why haven’t CRA regulations resulted in more low-income lending?
In part because regulators must balance two conflicting goals when enforcing
CRA: expanding low-income lending and assuring bank safety and soundness.
The Act requires that additional loans be made only to the extent “consistent
with safe and sound [bank] operation” (12 U.S.C. 2901). Thus, one would
expect regulators to be reluctant to push banks too far in providing support
for community development. Large increases in low-income lending imply
a substantial reduction in profits or even losses for the bank. Because bank
regulators are not only responsible for encouraging low-income lending but
also for enforcing safety and soundness requirements, they are understandably
loath to take steps that could undermine soundness.19



There is broad support for efforts to revitalize distressed low-income communities. A partnership of private and public interests as represented by CRA
is considered by many an ideal way to accomplish this social goal. While a
direct government transfer program might provide the needed funds, private
organizations bring a business acumen honed from experience operating in a
competitive marketplace. Furthermore, if private firms can contribute to community development, then government budgets are less burdened. But there
can also be serious problems in relying on private efforts mandated by legislation. In the case of CRA, a requirement to expand lending in low-income
communities may in some circumstances distort credit markets: projects for
which costs exceed benefits are undertaken, and projects are rejected when
benefits exceed costs. Further, CRA’s costs may result in banks losing business
to firms that are less efficient at providing deposit and lending services.
While some would like banks to play a greater role in community revitalization, a more aggressive low-income lending policy that further disadvantages banks relative to unregulated competitors would be hard to sustain.
CRA will likely survive in a more competitive economy as a tool to fight discrimination against low-income neighborhoods, but those who expect CRA to
play a growing role in community development funding may be disappointed.

19 Gunther (1999) discusses the conflict between encouraging low-income lending and promot-

ing safety and soundness. He provides evidence, at least for small banks, of the conflict between
enforcement of safety and soundness standards and CRA compliance.

R. L. Lacy and J. R. Walter: Price Theory and the CRA


Board of Governors. 2000a. “Money Stock and Debt Measures.” Federal
Reserve Bulletin 86 (June): Tables 1.21 and 1.32.
. 2000b. “Report on the Performance and Profitability of
CRA-Related Lending.” Report submitted to the Congress pursuant to
Section 713 of the Gramm-Leach-Bliley Act of 1999. Washington:
Board of Governors of the Federal Reserve System, July 17.
. Regulation BB Community Reinvestment. Code of Federal
Regulations, Title 12, c. II, part 228.
California Environmental Protection Agency, Air Resources Board. 2001.
“Amendments to the California Zero Emission Vehicle Program
Regulations: Final Statement of Reasons.” (December).
Canner, Glenn B., and Wayne Passmore. 1995. “Home Purchase Lending in
Low-Income Neighborhoods and to Low-Income Borrowers.” Federal
Reserve Bulletin 81 (February).
Cole, Rebel A., and John D. Wolken. 1996. “Bank and Nonbank
Competition for Small Business Credit: Evidence from the 1987 and
1993 National Surveys of Small Business Finances.” Federal Reserve
Bulletin 82 (November).
Community Reinvestment Act, U.S. Code, vol. 12, c. 30 (1977).
Davern, Michael E., and Patricia J. Fisher. 2001. “Household Net Worth and
Asset Ownership: Household Economic Studies: 1995.” U.S. Census
Bureau, Current Population Reports, Household Economic Studies,
Series P70-71. Washington: U.S. Government Printing Office (February).
Evanoff, Douglas D., and Lewis M. Segal. 1996. “CRA and Fair Lending
Regulations: Resulting Trends in Mortgage Lending.” Federal Reserve
Bank of Chicago Economic Perspectives 20 (November): 19–46.
Fama, Eugene F. 1985. “What’s Different About Banks?” The Journal of
Monetary Economics 15 (January): 29–40.
Federal Deposit Insurance Corporation. 1999. Quarterly Banking Profile.
Washington: Federal Deposit Insurance Corporation (Fourth Quarter).
Federal Financial Institutions Examination Council (FFIEC). 1997.
“Community Reinvestment Act: Examination Procedures for Large
Retail Institutions.” (April).


Federal Reserve Bank of Richmond Economic Quarterly

Gilbert, Alton R. 1986. “Requiem for Regulation Q: What It Did and Why It
Passed Away.” Federal Reserve Bank of St. Louis Review (February):
Gramlich, Edward M. 2001. “Preparing for CRA 2002.” Speech delivered at
the CRA and Fair Lending Colloquium, Boston, Mass. 23 October.
Gunther, Jeffrey W. 1999. “Between a Rock and a Hard Place: The
CRA-Safety and Soundness Pinch.” Federal Reserve Bank of Dallas
Economic and Financial Review 2: 32–41.
. 2000. “Should CRA Stand for Community Redundancy
Act?” Regulation 23 (3): 56–60.
Hahn, Thomas K. 1993. “Commercial Paper” In Instruments of the Money
Market, ed. Timothy Q. Cook and Robert K. LaRoche. Richmond:
Federal Reserve Bank of Richmond.
Joskow, Paul L., and Nancy L. Rose. 1989. “The Effects of Economic
Regulation.” In Handbook of Industrial Organization, vol. II., ed.
Richard Schmalensee and Robert D. Willig. Elsevier Science/
Kahn, Alfred E. 1990. “Deregulation: Looking Backward and Looking
Forward.” Yale Journal on Regulation 7 (Summer): 325–54.
Kennickell, Arthur B., Martha Starr-McCluer, and Brian J. Surette. 2000.
“Recent Changes in U. S. Family Finances: Results from the 1998
Survey of Consumer Finances.” Federal Reserve Bulletin 86 (January):
Lacker, Jeffrey M. 1995. “Neighborhoods and Banking.” Federal Reserve
Bank of Richmond Economic Quarterly 81 (Spring): 13–38.
Litan, Robert E., et al. 2001. “The Community Reinvestment Act After
Financial Modernization: A Final Report.” Study prepared for the U.S.
Department of Treasury (January).
Morrison, Steven A., and Clifford Winston. 1997. “The Fare Skies: Air
Transportation and Middle America.” Brookings Review 15 (Fall):
. 1998. “Regulatory Reform of U.S. Intercity
Transportation.” In Essays in Transportation Economics and Policy: A
Handbook in Honor of John R. Meyer, ed. Jose A. Gomez-Ibanez,
William B. Tye, and Clifford Winston. Washington: Brookings Institute.
Rhoades, Stephen A. 2000. Bank Mergers and Banking Structure in the
United States, 1980–98. Staff Studies 174. Washington: Board of
Governors of the Federal Reserve System.

R. L. Lacy and J. R. Walter: Price Theory and the CRA


Robertson, Ross M. 1995. The Comptroller and Bank Supervision: A
Historical Appraisal. Washington: The Office of the Comptroller of the
Temin, Peter. 1990. “Cross Subsidies in the Telephone Network after
Divestiture.” Journal of Regulatory Economics 2 (December): 349–62.
Winston, Clifford. 1998. “U.S. Industry Adjustment to Economic
Deregulation.” Journal of Economic Perspectives 12 (Summer): 89–110.

German Monetary History
in the Second Half of the
Twentieth Century:
From the Deutsche Mark to
the Euro
Robert L. Hetzel


tarting in January 2002, citizens of the European Monetary Union
(EMU) replaced their national currencies with the Euro, issued by the
European Central Bank (ECB). Europeans created a new pan-European
central bank as a symbol of a future united Europe. However, what historical
process explains the broad monetary policy of the ECB, that is, its objective of
price stability and its strategy for achieving that objective? The short answer is
that its founders designed the ECB to look like the Bundesbank. How then did
the Bundesbank evolve? To answer that question, I survey German monetary
policy in the second half of the twentieth century.
I divide this history into three main sections.1 The first treats the Bretton
Woods system of fixed exchange rates. The second treats the floating exchange
rate period that began in 1973. It chronicles the Bundesbank’s ultimate decision to accord primacy to reducing inflation rather than unemployment. The
last explains how the Bundesbank dealt with the pressures created by movement toward a single European currency.
The evolution of the Bundesbank into an institution now identified as a
modern central bank is fundamental to the article. A modern central bank
This article follows Hetzel (2002), which summarizes German monetary policy in the first half
of the twentieth century. The author gratefully acknowledges helpful comments from Michael
Dotsey, Martin M. Fase, Andreas Hornstein, Thomas Humphrey, Joachim Scheide, and Alex
Wolman. The views expressed in this article are those of the author and not necessarily those
of the Federal Reserve Bank of Richmond or the Federal Reserve System.
1 An examination of the monetary policies that central banks pursue requires a framework
for understanding their choice of objectives and the way that they achieve those objectives. The

Federal Reserve Bank of Richmond Economic Quarterly Volume 88/2 Spring 2002



Federal Reserve Bank of Richmond Economic Quarterly

determines the behavior of the price level exclusively through (indirect) control over the liabilities on its balance sheet (the monetary base). Conversely,
a modern central bank eschews control over the price level through interference with price setting in individual markets. Specifically, modern central
banks avoid controls such as exchange and capital controls, tiered exchange
rates, quantitative credit controls, moral suasion directed at the price setting
of private parties, direct wage and price controls, and so on.
Another theme of the article is how free markets restrict a country’s choice
of monetary arrangements. The system of fixed exchange rates required disruptive changes in the price level. Political pressures to avoid those changes in
turn produced pressures for exchange controls. Ultimately, Germany’s commitment to a free market economy pushed it to reject fixed exchange rates and
adopt floating exchange rates. The Bundesbank came to embody the modern conception of a central bank when, in the 1980s, it demonstrated how to
achieve price stability in a free market economy. That outcome contrasted
starkly with German monetary experience in the first half of the twentieth
century (Hetzel 2002).
To create the Bundesbank that came to epitomize a modern central bank,
Germany had to travel a long road. When West Germany came into existence
in 1949, the intellectual consensus in the West held that the depression arose
out of a grand failure of the free market model (Hetzel 2002). Similarly, very
few economists understood the price level as a monetary phenomenon, that
is, something under the control of the central bank. Instead, professional and
private opinion held that powerful labor unions and large corporations determined prices through their monopoly power. In this intellectual environment
and against the backdrop of staggeringly high unemployment during the depression, there was unanimous political opposition within postwar Germany
to an independent central bank (Buchheim 2002).
At the same time, there was a demand for monetary stability. That demand
arose out of the prewar experience of inflation, first hyperinflation and then the
suppressed inflation of the Third Reich. The memory was still recent of the
1948 currency reform that extinguished most of the savings held by Germans
in currency and bank deposits. In between inflations was the deflation of the
quantity theory provides such a framework. Modern central banks create only paper money, not
wealth. For that reason, ultimately all they can control is the price level—the money price of
goods. The quantity theory explains the behavior of the price level based on the way that central
banks create and destroy money.
A core implication of the quantity theory is that a central bank must choose between two
roles for the price level. With a system of fixed exchange rates, the central bank must allow the
price level to vary to produce the real terms of trade that equilibrates the balance of payments.
That is, the internal price level must vary to price the country’s exports in a way that achieves
balance of its international payments. Alternatively, with a system of floating exchange rates, the
price level varies to endow the nominal money stock with the real purchasing power the public
desires. The rate of inflation then depends upon the trend rate of growth of money the central
bank chooses.

R. L. Hetzel: From the Deutsche Mark to the Euro


depression. Germany accepted the postwar policy consensus that monetary
stability required pegged, infrequently adjusted exchange rates. Policymakers
attributed the depression in part to the competitive devaluations that countries
imposed to stimulate their economies and to the associated financial instability arising from the speculative capital flows those devaluations engendered
(Yeager 1976, 375). In response, the Western world designed the Bretton
Woods system.

In July 1944 at Bretton Woods, the victorious West designed a monetary
system intended to substitute stable exchange rates for the disruptive changes
in exchange rates that characterized the depression. However, in time that
system itself became a source of instability. The system became a dollar
standard in which U.S. monetary policy determined the inflation rates of the
other member countries. Also, the system of pegged but adjustable exchange
rates allowed those rates to move far from equilibrium. The resulting one-way
bets on exchange rate changes recreated the hot money flows that the designers
of the Bretton Woods system had hoped to banish. This section explains how,
over a two-decade period, Germany went from a pegged to a floating exchange
rate regime.

How Did the System Work?
With the end of price controls in June 1948, Germany moved dramatically
toward an internal free market. However, Germany continued to manage its
foreign trade through trade deals arranged bilaterally with foreign governments. In an open economy, managed trade requires significant government
control of economic activity. In order to move toward free trade, Germany
needed to abandon achievement of balance-of-payments equilibrium through
managed trade and restrictions on trade and capital controls.
Once Germany adopted the pegged exchange rates of Bretton Woods, free
trade required it to adjust its internal price level to price its export goods at
competitive international levels. After the war, like other European countries,
Germany experienced chronic trade deficits with the United States. Commentators assumed an indefinite dollar shortage. External trade balance with
unchanged exchange rates would have required deflation by Germany.
The postwar assumption that governments had a responsibility to manage
the economy to prevent high unemployment made such deflation impossible.
A major point of discussion below is how the Bretton Woods system worked in
the postwar period without forcing deflation on countries with trade deficits.


Federal Reserve Bank of Richmond Economic Quarterly

Faced with a requirement to deflate in order to achieve external balance, countries would undoubtedly have resorted to trade restrictions and capital controls.
Such recourse would have made impossible the postwar trade liberalization
that actually occurred.
The Bretton Woods system worked because of the behavior of its two
largest members, the United States and Germany. With its extraordinarily large
gold reserves, the United States could maintain an overvalued exchange rate
and lose gold for a long period without reacting.2 At the same time, Germany
maintained an undervalued exchange rate by recycling its trade surpluses to its
neighbors through capital exports and foreign aid. Not until the early 1970s did
the inherent instability of the Bretton Woods system and conflicting domestic
policies of its member countries cause the pegged exchange rate system to

A New Central Bank
After the war, Germany had no central bank.3 In its Western zones, the offices
of the former Reichsbank assumed some of the responsibilities of a central
bank. However, these banks could not issue currency and therefore constituted
only the empty shell of a central bank. The political objective of preventing
a return to a centrally organized banking system with interlocking links to
large corporations motivated U.S. policy exclusively. The British had a more
realistic attitude. They emphasized that economic integration required the
existence of a note-issuing bank that could assure settlement of transactions
on an economywide basis.
When the American and British occupation zones merged, British pragmatism prevailed. In March 1948, the Allies created the Bank deutscher L¨ nder
(BdL) to oversee the regional offices of the former Bundesbank. It possessed
the power to issue currency. With the replacement of barter with monetary
transactions after the June 1948 currency reform, the BdL became West Germany’s central bank. The BdL possessed a governing structure modeled after
the U.S. Federal Reserve System. The precedent of central bank independence
set by this structure became impossible to reverse when Germany reorganized
its central bank in 1957 to put it on a firm legal footing.
The Allied decree that established the BdL on 1 March 1948 required it “to
stabilize the currency.”4 The Bank emphasized that priority despite the rise in
2 Given domestic price levels, an overvalued exchange rate yields an excess demand for imports relative to exports. Similarly, an undervalued exchange rate yields an excess of exports over
imports. In the first case, a country with a fixed exchange rate loses official reserve balances. In
the second case, it gains reserves.
3 The material in this and the following paragraph is from Buchheim (1999, Sections 3 and
4 Holtfrerich (1999, 318). This paragraph draws on Holtfrerich (1999, Section 4).

R. L. Hetzel: From the Deutsche Mark to the Euro


Figure 1 German Unemployment Rate

Notes: Data are from “Bev¨ lkerung und Erwerbst¨ tigkeit im Deutschen Reich und in der
Bundesrepulik Deutschland” Bundesarbeitsblatt 7–8/1997, pages 110–11, Bundesanstalt
f¨ r Arbeit, Ministerium f¨ r Arbeit und Sozialordnung und eigene Berechnungen.

the unemployment rate from 4.2 percent in 1948 (Holtfrerich 1999, 328) to 11
percent in 1950 (Figure 1).5 From June 1948, the date of the currency reform,
to October 1948, the cost of living rose by 14 percent.6 The BdL responded
strongly to the inflation.7 The price level began to decline in 1949 and by June
1949 had reached its June 1948 level.
5 Giersch et al. (1992, Chapter 3) argue that the rise resulted from an influx of refugees
from former German territories and from increases in worker productivity.
6 The inflation indicated the continued presence of a monetary overhang despite the drastic
reduction in the units of circulating money. The inflation occurred despite the increase in demand
for money implied by the more than 50 percent increase from June through December 1948 in
the Allied index of bizonal industrial production (Buchheim 1999, 96).
7 It did so in a way determined by its quantitative (as opposed to interest rate) procedures
for the control of bank deposits and credit. The need for quantitative procedures came from the
fact that the discount rate of 5 percent set in June 1948 did not bind, as it was 2 percentage
points above market rates (Holtfrerich 1999, Figure 3). Commercial banks possessed large amounts
of excess reserves and did not borrow significantly from the BdL.
In fall 1948, the BdL tightened restrictions on credit extension by banks. The Allies wrote
off 80 percent of the deposits that remained frozen after the June reform but had been designated
for ultimate one-to-one conversion to deutsche marks (DMs). (They amounted to 50 percent of
the former bank deposits.) Government surpluses also reduced bank reserves.
By the end of 1949, reductions in banks’ excess reserves and in the discount rate to 4 percent
put banks into the BdL’s discount window. The resulting change in procedures from quantitative
credit controls to indirect control through the bank rate marked an important step toward market
rather than administrative rationing of credit.


Federal Reserve Bank of Richmond Economic Quarterly

Integrating Germany into the World Economy
The Allies maintained the old dollar-reichsmark exchange rate of 0.30, which
overvalued the deutsche mark (DM). Similar overvaluations of other currencies relative to the dollar resulted in an autarkic system of international trade
after World War II.8 Because their currencies were overvalued, the countries
of Europe managed the trading of their residents so that transactions would
balance bilaterally. By spring 1947, there were 200 bilateral agreements controlling trade in Europe alone. Importers had to obtain licenses, which limited
total imports country by country. Governments made their imports conditional
on another country’s acceptance of their exports because they feared running
short of the dollar reserves needed for essential food and fuel imports.
As a condition for aid, the United States insisted that European countries
replace bilateral trade deals with free trade and multilateral clearing arrangements. Backed by $350 million of Marshall Plan money, Western European
countries agreed in September 1950 to create the European Payments Union
(EPU).9 However, the EPU created no mechanism for eliminating overall payments imbalances.
Very quickly, the arrangements for the EPU came close to collapse because
of German balance-of-payments deficits. With the outbreak of the Korean War
in 1950, Germans, like Americans, tried to buy goods for fear inflation would
resurge (Hetzel 2001; Holtfrerich 1999, 334; Yeager 1976, 413). Germany
came under both foreign and domestic political pressure to control its imports
of raw materials through a system of central administration and to reimpose
price controls. “This was the last time that the essence of the liberal economic
system which West Germany had adopted in mid 1948 was actually put in
jeopardy” (Giersch et al. 1992, 101).
Germany’s Minister of Economics, Ludwig Erhard, refused to abandon
his free market reforms and predicted that his country’s trade deficit would
disappear. Despite opposition from Chancellor Adenauer, the Bundesbank
raised interest rates (Marsh 1992, 152). Erhard’s prediction came true. By
spring 1951, Germany began to run trade surpluses with the EPU. Germany’s
8 This paragraph and the next two summarize material in Yeager (1976, Chapter 21).
9 The multilateral clearing of EPU achieved a great simplification by removing the detailed

government interventions necessary for bilateral clearing. By making the members jointly responsible for the credit of each member, the EPU allowed consolidation of a member’s transactions
into a single overall net claim on the EPU.
Whether overall the Marshall Plan encouraged free trade is unclear. Its dollar payments, most
of which went to Britain, allowed countries to maintain overvalued exchange rates. Governments
possessed an incentive to maintain overvalued exchange rates because they lowered the cost of
food imports. In the post-depression intellectual environment of the time, “elasticity pessimism”
implied that devaluation would work only very slowly to reduce a trade imbalance. The trade
deficits created by overvalued currencies also provided European governments with the economically
perverse but politically attractive incentive to retain restrictions on trade (see Giersch et al. [1992,

R. L. Hetzel: From the Deutsche Mark to the Euro


free market reforms endured. German economic growth in the 1950s earned
the name of Wirtschaftswunder (economic miracle).

Maintaining Balance-of-Payments Equilibrium
The West succeeded in establishing a peaceful Europe after World War II,
where it had failed after World War I. Its success depended in part on the
establishment of a monetary system that did not entail the deflation of the late
1920s and early 1930s. German and U.S. behavior allowed movement toward
free trade with a system of pegged exchange rates while avoiding deflation.10
Immediately after the war, an undervalued dollar created a dollar shortage,
and the United States ran a huge balance of trade surplus. In part, the United
States recycled the resulting reserve inflows through unilateral transfers to the
rest of the world.11 Equally important, in 1949, the United States encouraged
its trading partners to devalue their currencies (revalue the dollar).12 By ceasing to overvalue their currencies, these countries eliminated pressures to either
deflate or resort to protectionism.
After a small deficit in 1951 and surplus in 1952, Germany began to run
persistent current account surpluses (Giersch et al. 1992, Table 28). Left alone,
the surpluses would have widened because of the increasing competitiveness
of German industry. After the war, Western European countries purchased
their capital goods exclusively from the United States. However, in the 1950s,
Germany replaced the United States as the major exporter of capital goods to
European countries (Giersch et al. 1992, 88–89; Yeager 1976, 486). Germany
10 The rejection of isolationism by the United States was also of central importance. U.S. aid
replaced punitive reparations. The resulting cooperation between European countries in the form of
the Organization for European Economic Cooperation (OEEC) facilitated the re-entry of Germany
into the European community. With the establishment of GATT, the United States encouraged an
open, multilateral trading system.
11 The sum of government and private unilateral transfers and capital outflows was $6.8 billion
in 1949. The figure fell to around $4 billion in the mid-1950s, but rose again to almost $7 billion
in 1957 (U.S. Historical Statistics, Part 2, Series U 1-25, “Balance of International Payments: 1790
to 1970”).
12 On 18 September 1949, after a sterling crisis and with U.S. encouragement, Britain devalued the pound by 30.5 percent. Thirty other countries, accounting for approximately two-thirds of
all world trade, followed in devaluing relative to the dollar (Yeager 1976, 444–45). Those devaluations left the dollar somewhat overvalued for most of the 1950s. Giersch et al. (1992, 93) present
a graph of the difference between the free market DM exchange rate on the Z¨ rich market and
the official exchange rate. According to this measure, the DM was about 20 percent overvalued
in early 1950. By the end of 1953, the overvaluation disappeared.
Starting in 1950 and for the remainder of that decade, the United States incurred a deficit
on current account (with the exception of 1956 and 1957, when the United States benefited from
special factors related to the Suez crisis). From 1950 through the end of the Bretton Woods system
in 1973, the United States persistently lost gold. (In 1949, the U.S. gold stock was $24.6 billion.
It declined steadily and reached $10.5 billion in 1972, the last full year of the Bretton Woods
system.) The willingness of the United States to allow reductions in its gold stock allowed other
countries to rebuild their reserves.


Federal Reserve Bank of Richmond Economic Quarterly

Figure 2 German Real GDP Growth

Notes: Data are from Ritschel and Spoerer (1997), Tables A.1, A.2, and A.3.

responded to balance-of-payments surpluses by liberalizing its trade faster
than other countries (Holtfrerich 1999, 331).13
After 1955, liberalization of its capital controls allowed Germany to become a significant exporter of capital (Yeager 1976, 490–96; Giersch et
al. 1992, Table 28). When the EPU disbanded in December 1958, Germany
was granting it significant amounts of credit to cover the deficits of other countries (Yeager 1976, 412). In the 1960s, Germany pursued a large-scale foreign
aid program and capital exports (Holtfrerich 1999, 377, 393).

An Independent Bundesbank
Although the system of pegged exchange rates left the BdL without goal independence, it still had to establish credibility for instrument independence.14
13 Germany became a “pioneer of European [trade] liberalization” in an effort to “become
a respected member of the Western world” (Giersch et al. 1992, 108–09). If Germany enjoyed
monopoly power in its exports of capital goods, reducing tariffs would reduce balance-of-payments
surpluses by weakening its terms of trade.
14 For a bank that needs to establish credibility, an exchange rate peg serves as a clear
objective, the accomplishment of which is evident to everyone. Because of the random variability
that relative price changes impart to short-term movements in the price level and the long lags
with which money affects the price level, central bank control of a price level objective becomes
evident only over a long period of time. There are then reasons for a new central bank to peg
its currency to that of a stable currency.

R. L. Hetzel: From the Deutsche Mark to the Euro


Figure 3 Price Levels

Notes: Annual observations of the natural logarithm with 1951 as the base. The source
of the data is Deutsche Bundesbank, ed. Geld und Bankwesen 1876–1975 (1976).

It had to do so despite the fact that “the Allies had established a totally independent central bank contrary to German wishes” (Buchheim 2002, 10). The
success of the BdL in maintaining the pegged exchange rate in an environment
of falling unemployment, strong real growth, and price stability created public
support for independence (Figures 1, 2, and 3). When in 1956 the Bundesbank increased its discount rate by 1 percentage point, Chancellor Adenauer
challenged the BdL publicly. He stated that “it is the little ones who will suffer
most. . . . [T]he guillotine falls on the man in the street, and that is what grieves
me so much” (Neumann 1999, 291). Public reaction, however, supported the
When in 1957 Germany replaced the Allied promulgation establishing
the BdL with the law establishing the Bundesbank, it had to respect public
support for an independent central bank (Buchheim 2002, 11ff).15 The 1957
Bundesbank Act created the Bundesbank and instructed it to “regulate the
amount of money in circulation and of credit supplied to the economy with
15 The Bundesbank also demonstrated its (instrument) independence later, in 1966 when output
stopped growing and the unemployment rate rose (Figures 2 and 3). Despite a public attack by
the Erhard government, the Bundesbank refused to lower its discount rate. Chancellor Erhard’s
government fell in November 1966. Because of the unsatisfactory behavior of wage growth and the
government budget, the Bundesbank then refused the demands of the new government of Chancellor
Kiesinger to cut interest rates (Holtfrerich 1999, 380–81).


Federal Reserve Bank of Richmond Economic Quarterly

the aim of safeguarding the currency” (Holtfrerich 1999, 318). The written
report of the Chairman of the Committee for Money and Credit of the German
parliament stated:16
The security of the currency. . . is the highest precondition for the retention
of a market economy, and hence in the final analysis that of a free
constitution for society and the state. . . . [T]he note-issuing bank must be
independent of these [political bodies] and subject only to the law.

However, the Act’s mandate for “safeguarding the currency” left unstated
whether the Bundesbank should stabilize the internal or external value of the

U.S. Inflation Destroys Bretton Woods
Germany had strong reasons for maintaining the peg of its currency with the
dollar at a fixed rate. The United States’ technological and manufacturing
supremacy coming out of World War II had made the dollar a symbol of
strength. A stable exchange rate with the dollar gave prestige to the mark.
Later, Germans associated exchange rate stability with the export boom that
powered postwar economic recovery. The export industries that benefited did
not want a revalued mark that would erode their profits.
However, the inflationary monetary policy the United States began in
the mid-1960s forced fundamental change on Germany both in its monetary
arrangements and in its intellectual environment. Maintenance of a fixed
exchange rate required Germany to match U.S. inflation. Initially, Germany
resorted to capital controls in a futile attempt to retain fixed exchange rates
and to avoid imported inflation. Eventually, Germany chose floating exchange
rates. Ten years of intellectual ferment then passed before the Bundesbank
used its newfound freedom to make price stability its overriding objective.
Like other countries in the 1970s, Germany experimented with aggregate
demand policies aimed at controlling inflation. In Germany, the unemployment rate rose far above the levels of the 1960s (Figure 1). Despite this fact,
inflation reached peaks of 7 percent in 1974 and 1982. Two results became
apparent. First, to control inflation, the central bank had to control money
growth. Second, high rates of money growth produced inflation, not low
The Bretton Woods system of pegged exchange rates required Germany to
allow its price level to rise along with that of the United States. Figure 3 shows
that over the period of the Bretton Woods system, U.S. and German price levels
16 Quoted in Stern (1999, 149).

R. L. Hetzel: From the Deutsche Mark to the Euro


Figure 4 Deutsche Mark–Dollar Exchange Rate

Notes: The source of the data is DRI-WEFA.

rose by the same amount.17 The increase in the rate at which prices rose in
the United States, beginning in 1965, ultimately destroyed the Bretton Woods
system. Germany was willing to accept 3 percent “adjustment inflation” as
a cost of the system (Holtfrerich 1999, 383). Moderate inflation was less
costly to the government than a confrontation with the export industries over
a revaluation of the DM. However, in the early 1970s, U.S. inflation pushed
the Bundesbank beyond its limit. Even so, abandonment of an exchange rate
peg and the move to a floating exchange rate came only after bitter debate and
a wave of inflation.
Figures 6 and 7 display graphically the dilemma Germany faced. The high
rates of money growth created by the Bundesbank’s defense of the mark-dollar
17 Given the similar rise in price levels, Germany’s terms of trade relative to the United
States rose by the amount of its revaluations. Germany revalued its currency relative to the dollar
by 5 percent in March 1961. Germany let the mark float upward in September 1969 and then
set a peg at a rate that revalued the mark by 9.3 percent (Figure 4). It again let the mark float
upward in May 1971. In December 1971 as part of the Smithsonian accords, Germany pegged
to the dollar at a rate that revalued the mark by 13.6 percent relative to its prior Bretton Woods
parity. Despite these revaluations, Germany’s terms of trade rose with the breakdown of Bretton
Woods in 1973 (Figure 5). That fact suggests that the capital controls imposed by the United
States in the 1960s kept the mark from becoming more undervalued (the dollar from becoming
more overvalued).


Federal Reserve Bank of Richmond Economic Quarterly

Figure 5 Real Exchange Rate

Notes: The real exchange rate is the ratio of the DM to dollar exchange rate multiplied
by the ratio of the U.S. CPI to the German CPI. 1951 equals 100.

exchange rate in the early 1970s produced high rates of inflation. Figure 6
plots quarterly observations of four-quarter growth rates of money and prices.
Figure 7 fits a step function to annualized quarterly growth rates of money
and prices. It highlights the lagged relationship between changes in money
growth and inflation through arrows that connect steps in money growth to
subsequent steps in inflation (see also Table 1).
In response to recession, in 1970 the Federal Reserve pushed down U.S.
short-term interest rates. Germany, with strong real output growth, maintained
a high level of short-term rates (Figure 2; von Hagen 1999, Figure 4). As a
result, large amounts of short-term capital flowed into Germany, while its
significant export of long-term capital ceased (Giersch et al. 1992, Table 28).
Net capital inflows forced the Bundesbank in May 1971 to buy large amounts
of dollars, including $1 billion in the last 40 minutes of trading on 5 May.
Germany let the mark float and then revalued it as part of the December 1971
Smithsonian agreement.
Opposing camps within the Bundesbank and the government debated
whether to stabilize the external or the internal value of the DM. Johnson
(1998, 70) quotes a Bundesbank official who characterized this debate as “a
Glaubenskrieg (religious war) of Wagnerian proportions.” Karl Klasen became Bundesbank president in January 1970. He favored capital controls to
preserve the foreign exchange value of the mark and credit controls to limit

R. L. Hetzel: From the Deutsche Mark to the Euro


Figure 6 Money Growth and Inflation in Germany

Notes: Quarterly observations of four-quarter growth rates of the CPI and M3. Data
are from DRI-WEFA. Heavy tick marks indicate fourth quarter. The gap in the money
growth series arises from deletion of the observations distorted by the discrete jump in
money that occurred with German unification in 1990.

the credit extension of German banks (Johnson 1998, 70–84; Solomon 1982,
In April 1972, in an arrangement called the Snake, most European countries agreed to limit fluctuations in their exchange rates more strictly than
provided for by the Bretton Woods system. When Britain abandoned the
Snake in June, speculators attacked the mark. Germany responded by adopting capital controls (Yeager 1976, 514).19 Again, as in 1950, Germany had to
18 In 1961, Bundesbank President Blessing had strongly opposed revaluation. It was Ludwig
Erhard, Minister of Economics and author of Germany’s free market reforms, who understood that
revaluation would allow Germany to maintain internal stability without external controls (Holtfrerich
1999, 375).
The essence of such controls is the implicit taxation of international investment flows, that
is, fiscal policy. Fiscal policy is properly the province of the government not the central bank.
Capital and credit controls are ultimately incompatible with central bank independence and with
free markets.
19 The Federal Reserve encouraged the Bundesbank (“Memorandum for the President,” Herbert Stein, 11 July 1972, Stein Box 49, “Nixon Presidential Materials Project” National Archives


Federal Reserve Bank of Richmond Economic Quarterly

Figure 7 Money Growth and Inflation in Germany

Notes: The dots are two-quarter moving averages of the current and preceding quarters’
annualized M3 growth. The solid line is a step function fitted to the quarterly growth
rates. Observations are omitted for M3 from 1990Q3 through 1991Q2. 1990Q3 includes
the discrete jump in M3 of 15 percent. The circles are two-quarter moving averages of
the current and preceding quarters’ annualized quarterly growth of the CPI. The dashed
line is a step function fitted to the quarterly growth rates. The arrows connect the money
step to the subsequent inflation step. Heavy tick marks indicate fourth quarter. See Table

decide whether to adopt monetary arrangements that would require a retreat
from free markets.
Overwhelmed by an inflationary American monetary policy, the Bretton
Woods system collapsed definitively in early March 1973. In February 1973,
the Bundesbank monetized foreign exchange inflows equal to 15 percent of
and Records Administration in College Park, MD): “In his letter of July 8 to you, Arthur Burns
describes the circumstances surrounding the German decision to impose exchange controls rather
than allow the mark to rise relative to the dollar. The German cabinet was apparently motivated by
the desire to minimize political difficulties—including difficulties for us in November. Dr. Burns
congratulated Dr. Klasen, head of the German Central Bank, on his victory in behalf of international monetary stability.”

R. L. Hetzel: From the Deutsche Mark to the Euro


Table 1 Money and Inflation Steps
Time Period


Averages of
M3 Growth

Time Period

Averages of
CPI Inflation





its monetary base (von Hagen 1999, 686). The Bundesbank’s purchase of 2.7
billion dollars on a single day, 1 March (3 percent of its monetary base), forced
a reluctant German government to float (Marsh 1992, 165).20 Henceforth, the
mark floated against the dollar (Figure 4). Although Germany finally chose
free markets and a floating exchange rate, its stubborn defense of pegged
exchange rates allowed the Bretton Woods system to continue long enough to
turn U.S. inflation into a worldwide economic boom and inflation.21
This experience made clear to the Bundesbank that a system of pegged exchange rates had stolen its independence. An exchange rate peg meant that the
Federal Reserve determined German monetary policy. Loss of independence
came from loss of control over money creation. Members of the Bundesbank
Council in their meetings described the Bundesbank as a “self-service store
for central bank money” (von Hagen 1999, 686).

Pegged Exchange Rates and Price Fixing
The Bretton Woods system contained an inherent contradiction: the attempt
to combine stable and adjustable exchange rates. Its architects wanted stable
20 For a fuller account, see Yeager (1976, 515–16) and Solomon (1982, Chapter XIII). Johnson (1998, 82–83) contains an account of the drama. Otmar Emminger, the Bank’s vice-president,
led the free market camp that favored a floating exchange rate and the indirect control of inflation
through monetary control. Helmut Schlesinger, a newly appointed Bundesbank Council member,
supported Emminger. By law, the German government set the external value of the mark. Backed
by an implicit threat of resignation by the Bundesbank’s Directorate, Emminger issued a challenge
to the government’s policy of exchange rate pegging. In the climactic meeting with Willy Brandt,
both Bundesbank President Klasen and Finance Minister Schmidt, who opposed floating, were in
the hospital and Emminger prevailed.
21 The economic boom and strong demand for energy made possible the sharp rise in oil
prices engineered by OPEC in 1973 and 1974. Hetzel (1999, Section 3) argues that excessive
money growth, not oil price shocks, caused the inflation of the 1970s and early 1980s.


Federal Reserve Bank of Richmond Economic Quarterly

exchange rates to eliminate what they saw as the destructive, competitive devaluations of the 1930s. The supposed failure of markets in the depression led
policymakers to believe that floating exchange rates would be destabilizing.
At the same time, its architects wanted adjustable exchange rates so that individual countries could equilibrate their balance of payments without recourse
to deflation or inflation. The resulting system delayed changes to maladjusted
exchange rates until a speculative crisis forced a large change.
Independent central banks that could sterilize international reserve flows
robbed the Bretton Woods system of an automatic market mechanism for
equilibrating the balance of payments. Governments faced an incentive to use
capital controls to reconcile the conflicting demands of external and internal
stability of their currency. Central bank discretion in a system of pegged
exchange rates created all the problems associated with government price
fixing. The pressure on government to impose capital controls politicized
monetary policy. Consider the 1969 revaluation of the DM.
In May 1968, the attempt to defuse riots in France led the French government to grant large wage increases to workers.22 De Gaulle’s refusal to devalue
the franc then set up a one-way bet for speculators: Although Germany would
not devalue the DM, it could well revalue it relative to the franc. In November
1968, large inflows of foreign currency into the Bundesbank prompted Germany to enact a “pseudo-revaluation” with a special tax on exports and tax
allowance on imports. When a leaked Bundesbank memorandum showed that
the Bundesbank had in fact not supported the measure, the head of the SPD
party accused it of having “stabbed its own government in the back.”
The Bundesbank imposed a non-interest-bearing reserve requirement of
100 percent on the growth of foreign deposits. In early May 1969, despite
heavy inflows of foreign exchange, the government, which had authority
over the foreign exchange parity of the DM, again decided against revaluation, “finally, unequivocally and eternally” as a government spokesman
put it (Holtfrerich 1999, 388; Yeager 1976, 509). The issue of whether to
revalue dominated the September 1969 elections. The new government of
Willy Brandt revalued the DM. Speculative capital flows then reversed as
more than DM 20 billion in hot money flowed out of Germany.



The March 1973 float of the mark gave the Bundesbank goal independence.
However, in the 1970s, the Bundesbank remained divided over how to exercise that independence. The 1957 law that created the Bundesbank required
it to “safeguard the value of the currency.” A history of inflation leading to
22 This paragraph and the next draw on Holtfrerich (1999, 384–89).

R. L. Hetzel: From the Deutsche Mark to the Euro


social discontent encouraged Germany to interpret that phrase to mean that
the Bundesbank had to safeguard the internal value of the mark. However,
Germany also had a history of unemployment leading to social discontent.
That history encouraged a monetary policy directed at holding down unemployment. In the 1970s, the Bundesbank had to contend with pressure from
the governments of Willy Brandt and Helmut Schmidt, which made unemployment the top priority.23

The Indecision of the 1970s
After the 1973 float of the mark, the Bundesbank concentrated on lowering
inflation and money growth fell (Figure 6). However, by the end of 1974,
the unemployment rate had risen significantly (Figure 1). Some Bundesbank council members supported the Brandt government’s policy goal of full
employment. Others, including Helmut Schlesinger and Otmar Emminger,
opposed an activist stabilization policy.24
This division appeared in a disagreement over how to use the newly invented money target. In 1973, the Bundesbank had adopted a target for “central
bank money.”25 In December 1974, it began annual public announcements
of the target. The hawks wanted to use money targets to ratchet down money
growth and inflation over time. The doves wanted to use them to reassure
the public that temporarily expansionary monetary policy would not become
inflationary (von Hagen 1999, 425).
In a victory for the doves, toward the end of 1974, the Bundesbank began
to lower interest rates. It also announced a generous target of 8 percent for
money growth. Beginning in 1977, the mark began to appreciate strongly
against the dollar. That appreciation made the Bundesbank unwilling to raise
interest rates (von Hagen 1999, 423; Baltensperger 1992, 442). Monetary
policy remained inflationary throughout the remainder of the 1970s (Figures
6 and 7). The Bundesbank regularly overshot its already expansionary money
targets. In 1978, growth of central bank money exceeded its average target of
8 percent by 3 percentage points.
The decision made in March 1973 to float the mark gave the Bundesbank
goal independence and made possible the path that would lead to the Bundesbank of the 1980s: the independent, monetarist pillar of West German society.
23 Finance minister Karl Schiller resigned in June 1972 over the refusal of the government
to allow a float of the mark against the dollar. Willy Brandt replaced him with Helmut Schmidt,
who later replaced Brandt as Chancellor. Schmidt used the slogan, “Five percent inflation is better
than five percent unemployment” (Johnson 1998, 78).
24 This paragraph and the following one draw on von Hagen (1999, 686) and Johnson (1998,
25 This monetary aggregate was akin to the monetary base adjusted for changes in reserve


Federal Reserve Bank of Richmond Economic Quarterly

However, in the 1970s, other paths beckoned. Germany and other Western
countries undertook the Keynesian experiment of aggregate demand management. The failure of such policies to “buy” low unemployment through high
inflation changed the intellectual and political environment.
In that new environment, countries were willing to assign the control
of inflation to their central banks. Japan had steadily pursued a policy of
a gradual return to price stability since 1974. In 1979, with the election of
Margaret Thatcher as prime minister, Britain began to disinflate. In that same
year, Federal Reserve Chairman Paul Volcker led the United States down the
path of disinflation. The stagflation of the 1970s also made Germany willing
to forswear activist macroeconomic policies and to assign to the Bundesbank
the goal of price stability (Dyson and Featherstone 1999, 747, 752).

The EMS Threatens Bundesbank Independence
In the postwar period, European economic integration became a strategy for
anchoring Germany within a democratic Europe (Arestis, McCauley, and
Sawyer 1999, 1–4). Wim Duisenberg (1999, 4), first head of the European
Central Bank, expressed the goal using Thomas Mann’s phrase: “A European
Germany, rather than a German Europe.”
In December 1969 at a conference at The Hague, governments of the
European Economic Community (EEC) agreed on European monetary union
as a goal. After the breakdown of Bretton Woods in 1973, Germany wanted
to make the Snake into a joint float of European currencies against the dollar.
One reason was that when capital flowed out of the United States in response to
inflationary worries, it primarily went to Germany. The mark then appreciated
not only relative to the dollar but also relative to Germany’s major European
trading partners, and German exports suffered. However, European countries
were not ready to sacrifice independent national monetary policies. In 1977
and 1978, the dollar again depreciated strongly (Figure 4). The corresponding
appreciation of the mark gave Germany an incentive to revive the Snake once
In summer 1978, Germany’s Chancellor Helmut Schmidt and France’s
President Valery Giscard d’Estaing agreed to link their currencies within a
European Monetary System (EMS).26 Their motivations foreshadowed those
that would later lead to the EMU. Germany needed its foreign policy interests
to be identified with an aspiration to build a united Europe. Other countries
would then be more receptive to German diplomatic initiatives, especially
26 The EMS’s “exchange rate mechanism” of fixed parities is referred to as ERM.

R. L. Hetzel: From the Deutsche Mark to the Euro


toward Eastern Europe.27 Schmidt said later that the EMS was part of a
“grand strategy for integrating Europe” (Marsh 1992, 202).
France wanted to build pan-European institutions to ensure that its influence within Europe would remain on par with that of an economically
dominant Germany. France was confident that its civil servants would assure
attention to French interests within such institutions. However, because the
EMS did not create a single central bank for all of Europe, it left unanswered
the roles of the Bundesbank and the Banque de France in maintaining the exchange rate peg. The Bundesbank, however, did not wait for Bonn and Paris
to define its role.

The Bundesbank Defines Itself
The launch of the EMS in March 1979 initiated the Bundesbank policy of
setting money targets to achieve price stability (von Hagen 1999, 433–36).
The EMS had the potential to recreate the experience of Bretton Woods, with
the Bundesbank forced to create money and inflation by supporting a weak
franc instead of a weak dollar. If France wanted the EMS, the Banque de
France would have to subordinate its monetary policy to pegging the franc
to the mark. Karl Otto P¨ hl, Bundesbank president from 1979 to 1991, said,
“The Bundesbank turned the original concept [of the EMS] on its head by
making the strongest currency the yardstick for the system” (Marsh 1992,
Starting in 1979, the Bundesbank began to pursue the goal of price stability. Not only the EMS, but also inflation motivated the change in policy.
From a low of 2.7 percent in 1978, CPI inflation rose to 6.3 percent in 1981
(Figure 6). The Bundesbank lowered its monetary target range from 1979 (6
to 9 percent) through 1985 (3 to 5 percent). Only in one year of this period,
1983, did money growth slightly exceed the target range. Money (M3) growth
fell from 10 percent in the 1970s to 6 percent in the 1980s. By retaining price
stability as its primary objective despite the high unemployment rates of the
1980s, the Bundesbank gained credibility for its policy of price stability.

The Bundesbank as Guarantor of Stability
From modern Germany’s postwar inception in 1949, its polity has been corporatist: the major organized groups—political parties, corporations, and trade
unions—determine the economic and political consensus. By the 1980s, this
27 The most complete account of the history of European monetary union is in Dyson and
Featherstone (1999) and Connolly (1995).


Federal Reserve Bank of Richmond Economic Quarterly

framework for achieving consensus had lost its balance. The Bundesbank
restored balance by joining that framework as the representative of stability.
The German corporate consensus exercised a dramatic effect on the labor
market in the 1950s and 1960s (Giersch et al. 1992, Chapter 4, Section A).
The labor unions kept real wages below their market clearing value to help
Germany regain its prominence as one of the world’s great exporters of manufactured goods. An influx of foreign workers met the resulting labor shortage.
In the early 1950s, the unemployment rate fell slowly because of the problems in absorbing German-speaking immigrants. In the 1960s, however, it
generally remained well below 1.5 percent (Figure 1).
As described above, in the latter part of the 1960s under the Bretton Woods
system, the rise in inflation in the United States rendered the dollar overvalued
(the mark undervalued).28 The profits of German corporations soared because
of the export boom set off by the undervalued mark. At the same time, imported
inflation eroded the real value of nominal wage contracts. CPI inflation, which
had averaged 1.6 percent in 1967, 1968, and 1969, rose steadily until it reached
7 percent in the early 1970s. In response, labor unions broke the corporatist
social contract and launched a wave of wildcat strikes in autumn 1969.29
Pushed by “shop floor radicals,” major unions like chemicals and autos went
on strike in the early 1970s (Johnson 1998, 72–73, 90–95).
The postwar German consensus that had produced the Wirtschaftswunder
continued to erode in the 1970s. In 1978, the printers’ union went on strike
to prevent the introduction of labor-saving technology. Also in the late 1970s,
the Social Democratic Party began to identify with the program of labor. This
identification “endangered the consensual pillars of corporatism” (Giersch et
al. 1992, 214–16).
With the full-employment pledges by the Brandt and Schmidt governments, with labor unions desirous of large wage increases, and with corporations reluctant to allow appreciation of the mark, the Bundesbank became the
member of the corporatist framework defending “stability”—price stability
and balanced budgets. Moreover, when the Bundesbank agreed on the primacy of maintaining the DM’s internal value, it could represent the general
public desire for economic stability. As Otmar Emminger said in his inaugural
speech upon becoming Bundesbank president in 1977, “Monetary stability is
linked up with general social stability—and with political stability” (Marsh
1992, 37).
28 Giersch et al. (1992, 164–66) document the sharp rise in the German trade surplus.
29 Giersch et al. (1992, 154–58) discuss reasons for the “sudden switch of union behavior

from moderation to aggressiveness in the late 1960s and early 1970s.” Holtfrerich (1999, 387)
quotes Schiller (Minister of Economics) as arguing for a DM revaluation in 1969 to eliminate the
threat to “social symmetry” produced by the surge in corporate profits. Rising inflation that eroded
the value of collective wage agreements was a social “bomb.”

R. L. Hetzel: From the Deutsche Mark to the Euro


Marsh (1992, 145) writes, “If it [the Bundesbank] feels inflationary pressures are getting out of hand, the central bank reserves the right to confront the
politicians, industrialists, and trade unionists who exert the main influence on
corporate Germany.” However, the Bundesbank could not confront the unions
directly. To do so would come “dangerously close to compromising their constitutionally guaranteed right to autonomous wage negotiations” (Marsh 1992,
145). The Bundesbank could not on its own conduct a “disguised incomes
policy,” that is, tell the unions what wage increases to negotiate (Johnson 1992,
92, 94). The Bundesbank could, however, use its money targets to make its objective of price stability credible and thus exercise indirect influence over wage
negotiations. Those procedures became “an integral component of German
‘stability culture.’”(Schmid 1996, 42).30
The Bundesbank’s stability policy succeeded because of widespread public support.31 Capie and Woods (2001), in their review of Fifty Years of the
Deutsche Mark, cite Richter (1999, 562):
30 The Bundesbank used a quantity theory framework to derive a target for money growth.
There are many descriptions of these procedures. See, for example, Schlesinger (1984) and Schmid
(1996). The Bundesbank began by setting a target for growth in nominal aggregate demand (nominal output). The target for growth in nominal demand had two components. One was an estimate
of the trend rate of growth of real output. (By using trend growth, the Bundesbank sent the message that monetary policy was not an appropriate instrument for countercyclical aggregate demand
management.) The other component was “unavoidable” inflation.
In 1986, the Bundesbank restored price stability and discarded the idea of “unavoidable inflation” for an inflation target of 2 percent or less. The Bundesbank worked hard to achieve a
consensus for its inflation target from government, labor, and business (Schlesinger 1984). The
Bundesbank then set a money growth target equal to the target for nominal output growth minus estimated growth of monetary velocity. Clarida, Gali, and Gertler (1998) estimate a monetary
policy reaction function for Germany, which they interpret as evidence of inflation targeting.
31 Baltensperger (1999, 462–63) reviews “the often highly controversial and turbulent debate”
over monetary policy in the early 1980s, but concludes, “all in all, Bundesbank policy enjoyed
broad public support.” Why this support despite the sustained rise in the unemployment rate over
the period from 1973 through 1983 (Figure 1)? Connolly (1995, 33, 301) contends that Germans
liked making the EMS into an “undeclared DM-zone” and having other currencies devalue relative
to the DM. More important, the willingness of most major countries to assign to their central
banks primacy for a price stability objective attests to the importance of a fundamental change in
the political and intellectual environment. Kitterer (1999, 192) points to the “failure of demand
management despite a sharp rise in the public sector deficits” and to “the simultaneous sluggish
growth and high inflation of the period.” Countries became disillusioned with the ultimate ineffectiveness and the inflationary consequences of the aggregate demand policies of the 1970s. Within
Germany, the more conservative environment appeared with the program of fiscal consolidation of
the CDU/CSU and FDP coalition of Helmut Kohl that replaced the SPD/FDP coalition of Helmut
Schmidt in October 1982 (Giersch et al. 1992, 192–96).
In the postwar period, the German government had become increasingly leftist and interventionist (Giersch et al. 1992, 125). Although the government had always regulated the economy,
“public subsidization and heavy legal regulation of economic activity. . . took on a new qualitative
dimension in the 1970s and 1980s” (Giersch et al. 1992, 216). For a while in the 1970s, Keynesian aggregate-demand policies had offset the rise in unemployment produced by the resulting
increased inflexibility in labor markets. However, inflation vitiated the effectiveness of those policies. By the 1980s, “public opinion and the vast majority of academic economists. . . supported the
Bundesbank’s line, if only because there seemed to be no realistic alternative” (Giersch et al. 1992,


Federal Reserve Bank of Richmond Economic Quarterly
The German public. . . after having lost their savings twice within 25
years [1923 and 1948], definitely wanted a stable currency. . . . No Bonn
government in its right mind would have. . . put the Bundesbank under

The Bundesbank as European Central Bank
By 1980, almost all the major industrial countries had rejected the activist policies of the 1970s that had led to inflation. France alone retained expansionary
fiscal and monetary policies. In May 1981, Fran¸ ois Mitterand became presc
ident of France. He pursued a program of government intervention in the
economy and expansion of aggregate demand. Capital flowed out of France
and the franc weakened. When the Bundesbank refused to lower interest rates
and inflate to support a weakened franc, France had to devalue. In a series
of devaluations ending March 1983, the value of the French franc fell by 30
percent against the DM.
With inflation and a weak currency, France could not exercise the same
leadership within Europe as Germany.32 After March 1983, France followed
conservative fiscal policies, and the Banque de France gave priority to preserving the parity of the franc with the DM. The combination of Bundesbank
commitment to internal price stability and the peg of the franc to the DM
determined the character of the EMS. Continental Europe gained a central
bank—the Bundesbank—when other countries chose to peg to the DM even
at the expense of their own domestic monetary policies (Baltensperger 1999,
440). A stability-oriented Bundesbank policy of monetary targeting designed
to achieve price stability provided a nominal anchor for the EMS. At the same
time, de facto establishment of the Bundesbank as the European central bank
gave France an incentive to regain influence over monetary policy by creating
a de jure European central bank.

Backsliding with the Louvre Accord
In 1987, the Louvre Accord initiated expansionary and ultimately inflationary
monetary policies among the world’s major industrial countries.33 The Louvre
Accord and simultaneous EMS problems due to weakness in the franc pushed
the Bundesbank away from its stability-oriented policy (Baltensperger 1999,
466–75). In January 1987, labor unrest in France weakened the franc, and the
Bundesbank intervened in the foreign exchange market to prop up the franc.
32 Comments on French politics reflect the themes developed in de Boissieu and Pisani-Ferry
33 The political problem was the rise in U.S. protectionism produced by a large U.S. current
account deficit. The Plaza Accord of September 1985 had attempted to use coordinated intervention

R. L. Hetzel: From the Deutsche Mark to the Euro


Already by 1986 the Bundesbank had significantly overshot its target range for
money. Nevertheless, in early 1987, the Bundesbank reduced the repurchase
rate. France still had to devalue the franc.
The German money stock continued to overshoot its target range through
1987. In response, early in October, the Bundesbank nudged its repurchase
rate up slightly. On Friday 16 October, U.S. Treasury Secretary James Baker
criticized Germany for backing off its pledge to stimulate its economy (Connolly 1995, 40). On Monday 19 October, the U.S. stock market crashed.34
Concerned that the fall in stock prices would depress economic activity, central banks lowered interest rates. In 1988 in Germany, money again
overshot the top of its target range. The CPI, which had remained basically
unchanged from 1985 through 1988, began to rise sharply in 1989. The
Bundesbank’s attempt to resist an appreciation of the mark led to excessive
money growth and inflation just as it had in the early and the late 1970s. The
Bundesbank kept its repurchase rate basically steady in 1986 and lowered it in
1987 despite significant overshoots in its money targets in both years. German
CPI inflation climbed steadily from −0.1 percent in 1986 to 4 percent in 1992
(Figure 6).



German Chancellor Kohl and French Prime Minister Mitterand put the EMU
on the agenda for discussion in 1988 and 1989. “But, once German unification
was set under way, the EMU became endowed with a new significance: as a
test of the political resolve of a unified Germany to bind itself into Europe”
(Dyson and Featherstone 1999, 369). The Bundesbank accepted the decision
in the foreign exchange market to depress the value of the dollar. Policymakers hoped that a
depreciated dollar would lower the U.S. deficit by stimulating U.S exports. However, a steady
depreciation of the dollar (Figure 4), which had already begun in early 1985, failed to lower the
U.S. current account deficit.
The U.S. Treasury then lobbied other countries to expand domestic demand as a way to
reduce their trade surpluses and lower the U.S. trade deficit. The agreement reached at Paris in
February 1987 required Japan and Germany to stimulate their economies to increase imports while
the United States reduced its fiscal deficit. Other countries saw the U.S. fiscal deficit as the cause
of the U.S. current account deficit. The Gramm-Rudman-Hollings agreement to balance the U.S.
budget gave credibility to the U.S. side of the Louvre agreement.
At the time, worldwide stimulus appeared acceptable. The disinflationary monetary policies
of the first half of the 1980s had tamed inflation everywhere. Together with the transitory effect
of the fall in oil prices, virtual price stability had emerged in 1986. Also, the desire of Japan and
Germany to prevent further appreciation of their currencies against the dollar encouraged them to
reduce interest rates. For a discussion of this period, see James (1996, 433–53), Solomon (1999,
21–29), and Volcker and Gyohten (1992, 248–58).
34 Consensus is difficult to establish over the causes of a one-time event like the crash, but
the Louvre Accord must have been one of them. It promised an end to dollar depreciation through
coordinated government policies. The public rift between the U.S. Treasury and the Bundesbank
cast doubt on the viability of that coordination and, therefore, on the value of the dollar. With
the value of the dollar suddenly open to question, foreign investment in the United States became
less attractive. The fall in the U.S. stock market quickly spread to foreign markets.


Federal Reserve Bank of Richmond Economic Quarterly

to replace it with a European central bank. At the same time, the Bundesbank
worked to ensure that its successor would continue its policy of price stability.
That continuity required an explicit mandate for price stability with the force of
a treaty among countries. To bequeath the Bundesbank’s credibility to the new
ECB, the Bundesbank also lobbied for the replication of its own institutional
structure. Finally, the Bundesbank pursued a monetary policy that would
enable the new central bank to begin operation in an environment of price
stability. To do so, it had to undo the post-Louvre inflation of about 5 percent
(Figure 6). That task, which took place in an extremely difficult political
environment, constituted one of the great successes of central banking.

German Reunification
The Berlin Wall fell in November 1989. On 6 February 1990, Chancellor
Kohl decreed that West Germany would exchange the DM for the ostmark at a
ratio of one to one, which compared with a free market exchange rate of 7 to 1
(Marsh 1992, 178). Monetary union on 1 July 1990 came before reunification
on 3 October 1990.
The decision to reunify required immediate monetary reform. East Germany knew that West Germany would exchange ostmarks for DMs not at their
black market rate, but rather at a politically determined rate. Chancellor Kohl,
mindful of the symbolism of the one-for-one exchange in 1948 between the
old reichsmarks and the new DMs, decided to exchange East for West German currency at a one-for-one rate.35 East Germany could therefore obtain
resources just by printing ostmarks.36
35 However, while the 1948 reforms had allowed the market to determine relative prices,
the 1990 reform superseded the market. To avoid an inflow of East German workers into West
Germany, Chancellor Kohl decided to raise real wages in East Germany to move them toward parity
with those of West Germany (Marsh 1992, 183, 187). The West German government converted East
German wage rates one for one into DM wage rates, despite the argument of the Bundesbank for
a two-to-one conversion (Streit 1999, 660–62). Moreover, Germany granted East German workers
large pay raises after unification. “Following monetary union with West Germany in 1990, the real
wage of East German workers rose 83%” (Hunt 2001, 190).
The attempt to move toward real wage parity with West German workers conflicted with
economic reality. East German workers were less productive than West German workers. In 1991,
productivity in East Germany (output per worker) was only 30 percent of the West German level
(Marsh 1992, 171). The economics of that political decision meant that Germany had to raise
the capital-labor ratio in East Germany. However, the effort to make East Germany into a capitalintensive economy like West Germany, when the low productivity of its workers demanded a laborintensive economy, created large-scale unemployment in East Germany.
36 In March 1991, Kohl faulted Bundesbank president P¨ hl publicly for the latter’s criticism of
the terms of monetary union with East Germany. Marsh (1992, 147) commented on the relationship
between the Chancellor and the Bundesbank president: “Bound by the common desire to maintain
confidence in the conduct of state affairs, the two are condemned to harmonious coexistence.”
Lacking a harmonious working relationship with Kohl, P¨ hl resigned on 15 May 1991.

R. L. Hetzel: From the Deutsche Mark to the Euro


Maastricht and the Birth of the Euro
The same political forces that had brought about the EMS now led to the EMU.
In the early 1980s, Germany wanted to undertake its diplomatic initiatives
within the context of building a united, democratic Europe, thereby lessening
other countries’ fears of a Europe dominated by Germany.37 France wanted
to constrain future German influence within Europe through pan-European
organizations (Marsh 1992, 198).
After the EMS crisis of March 1983, France abandoned its policy of aggregate demand expansion to lower unemployment in an attempt to remain
within EMS with no further devaluations. France thus fulfilled a condition
necessary for movement toward monetary union with Germany. These actions
required a prior fundamental decision by President Mitterrand to reshape his
presidency by abandoning the agenda to substitute socialism for a free market
economy and replacing that agenda with construction europ´ enne (Le Monde,
19 May 2001, 1; Dyson and Featherstone 1999, 199).38
In 1986, Europe moved closer to economic integration when it passed
the Single European Act, which required the abolition of all remaining trade
impediments within the European Union. Supporters of the European Union
wanted Europe to possess an economic and political stature comparable to that
of the United States. They talked of Eurosclerosis and believed that European
integration would make European firms competitive with American firms.
France especially wanted to create a European economic bloc that would rival
the United States in economic influence (Boissieu and Pisani-Ferry 1999, 68).
In 1988, Hans-Dietrich Genscher, the German foreign minister; Edouard
Balladur, the French Finance Minister; and Jacques Delors, the president of the
European Commission (EC), began the process that would lead to the creation
of a European central bank. Behind them stood Chancellor Kohl and President
Mitterand. The European Council met in Hanover in June 1988 and set up
the Delors Committee to devise a plan for a single currency. The Committee
delivered its report at the Madrid European Council Meeting in June 1989.39
37 For example, Le Monde (1 May 2001, 8) wrote: “It is for Germany to regain the power
that it lost in 1945. However, because of its past, in order not to scare its partners, it can only
do so through Europe.”
38 “When President Fran¸ ois Mitterrand opted in 1983 for what came to be called the franc
fort policy, he was effectively discarding most of his election commitments to old-fashioned socialism in France” (Financial Times 31 July–1 August 1993, 2). (Note that in French the pronunciation
of the term “franc fort” is the same as “Frankfurt,” the home of the Bundesbank.)
The extraordinary depth of this commitment appeared in France’s willingness to alter fundamentally the dirigiste character of its economy. At the beginning of Mitterrand’s presidency,
nationalized firms made up a quarter of the French economy, with the remainder heavily regulated
or subsidized. France was protectionist and retained capital controls. European integration required
economic and financial liberalization. Within a decade, France moved dramatically away from its
traditional interventionist model of government control (de Boissieu and Pisani-Ferry 1999, 56–57).
39 Connolly (1995) and Vanthoor (1999) provide this chronology. The most detailed account
is in the early chapters of Dyson and Featherstone (1999).


Federal Reserve Bank of Richmond Economic Quarterly

The conviction that the EMU would advance the political unification of Europe
united the participants (Dyson and Featherstone 1999, 273).
The unification of Germany provided the impetus for governments to
make the hard political decisions necessary for the realization of the EMU. A
reunified Germany would not only be stronger economically, but also would
lie in the center of a Europe rejuvenated by the fall of communism. Possessed
of a sense of history, Chancellor Kohl wanted to shape two great historical
movements: German reunification and European federation. With the fall of
the Berlin Wall in November 1989, those movements came together.
Kohl wanted a reunified Germany, but not a Germany that Europe would
fear as a bully. For Kohl, European monetary union was the instrument that
would bring reunification and European federation together.40 He said, “Political union and economic and monetary union are inseparably linked. The
one is the unconditional complement of the other” (Marsh 1992, 211).
At Maastricht in December 1991, members of the European Union signed
the Treaty on European Union, which laid down conditions for membership
in the EMU. Monetary union required that Germany forsake the DM—the
symbol of everything that it had done right after World War II. German public
opinion could accept such a sacrifice only if EMU members adopted the German model of stability symbolized by the Bundesbank (New York Times 15
March 2000, C1). France acceded for the sake of its foreign policy objectives
of countering U.S. worldwide hegemonic influence and German European
hegemonic influence (Dyson and Featherstone 1999, 252; Marsh 1992, 204).
In 1990, Bundesbank president P¨ hl chaired the Committee of EC Central
Bank Governors that drafted the ECB statute, and the Bundesbank prepared
the single draft for negotiations. The Bundesbank worked to preserve its
Stabilitatspolitik.41 It replaced the ambiguous reference in the Bundesbank
Law of 1957 to “safeguarding the currency” with the explicit language, “the
primary objective of the ESCB shall be to maintain price stability.”42 Furthermore, it included the statement that “the ESCB shall act in accordance
40 Dyson and Featherstone (1999, 307–08) write that German unification provided Kohl with
“an opportunity to make an historic contribution to unifying Europe and to reinvent himself as an
historic Chancellor. . . . His European vision was a generational as well as personal matter. It was
bound up with a notion of a special historical responsibility to create a Europe that would never
again experience the horrors of 1914–18 and 1933–45.”
41 Realization of the EMU required convincing Germans that its members would accept the
German stability culture of fiscal responsibility. To be eligible to join the EMU, countries had to
meet convergence criteria that included guidelines on inflation and budget deficits (Connolly 1995,
79; Vanthoor 1999, 131; Dyson and Featherstone 1999, 3). By making the EMU potentially into
a club of outsiders and insiders, Maastricht created an enormous incentive to meet these criteria.
The Bundesbank’s role was then to make certain that governments did not relax the convergence
42 The ESCB is the European System of Central Banks. All members of the EU belong to
it. The ECB is the European Central Bank, and only a subset of EU countries are members.

R. L. Hetzel: From the Deutsche Mark to the Euro


Figure 8 M3 Velocity

Notes: Observations are quarterly values of the logarithm of German velocity. Velocity is the ratio of nominal GDP to M3. Until 1990Q2, figures are for West Germany.
Thereafter, they are for unified Germany. The slope of the trend line fitted from 1969Q1
through 1990Q2 is −1.4. From 1990Q3 through 1998Q4, the slope is −2.3. Heavy tick
marks indicate fourth quarter.

with the principle of an open market economy with free competition, favoring
efficient allocation of resources” (Dyson and Featherstone 1999, 387–89).43

The Breakdown of the EMS
German reunification created economic as well as political shock waves. The
increased demand for capital investment in a united Germany required Germany to change from a capital exporter into a capital importer. To provide the
additional resources needed in Germany, Germans would have to buy more
from foreigners, who would in turn have to buy less from Germans. In 1989,
43 This statement required monetary arrangements compatible with free markets. Such arrangements imply control of the price level through monetary control rather than through government
intervention in the marketplace in the form of incomes policies. Monetary control cannot come
from selected credit control and government credit rationing. Such arrangements require freely
floating exchange rates rather than pegged exchange rates maintained by capital controls.


Federal Reserve Bank of Richmond Economic Quarterly

Germany’s current account surplus was almost 5 percent of GDP. In 1991,
Germany moved to a current account deficit equal to 1 percent of GDP (Whitt
1994, 23). This reversal required that prices in Germany rise more than the
prices of its trading partners. The required relative change in international
price levels could have happened automatically through a revaluation of the
DM, as desired by the Bundesbank; however, France refused the required
mirror devaluation of the franc. Why?
The enhanced role of a reunified Germany in Europe made the French
objective of intertwining the destinies of Germany and a united Europe all the
more pressing. Accordingly, the goal of European monetary union took on
greater urgency. France wanted the same prestige as Germany in negotiations
over the design of the ECB, not to be treated as a weak currency country
(de Boissieu and Pisani-Ferry 1999, 63–64). Monetary union would also be
less palatable in Germany if the most important currency after the DM, the
franc, was seen as weak. Moreover, at considerable political cost, France had
pursued restrictive economic policies since 1983 to prevent franc devaluations,
and it did not want its sacrifice to have been in vain.
The alternative to revaluing the mark was either inflation in Germany or
disinflation by Germany’s EMS partners. In the event, both occurred to some
extent, along with devaluations by some EMS members. The initial unwillingness of other countries to devalue required them to disinflate. Immediately
following reunification, this bitter medicine appeared to work and the currency
parities of the EMS held. However, the EMS came apart in 1992 and 1993.
In 1992, German money growth exceeded the 5 1/2 percent top of the
Bundesbank target range, and in July the Bundesbank raised its discount rate.
The discount rate did not affect the money market rate, which was determined
by the repurchase rate. However, the rise was enough to make the financial
markets doubt the ability of other EMS countries to maintain the punishing
level of interest rates necessary to prevent the depreciation of their currencies.
In September 1992, the markets forced Britain and Italy out of the EMS, and
Spain, Portugal, and Ireland devalued.44
One can understand the unyielding actions of the Bundesbank in this
period in the context of the decision taken at Maastricht the previous December
to proceed with monetary union. In 1992, CPI inflation in Germany was
4 percent (Figure 6). Although the Bundesbank might have to relinquish
control over monetary policy to the new ECB, it was going to do so in a way
that preserved its stability culture. The convergence criteria for joining the
EMU, which enshrined fiscal rectitude and price stability, would mean little
44 Estimates put the Bank of England’s losses from intervention in the foreign exchange
market at approximately six billion dollars. George Soros gained one billion dollars. The figures
are, respectively, from Pringle (1996, 123) and Hanke and Walters (1994, 141).
The name given to the day Britain left the ERM, Black Wednesday, reveals the drama of
the events. The Financial Times (9–10 March 1996) wrote later: “[John] Major [British Prime

R. L. Hetzel: From the Deutsche Mark to the Euro


if Germany itself did not itself enter into monetary union with price stability
(Marsh 1992, 217). In 1992, the Bundesbank could not ignore the significant
overshoot in its money target, regardless of the consequences for the EMS.
During the fall 1992 exchange rate crisis, the franc held. Michel Sapin,
the French finance minister, reminded the markets that speculators had been
beheaded during the French Revolution. More important, the Banque de
France followed a highly restrictive monetary policy. (The degree of restriction
appears in Figure 9 in the sharp increase in the interest rate difference between
France and Germany.) Restrictive monetary policy in France led to weakness
in the French economy in 1993.45
In July 1993, financial markets forced France to allow the franc to depreciate by 6 percent relative to the central rate within the EMS exchange rate
system. Speculators assumed that France and the other countries forced to
devalue would lower interest rates to stimulate their economies. However,
France, Denmark, and Belgium maintained the existing high level of interest
rates. These countries feared that a “competitive devaluation” would split Europe into weak and strong currency blocks and create protectionist pressures
(Financial Times 3 August 1993, 2; Boissieu and Pisani-Ferry 1999, 78). By
maintaining a restrictive monetary policy, France returned the franc in 1996
close to the central EMS rate of 3.35 francs to the mark. The small difference
between French and German interest rates showed that the mark/franc peg had
become credible (Figure 9).
France restored the external value of the franc by maintaining a lower
inflation rate than Germany. Through 1990, France had higher inflation. In
Minister] now fell into a perilous trap. Like Winston Churchill and Harold Wilson before him,
he treated sterling’s exchange rate as a badge of national pride. . . . Interest rates had been raised
to 15 percent. More than $30 billion had been thrown in vain at the markets, all but exhausting
the reserves. The government’s economic policy lay in ruins. Major’s political reputation was
in shreds, his party torn asunder. The failure robbed the Prime Minister of the authority of his
Norman Lamont, Chancellor of the Exchequer during the ERM crisis, blamed the Bundesbank for Britain’s forced departure. He believed that the Bundesbank should have lowered German
interest rates to help defend the pound (Lamont 1999–2000). England’s Prime Minister Major and
Chancellor Lamont were especially unhappy over an interview in the German newspaper Handelsblatt reporting a statement by Helmut Schlesinger, Bundesbank president, that “further devaluations
are not excluded.” That newspaper article precipitated the final crisis (Frowen 1999–2000).
45 In response to this weakness, the French lowered interest rates. Initially, the franc held
against the mark. However, in July, the French economic institute, INSEE, predicted that in 1993
French GDP would fall 0.7 percent (Stanley 1993, 3). Speculators bet that a weak French economy
would force the Banque de France to lower rates further and abandon the franc-mark parity. At
the same time, the Bundesbank, concerned about a pickup in money growth, refused to cut its
discount rate. The unwillingness of the Bundesbank to lower rates to defend the EMS weakened
the credibility of the EMS and speculators attacked the franc.
The New York Times (31 July 1993, 47) wrote: “The attack on the franc came after a
decision Thursday by the Bundesbank to fight German inflation by refraining from a cut in its
discount rate. . . . For France and Mr. Balladur [French Prime Minister], the options look bleak:
either devalue, which would amount to a political humiliation after vows of resistance, or try to
hang on, at potentially devastating further cost to the economy. . . . Already last week, the Bank of
France was obliged to raise overnight rates to 10 percent from 7.75 percent.”


Federal Reserve Bank of Richmond Economic Quarterly

Figure 9 German and French Money Market Rates

Notes: Observations are monthly averages of money market rates on day-to-day money
for Germany and France. Heavy tick marks indicate December observations.

the subsequent four years, it maintained an inflation rate somewhat more than
1 percentage point lower than Germany. By 1995, France had come close to
achieving price stability with an inflation rate of about 1.5 percent. In contrast,
from 1991 to 1994, Germany had an inflation rate of about 3.5 percent.

Was Bundesbank Policy Inflationary or
A characterization of German monetary policy in the early 1990s turns on
whether the Bundesbank pursued an expansionary policy to relieve the exchange rate stresses in the EMS. Alternatively, did it attempt to reestablish
price stability in an effort to help the new ECB begin operation in an environment conducive to the establishment of credibility? The answer depends upon
how one assesses the high rates of money growth in the years 1992, 1993, and
1994 (Figures 6 and 7).
To begin, the inflation of the early 1990s derived from the expansionary
character of earlier monetary policy. On 1 July 1990, monetary union increased the monetary aggregate M3 by 15 percent. However, unified German

R. L. Hetzel: From the Deutsche Mark to the Euro


GDP increased only 8 percent. At the time, the Bundesbank put the resulting monetary overhang at 5 percent (Baltensperger 1999, 479). This increase
in M3 combined with the earlier rapid increases begun in 1987 explains the
inflation rates of the early 1990s.
The Bundesbank did not pursue a stimulative monetary policy as it would
have had to do in order to prevent the depreciation of the franc. In August
1992, in response to strong economic growth, the Bundesbank had raised the
repurchase rate to almost 10 percent. Only when signs of recession began to
appear did it lower rates significantly. When in July 1993 the franc fell to the
level requiring central bank intervention, the Bundesbank did not lower its
rates. Newly appointed Bundesbank president Hans Tietmeyer rejected the
demand “that Germany must immediately abandon its monetary sovereignty”
(Connolly 1995, 324).46
Low real growth accompanied restrictive monetary policy. Annualized
real GDP growth in Germany averaged only 0.8 percent over the years 1992
through 1997. In return, by 1998, German CPI inflation averaged less than 1
percent. French CPI inflation was also less than 1 percent in 1998. The social
cost was high. In France, the unemployment rate was 9 percent in 1990 and
12 percent in 1993. In Germany, the unemployment rate reached 12 percent
in 1998 (Figure 1).
Nevertheless, together, the Bundesbank and the Banque de France bequeathed a priceless gift to the new ECB. They created virtual price stability.
If in 1999 the new ECB had put Europe through a recession in order to lower
inflation, it might not have survived as an institution.



Can the ECB summon wide support for a policy of price stability? An affirmative answer will require that Europeans accept the ECB and its objective

46 Figures 6 and 7 reveal high money growth in the early 1990s. However, that money
growth corresponded to an unusually high demand for money associated with turbulence in currency
markets and the view that the mark was immune from devaluation. It seems likely that the probable
division of Europe into strong and weak currency blocks after the breakup of the ERM led to a
general substitution into the DM. The DM also became a “parallel currency” in some East European
countries (Baltensperger 1999, 482). Figure 8 shows the behavior of M3 velocity. Adjusted for
trend, the demand for M3 (measured by the inverse of velocity) was reasonably stable through
mid-1989. Velocity rose after unification. However, as indicated by the more steeply sloped trend
line, money demand grew unusually fast beginning in 1992.
The high money growth step from 1991Q3 through 1994Q2 does not lead to a subsequent
high inflation step—the only exception in the figure (Figure 7). Instead, inflation fell steadily
beginning in 1993. Although the Bundesbank allowed M3 growth to exceed its target range in
1992 and 1993, the overshoot failed to compensate for the increased money demand. Starting in
1994Q3, the Bundesbank maintained trend M3 growth below 5 percent—a historically low rate.


Federal Reserve Bank of Richmond Economic Quarterly

of price stability as part of a constitutional framework. Such a framework limits government discretion. The clearest example of constitutional limitations
on the sphere of government action is the protection of fundamental human
rights. Freedom of speech is not subject to majority vote as part of the normal
political process.
The alternative to making monetary policy part of the constitutional framework is to make the purchasing power of money subject to ongoing democratic
debate. Just as they do for the farm subsidies of the European Union, constituencies would organize and lobby on behalf of the objective of “low” unemployment rather than price stability. If Europeans come to believe monetary
policy should be part of the democratic process rather than a constitutional
framework, they will see the ECB as undemocratic and elitist. Political attacks will then erode its legitimacy. To maintain its independence and support
for the objective of price stability, the ECB must explain why a constitutional
framework should constrain monetary policy.47
The ECB can point out that in the 1970s, Europe pursued policies meant
to manage real aggregate demand and achieve socially desirable low unemployment rates. Political pressures to reconcile conflicting objectives for low
unemployment and price stability created a demand for incomes policies that
would control the price setting of private markets. Such pressures threatened
both central bank independence and free markets.48
The Maastricht Treaty appropriately specified price stability as the objective for the ECB. However, the EMU’s geographical composition derives from
a political, not an economic, consideration. Europeans desired an EU-wide
symbol that would promote an EU-wide identity. The EMU is a mechanism
to promote the political unification of Europe. But imposition of a common
monetary policy on an economically diverse area entails a cost.
EMU member countries experiencing an adverse change in their terms of
trade with other member countries will experience deflation.49 Increasing the
overall EMU-wide inflation rate to deal with periodic regional deflationary
stresses will not avoid the need for real economic adjustment. Regional deflations will be the price incurred for a common symbol. They will not imply
that the EMU or its objective of price stability is inappropriate.
47 Otto Pfleiderer (quoted in Buchheim 2002, 3), later a member of the BdL, expressed the
alternative view in 1946. “Only the government itself could in a democracy bear responsibility
for the principal measures of monetary policy. . . which is an important part of general economic
policy. . . . There is the danger that the central bank would develop as a kind of second government.
As such it would be able to counteract the economic policy aims of the responsible government.”
48 A commitment to EU-wide free markets is the necessary precondition for the economic
integration of Europe. Modern central banks like the Bundesbank control the price level through
management of their balance sheet. They forswear the direct government interference in markets
entailed by incomes policies.
49 A specific example can clarify the issue. After the creation of the EMU in January 1999,
the Euro depreciated due to outward capital flows, especially to the United States. However, for

R. L. Hetzel: From the Deutsche Mark to the Euro


Europeans now have the opportunity to make the Euro the kind of symbol
that the DM was for Germany in the postwar period.50 Europeans can now
create the right future for the Euro to represent. The Euro can then become
the symbol of a prosperous, democratic, and unified Europe.

Arestis, Philip, Kevin McCauley, and Malcolm Sawyer. 1999. “From
Common Market to EMU: A Historical Perspective of European
Economic and Monetary Integration.” Working Paper 263, the Jerome
Levy Economics Institute.
Baltensperger, Ernst. 1999. “Monetary Policy under Conditions of
Increasing Integration (1979–96).” In Fifty Years of the Deutsche Mark,
ed. Deutsche Bundesbank. Oxford: Oxford University Press.
Bernholz, Peter. 1999. ‘The Bundesbank and the Process of European
Monetary Integration.” In Fifty Years of the Deutsche Mark, ed.
Deutsche Bundesbank. Oxford: Oxford University Press.
Board of Governors of the Federal Reserve System. 1976. Banking and
Monetary Statistics: 1941–1970.
Buchheim, Christoph. 1999. “The Establishment of the Bank Deutscher
L¨ nder and the West German Currency Reform.” In Fifty Years of the
Deutsche Mark, ed. Deutsche Bundesbank. Oxford: Oxford University
. 2002. “How Did the Bundesbank Become Independent?”
Manuscript, University of Mannheim.
countries enjoying capital inflows within the EMU, that depreciation was not appropriate. For
example, Ireland and the Netherlands experienced above-average EMU inflation rates, which kept
their terms of trade with the rest of the world from deteriorating. The analogue to these countries
for the United States was Argentina, which had a monetary union with the United States through
its currency board. Due to capital inflows, the United States experienced an appreciation in its
terms of trade. Because that appreciation was not appropriate for Argentina, which experienced
capital outflows, it had to deflate.
50 Germany endowed its 1948 currency reform with vitality by combining it with a vast
deregulation, namely, an end to price controls. For Europe today, the analogous measure would
be to deregulate its labor markets. Milton Friedman (1953, 187) identifies the major problem that
governments must confront in a world of ongoing change: They must ignore “the urge for security
that is so outstanding a feature of the modern world and that is itself a major source of insecurity
by promoting measures that reduce the adaptability of our economic systems to change without
eliminating the changes themselves.”


Federal Reserve Bank of Richmond Economic Quarterly

Bureau of the Census, U. S. Department of Commerce. 1975. Historical
Statistics of the United States, Colonial Times to 1970, Part 2.
Clarida, Richard, Jordi Gali, and Mark Gertler. 1998. “Monetary Policy
Rules in Practice: Some International Evidence.” European Economic
Review 42 (June): 1033–67.
Connolly, Bernard. 1995. The Rotten Heart of Europe. London: Faber and
de Boissieu, Christian, and Jean Pisani-Ferry. 1999. “The Political Economy
of French Economic Policy in the Perspective of EMU.” In Forging an
Integrated Europe, ed. Barry Eichengreen and Jeffry Frieden. Ann
Arbor: The University of Michigan Press.
Deutsche Bundesbank. 1992. “Commentaries.” Deutsche Bundesbank
Monthly Report 44 (August).
Duisenberg, Willem F. 1999. “The Past and Future of European
Integration—A Central Banker’s View.” The Per Jacobsson Lecture,
Dyson, Kenneth H. F., and Kevin Featherstone. 1999. The Road to
Maastricht: Negotiating Economic and Monetary Union. Oxford:
Oxford University Press.
Financial Times. 1931. “The Nightmare Facing French Policymakers:
Pursuing Rational Economics Means Political Humiliation.” 31 July–
1 August.
. 1993. “Welcome Flexibility, Goodbye Simplicity.” 3
. 1996. “The Countdown to Meltdown.” 9–10 March.
Friedman, Milton. 1953. “The Case for Flexible Exchange Rates.” In Essays
in Positive Economics, ed. Milton Friedman Chicago: The University of
Chicago Press.
Frowen, Stephen. 1999–2000. “Unjustified British Critique of the
Bundesbank.” Central Banking 10: 40–45.
Giersch, Herbert, Karl-Heinz Paqu, and Holger Schmieding. 1992. The
Fading Miracle: Four Decades of Market Economy in Germany.
Cambridge: Cambridge University Press.
Hanke, Steve H., and Sir Alan Walters. 1994. “Easy Money.” Forbes, 31
Hetzel, Robert L. 1999. “Japanese Monetary Policy: A Quantity Theory
Perspective.” Federal Reserve Bank of Richmond Economic Quarterly
85 (Winter): 1–25.

R. L. Hetzel: From the Deutsche Mark to the Euro


. 2002. “German Monetary History in the First Half of the
Twentieth Century.” Federal Reserve Bank of Richmond Economic
Quarterly 88 (Winter): 1–35.
Holtfrerich, Carl-Ludwig. 1999. “Monetary Policy under Fixed Exchange
Rates (1948–70).” In Fifty Years of the Deutsche Mark, ed. Deutsche
Bundesbank. Oxford: Oxford University Press.
Hunt, Jennifer. 2001. “Post-Unification Wage Growth in East Germany.” The
Review of Economics and Statistics 83 (February): 190–95.
James, Harold. 1996. International Monetary Cooperation Since Bretton
Woods. Washington: International Monetary Fund.
Johnson, Peter A. 1998. The Government of Money: Monetarism in Germany
and the United States. Ithaca: Cornell University Press.
Keynes, John M. [1923] 1971. “A Tract on Monetary Reform.” In The
Collected Writings of John Maynard Keynes, vol. 4. London: Macmillan.
Kitterer, Wolfgang. 1999. “Public Finance and the Central Bank.” In Fifty
Years of the Deutsche Mark, ed. Deutsche Bundesbank. Oxford: Oxford
University Press.
Lamont, Norman. 1999–2000. “Black Wednesday—The Controversy
Continues.” Central Banking 10: 65–69.
Le Monde. 2001. “Le 10 Mai 1981, La Gauche et ‘Le Monde.’” 19 May.
Marsh, David. 1992. The Most Powerful Bank: Inside Germany’s
Bundesbank. New York: Random House.
Neumann, Manfred J. M. “Monetary Stability: Threat and Proven
Response.” In Fifty Years of the Deutsche Mark, ed. Deutsche
Bundesbank. Oxford: Oxford University Press.
New York Times. 1993. “Attack on Franc Threatens System.” 31 July.
. 2000. “In the Midst of Upheaval, Yet out of Sight: Horst
Koehler.” 15 March.
Pringle, Robert. 1996. “Black Wednesday Rides Again.” Central Banking 6
(Spring): 121–23.
Richter, Rudolf. 1999. “German Monetary Policy as Reflected in the
Academic Debate.” In Fifty Years of the Deutsche Mark, ed. Deutsche
Bundesbank. Oxford: Oxford University Press.
Schlesinger, Helmut. 1984. Luncheon address at symposium, Federal
Reserve Bank of Kansas City, 1–3 August, in Jackson Hole, Wyoming.
Extract from press articles, Deutsche Bundesbank, 10 August 1984.


Federal Reserve Bank of Richmond Economic Quarterly
. 1992. Speech at the Fiftieth Anniversary of the Bank of
Thailand, 11 December. Reproduced in press articles, Deutsche
Bundesbank, 15 December.

Schmid, Peter. 1996. “Monetary Targeting.” In Monetary Policy in
Transition in East and West: Strategies, Instruments and Transmission
Mechanisms, ed. Austrian National Bank.
Solomon, Robert. 1982. The International Monetary System, 1945–1981.
New York: Harper & Row.
. 1999. Money on the Move: The Revolution in International
Finance since 1980. Princeton, N.J.: Princeton University Press.
Stanley, Stephen. 1993. “ERM Collapse.” Federal Reserve Bank of
Richmond Pre-FOMC Memo, 10 August.
Streit, Manfred E. 1999. “German Monetary Union.” In Fifty Years of the
Deutsche Mark, ed. Deutsche Bundesbank. Oxford: Oxford University
Stern, Klaus. 1999. “The Note-Issuing Bank within the State Structure.” In
Fifty Years of the Deutsche Mark, ed. Deutsche Bundesbank. Oxford:
Oxford University Press.
Vanthoor, Wim F. V. 1999. A Chronological History of the European Union
1946-1998. Cheltenham, UK: Edward Elgar.
Volcker, Paul A. and Toyoo Gyohten. 1992. Changing Fortunes: The World’s
Money and the Threat to American Leadership. New York: Random
von Hagen, J¨ rgen. 1999. “A New Approach to Monetary Policy (1971-8).”
In Fifty Years of the Deutsche Mark, ed. Deutsche Bundesbank. Oxford:
Oxford University Press.
. 1999. “Money Growth Targeting by the Bundesbank.”
Journal of Monetary Economics 43: 681–701.
Wall Street Journal. 1993. “German Stance on Rates Sends ERM to Brink.”
4 August.
Whitt, Joseph A. 1994. “Monetary Union in Europe.” Federal Reserve Bank
of Atlanta Economic Review 1 (January/February): 11–27.
Yeager, Leland B. 1976. International Monetary Relations: Theory, History
and Policy. New York: Harper & Row.

Towards a Theory of
Capacity Utilization:
Shiftwork and the
Workweek of Capital
Andreas Hornstein


mong the large number of economic indicators that provide information on the current or future state of the economy, the index of capacity
utilization (CU) published by the Federal Reserve Board is one of the
more prominent. Low levels of the CU index number tend to be associated
with below-average aggregate activity, and high levels are supposed to indicate inflationary pressures (Corrado and Mattey 1997).1 For the most part,
CU is an empirical concept that is only loosely related to economic theory,
and until recently it has not played an important role in models of the business
cycle. There are various interpretations of what CU means, but in this article I address one particular aspect of CU from the point of view of standard
production theory. I first review some empirical evidence on the workweek
of capital, a measure that makes the concept of CU operational. I then extend standard production theory to incorporate the workweek of capital into
the neoclassical growth model. Finally, I argue that recent attempts to use
variations in the workweek of labor in order to get CU-adjusted measures of
short-term productivity growth are potentially misguided, since the workweek
of labor is not an unbiased measure of the workweek of capital.
The Board’s CU index is defined as the ratio of actual output to potential
output. Potential output reflects “sustainable practical capacity, defined as the
The author would like to thank Mike Dotsey, Margarida Duarte, Tom Humphrey, and Yash
Mehra for helpful comments. The views expressed in this paper are those of the author and
do not necessarily represent those of the Federal Reserve Bank of Richmond or the Federal
Reserve System.
1 Finn (1995) argues that the CU index is not particularly useful in forecasting future inflation


Federal Reserve Bank of Richmond Economic Quarterly Volume 88/2 Spring 2002



Federal Reserve Bank of Richmond Economic Quarterly

greatest level of output each plant in a given industry can maintain within the
framework of a realistic work schedule, taking account of normal downtime
and assuming sufficient availability of inputs to operate machinery and equipment in place” (Corrado and Mattey 1997, 152). The capacity measures are
limited to the manufacturing industries, mining, and electric and gas utilities,
and they are based on the Survey on Plant Capacity, which is produced by
the U.S. Census Bureau in the fourth quarter of each year. Capacity measures
for quarters are obtained by smooth interpolation of the fourth-quarter numbers. Given the smooth interpolation of capacity and the volatility of output,
movements in the CU index are mainly due to movements in actual output.
The standard theory of production views the quantity of output produced
as a function of the quantities of inputs used. How does the concept of CU fit
into this theory? I focus on the “utilization” aspect of CU and disregard the
“capacity” aspect.2 When we measure inputs to production we usually consider the capital stock, total hours worked by production workers/employees,
and the quantities of intermediate inputs (materials and energy) used. We
implicitly assume that the input services provided by capital are proportional
to its accumulated stock. Looking at the production of a plant, we can see
that its output obviously depends on the extent to which it uses its existing
capital stock: how many machines are running and for how long? That is, the
plant can vary the service flow per unit of capital, and this input variation is
not covered in the usual input measures. This opportunity to vary the flow of
services from capital creates a problem for productivity measurement when
we want to attribute movements in output to movements in inputs and changes
in productivity.
The workweek of capital is supposed to capture the service flow of the
capital stock, which is proportional to the average duration of time for which
a unit of capital is operated. The workweek of capital is different from the
workweek of labor, which is the average duration of time a unit of labor
(worker) is employed. To the extent that labor and capital are complementary
inputs (for example, a certain number of workers are needed to operate a
machine), the workweek of capital and the workweek of labor are related,
but they need not be the same. For instance, if a plant is operating multiple
shifts, then the workweek of capital will be a multiple of the workweek of
labor. Furthermore, if the extent to which a plant uses shiftwork changes over
the cycle, the cyclical behavior of the workweeks of capital and labor will be
2 The capacity concept implies that there is a maximal output level in the short run, the
“capacity constraint,” and that this level cannot be exceeded no matter how many variable inputs
are hired. Economic variables will respond differently to changes in the environment, depending on
whether the capacity constraint is binding or not. This behavior can introduce a nonlinearity into
observed economic relations. For a recent model with occasionally binding capacity constraints,
see Hansen and Prescott (2000).

A. Hornstein: Shiftwork and the Workweek of Capital


In the remainder of the article I first review evidence on the workweek
of capital from micro- and macrostudies. I then describe a simple model
with variable employment in late shifts. Finally, I discuss the implications of
variations in shiftwork for the measurement of productivity changes.



The workweek of capital varies widely across industries and across plants. The
average length of the workweek of capital in an industry depends on common
elements of the production processes used by different plants in an industry.
Within an industry, plants deviate from the industry average in response to
variations in demand across plants because production cannot be reallocated
between plants. Depending on the structure of the production process, there
are limits on the extent to which firms can vary the workweek of capital.
Microevidence indicates that plants can adjust the workweek of capital along
a number of margins. Aggregating plant level data to get industry data shows
that the workweek of capital varies substantially over time and indeed is more
volatile than the workweek of labor. It also appears that at the aggregate level
a substantial fraction of the capital workweek’s volatility is due to movements
in the share of the labor force that works on late shifts.
How long do plants operate in a quarter and how do they change the
duration for which they operate? Mattey and Strongin (1997) answer this
question based on plant level data on actual and capacity hours worked per
quarter from the Survey on Plant Capacity. They break down total operating
hours in a quarter as follows:
W eeks Days Shif ts H ours
Quarter W eek
Shif t
Note that as long as plants do not operate 24 hours a day for every day of the
quarter, there is scope for variation in the workweek of capital. Mattey and
Strongin (1997) find that in their sample 35 percent of all plants do not operate every week, 62 percent do not operate every day of the week, 20 percent
have only one shift, and 13 percent do not work overtime. Plants design their
operating margins in order to vary the workweek of capital. For example, in
some industries almost all plants seem to operate essentially every hour of the
quarter. Within other industries the workweek of capital varies substantially
across plants. Mattey and Strongin (1997) classify industries according to operating margins, and they distinguish between “continuous process” industries
and “variable workweek” industries. The two classifications do not exhaust
their sample.
An industry is classified as continuous process if its plants report that at
capacity they essentially operate every hour of the quarter. Continuous process
industries contain one-fourth of all plants in the sample. Even within this class,
not all plants actually operate at full capacity: 20 percent shut down for five


Federal Reserve Bank of Richmond Economic Quarterly

weeks in a quarter, 11 percent work only five days a week, and 9 percent work
only two shifts. For continuous process plants, output variations take place
mainly through the variation in material inputs use and not through variations
in the workweek of capital. Petroleum is an example of a continuous process
An industry is classified as variable workweek if individual plants in that
industry show substantial variation in their actual workweek over time. Most
plants in these industries (85 percent) operate at least 12 weeks per quarter
and five to six days a week (93 percent), so that differences in the workweek
of capital across plants are evident mainly in the number of shifts operated:
27 percent operate one shift, 40 percent operate two shifts, and 32 percent
operate three shifts. A substantial share (16 percent) of these plants also use
overtime as an option.
Obviously the boundaries between continuous process and variable workweek industries are not clear cut. For example, the steel industry appears to
be close to the continuous process ideal: an individual blast furnace is a continuous process technology. Yet Bertin, Bresnahan, and Raff (1996) argue
that for iron production, the relevant unit of observation is a plant that may
operate several blast furnaces, and therefore the average time its furnaces are
operated (workweek) is the more relevant measure. Another example is an
apparel workshop that can vary the average time its sewing machines are used
along two margins: (1) the length of time of time an individual machine is
used, and (2) the number of machines actually operated.
The prototypical example of a variable workweek industry is automobile
production, particularly its assembly plants. Bresnahan and Ramey (1994)
study 50 U.S. automobile plants from 1972–1983. They find that variations
in the workweek of capital are mainly due to weekly shutdowns related to
model changeovers, inventory adjustment, and holidays. Individual plants
only infrequently change the number of shifts they operate. Nevertheless,
since shift changes have a huge impact on output, variations in the number of
shifts operated make a substantial contribution to output volatility (about 25
percent). Hall (2000) in a study of 14 Chrysler plants from 1990–1994 also
finds that most of the variation in an individual plant’s workweek is due to
short-term shutdowns rather than to variations in the number of shifts operated.
Even though at the individual plant level the shift margin does not appear to
be important, it can still be important at the industry level if the relative shares
of plants operating at different shifts change systematically over the cycle.
What are the implications of plant-level patterns for industrywide movements in the workweek of capital? Beaulieu and Mattey (1998) construct
capital workweek series for two-digit Standard Industrial Classification (SIC)
industries in the manufacturing sector for the period 1974–1992. They find
that the capital workweek is both longer and more volatile than the workweek
of labor. For the overall manufacturing sector, the mean workweek of capital

A. Hornstein: Shiftwork and the Workweek of Capital


Figure 1 Workweek of Capital and Workweek of Labor in

Notes: The average workweek of capital for manufacturing is from Beaulieu and Mattey
(1998). The average hours worked are total hours worked divided by the number of
employees for manufacturing only. For both series, the growth rate is calculated from
the fourth quarter in the previous year to the fourth quarter of the current year.

is 97 hours. In Figure 1, I plot the fourth-quarter to fourth-quarter percentage
growth rates of the workweek of capital and average hours worked in manufacturing.3 We can see that the workweek of capital is substantially more volatile
than average hours worked, and although the two series tend to move together,
the fit is not very tight: the correlation coefficient is about 0.6. Beaulieu and
Mattey (1998) also find significant differences in the statistical properties of
the workweek of capital across industries. The mean workweek of capital
ranges from 44 hours in apparel to as high as 156 hours for the continuousprocess-type petroleum refining industry. The volatility of the workweek of
capital is also quite different across industries; for example, the standard deviation of percentage changes in the workweek of capital ranges from a high
of 10.0 in primary metals to a low of 3.0 for chemicals and petroleum.
3 The SPC is undertaken in the fourth quarter of each year, that is, the workweek of capital

refers only to that quarter. For consistency I have used the same procedure for average hours


Federal Reserve Bank of Richmond Economic Quarterly

Unfortunately, the work by Beaulieu and Mattey (1998) is limited to the
manufacturing sector. Shapiro (1996) constructs an alternative measure of
the workweek of capital based on the employment pattern of shiftwork in the
Current Population Survey, and his work covers manufacturing and nonmanufacturing industries of the economy. With respect to the manufacturing sector,
Shapiro (1996) suggests that the capital workweek is shorter (52 hours) and
only half as volatile as Beaulieu and Mattey (1998) argue. Like Beaulieu
and Mattey (1998), Shapiro (1996) also observes substantial variation of the
mean and volatility of the capital workweek across industries in manufacturing. With respect to nonmanufacturing industries, Shapiro (1996) finds
that the capital workweek is only 44 hours, which is close to the workweek
of labor, and that the capital workweek is substantially less volatile than in
Why does the capital workweek tend to be more volatile than the workweek of labor, especially in the manufacturing sector? Shapiro (1996) points
to variations in the extent to which shiftwork is used in production. He finds
that in overall manufacturing about 25 percent of all production workers are
working late shifts.4 Furthermore, in each industry the late-shift share of employment is quite volatile and tends to increase with overall employment. In
particular, Shapiro estimates that a 1 percent increase of employment increases
late-shift employment by 1.5 percent. For nonmanufacturing industries, where
the capital workweek is less volatile, late-shift work is not as prevalent, and
for most service industries the late-shift employment share tends to decline
with overall employment.

I now construct a simple model where capacity utilization is reflected in the
workweek of capital and the workweek of capital is closely related to the
share of late-shift work in total employment. This model builds on work by
Kydland and Prescott (1991), Bils and Cho (1994), and Hall (1996). I argue
that there are systematic differences between the workweek of capital and the
workweek of labor. The model serves as an illustration only, and I do not
provide a complete analysis of all of its properties in this article.

Shiftwork in Production
In the standard model of production, we view output y as a function of the
capital stock k and total hours worked nh:
y = zk α (hn)1−α , 0 < α < 1,


4 The employment share of late-shift work ranges from 4 percent in apparel to 40 percent
in tobacco.

A. Hornstein: Shiftwork and the Workweek of Capital


where n is employment and h is average hours worked per worker (that is,
the workweek of labor), and z denotes productivity.5 We usually assume
that output increases as inputs increase, the marginal product of each input is
positive, and the marginal product of each input is declining. Furthermore,
production is constant-returns-to-scale (CRS) in the capital stock and total
hours worked: if we double the capital stock and total hours worked, then
output doubles. This structure assumes that the contribution of the input
capital is proportional to the stock of capital.
We can allow for variations in the utilization of capital through changes
in the workweek of capital, assuming that labor works a single shift:
y = zk α n1−α h = z (hk)α (hn)1−α .


We now assume that production per unit of time is CRS in the capital stock
and workers employed. Total output is then proportional to the hours capital
and workers are employed. The relevant inputs for this production structure
are the services capital and labor provide, and these services are proportional
to the hours worked. Furthermore, the production structure continues to be
CRS with respect to the capital and labor services employed. This production
structure has been used by Kydland and Prescott (1991), Bils and Cho (1994),
and Hall (1996). Note that for this production structure, the workweeks of
capital and labor are the same.
What happens if we allow for more than one shift and if we can vary the
relative employment levels on the two shifts? I consider the case where the
economy can operate the capital stock with two employment shifts. That is, in
any given period the existing capital stock can be used twice in production. For
this case I assume that production takes place with machines and that machines
and workers are complementary, meaning that if a worker is matched with a
machine containing k units of capital, then output per worker per unit of time
˜ α . Assuming that all machines are the same, the number of machines m
is zk
is limited by the available total capital stock, mk ≤ k. Given the available
machines, the economy can employ n1 ≤ m workers, each working for h1
hours on the first shift, and output from the first shift is k α h1 n1 . The same
machines can be used on the second shift with employment n2 ≤ m and shift
length h2 , with corresponding output k α h2 n2 . The sum of shift lengths is
bounded by the duration of a period, h1 + h2 ≤ h. Total production in a
period is then
y = zk α (n1 h1 + n2 h2 ) .
For an efficient allocation of capital, all capital is used in machines and for
at least one shift all machines are employed. An efficient allocation means that
5 For concreteness I have assumed that the production function is Cobb-Douglas. All the
arguments apply for a general concave constant-returns-to-scale production function.


Federal Reserve Bank of Richmond Economic Quarterly

for a given employment decision, the vector (k, m) maximizes output. Since
the marginal product of capital is positive, and there is no additional cost
of building a machine besides the amount of capital it contains, it is always
efficient to use all available capital, that is, km = k. For the same reason, the
efficient number of machines is equal to the maximum of the two employment
levels. Without loss of generality let n1 ≥ n2 , in which case we need at least
m = n1 machines. It is not efficient to create more machines, because that
would only reduce the capital-labor ratio and therefore reduce output for a
given employment decision. This argument simplifies the representation of
the production function to
y = z (k/n1 )α (n1 h1 + n2 h2 ) , with h1 + h2 ≤ h
= z (uk)α (hn)1−α .


In this economy, the workweek of capital (that is, the average duration a unit
of capital is operated) is u = (n1 h1 + n2 h2 ) /n1 . The workweek of labor (that
is, the average duration a worker is employed) is h = (n1 h1 + n2 h2 )/n, where
n is total employment n = n1 + n2 . Variations in the employment share of the
first shift ω = n1 /n drive a wedge between the workweek of capital and the
workweek of labor:
u = h/ω.


I make one more assumption regarding the dynamic structure of production. Specifically, I assume that it is costly to change the capital-labor ratio
in machines. For simplicity, I choose an extreme case where the capital-labor
ratio is fixed at the beginning of the period. This means that the capital-labor
ratio cannot be adjusted in response to new information on the state of the
economy. The capital-labor ratio, however, can be adjusted at no cost at the
end of a period. This assumption essentially makes employment on the first
shift a predetermined variable. On the other hand, variations of employment
on the second shift allow the economy to respond more flexibly to contemporaneous shocks. This feature of the economy will give rise to variations in the
employment ratio of the second shift, and it will increase the volatility of the
workweek of capital relative to the workweek of labor.
The remainder of the production structure is standard:
y =c+x+g
k = (1 − δ) k + x
ln g /g = ρ g ln (g/g) + εg
ln z = ρ z ln z + εz .
Output can be used for private consumption c, government spending g, and
investment x. Investment augments the capital stock, which depreciates at a
constant rate δ. Primes denote the next period’s values. Government spending
is financed by lump-sum taxation, and log-deviations from its mean g follow an

A. Hornstein: Shiftwork and the Workweek of Capital


AR(1) process with autocorrelation coefficient ρ g . Productivity also follows
an AR(1) process with autocorrelation coefficient ρ z . The disturbance terms
ε g and ε z in the government spending and productivity equations are i.i.d. and
uncorrelated with each other.6
Capital is not always utilized to the fullest extent. In our setup full capital
utilization means that both shifts use all available machines, n1 = n2 , and
machines are continuously operated through the period, h1 + h2 = h. Capital
will not be fully utilized if there are increasing marginal costs to the utilization
of capital. One way to model these costs of capital utilization is to assume that
the rate at which capital depreciates increases as capital utilization increases
(see Greenwood, Hercowitz, and Huffman [1988], Burnside and Eichenbaum
[1996], and Basu and Kimball [1997]). In my setup there is another reason why
capital would not always be operated at full capacity. Since capital utilization
is tied to the use of labor, higher capital utilization requires more employment
at less desirable times and longer work hours. If there is a wage premium
for work at extended hours and that wage premium is sufficiently high, then
capital may never be used at full capacity. I now describe preferences that
give rise to a wage premium for shiftwork and overtime work.

Preferences and the Wage Premium
There is an infinitely-lived representative household with a large number of
members. In any time period the household can send n1 of its members to the
first shift, where they will work h1 hours, and it can send n2 of its members
to the second shift, where they will work h2 hours. The number of employed
household members cannot exceed the total number of household members,
n1 + n2 ≤ n. The household’s expected utility from a random consumption
and labor supply process is



β t log ct −



+ ψ i nit it



with discount rate β ∈ (0, 1), and σ i , ψ i , γ , φ ≥ 0.7
The household is assumed to maximize the expected present value of
utility subject to budget constraints. The household purchases consumption
goods, saves, and supplies its labor. The market wage rates for employment
on the two shifts are given by the wage functions w1t (h1t ) and w2t (h2t ). The
6 My treatment of government spending follows Hall (1996). In the last section I will discuss
some issues in the measurement of productivity within the context of the capacity utilization model.
The methods I discuss there use Instrumental Variable (IV) techniques, that is, they require the use
of a variable which is exogenous to productivity but affects production decisions in the economy.
In my model economy, government spending is such an instrumental variable.
7 These preferences are based on those described by Bils and Cho (1994).


Federal Reserve Bank of Richmond Economic Quarterly

household’s period budget constraint is then
ct + at+1 ≤ Rt at + w1t (h1t ) n1t + w2t (h2t ) n2t ,
where Rt is the return on asset holdings at . Note that the household can choose
only which shifts to work, but not how long to work on each shift. On the
other hand, when firms make their employment decisions, I assume that they
choose employment in each shift and the length of each shift, given the wage
functions they see in the labor market. The cost minimization problem of a
firm is
min w1 (h1 ) n1 + w2 (h2 ) n2
hi ,ni

s.t. y = z (k/n1 )α (n1 h1 + n2 h2 ) ,
0 ≤ n2 ≤ n1 .
We can define a competitive equilibrium for this economy, and it turns out to
be the solution to the planning problem where we choose an allocation that
maximizes the household’s utility subject to the constraint that the allocation
is feasible.8
The optimal employment decision by a household implies that the marginal
disutility of employment at a given shift length is equal to the wage for a shift
of that length:
σi γ
wit (hi ) =
nit + ψ i hit / (1 + φ) ,
where λt is the Lagrange multiplier on the period budget constraint. In an
equilibrium we can take this condition as the definition of the wage function.
We can see that for the same employment levels and shift lengths, work on the
second shift will require a higher wage than work on the first shift if σ 1 ≤ σ 2
or ψ 1 ≤ ψ 2 , which means that work on the first shift creates less disutility
than work on the second shift.

A Quantitative Evaluation of the Model
What does this model say about the behavior of the workweeks of capital
and labor? In particular, does it predict that the workweek of capital is more
volatile than the workweek of labor and that the late-shift employment share
is procyclical? The model is sufficiently complicated that analytical characterizations are not feasible. I therefore parameterize the model, obtain a
numerical solution, and calculate the response of the workweeks of capital
and labor to a productivity shock.
8 In a more general formulation we have shifts of different lengths with wages depending on

the length of the shift. Nevertheless, in an equilibrium, the household and firms would choose to
operate each shift at one particular length. See Hornstein and Prescott (1993).

A. Hornstein: Shiftwork and the Workweek of Capital


The main difference between the model described above and the standard
growth model relates to the description of employment, in particular how
labor market variables enter preferences and production. With respect to
preferences, hours worked and employment are separate arguments in the
utility function; furthermore, there are two types of employment (shifts). For
the specification of the employment and hours elasticities, I follow Bils and
Cho (1994) and select φ = 2 and γ = 1.6.9 I calibrate the scale parameters on
the disutility of work based on assumptions on the relative steady state values
of employment and hours worked for the two shifts.
Total employment is normalized at one, and I assume that 20 percent of
total employment is in the second shift, n1 = 0.8 and n2 = 0.2. As stated
above, Shapiro (1996) reports a mean of 25 percent for late-shift employment
in manufacturing, but manufacturing represents only a subset of the economy,
and late-shift work is less prevalent outside manufacturing. Since I do not
have any information on the relative length of shifts, I simply assume that
in the steady state both shifts are of equal length, which I normalize to one,
h1 = h2 = 1. The two assumptions on relative employment and shift length
imply that the workweek of capital is 25 percent longer than the workweek
of labor. The calibrated capital workweek is substantially shorter than what
Shapiro (1996) reports for the capital workweek in manufacturing based on
Survey on Plant Capacity data, but it is comparable to his capital workweek
estimates based on the Consumer Population Survey. The assumptions on
steady state employment and hours worked and the assumptions on the elasticity parameters together determine the scale coefficients σ i and ψ i in the
utility function.
We can evaluate the parameterization of labor supply based on the implied
shift premium and labor supply elasticities. First, the implied steady state shift
premium of the second shift is quite high, about 70 percent. This premium is
substantially higher than the 10 percent shift premium Bils (1995) argues for
or the 20 percent night-shift premium Shapiro (1996) suggests. Second, from
the equilibrium wage function the implied elasticity of shift employment and
hours worked to changes in the wage rate are10
∂w (h) n
=γ γ
∂n w (h)
n + ψh1+φ / (1 + φ)
ψh1+φ / (1 + φ)
∂w (h) h
= (1 + φ) γ
1/ηh ≡
∂h n
n + ψh1+φ / (1 + φ)
1/ηn ≡

9 I should note that the argument which Bils and Cho (1994) make for these particular pa-

rameter values is not strictly applicable to my model since their interpretation of the employment
and hours worked variables is different from mine.
10 These are wealth compensated supply elasticities since the Lagrange multiplier (marginal
utility of consumption) is taken as constant.


Federal Reserve Bank of Richmond Economic Quarterly

For the given parameterization of preferences these labor supply elasticities
ηn1 = 1.46; ηn2 = 0.94; ηh1 = 0.58; and ηh2 = 1.00.
The labor supply elasticities are relatively low compared to other specifications used in dynamic general equilibrium models, where supply elasticities
around 2 are more common. The labor supply elasticities in this model and
in a standard growth model are, however, not directly comparable for two
reasons. First, standard dynamic general equilibrium models usually do not
distinguish between the labor supply elasticity of employment and the labor
supply elasticity of hours worked per worker as I have done. Second, supply
elasticities are usually stated in terms of percentage response of hours to a
percentage change in the hourly wage rate.12
The remaining parameter values are comparable to those used in other
studies. I choose the time discount factor β = 0.99 such that the annual steady
state interest rate is 4 percent, the depreciation δ = 0.02 such that the annual
depreciation rate is 8 percent, and the capital coefficient α = 1/3 such that the
capital income share is one-third. The parameterization of productivity and
government spending is based on Hall (1996). For the productivity process
I assume that ρ z = 0.97 and the standard deviation of innovations to the
productivity process is 0.006. For the government spending process I assume
that ρ g = 0.97 and the standard deviation of innovations to the productivity
process is 0.009. The steady state share of government spending in output is
g/y = 0.18.
Figure 2 displays the responses of the workweeks of capital and labor to
a percentage point deviation of productivity from the steady state value. We
can see that on impact, the workweek of capital increases more than does the
workweek of labor. Since employment on the first shift is predetermined, the
economy cannot use this margin when it responds to the contemporaneous
productivity shock. The economy is, however, free to increase employment
11 Some algebra shows that the steady state values of these labor supply elasticities are

1/ηn1 = γ [ρ/ (1 − ρ) − αu/ h1 ] / [1 − αu/ h1 ] ,
1/ηn2 = γ ρ/ (1 − ρ) ,
1/ηh1 = 1/ (1 − αu/ h1 ) ,
1/ηh2 = 1.
12 A more appropriate procedure would define the labor supply elasticity with respect to
average wage changes. This does not affect the measure of employment supply elasticity, but it
does change the measure of hours elasticity:

1/ηh ≡

∂ [w (h) / h]
= 1/ηh − 1.
w (h) / h

According to this alternative measure, the hours supply elasticity is actually quite high.

A. Hornstein: Shiftwork and the Workweek of Capital


Figure 2 Workweek of Capital and Workweek of Labor in Model

on the second shift, which increases the employment share of late-shift work.
This increase in the employment share of the second shift in turn amplifies the
response of the workweek of capital.
The model does not generate persistent differences between the workweek
of capital and the workweek of labor. That is to say, only unanticipated
shocks create a divergence between the two workweek definitions. After the
first period, when employment on the first shift can be adjusted again, the
economy attains its target late-shift employment ratio. This feature of the
model is due to the particular assumption on adjustment costs for the capitallabor ratio, namely infinite adjustment costs at the beginning of the period and
zero adjustment costs at the end of the period. Alternatively, we could assume
that any changes in the production structure, specifically the capital content
of machines, involve some resource costs (see, for example, Bils and Cho
[1994]). With this alternative assumption, there will be persistent deviations
between the workweek of capital and the workweek of labor in response to a
productivity shock.
Another way to evaluate the role of shiftwork in the model is to compare
its business cycle properties to those of the U.S. economy and other models
without shiftwork. The business cycle properties are defined with respect
to Hodrick-Prescott filtered time series of output, consumption, investment,
capital stock, total hours worked, employment, and the workweek of labor
(see columns 1 and 5 of Table 1). For our purposes it is of interest to note


Federal Reserve Bank of Richmond Economic Quarterly

Table 1 Business Cycle Properties of U.S. Data and Model Economies
Percentage Standard

Capital Stock
Total Hours
of Labor
of Capital

Correlation with Output























Notes: U.S. Data are from Bils and Cho (1994) and they cover the time period 1955:III–
1984:I. Model I is the one-shift model with employment not predetermined; Model II is
the one-shift model with employment predetermined; Model III is the two-shift model
with employment of the first shift predetermined. All time series are detrended with the
Hodrick-Prescott filter. The model statistics are based on 100 simulations where each
simulation consists of 30 years of quarterly observations.

that employment is more volatile than the workweek of labor (average hours
worked), and employment is slightly more correlated with output than is the
workweek of labor. I do not have quarterly observations on the workweek
of capital, but in the previous section I note that for annual growth rates, the
workweek of capital is more volatile than the workweek of labor and that the
workweeks of capital and labor are only weakly correlated.
To evaluate the contribution of shiftwork as modeled in this article, I
consider three models. The first and second models assume that there is
a distinction between employment and the workweek of labor, but that all
employment is in one shift only. Preferences and production are as previously
described, with the restriction that n2 = h2 = 0. For Model I, employment
and the workweek of labor are determined at the beginning of the period,
after the productivity and government spending shock have been observed.
For Model II, employment is determined before observations on the current
productivity and government spending shock are available. Model I is similar
to Kydland and Prescott (1991) and Bils and Cho (1994), whereas Model II is
similar to Burnside, Eichenbaum, and Rebelo (1993) and Hall (1996) in that
part of the employment decision is predetermined. Finally, Model III is the
economy with shiftwork as described above.

A. Hornstein: Shiftwork and the Workweek of Capital


Table 2 The Workweeks of Capital and Labor, Annual Growth Rates
Std. Dev.
Model III



Std. Dev.


Corr. ( u, h)

Notes: For U.S. data, see Figure 1.

For each model I generate 100 random samples, each with 30 years of quarterly observations. The artificial time series are detrended with the HodrickPrescott filter, and I calculate the average standard deviations and correlations
with output of the detrended series. The results are listed in Table 1. We can
see that given the exogenous disturbances, the model economies are not as
volatile as the U.S. economy. The model economies capture the fact that investment is more volatile than consumption, but consumption tends to be too
smooth. The model economies also capture the fact that employment is more
volatile than the workweek of labor, but overall labor is too smooth relative
to the U.S. economy. Concerning the comovement with output, we see that
in the models employment is more closely correlated with output than it is in
the U.S. economy.
In Model III, the workweek of capital is indeed more volatile than the
workweek of labor for detrended quarterly data. Since quarterly U.S. data on
the workweek of capital are not available, I have calculated the standard deviations and correlations for the annual percentage growth rates of the workweeks
of capital and labor (see Table 2). We can see that in the U.S. manufacturing
sector, the workweek of capital is relatively more volatile than the workweek
of labor. Furthermore, the relationship between the workweek of capital and
the workweek of labor is much tighter in the model than in the data. While
the model captures the qualitative features of the workweek of capital, it does
not come close yet to replicating its quantitative properties.



I now study the implications of variable shiftwork for the measurement of
productivity changes. Since productivity change is defined as output changes
that cannot be attributed to input changes, unobserved input movements obscure our measures of productivity change. Changes in capital services, such
as changes in the workweek of capital, represent important input movements
that are not reflected in our standard measures of inputs. Recently, Basu and
Kimball (1997) have argued that unobserved variation in the utilization of


Federal Reserve Bank of Richmond Economic Quarterly

inputs is related to the observed variation in the workweek of labor.13 They
use this relationship to obtain a utilization-corrected measure of productivity
change. There are two potential problems with the approach of Basu and Kimball (1997). First, their procedure requires that the workweek of capital be
strictly proportional to the workweek of labor. If this is not true, as suggested
by the available evidence and theory on the workweek of capital, then their
procedure does not necessarily generate unbiased estimates of the volatility
of productivity change. Second, even if the estimates of the volatility of productivity are unbiased, they may not be precise because the estimates rely on
instrumental variables that may be quite poor.
Consider the standard production function (1), which takes as inputs the
capital stock and total hours worked. We can measure productivity growth
through the Solow residual, which defines productivity growth as output
growth less input growth weighted by the output elasticities of inputs:
ˆ ˆ
zm ≡ y − α k − (1 − α) h + n = z.
The Solow residual is an operational concept because in a competitive equilibrium with CRS production, we can identify the elasticities with the factor
income shares of inputs. Consider now the production function (2) with a
variable workweek of capital but only one shift. If we continue to assume
that the relevant inputs are the stock of capital and total hours worked, then
measured productivity growth no longer reflects true productivity growth:
zm = z + α h.
However, a simple correction of the Solow residual, made by subtracting the
growth rate of average hours worked weighted by the factor income share of
capital, retrieves the true productivity change:
ˆ ˆ
zm − α h = z.
Empirically, this simple correction does not deliver measures of true
productivity change. Suppose you have variables—call them instrumental
variables—that on a priori grounds are considered to be independent of true
productivity change. On these grounds, the instrumental variables should be
uncorrelated with the Solow residual corrected for average hours worked. In
empirical applications, however, it turns out that the Solow residual corrected
for average hours worked remains correlated with these instrumental variables.
However, in an instrumental variables regression of the measured Solow residual on average hours growth, Basu and Kimball (1997) find that the coefficient
on average hours growth is around one, larger than the factor income share
of capital, which is substantially less than one. They argue that the relatively
large coefficient on average hours worked reflects other unobserved input utilization, which is strictly proportional to average hours worked. Furthermore,
13 See also Basu, Fernald, and Kimball (2000).

A. Hornstein: Shiftwork and the Workweek of Capital


they argue that once they correct the Solow residual for movements related
to movements in average hours worked, they can recover exogenous movements in productivity. By contrast, I argue here that this contention is not true
when the workweek of capital and the workweek of labor are not perfectly
Consider the production function (3) of the two-shift model described
above. With this production structure, measured productivity growth based
on changes in the capital stock and total hours worked is
zm = z + α u.


Again, the true productivity disturbance can be recovered by correcting for
the workweek of capital. Shapiro (1996) argues that industry Solow residuals
that are corrected for the workweek of capital are essentially uncorrelated with
instrumental variables. A problem with this approach is that only a limited
number of observations on the workweek of capital are available. Shapiro
(1996) uses the workweek of capital numbers constructed by Beaulieu and
Mattey (1998), and this sample is limited to the years 1974–1992. Given
the limited availability of direct observations on the workweek of capital, the
argument of Basu and Kimball (1997) for the use of average hours worked
as a proxy for different forms of capacity variation is attractive. Especially
so since, as they argue, average hours worked not only covers variations in
the workweek of capital, but also variations in capital utilization that are not
related to corresponding changes in the worktime of labor.
Basu and Kimball (1997) suggest estimating the regression equation
zm = b h + e


using instrumental variable techniques.14 Let q denote an instrumental variable that is uncorrelated with the true productivity shock, then the two-stage
instrumental variable estimator of b is

E zm q
ˆ ˆ
E hq


E q αu + z
ˆ ˆ ˆ
E qh

given the true relationship (8). If we suppose for the moment that changes in
the workweek of capital are proportional to changes in the workweek of labor,
u = µh, then the estimator simplifies to
b = αµ.
14 Basu and Kimball’s (1997) approach is actually somewhat more complicated in that they
consider the possibility of noncompetitive behavior. They estimate the equation

y = γ m + bh + e,
where γ is the average markup of price over marginal cost. I assume competitive behavior, that
is, γ ≡ 1, and only the coefficient b must be estimated.


Federal Reserve Bank of Richmond Economic Quarterly

Figure 3 Productivity Volatility Implied by Workweek Volatility

Notice that the estimator does not have a structural interpretation, since µ
reflects a relation between two endogenous variables that depends on elements
of a fully specified equilibrium model. Nevertheless, correcting the Solow
residual recovers the true changes in productivity
ˆ ˆ
¯ˆ ˆ
zm − bh = αµh + z − bh = z.
The only problem with this approach is that the above model of the workweek of capital predicts that for a reasonable specification of the production
structure, the changes in the workweek of capital are not strictly proportional
to changes in the workweek of labor. Furthermore, in my review of the empirical evidence on the workweek of capital, I have shown that the relation
between the workweek of capital and the workweek of labor is not very tight;
the correlation coefficient between their respective percentage changes is only
If the workweek of capital is not tightly related to the workweek of labor,
then the average hours-corrected estimates of the volatility of productivity
disturbances obtained by Basu and Kimball (1997) are not unbiased. Suppose
that the relation between the workweek of capital and the workweek of labor is
ˆ ˆ
u = µh + v, where v is an endogenous movement in the workweek of capital
that is orthogonal to the workweek of labor. Since v is endogenous, we would
expect it to be correlated with the instrumental variable q, E v q = 0. The

A. Hornstein: Shiftwork and the Workweek of Capital


estimated parameter and the hours-corrected Solow residual are then
b = αµ +

E vq
E hq

¯ˆ ˆ
and zm − bh = z + v − bh .
ˆ ¯ˆ

Notice that the hours-corrected Solow residual no longer provides an estimate
of true productivity movements.
Can we say anything about the volatility of the true productivity shocks
conditional on what we know about the workweek of capital? From equation
(8) we can write the variance of the Solow residual σ 2m as a function of the
variance of the true productivity shock σ 2 , the variance of the workweek of
capital σ 2 , and the correlation of true productivity shocks and the workweek
of capital ρ zu . This expression defines an implicit equation in the volatility of
the true productivity shock conditional on the volatility of the measured Solow
residual, the volatility of the workweek of capital, and an assumption on the
correlation coefficient between the true productivity shock and the workweek
of capital:
0 = σ 2 + 2αρ zu σ u σ z + α 2 σ 2 − σ 2m .
Consider now the manufacturing sector. From Beaulieu and Mattey
(1998), the volatility of the workweek of capital for the time period 1974–1992
is 2.9 percent, and from Basu, Fernald, and Kimball (1999), the volatility of
the Solow residual is σ zm = 3.1 percent. Because Basu, Fernald, and Kimball
(1999) estimate the Solow residual from gross output data, and intermediate
inputs make up a substantial share of total payments to inputs, I choose a
capital coefficient α = (1/3) (1/2). In Figure 3, I plot the implied volatility
of the true productivity shock for values of the correlation coefficient between
negative one and positive one. The two horizontal lines indicate the volatility
of the unadjusted Solow residual and the hours-worked-adjusted Solow residual from Basu, Fernald, and Kimball (1999) for the manufacturing sector. We
can see that the workweek-of-labor-corrected Solow residual underestimates
(overestimates) the true volatility of the productivity shocks when the workweek of capital and the true productivity disturbance are weakly (strongly)
correlated. The critical value for the correlation coefficient is 0.6. Since we
expect a relatively strong positive correlation between the capital workweek
and productivity disturbances, the actual bias might not be very large.
Suppose that Basu and Kimball’s (1997) estimates of the volatility of production are unbiased. Can we say anything about how precise these estimates
are? This question is relevant since the estimate of the coefficient b in equation
(9) and the estimated productivity change zm − bh are based on instrumental
variables that are quite poor and the sample size is quite small (only 40 years).
We can evaluate the uncertainty surrounding the estimates using the workweek of capital model described above. This model captures the qualitative
features that the workweek of capital is more volatile than the workweek of


Federal Reserve Bank of Richmond Economic Quarterly

Table 3 Ratio of Estimated to True Volatility of Productivity Growth
Sample Size
40 years
200 years
400 years


Std. Dev.



labor and that the two workweeks are not perfectly correlated. In order to
evaluate the uncertainty about the estimates, I generate 1,000 samples of 40
years of quarterly observations for the model. For each sample I construct
annual data from the quarterly data and then estimate equation (9) with the
annual data. In the model, government spending is exogenous and affects
other endogenous variables, that is, it can be used as an instrumental variable.
For the estimation of equation (9), I use contemporaneous and lagged growth
of annual government spending as instruments. I then use equation (9) to
calculate the volatility of estimated productivity growth rates. The model tells
me the volatility of the two productivity growth rates. In Table 3, I display the
means and standard deviations of the ratio of estimated to true productivity
volatility across the samples. We can see that for a small sample of 40 years, on
average we underestimate the true productivity volatility. More important, the
standard deviation on the estimates is very large: the two-standard deviation
error band for the ratio reaches from 0.18 to 1.58.
Given the small sample of available data, the estimate of true productivity
volatility is very imprecise. Basu and Kimball (1997) increase the sample
size by using industry data rather than aggregate data. They assume that in
equation (9), the coefficient b is the same across industries, and then they
pool industry data. For manufacturing they pool durable-goods-producing
industries (ten industries) and nondurable-goods-producing industries (seven
industries). I replicate the industry pooling approach by assuming that each
industry represents another 40 years of observations. With more observations
the estimate of productivity volatility appears to be unbiased (see Table 3). This
result should not be too surprising since the differences between the workweek
of capital and the workweek of labor are not quantitatively important in the
model, as opposed to the data. We might therefore expect that the use of the
workweek of labor rather than the workweek of capital in equation (9) would
not generate a large bias in the estimate of the volatility of productivity. Even
with a larger data set, however, the estimate of productivity volatility is not
very precise: the two-standard deviation error band for the ratio still ranges
from 0.79 to 1.27.

A. Hornstein: Shiftwork and the Workweek of Capital



Variations in capital utilization as measured by the workweek of capital are
large; indeed, the workweek of capital is substantially more volatile than the
workweek of labor. This observation suggests that for output fluctuations,
short-term variations in the utilization of the capital input are at least as important as short-term variations in the utilization of the labor input. Yet, official
statistics are collected only for variations in the workweek of labor, not for
the workweek of capital. Improved measurement of the workweek of capital
is clearly called for (Shapiro 1996). Improved data would allow for a better
assessment of the role of productivity disturbances.

Basu, Susanto, and Miles S. Kimball. 1997. “Cyclical Productivity with
Unobserved Input Variation.” NBER Working Paper 5915.
, John Fernald, and Miles S. Kimball. 1999. “Are
Technology Improvements Contractionary?” Manuscript, University of
Beaulieu, J. Joseph, and Joe Mattey. 1998. “The Workweek of Capital and
Capital Utilization in Manufacturing.” Journal of Productivity Analysis
10 (October): 199–223.
Bertin, Amy L., Timothy F. Bresnahan, and Daniel M. G. Raff. 1996.
“Localized Competition and the Aggregation of Plant-Level Increasing
Returns: Blast Furnaces, 1929–1935.” Journal of Political Economy 104
(April): 241–66.
Bils, Mark. 1995. “Measuring Returns to Scale from Shift Practices in
Manufacturing.” Manuscript, University of Rochester.
, and Jang-Ok Cho. 1994. “Cyclical Factor Utilization.”
Journal of Monetary Economics 33 (April): 319–54.
Bresnahan, Timothy F., and Valerie A. Ramey. 1994. “Output Fluctuations at
the Plant Level.” Quarterly Journal of Economics 109 (August):
Burnside, Craig, and Martin Eichenbaum. 1996. “Factor-Hoarding and the
Propagation of Business-Cycle Shocks.” American Economic Review 86
(December): 1154–74.
, and Sergio Rebelo. 1993. “Labor Hoarding and the
Business Cycle.” Journal of Political Economy 101 (April): 245–73.


Federal Reserve Bank of Richmond Economic Quarterly

Corrado, Carrol, and Joe Mattey. 1997. “Capacity Utilization.” Journal of
Economic Perspectives 11 (Winter): 151–67.
Finn, Mary G. 1995. “Is ‘High’ Capacity Utilization Inflationary?” Federal
Reserve Bank of Richmond Economic Quarterly 81 (Winter): 1–16.
Greenwood, Jeremy, Zvi Hercowitz, and Gregory W. Huffman. 1988.
“Investment, Capacity Utilization, and the Real Business Cycle.”
American Economic Review 78 (June): 402–17.
Hall, George J. 1996. “Overtime, Effort, and the Propagation of Business
Cycle Shocks.” Journal of Monetary Economics 38 (August): 139–60.
. 2000. “Non-convex Costs and Capital Utilization: A Study
of Production Scheduling at Automobile Assembly Plants.” Journal of
Monetary Economics 45 (June): 681–716.
Hornstein, Andreas, and Edward C. Prescott. 1993 [1990]. “The Firm and
the Plant in General Equilibrium.” In General Equilibrium, Growth, and
Trade. Volume 2. The Legacy of Lionel McKenzie, ed. Robert Becker
et al. San Diego, CA: Academic Press: 393–410.
Kydland, Finn E., and Edward C. Prescott. 1991. “Hours and Employment
Variation in Business Cycle Theory.” Economic Theory 1 (January):
Mattey, Joe, and Steve Strongin. 1997. “Factor Utilization and Margins for
Adjusting Output: Evidence from Manufacturing Plants.” Federal
Reserve Bank of San Francisco Economic Review 2: 3–17.
Shapiro, Matthew D. 1996. “Macroeconomic Implications of Variation in the
Workweek of Capital.” Brookings Papers on Economic Activity 2:

Can Risk-Based Deposit
Insurance Premiums
Control Moral Hazard?
Edward Simpson Prescott


alls for deposit insurance reform regularly sound the refrain to make
deposit insurance premiums more risk based.1 Those who support
such a change believe that risk-based premiums will discourage insured banks from taking excessive risk because a bank facing higher premiums
will think twice before undertaking a risky activity.
This logic seems impeccable: Let banks face the true cost of risk and
they will appropriately balance the tradeoff between risk and return. While
seemingly correct from the standard perspective of price theory, this argument
requires the deposit insurer to be able to observe the risk characteristics of a
bank’s investment portfolio. There are good reasons to think that this is not the
case; it is hard for outsiders to evaluate a bank loan or a complicated portfolio
of financial derivatives. Under these conditions, risk-based deposit insurance
premiums are not enough to control moral hazard. Instead, other devices
such as performance-based insurance payments and supervisory monitoring
are needed as well.
When one party to a transaction has information that the other party does
not have, economists describe the transaction as one with private information.
Various types of information may be private, but I am concerned with a payoffrelevant action. This model is sometimes referred to as the moral-hazard or
hidden-action model. In this article, the action that may be hidden from others
is the risk characteristics of a bank’s investment decisions. The economic literature on moral hazard emphasizes the importance of state-contingent payments
The author would like to thank Huberto Ennis, Tom Humphrey, John Walter, Roy Webb, and
John Weinberg for helpful comments. The views expressed in this article do not necessarily
represent the views of the Federal Reserve Bank of Richmond or the Federal Reserve System.
1 For recent examples, see FDIC (2000) or Blinder and Wescott (2001).

Federal Reserve Bank of Richmond Economic Quarterly Volume 88/2 Spring 2002



Federal Reserve Bank of Richmond Economic Quarterly

for giving people the right incentives.2 A simple example of state contingencies is a salary plus a commission. Sales representatives are frequently paid
this way to give them an incentive to work hard. In contrast, risk-based deposit
insurance premiums are not state contingent. They are entirely ex ante. As
we will see, this limits their usefulness as a tool to control moral hazard.
I have three goals in this article. The first goal is to show what riskbased deposit insurance premiums can and cannot do. Risk-based premiums
are useful for preventing transfers between different risk classes of banks, but
they cannot control moral hazard. This idea is not new. It appears to be widely
known among banking economists, but it rarely seems to have been formally
expressed.3 The second goal is to illustrate how state contingencies in deposit
insurance payments can be used to control moral hazard. As indicated above,
this illustration will use a model with private information. The final goal is
to formally develop a role for supervisory activities like safety and soundness
exams. These exams are modeled as a costly means for reducing the amount
of private information between the deposit insurer and the bank.4 Most of the
literature on bank regulation takes the amount of private information as given.
But as long as these supervisory activities reduce private information, they
play a crucial role in any well-designed deposit insurance system.
The ideas in this article can be expressed with an analogy to an insurance
contract. In dealing with different risks, insurance companies do more than
adjust premiums. They also alter deductible amounts, copayment rates, and the
probability of inspections. These contractual features are designed to prevent
the insured from altering the risks it faces in a way that is detrimental to the
insurance company, while still providing a degree of insurance. Of course,
these characteristics of the insurance contract change with the risks, so in that
sense well-designed deposit insurance contracts are risk based. Nevertheless,
the premium level is not the only thing that changes. The analogy carries
through to deposit insurance, which is why a well-designed deposit insurance
system needs to do more than make premiums risk based.

There is a deposit insurer who insures the depositors of one bank. The insurer
is risk-neutral and has access to outside funds, so it has enough resources to
cover its exposure. For simplicity, I assume that the bank is fully funded by
2 For a survey of moral-hazard models, see Hart and Holmstrom (1987) or Prescott (1999).
3 One exception is John, John, and Senbet (1991), and there are probably others as well.
4 There is a literature on costly monitoring and auditing. Examples include Townsend (1979)

and Dye (1986).

E. S. Prescott: Deposit Insurance Pricing


deposits.5 I also ignore any liquidity or payment services provided by deposits.
For my purposes, it is sufficient to treat deposits as just another form of debt.
These deposits are fully insured and pay a gross rate of return of one.
The bank has access to several investment strategies. Each strategy requires one unit of capital to be invested. I assume that because of investment
indivisibilities, the bank can engage in only one strategy at a time. The return
r of each investment strategy i is uncertain. The probability distribution of returns for a given investment strategy is written f (r|i). For simplicity, I assume
that only a finite number of returns are possible. The bank is risk neutral but
has limited liability. If the investment’s return is less than one, the depositors
receive everything produced by the bank plus enough of a payment from the
deposit insurer that they receive the guaranteed gross return of one. If the
return is greater than one, depositors receive a payment of one, any charges
imposed by the deposit insurer are paid by the bank, and the bank keeps the
remainder (if any) of its return.
The objective in this economy is to design the deposit insurance scheme so
that the bank chooses the highest net present value investment project. Because
of deposit insurance, however, meeting this objective is not straightforward.
In the following sections, I work through the following three variations on the
1. In the first variation, I assume that the deposit insurer observes the
bank’s investment strategy. Risk-based premiums are sufficient to control risk in this case.
2. In the second variation, I assume that the deposit insurer no longer
observes the bank’s investment strategy. This is the hidden-action or
moral-hazard model. Risk-based premiums do not control moral hazard
in this case and state-contingent payments are needed.
3. In the final variation, I develop a role for safety and soundness exams.
The deposit insurer may spend resources that reduce (but do not eliminate) private information. In the example, the optimal deposit insurance
system requires an exam in addition to state-contingent payments.

Full Information
In this section, I assume that the bank’s investment decision is observed by
the deposit insurer. In this case, economists say there is full information. It is
under full-information conditions that risk-based deposit insurance premiums
can succeed.
5 For a related analysis of capital regulations, see Marshall and Prescott (2001) and Prescott



Federal Reserve Bank of Richmond Economic Quarterly

Table 1 Probability Distribution of Returns










Notes: Probabilities and expected return of each investment strategy. The row labeled is
corresponds to the high-mean, low-risk strategy, while the row labeled ir corresponds to
the low-mean, high-risk strategy. The last column lists the expected return or mean.

I illustrate this point with a simple example. Assume that the bank can
choose between two investment choices. One of these choices is a low-risk,
high-mean strategy, is , while the other is a high-risk, low-mean strategy, ir .6
There are three possible returns: a low one of 0.9, a medium one of 1.05, and
a high one of 1.2. Table 1 lists the probability distribution of returns f (r|i)
as well as the expected return.
The socially desirable investment strategy is is . Its expected output is
higher than that of the risky investment strategy ir . The distribution of returns
also differs between the two strategies. The safe strategy usually produces the
medium return of 1.05, while the risky strategy is much more likely to produce
either low or high returns.

Without Deposit Insurance

Without deposit insurance, the market prices deposits to reflect risk. If the
risk-free rate on deposits were zero and the bank took investment strategy is ,
the depositors of the bank (assumed to be risk neutral) would require that the
deposits pay 1.011 if the bank is solvent. This would give depositors an expected payoff of 0.1(0.9)+0.9(1.011) ≈ 1.0, which is equal to their expected
payoff if they invested in risk-free assets. Alternatively, if the bank took investment strategy ir , a similar analysis would find that depositors would require
a payment of approximately 1.0429 to compensate them for the increased
chance of the low return.
6 Restricting the bank to two investment strategies is done mainly for expository purposes.
Marshall and Prescott (2001) study a model where the bank can choose both the mean and variance
characteristics of its loan portfolio. They find that the two investment strategies that mattered
the most for deposit insurance are the low-risk, high-mean strategy and the high-risk, low-mean
strategy. Restricting the investment strategies to these two choices is a stand-in for the more
complicated problem.

E. S. Prescott: Deposit Insurance Pricing


The bank’s payoff is the difference between its return and its payment to
depositors. In either case, the expected gross return to depositors is 1.0, so
the bank’s expected payoff is
E(r|i) − 1.0.


Faced with this tradeoff, the bank would take the socially desirable investment
strategy, is , because E(r|is ) − 1 > E(r|ir ) − 1.
With Deposit Insurance

Improperly priced deposit insurance may distort the bank’s preference-ordering
over these choices. To see this distortion, consider the situation where the deposit insurance premium is independent of the bank’s investment strategy.7
Because of deposit insurance, depositors always receive 1.0. With limited liability, the bank’s payoff function is max{r − 1 − p, 0}, where 1 is the payment
to depositors and p is the premium.8 When the premium is set to zero, the
bank’s expected utility is
f (r|i)(r − 1.0) = E(r|i) − 1.0 +

f (r|i)(1.0 − r).



Compared with equation (1), the bank’s payoff without deposit insurance,
the bank’s utility under deposit insurance contains an additional term. This
additional term is sometimes referred to as the value of the deposit insurance
put option. It can be considered a put option because it allows the bank to
dump its liabilities on the deposit insurer at a strike price of zero. It is valuable
because with deposit insurance, risk is not reflected in the price of deposits.
The lower rate paid on deposits leads to an increased payoff to the bank,
the amount of which is the additional term. In essence it is a transfer from
the deposit insurer to the bank; it also illustrates why underpriced deposit
insurance can lead to a taste for risk. This last term increases as the expected
transfer from the deposit insurer increases.
This taste for risk matters in the example. If premiums are set to zero,
the bank prefers the risky strategy despite the higher expected return of the
safe strategy. In particular, the return to the bank of the risky strategy is
0.3(0.0) + 0.3(0.05) + 0.4(0.2) = 0.095, while the corresponding return of
the safe strategy is only 0.1(0.0) + 0.6(0.05) + 0.3(0.2) = 0.09.
7 For early work identifying the risk-taking incentives created by deposit insurance, see Merton
(1977) and Kareken and Wallace (1978).
8 In practice, banks pay any premiums before investing the funds. Throughout this article I
assume premiums are paid after the fact and use as our operational definition of a premium a
constant payment that is made subject to limited liability. This assumption is made because I do
not want to worry about how the deposit insurer invests the premiums it collects. The assumption
does not alter the results.


Federal Reserve Bank of Richmond Economic Quarterly

Risk-based premiums can deal with these perverse incentives but only if
the deposit insurer observes the investment strategy taken by the bank and
makes the premiums dependent on it. Let the insurer index premiums by the
bank’s risk strategy, pi , and set premiums to be actuarially fair.9 The premium
level for a given investment strategy i must satisfy
f (r|i)pi +

f (r|i)(r − 1.0) =

f (r|i)(1.0 − r). (3)


The left-hand side is the expected value of collected premiums. The second
term on the left-hand side reflects the amount of funds collected by the insurer
if the bank produces enough to pay depositors but not enough to pay the full
amount of the premium. The right-hand side of equation (3) is the expected
transfer made by the deposit insurer to depositors. Later it will be convenient
to write (3) as
f (r|i)pi =

f (r|i)(1.0 − r).

Under this actuarially fair, risk-based premium schedule, the bank’s expected payoff is
(r − 1.0 − pi ) = E(r|i) −

f (r|i)r −

f (r|i)1.0

f (r|i)pi


= E(r|i) −

f (r|i)r −

f (r|i)1.0

f (r|i)(1.0 − r)


= E(r|i) − 1.0.


This equation is identical to equation (1), which describes the expected payoff
to the bank under the no deposit insurance case. There is equivalence because
in the risk-based deposit insurance premium case, the premiums are set to
exactly offset the expected payments made by the deposit insurer. In the
context of equation (2), the premiums paid exactly offset the value to the bank
of the deposit insurance put option. Consequently, just as in the no deposit
insurance case, the bank will choose the safe investment strategy because it
has the highest expected return.
9 Analysis of deposit insurance usually operates under the assumption that actuarially fair
deposit insurance is desirable. This mode of operation is based on the view that transfers to or
from taxpayers are undesirable. For a deposit insurance model that argues that this view may be
incorrect, see Boyd, Chang, and Smith (2001).

E. S. Prescott: Deposit Insurance Pricing


In the numerical example, the actuarially fair deposit insurance premium
for investment strategy is is 0.011. (Recall that in this article the premium is
being assessed after the return is realized, and to be consistent with limited
liability the bank cannot pay its premium if it produces the low return of 0.9.)
The corresponding premium for the ir investment strategy is 0.0429. With
these investment-dependent premiums the expected payoff to the bank of is
is 0.08, while the corresponding payoff to the bank if it takes ir is 0.065.
Consequently, with risk-based deposit insurance premiums, the bank chooses
the socially desirable investment.
This example illustrates the argument behind risk-based deposit insurance
premiums. Risk-based premiums control risk because premiums can be made
explicitly on the investment strategy, and if they are set to keep deposit insurance fairly priced, the bank faces the true costs of its investment decision.
But this result depends on the insurer being able to ascertain just how risky
a strategy the bank is taking, which it must be able to do in order to set the
premiums properly. It is by no means clear, however, that assessing the bank’s
strategy is an easy task. As I mentioned earlier, the quality of a bank loan may
be hard to determine, let alone the quality of an entire portfolio. Just witness
the enormous debate and controversy over how to make the Basle capital regulations reflect risk more accurately.10 In the next section, I will illustrate just
how important the full-information assumption is and how the conclusions
change when it is dropped. Those results will form the basis for my argument
that risk-based premiums alone cannot control moral hazard.

Private Information
To illustrate the second variation on the environment, where the bank’s investment strategy is private information, let us continue with the numerical
example. The deposit insurer sets a risk-based premium of 0.011 if the bank
takes the safe strategy and 0.0429 if it takes the risky strategy. But to implement this policy, the insurer has to know which strategy the bank takes. For
the reasons described above, this knowledge is not easy to ascertain. What if
the bank claims it is taking the safe strategy but is actually taking the risky
I can evaluate this possibility by setting the premium to 0.011, that of the
safe strategy, and evaluating the expected payoff to the bank if it takes the risky
strategy. Its payoff in this case is 0.3(0) + 0.3(1.05 − 1 − 0.011) + 0.4(1.2 −
1 − 0.011) = 0.0872. This expected payoff is greater than 0.08, which is
10 The 1988 Basle Accord assigned risk weights to different classes of assets and then set
a minimum capital requirement based on the sum of these risks. There has been widespread
dissatisfaction with the Accord because all loans of a particular class, such as Commercial and
Industrial loans, are treated as equally risky. A major reconsideration of the Accord is underway
right now, and the proposals for reform are based on trying to better ascertain risks at the level
of individual loans.


Federal Reserve Bank of Richmond Economic Quarterly

what the bank would get if it took the safe strategy. This evaluation suggests
that the insurer cannot use the risk-based premium schedule analyzed above
to implement is .
Unlike in the previous section, the insurer does not observe the bank’s
investment strategy and the bank is therefore able to say that it is taking one
strategy while it is really taking a different one. Economists say there is private information when information relevant to a transaction or a contractual
arrangement is known to only one of the participants. In the context of deposit insurance pricing, private information puts limits on the types of pricing
schemes that can be used. Economists deal with these limits by requiring contracts, or in this case pricing schemes, to be incentive compatible. A deposit
insurance pricing scheme and an investment strategy are incentive compatible
if under the scheme it is in the bank’s best interest to take the investment strategy. In contrast, there is no such requirement in the full-information case. If
the bank changes its strategy, the premium level can change with it.
As the above analysis indicates, a fixed premium and the socially desirable investment strategy is are not incentive compatible. The insurer can do
better, however, if it does not restrict itself solely to premiums but also allows
payments to depend on the realized return. More formally, I write these payments as p(r). A deposit insurance premium is a special case of this function
in which p(r) equals a constant.11 With this notation, I can more formally
define incentive compatibility.
Definition 1 A deposit insurance price system p(r) and investment strategy
i is incentive compatible if for all alternative investment strategies i
f (r|i) max{r − p(r) − 1.0, 0} ≥

f (r|i ) max{r − p(r) − 1.0, 0}.

In words, this definition says that for a given deposit insurance price system
p(r), the expected payoff a bank receives from taking investment i must be
more than it would receive if it took any other possible investment strategy i .
For example, the safe investment strategy is is not incentive compatible when
the fixed premium is set to 0.011. The risky investment strategy ir , however,
is incentive compatible for that same premium.
With private information, state-contingent payments may improve upon
risk-based premiums (which are not state contingent). To see this, consider
the following deposit insurance pricing scheme. If the bank produces the high
return, charge it 0.053, and if it produces the middle return, rebate to it 0.01.
Of course, no payments are made if the bank produces the low return since
the bank fails in this event.
11 Technically, in this article p(r) is only a constant when the bank has enough funds to
pay the premium.

E. S. Prescott: Deposit Insurance Pricing


The safe investment strategy is incentive compatible for this deposit insurance pricing system. If the bank chooses the safe investment strategy, it
receives 0.08. (The number is unchanged from above since the price schedule
was chosen to be actuarially fair.) Furthermore, incentive compatibility holds
because the expected payoff to the bank from taking the risky strategy is now
only 0.077.
This effect can be seen more formally through an analysis of the likelihood
ratios. In moral hazard problems with recommended strategy i, the likelihood
ratio for a given return r is the probability of r, given alternative investment
strategy i divided by the corresponding probability if the recommended strategy was taken. More formally, the ratio is f (r|i ) . Examination of the incentive
f (r|i)

constraint reveals the following. If p(r) is set high when p(r|i ) is high, a bank
that takes i is punished relatively more than a bank that takes the desired i.
Similarly, if p(r) is set low (or even negative) when this fraction is low, a bank
that takes i is rewarded relatively less than a bank that takes the desired i.
In this example, the likelihood ratio (when i = is and i = ir ) is high
for the high return and low for the middle return. This property of the ratio
generates the seemingly paradoxical result that the payment is higher if the
highest return is produced.12 But in this example, a low payment for the
high return would give the bank too much of an incentive to take the risky
investment strategy.13 Finally, it is worth noting that the likelihood ratio is
high for the low return as well, but because of limited liability the bank cannot
make payments to the insurer.
Figure 1 illustrates why this pricing scheme is effective. The solid line
depicts the payoff to the bank if it faces a fixed premium. The dashed line with
the stars reports the payoff from a pricing schedule that collects all payments
from the bank when the bank does very well. Notice how the shapes of the
two functions differ. The solid line is convex, which means it rewards risktaking.14 The dashed line with the stars, while convex in portions, is basically
a concave function. It does not reward risk-taking.
The lesson of this example is that risk-based premiums cannot control
moral hazard on their own. Private information requires richer deposit insurance pricing schemes that take advantage of state-contingent pricing. This
is not to say that risk-based premiums are not useful but that they are only
one component of the entire deposit insurance price system. For example, if
12 For similar results in the context of bank capital regulations, see Marshall and Prescott
(2001) or Prescott (2001).
13 One potential problem with this pricing scheme is that high returns could also reflect
innovation. High payments for high returns would then have the undesirable effect of punishing
innovation. The proper balance of these considerations is an open research question.
14 In the full-information case, this shape did not cause the bank to prefer the risky investment
because the premium level could change with investment strategy. Under private information, the
premium does not change with the strategy so the convex shape becomes a problem.


Federal Reserve Bank of Richmond Economic Quarterly

Figure 1 Bank’s Payoff as a Function of the Return

Notes: The solid line depicts the bank’s payoff as a function of the return if it pays a
fixed premium of 0.02. The dashed line with the stars represents the bank’s payoff for a
deposit insurance pricing system that charges no premium but requires a payment if the
bank produces a return greater than 1.1. For both payoff functions, the horizontal portion
reflects limited liability. Because of limited liability, a bank facing a fixed premium has
a convex payoff function. Payoff functions with this shape create a taste for risk. (To
see this draw a line between a return on the horizontal portion of the payoff function
and a return on the increasing portion. Randomizing over these two returns is preferred
to the certain production of the expected amount.) A bank that faces the alternative price
schedule has a payoff function that is almost concave, with only a portion being convex.
Concave payoff functions create a distaste for risk.

some investment decisions are easy to observe, like the class of investments
a bank specializes in, then the analysis will contain elements of both the full
information and private information models. In this case, there could be one
pricing scheme for banks that specialize in real estate lending and another
pricing scheme for banks that hold safe assets like Treasuries. The real estate lending bank might face high premiums plus state-contingent payments,
while the Treasury-holding bank might face low premiums and relatively nonstate-contingent payments. The pricing scheme is risk based as advocated by
proponents of risk-based deposit insurance premiums, but, as my analysis
suggests, the pricing scheme would also be state contingent.

E. S. Prescott: Deposit Insurance Pricing


Changing the Information Structure
The previous analysis focused on how a price system with state-contingent
pricing could improve upon narrow risk-based premium systems. Indeed, the
state-contingent price system was successful at implementing the safe, socially
desirable investment strategy. The example should not be taken, however, to
mean that state-contingent pricing can control all of the moral hazard created
by deposit insurance. In many moral hazard problems, the best incentivecompatible contract only partially mitigates the moral hazard.
In this section, I consider the third variation on the environment by providing a private information environment where the insurer can take some costly
action that lets it observe some of the private information. This analysis can
be used to form the basis for analyzing numerous supervisory activities like
safety and soundness exams, audits, and off-site surveillance. As we will see,
these activities can play a crucial role in a well-designed deposit insurance
pricing system.
To illustrate this principle, I return to the example used in the above section.
Now, however, I assume that it costs the bank effort and resources to screen
its investment portfolio in order to identify the is investment strategy. If the
bank does not supply this effort, it cannot take the is strategy. The effort cost
translates directly into a utility loss to the bank that corresponds to a drop in
its payoff of 0.05 units. This loss is not affected by limited liability. The idea
is that this loss corresponds to effort by bank management. The bank can
choose not to supply the screening effort. If it takes this route, it saves utility
but must choose investment strategy ir . As before, I assume that the socially
desirable investment strategy is for the bank to take is .15
The incentive problem here is more severe than in the previous example.
Before it was only necessary to worry that the bank might take the risky
strategy. Now, however, it is also necessary to worry that the bank might not
screen its portfolio and then take the risky strategy by default. If it does not
screen its portfolio, it saves on the utility cost of 0.05. This additional saving
is important for the incentive constraints. In particular, the safe investment
strategy cannot be implemented with the deposit insurance pricing schedule
examined above. Furthermore, this strategy cannot be implemented for any
actuarially fair deposit insurance pricing scheme.16
15 In making this assumption, I am ignoring the utility cost to the bank in my welfare
calculation. This assumption keeps the problem simple.
16 For the example, an actuarially fair pricing scheme must satisfy

0.6p(rm ) + 0.3p(rh )



where p(rm ) is the payment made if the medium return is generated and p(rh ) is the payment
made if the high return is generated. The right-hand side is 0.01 because that is the expected
payment made by the deposit insurer to the depositors.
For is to be incentive compatible, the pricing scheme must satisfy the incentive compatibility


Federal Reserve Bank of Richmond Economic Quarterly

What is the insurer to do? Let us make one last addition to the environment
and allow the insurer to spend 0.02 units examining the bank. By examining
the bank, the insurer does not observe which investment strategy the bank
takes, but it can tell if it expended the effort to properly screen the projects.
Observing this effort could be interpreted as examiners checking bank lending
procedures or resources devoted to risk management.
If the insurer examines the bank, the problem is identical to that of the
previous section except that now the insurer also has to make up the examination cost of 0.02 units from its pricing scheme. It can recover these funds
by setting the rebate to zero and raising the charge on the high return to 0.10.
Under this deposit insurance pricing and inspection system, it is incentive
compatible for the bank to screen and then take the safe investment strategy.
The exam prevents the bank from not screening and once it screens, the statecontingent payments convince the bank to take the safe investment strategy.
Finally, the deposit insurance price system is actuarially fair (including examination costs), so no resources are transferred in or out of the banking system
in expectation.
The key feature of this example is the way in which the examination policy
changes the information structure of the bank. In this example, the information is revealed in a straightforward manner. More generally, examinations
or other types of supervisory monitoring may only reveal signals that are partially correlated with the true action. Or, supervisors may want to use the
information they receive from inexpensive information gathering methods,
like balance sheet observations, to decide whether or not they should gather
more information using more costly methods like on-site exams. All these
possibilities can be added to the framework developed in this article.



This article argues that risk-based deposit insurance premiums alone cannot
control moral hazard in deposit insurance. The examples demonstrate how
richer procedures with more complicated pricing schedules and examination
procedures can be more useful than risk-based deposit premiums. The critical
factor in the analysis is private information.
Interesting parallels to the analysis exist in markets without government
insurance. As was discussed earlier, insurance contracts include deductibles
and copayments and may allow for audits to control moral hazard.17 Banks

−0.3p(rm ) + 0.1p(rh ) ≥ 0.055.

Furthermore, the payments are subject to limited liability, which means that p(rm ) ≤ 0.05 and
p(rh ) ≤ 0.2. A simple graph reveals that there is no pair (p(rm ), p(rh )) that satisfies these four
17 Experience rating is an important tool used by insurance companies that was not addressed

E. S. Prescott: Deposit Insurance Pricing


also take several actions to mitigate the private information of their borrowers.
For example, they regularly impose covenants on their borrowers’ actions and
they often list conditions under which they can call a loan.18 Just as there is
more to the price of a bank loan than the interest rate, there is more to pricing
deposit insurance than insurance premium levels.

Black, Fisher, Merton H. Miller, and Richard A. Posner. 1978. “An Approach
to the Regulation of Bank Holding Companies.” Journal of Business 51:
Blinder, Alan S., and Robert F. Wescott. 2001. “Reform of Deposit
Insurance: A Report to the FDIC.” (March).
Boyd, John H., Chun Chang, and Bruce D. Smith. 2001. “Deposit Insurance:
A Reconsideration.” Manuscript, Carlson School of Management,
University of Minnesota.
Dye, Ronald A. 1986. “Optimal Monitoring Policies in Agencies.” RAND
Journal of Economics 17: 339–50.
Federal Deposit Insurance Corporation. 2000. “Options Paper.” (March).
Hart, Oliver D., and Bengt Holmstrom. 1987. “The Theory of Contracts.” In
Advances in Economic Theory: Fifth World Congress, ed. Truman F.
Bewley. Cambridge: Cambridge University Press: 71–155.
John, Kose, Teresa A. John, and Lemma W. Senbet. 1991. “Risk-Shifting
Incentives of Depository Institutions: A New Perspective on Federal
Deposit Insurance Reform.” Journal of Banking and Finance 15:
Kareken, John H., and Neil Wallace. 1978. “Deposit Insurance and Bank
Regulation: A Partial-Equilibrium Exposition.” Journal of Business 51:
Keeley, M. C. 1990. “Deposit Insurance, Risk, and Market Power in
Banking.” American Economic Review 80: 1183–1200.
in this article. Indeed, an experience rating may be a partial substitute for state-contingent payments
by the bank. This omission was made for simplicity; static models are a lot easier to work with
than dynamic ones. Nevertheless, this tool may be very important and deserves to receive more
attention than it has received from the bank regulation literature.
18 For examples and further discussion of the parallels between private lending and how bank
regulation should be structured, see Black, Miller, and Posner (1978).


Federal Reserve Bank of Richmond Economic Quarterly

Marshall, David A., and Edward S. Prescott. 2001. “Bank Capital Regulation
With and Without State-Contingent Penalties.” Carnegie-Rochester
Conference on Public Policy. Forthcoming.
Merton, Robert C. 1977. “An Analytic Derivation of the Cost of Deposit
Insurance Guarantees.” Journal of Banking and Finance 1: 3–11.
Prescott, Edward S. 1999. “A Primer on Moral-Hazard Models.” Federal
Reserve Bank of Richmond Economic Quarterly 85 (Winter): 47–77.
. 2001. “Regulating Bank Capital Structure to Control
Risk.” Federal Reserve Bank of Richmond Economic Quarterly 87
(Summer): 35–52.

Federal Reserve Bank of St. Louis, One Federal Reserve Bank Plaza, St. Louis, MO 63102