Full text of Economic Perspectives (Federal Reserve Bank of Chicago) : Fourth Quarter 1998, Volume XXII, Issue 4

View original document

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

Federal Reserve Bank
of Chicago
Fourth Quarter 1998

Economic.

perspectives

Credit derivatives: Just-in-time provisioning for loan losses

Assessing the condition of Japanese banks:
How informative are accounting earnings?

Foreign growth, the dollar, and regional
economies, 1970-97

The business cycle: It’s still a puzzle

Index for 1998

perspectives

President
Michael H. Moskow

Senior Vice President and Director of Research
William C. Hunter

Research Department
Financial Studies
Douglas Evanoff, Vice President

Macroeconomic Policy
Charles Evans, Vice President
Microeconomic Policy
Daniel Sullivan, Vice President
Regional Programs
William A. Testa, Vice President

Administration
Vance Lancaster, Research Officer

Economics Editor
David Marshall
Editor
Helen O’D. Koshy
Production
Rita Molloy, Kathryn Moran, Yvonne Peeples,
Roger Thryselius, Nancy Wellman
Economic Perspectives is published by the Research De
partment of the Federal Reserve Bank of Chicago. The views
expressed are the authors’ and do not necessarily
reflect the views of the Federal Reserve Bank of Chicago
or the Federal Reserve System.

Single-copy subscriptions are available free of charge. Please
send requests for single- and multiple-copy subscriptions, back
issues, and address changes to the Public Information Center,
Federal Reserve Bank of Chicago, P.O. Box 834, Chicago,
Illinois 60690-0834, telephone 312-322-5111
or fax 312-322-5515.
Economic Perspectives and other Bank
publications are available on the World Wide Web
at http:Avww.ffbchi.org.

Articles may be reprinted provided the source is credited
and the Public Information Center is sent a copy ofthe
published material. Citations should include the following
information: author, year, title of article, Federal Reserve
Bank of Chicago, Economic Perspectives, quarter, and
page numbers.
ISSN 0164-0682

Contents

Fourth Quarter 1998, Volume XXII, Issue 4

Credit derivatives: Just-in-time provisioning for loan losses
James T. Moser

Credit derivative contracts offer a new route for managing counterparty exposures.
This article discusses two formats of these contracts. The contracts have potential for
providing portfolio managers with a cost-effective, just-in-time source of liquidity.

Assessing the condition of Japanese banks:
How informative are accounting earnings?
Hesna Genay

This article examines the accounting and stock market performance of Japanese banks
from 1991 to 1997. Overall, the results indicate that the accounting, disclosure, and
regulatory practices of Japanese banks have driven a wedge between their accounting
and stock market returns in recent years and, furthermore, that regulatory forbearance
might have become a more important source of value to shareholders than the value
of assets in place.

Foreign growth, the dollar, and regional economies, 1970-97
Jack L. Hervey and William A. Strauss

International markets are an important contributor to U.S. economic activity. U.S.
regions have varying exposure to the influences of international markets—foreign
demand or exchange rate movements. Still, the overriding determinant of regional
economic growth is the state of the domestic economy.

The business cycle: It’s still a puzzle
Lawrence J. Christiano and Terry J. Fitzgerald

The business cycle is characterized by contractions and expansions in economic
activity that are synchronized across a broad range of sectors. The authors provide
evidence to document this, and survey some of the theories that have been proposed
to explain it. Although much progress has been made, research in this area is still at
an early stage.

Index for 1998

Credit derivatives: Just-in-time
provisioning for loan losses

James T. Moser

Introduction and summary
Risk managers use a “peeling an onion” analogy to
illustrate their prioritization of risk management activ
ities. The resulting priorities have produced the con
tracting innovations needed to manage the outer
layers of this risk onion. These tools are derivative
contracts whose values are driven by changes in
interest rates, equity prices, and foreign exchange
rates. Having dealt with these outer layers, today’s
risk managers are paying increasing attention to the
inner layers of the onion, most especially credit risk.
Furthermore, globalization of the financial markets is
increasing diversification opportunities. To remain
competitive in the global marketplace, financial insti
tutions whose borrowers are concentrated in certain
business or geographic sectors are seeking methods
to improve their diversification of credit exposures.
The efforts of risk managers are proceeding on
two fronts. First, they are developing methods to mea
sure credit risk exposures. Three of the better known
procedures for measuring credit exposures are the
Expected Default Frequency metric developed by
KMV,1 J. P. Morgan’s CreditMetrics,- and Credit
Suisse’s CreditRisk i ,3 Second, risk managers are en
gineering derivative contracts to enable transference
of credit risk exposures.4 This article examines some
of these contracts and compares this new risk man
agement route with a traditional route for managing
loan loss exposures.
Descriptions of growth prospects for the credit
derivatives market in terms such as “the next interest
rate swap” stem from a confluence of events. Smithson
(1997) points out that the first steps came as overthe-counter (OTC) derivatives dealers began to recog
nize the need to manage their credit exposures to one
another. This recognition led to efforts to quantify and
then to create structures controlling credit risk expo
sures. One such structure is the derivative product

company (DPC), in which derivative contracts are
booked in a subsidiary that then books an offsetting
position with its parent.5 Such structures shift broad
market exposures, most often to interest rates, from
the subsidiary to the parent firm while retaining credit
exposures to original counterparties at the subsidiary
level. These structures are motivated by the need to
raise the credit ratings of OTC dealers and improve
their ability to compete for business. DPC structures
isolate credit risk from other risk sources. This enables
institutions to allocate capital directed at credit risk
concerns. In addition, DPC structures motivate special
ization in credit risk management.
Recently, attention has focused on transferring
credit risk from one party to another using credit
derivative contracts. Various contracting schemes are
now labeled credit derivatives. The common feature
of these risk management tools is that they retain
assets on the books of originating institutions, while
transferring some portion of the credit exposure inher
ent in these assets to other parties. This accomplishes
several objectives. Originating institutions have a
vehicle that transfers credit risk without requiring the
sale of the asset. When asset sales weaken an insti
tution’s relationships with its borrowers, a vehicle
transferring only the credit exposure permits the insti
tution to retain its relationship. In addition, the ability
to reshape credit exposures through derivatives can

James T. Moser is an economic adviser at the Federal
Reserve Bank of Chicago. The article has benefited
from conversations with Conrad Bahlke, Eli Brewer,
Nicola Cetorelli, Hesna Genay, Philipp Hartmann,
Donald Hester, Allison Holland, John Kambhu, Steve
Kane, and Tom Nohel. The author is especially grateful
to Charles Smithson who kindly provided an excellent
primer on credit derivatives and to David Marshall who
carefully read and commented on the earlier drafts.

Economic Perspectives

be used to improve diversification. For example, an
institution with loan concentrations in a problem
industry can lessen credit exposures by swapping
its exposures in the problem industry for credits
from a broader borrowing segment. Thus, following
an oil price decline, the credit exposures from loans
to oil exploration firms may be regarded as excessive.
A credit risk swap reduces the institution’s concen
tration in these firms to achieve a more diversified
loan portfolio.
Current regulatory policy toward credit derivatives
does not recognize their risk reducing potential. In
stead, it emphasizes their potential use as risk increas
ing instruments.6 Consequently, users receive only
limited relief from regulatory capital requirements.
Relief from regulatory capital requirements is available
when credit derivatives are used to hedge assets held
in bank trading books. For assets held in banking
books, regulatory capital relief is less generous, limited
to instances when the credit derivative gives a oneto-one match with the loss experience of individual
banking-book positions. This treatment cannot be
applied to portfolio positions held within the banking
book. In addition, regulatory capital can be required
for the credit derivative itself. If holdings of regulatory
capital are costly, banks will generally find this treat
ment restricts their use of these contracts. In contrast,
banks’ holdings of provisions against loan losses,
a traditional method for managing credit risk, can be
used to fulfill their tier two capital requirements.
In view of the potential for more cost-efficient
management of credit risk, current regulatory policy
toward credit derivatives needs reexamination. In this
article, 1 compare the outcome from a credit derivative
contract with that of loan loss provisioning, the more
traditional method of managing credit risk. The com
parison illustrates that under some circumstances,
credit derivatives obtain the same economic outcome
and, in these circumstances, can be afforded regula
tory treatment similar to that of the traditional risk
management method.

Review of credit derivatives
The variety of credit derivative contract forms
can obscure the common role of these contracts as
mechanisms to transfer credit risk between counter
parties and the returns for bearing this category of risk.
The British Bankers’ Association (BBA) surveyed
the London market in 1996. The credit derivatives it
encountered fell into four categories. Below, 1 review
two of the more important contracting formats en
countered by the BBA survey: total return swaps
and credit swaps.

Federal Reserve Bank of Chicago

Total return swaps
Figure 1 depicts the payment flows for a total
return swap. The swap exchanges the payment con
figurations of two counterparties—actual payments
made between the two counterparties being the net
of the respective payment configurations. The total
return payor pays out based on the return from its
holdings of a risky debt obligation or a portfolio of
risky debt obligations. Total return for risky debt is
the sum of an interest income stream and changes in
the market value of the debt. The risk of these returns
is the variability in this sum. Of particular interest for
credit risk managers are bond defaults and changes in
the prospects for subsequent default. Box 1 describes
the relationship between changes in default prospects
and credit risk. Clearly, if a bond defaults, returns from
the bond are affected by a curtailment of interest pay
ments. In addition, the value of the debt will be affect
ed by market assessments of value recovered through
bankruptcy proceedings. Prospects for future default
on the obligation are typically characterized as ratings
changes. Yields for risky debt adjust according to
changes in these prospects, rising when payment
prospects worsen.

The counterparty to a total return swap, the total
return receiver, bases what it pays on the returns from
a default-free obligation less the negotiated compen
sation for taking on exposure to the risky debt. It
receives the return from the underlying risky debt. The
result of the swap is that the total return payor obtains
the income stream appropriate for a default-free obliga
tion and the total return receiver obtains the income
stream appropriate for holdings of risky debt. The
reconfiguration of income streams is accomplished
contractually rather than by exchanging ownership
of the respective debt obligations.7
Payments based on principal repayment are
typically omitted in this contract format. Thus, the
risk reduction for total return payors is largely the
income loss from ratings downgrades, rather than the
amounts recovered from defaults. Since vehicles to
manage losses from changes in interest rates are well

BOX 1

Sources of credit risk

Panel A of the figure below illustrates the payout
at maturity of a risky debt obligation. Points to the
right of the “kink” represent the promised pay out
of the bond. When the firm’s value exceeds the value
of its promised payments, bondholders receive the
full value of the promised amount. To the left of the
kink, the ow ners of the firm default, ceding owner
ship to the firm’s debt holders.

known, 1 proceed with the assumption that the inter
est rate exposure of the risky debt obligation is fully
hedged. This allows me to focus on the value fluctua
tions from changes in default risk.8
Suppose a bank’s holding of single-A, floating-rate
debt pays 200 basis points over the reference rate. If
the reference rate is 8 percent, then the borrower is
obligated to pay 10 percent for that period. A credit
rating downgrade of the borrower that decreases the
price for that debt by 8 percent implies a total return
of 2 percent. The receiver of the total return swap is
due to be paid 2 percent for that period. The total return
payor is due to receive the 8 percent reference rate
less a spread amount of say 25 basis points, totaling
7.75 percent for the period. Payments are the net of
these amounts, so the total return payor receives 5.75
percent. Combining this receipt with the 2 percent
obtained from the payor’s debt holding gives a return
of 7.75 percent. Therefore, the payor locks in a 7.75
percent return. The appeal for the total return receiver

Panel B charts the probability for each possible
value of the firm. The filled-in bars represent the
distribution of probabilities based on initial infor
mation. The most probable outcome is well above
the promised payment amount as indicated in panel
A and the probability of a zero outcome approaches
zero. The lighter-shaded bars represent a revised
distribution of probabilities such as might occur
after the release of negativ e new s about the firm’s
future prospects. The most probable outcome is
shifted dow nw ard to just about the level of the
promised payment amount and a zero outcome is
a nonzero probability event.
Combining these probabilities and their re
spective outcomes, one can calculate an expected
(probability-weighted) payment amount. Visual
inspection (correctly) suggests that the expected
payment amount declines with the revised proba
bilities. To understand credit risk, consider that the
amounts in panel A are contractually determined.
Bond ratings are a rank-order measure of the prob
ability that the firm’s ability to meet its debt obli
gation will change; higher rankings imply less
likelihood of a change w ithin a given period of
time. Hence, a rating downgrade implies a capital
loss because it is more likely that the firm w ill be
unable to meet its debt obligation.

in the swap arrangement is the ability to participate in
the return stream of the underlying debt obligation
without investing in the bond itself.
As demonstrated, the total return swap increases
cash flow certainty. The traditional bank management
strategy achieves a similar end. Provisioning that
invests some assets in default-free securities also
achieves a lower bound for default losses. An important
distinction is that the provisioning strategy maintains
an inventory of liquid assets, while the credit derivative
strategy delivers cash flows as losses are realized.

Credit swaps
Compared with the total return swap, the contin
gent payout feature of credit swap contracts comes
closer to matching features usually associated with
insurance contracts. As displayed in figure 2, fixed
payors insure against credit events by making periodic
payments of a fixed percentage of the loan’s par value.
On occurrence of a predefined credit event such as a

Economic Perspectives

loan default, the contingent payor makes a payment
compensating the insured for part of its loss. Otherwise,
the contingent payor pays zero.
Taking the defined credit event to be default on
a debt obligation, a credit swap might be structured as
follows. As before, suppose floating-rate debt rated
at single A pays 200 basis points over its reference
rate. The holder of this debt negotiates a credit swap

to insure against loss due to default. The debt holder
is a fixed payor in the contract, paying 10 basis
points per period to the contingent payor. Should the
debt issuer default on the obligation, the fixed payor
receives a preset payment. Otherwise, the contingent
payor pays out zero. The payment offsets the loss in
curred due to the default. Contracts can be structured
in many ways, for example, payment of a fixed amount
on default or payment proportional to loss amounts.
In the case where the credit swap pays the differ
ence between the loan principal value and the recovered
amount, the credit swap limits the loss for the defined
credit event to the value of loan principal. An invest
ment policy combining default-free securities with
risky debt can replicate this lower bound for loss.
Thus, traditional loan loss provisioning combined
with investing provisions in default-free securities can
duplicate the benefits of a credit derivative. The differ
ence is that the credit derivative delivers cash flows on
a just-in-time basis, while the provisioning strategy
retains cash inventories. Once credit derivatives are
understood as an alternative to traditional provisioning
and investing methods, the choice between the two
alternatives is one of cost effectiveness.

Simple model for choosing between
credit risk management tools
Credit derivatives fulfill purposes similar to those
achieved through traditional methods of credit risk
management. Suppose a bank decides its exposure to
credit risk is excessive. To lessen its exposure, it rein
vests some of its cash flow in default-free securities
such as Treasury bills. These investments will be

Federal Reserve Bank of Chicago

labeled provisions for loan losses.9 In making this
decision, the bank foregoes other lending opportunities.
Therefore, its opportunity cost from the credit risk
management decision is the foregone return from ex
tending loans. I compare the bank’s use of funds for
the loss provision and the credit derivative. When
the credit derivative can be had at a lower cost than
the funds outlay for a loss provision, the bank has an
opportunity to extend its loan portfolio. The expected
return from investing this difference in loans can exceed
the opportunity loss when banks invest in low-risk,
low-return assets.

Opportunity cost comparison of credit derivatives
and loan provisions
In one period a loan currently valued at L will
have one of two values. In the up state the borrower
repays the loan, giving the lender proceeds of uL. In
the down state, the borrower defaults on the loan and
the lender recovers the fraction d of the amount due
from the borrower. A one-period risk-free investment
can be made that returns r dollars for every dollar in
vested in the current period. It is natural to stipulate
that in the up state the loan pays more than its current
value and in default it pays less than its current value,
so uL> L >dL. Further, since the loan is risky, its return
in the up state is larger than a parallel investment at
the risk-free rate, so uL> rL.
I assume an insurance contract can be purchased
that pays the difference between the face value of the
loan and its recovery value when the down state occurs
(more on this later). The price of this contract is based
on the current price of the loan, the payoffs in the up
and down states, and the risk-free rate of interest. I
label this contract /(/.. u, d, r). I consider two invest
ment strategies, provisioning and credit derivatives
(as shown in table 1).
Strategy 1 is the provisioning strategy. If the loan
defaults, the loss will be 1 - d dollars per dollar of
loan value for a total loss of (1 - d)L dollars. Investing
the amount (1 -d)L/r at the risk-free rate, the oneperiod payoff from provisioning is (1 - d)L no matter
which state occurs. The portfolio includes the loan
that pays off uL in the up state and dL in the down
state. In the down state, proceeds from the provision
ing investment match the loss realized on the loan.
Therefore, the bank has prefunded the loss and
locked in L, the face value of the loan. In the up state,
the bank realizes gains on both the loan and the pro
visioning investment.
Strategy 2 uses a credit derivative contract to in
sure against cash flow disruption. Like the provision
ing strategy, proceeds from the credit derivative match
the loan loss realized when the down state occurs.

There are two other ways in which credit
derivatives can potentially add value, first
Comparison of outcomes from provisioning
by
improving the efficiency of capital allo
and credit insurance contract
cations and, second, by acting as a form
Payout at time f + 1
of reinsurance.
Investment made at time f
Up state
Down state
Regulators require that banks retain
4 percent tier one capital holdings against
1) (1 - d)L/r at riskless rate r
risk-weighted assets. In strategy 1 the
plus the loan L
(1 - d)L + uL
(1 - d)L + dL
bank must hold capital to support both
2) Purchase insurance
the risky loan and the default-free security.
contract l(L,u,d,r)
Strategy 2 also includes the loan asset,
plus the loan L
0 + uL
(1 - d)L + dL
but
replaces the security investment with
Difference 1-2:
0
a credit derivative. When the capital re
quired to support this asset configura
tion
is less than in strategy 1, additional
With respect to the down state the bank is indifferent
capital
is
freed
up
to support further lending activity.
between the two strategies. Should the up state occur,
Is
this
a
plausible
scenario?
Consider that regulatory
the bank realizes uL from the loan but proceeds from
agencies
require
capital
holdings
against interest rate
the credit derivative contract are zero. Comparing
risk.
The
investment
in
the
default-free
security in
strategies 1 and 2 in the up state, the difference is the
creases
interest
rate
risk
and
requires
that
capital be
amount of the loan loss.
held.
Therefore,
the
bank
’
s
avoidance
of
credit
risk
The bank’s decision requires comparing the time Z
increases
its
capital
requirement
for
interest
rate
risk.
costs of its alternatives to obtain the up state outcomes.
The
alternative,
an
insurance
contract,
creates
no
To facilitate the comparison, 1 stipulate the existence
additional interest rate risk, therefore credit risk is
of additional lending opportunities matching those of
managed
on par with that obtained by the security in
the loan considered above. The expected return from
vestment
but with a smaller required capital outlay.
these lending opportunities is denoted rL. The bank
This
rationale
is similar to that for the DPC structure
has three investment alternatives: provisioning, lending,
described
earlier.
In both cases, isolating credit risk
and insuring. It can fund its provisioning account with
from
broad-market
risks, such as interest rate risk,
an outlay of (1 - d)L/r. If the cost of the insurance
enables
more
efficient
capital allocations.
contract /(/.. w, d, r) exceeds (1 -d)L/r, the bank rules
Finally,
credit
derivatives
can be seen as a form
out the insurance contract. This is because the out
of
reinsurance.
11
Reinsurance
markets
exist to shift risks
comes from insuring and provisioning are identical in
between
intermediaries.
These
markets
become neces
the down state and the up state return from provision
sary
when
geographic
or
other
restrictions
prevent
ing dominates insuring for positive rates of interest.
intermediaries
from
maintaining
sufficiently
well-diver

When the cost of the insurance contract is equal
sified
portfolios.
For
example,
a
Florida
insurance
firm
to or below the outlay required for the provisioning
has excessive exposure to hurricane damages and a
alternative, the bank weighs the risk-adjusted expected
California insurance firm has excessive exposure to
return from investing in loans earning the loan rate rL
earthquake damages. A reinsurance contract exchang
against the return from its provisioning alternative. The
ing
their respective exposures improves the financial
investable amount in loans is [(1 - d)L/r-I(L, u, d, /')]•10
performance
of both firms by increasing the diversifi
The bank then chooses the larger of the risk-adjusted
cation
of
each
contract participant. Diamond (1984)
expected payouts from the two strategies. Since in
shows
that
derivative
contracts used to control expo
creasing its loans potentially improves diversification
sure
to
common
risks
enable institutions to improve
of the bank’s loan portfolio, the risk increase from
their
diversification
and
lower certain costs.12 These
new loans can be negligible.
reductions
shift
the
margin
for loans downward, in
Thus far, the comparison demonstrates that two
creasing
the
level
of
loans
taken
by the intermediary.
inventory management methods can fulfill risk man
The
reinsurance
aspect
of
credit
derivatives may
agement requirements. The loan provisioning strate
provide
an
additional
and
possibly
more efficient
gy corresponds to a static inventory by choosing
mechanism
for
achieving
diversification.
inventory levels in anticipation of future liquidity
needs. The credit derivative strategy corresponds to
Cost ofinsurance
a just-in-time inventory management style by con
Cox, Ross, and Rubinstein (1979) employ risktracting for deliveries as needs for liquidity arise.
adjusted probabilities to compute the expected payoff
TABLE 1

Economic Perspectives

from an option contract. The risk adjustment is ob
tained by choosing probabilities that are consistent
with an arbitrage replicating the value of the option
from investments in the underlying asset and a safe
asset. Since the arbitrage is riskless, the expected
payoff from the option is discounted at the rate for
the safe asset. Recognizing that the insurance con
tract above can be construed as a put option, the value
of the credit derivative can be obtained using the
binomial approach developed by these authors.
Considering the insurance as a one-period con
tract remains useful. Further, assume that the loan being
insured is a one-period loan that matures on the same
date as the option. Restricting the insurance policy in
this way avoids the need to incorporate the covariance
between the riskless rate and the rate for risky debt.
Therefore, attention is focused entirely on the credit
risk aspects of the loan rather than on any interest
rate risk. Under these conditions, the price of the in
surance contract is
I(L,u,d,r)

r‘(L,u,d,r)--------H
u-d

(L,u,d,r)------ ! r,
u-d

where /“() and / '() are, respectively, the payoffs from
the insurance contract in the up state and down state.
Adding to the comparison between credit derivatives
and loan provisions, the pricing model offers insight
into the effect of interest rates on the credit derivative
decision. As the level of rates for the safe asset rises,
the level of funding required to provision against
losses falls. In addition, the price paid for insurance
declines. The rate of decline in the price paid for in
surance is greater.13 This implies that as interest rates
rise, the credit derivative alternative becomes increas
ingly attractive vis-a-vis the provisioning alternative.
This pricing model assumes that the outcomes
for loans are not influenced by the purchaser of the
insurance contract. More likely, insurance contracts
will have greater appeal when the insured has a higher
expectation of loss than the insurer. These informa
tion asymmetries, or adverse selection problems,
imply that a premium will be charged for insurance
contracts that fail to protect the insurer against her in
formation disadvantages. Denoting this adverse selec
tion premium p, the price of insurance is I(L, h. d, r) + p.
Smith and Warner (1979) show that joint benefits
give the insurer and the insured an incentive to
minimize adverse selection premia. My results sug
gest that the common interests of these counter
parties lead to contracts that reduce the bank’s
opportunity cost by freeing up additional funds
for lending.

Federal Reserve Bank of Chicago

However, resolving adverse selection problems
is not without cost. Contracts structured on state
variables determined outside the firm, such as a stan
dard reference rate, can bypass adverse selection
problems. However, use of a standard reference rate
introduces basis risk. Basis risk for a credit derivative
exists when the correlation between the drivers that
determine payments due on credit derivatives does
not match the loss experience for the insured debt.
For example, a lender holding a loan issued by a specific
corporation may find that the returns of a security
within the same industry generally reflect the pros
pects of defaults within that industry. Such a security
is likely to resolve the adverse selection problems.
However, credit problems that are unique to the indi
vidual firm will not be reflected in the reference security
so payments based on the reference security may not
cover losses on the loans to the individual firm. So,
the resolution of adverse selection problems is ob
tained at the cost of mismatches between payments
on the credit derivative and loan performance. This
situation introduces a margin between the cost of im
perfect loss protection and premia paid for adverse
selection problems. Understanding this margin enables
an improved prediction of the types of credit derivative
contracts that are most likely to succeed.

Rationalesfor loan provisioning
Kwan (1997) describes loan loss provisioning
as a contra asset account. The size of the account is
maintained at the level of losses the bank expects to
realize. The size decision affects earnings in two
ways. First, when a bank increases its provisions, it
defers recognition of earnings. This has tax implica
tions, reducing current taxable income. Later, as loan
losses are realized, the provisioning account is written
down and the previously deferred earnings are recog
nized along with the loan loss. Because the recog
nized loss amount and the now-recognized deferred
earnings net to zero, loan losses reduce taxable income.
Second, to the extent that earnings performance
signals actual cash flow performance, then bank
managers have incentives to manage earnings levels.
For example, when the level of earnings may incorrectly
signal future prospects, managers can adjust earnings
to prevent unwarranted stock price changes. More
straightforwardly, earnings figures will be managed
when earnings are used to gauge the performance of
bank managers.
Here, I construe loan loss provisioning as follows.
The bank manages its exposure to credit risk by in
suring that it has access to cash sufficient for its
operating requirements. It can accomplish this by in
vesting in assets that can be readily sold to obtain

needed cash or, as previously discussed, using a
credit derivative to insure its access to cash. Consider
a bank constrained from using a credit derivative that
is choosing the portion of its earnings to be paid out
as dividends. A large dividend payout reduces cash
available for investment in default-free securities. By
reducing its payout, it can increase its holdings of
liquid assets. These asset holdings can be thought
of as liquidity buffer stocks. Absent these sources
of liquidity, the bank becomes more likely to be
forced to meet its obligations through the sale of its
less liquid loans.
The adverse selection premium described earlier
amplifies the value of maintaining these buffer stocks.
Banks unable to provide credible signals for their
valuations of loans put up for sale will generally find
that these loans must be sold at a discount to the
bank’s assessed valuations. The difference between
the market price and the bank’s valuation is the adverse
selection premium, which compensates purchasers for
the risk that the bank is selling its weakest loans. Such
revenue shortfalls can impair the ability of the bank
to meet its financial obligations. To avoid this outcome,
the bank can sell inventories of liquid assets without
a discount and use the proceeds to fund its other
obligations. Then the bank faces an inventory problem.
It must maintain an inventory of liquid assets suffi
cient to meet its future loan loss experience. However,
investments made in this inventory generally yield a
lower return than the bank’s other uses for its fiinds.
So, the bank incurs an opportunity loss for maintain
ing an inventory of loan loss reserves. The previous
section showed that credit derivatives mitigate this
opportunity loss in certain circumstances.
In this sense, the credit derivative strategy can
be construed as dynamically provisioning against
loan losses. Contrast this with the static inventory
allocation represented by loan loss provisions. With
credit derivatives, the bank maintains an off-balancesheet position that delivers funds as the needs arise,
rather than maintaining a funds inventory. The justin-time arrival of funds via a credit derivative contract
fulfills the need for immediate funds to meet financial
obligations. Like manufacturing firms that adopt justin-time inventory systems, banks may find this a
cost-efficient solution to funding their operations.
The value of this alternative inventory method
should be included in the franchise value of the in
stitution. When claims against this franchise value
are limited to the bank’s owners, bank managers act
for the owners in their inventory decisions. These
agents add value when their allocation decisions use
credit derivatives to reduce the opportunity cost of

carrying inventories of lower-yielding liquid assets in
place of higher-yielding loans.

Policy implications
The conclusions outlined in this article have
implications for the regulatory policy afforded to
credit derivative contracts. Below, I describe current
regulatory policy on capital requirements. See
Watterson and Bahlke (1997) for a more comprehensive
treatment of the legal and regulatory issues involved
in credit derivatives.

Regulatory policy toward credit derivatives
Regulatory capital is broken into tiers. Tier one
capital, required to be no less than 4 percent of riskweighted assets, is an institution’s net worth.14 Tier
two capital includes these items plus other market
issuances, but also includes provisions for loan losses
subject to two limitations. The first limitation is that
loan loss provisions included as capital cannot exceed
1.25 percent of gross risk-weighted assets. The second
is that the total value of these provisions cannot
exceed that of all other forms of tier two capital. With
tier two capital requirements at 8 percent of risk-weight
ed assets, loan loss provisions are an important com
ponent of regulatory capital Proponents of RAROC
(risk-adjusted return on capital) and similar mechanisms
argue that, on correctly risk-adjusted bases, tier two
capital levels generally should be around 5 percent.
This implies that institutions presently having excess
balances of liquid assets are bearing a large cost for
holding these balances. One can expect banks to seek
to lower their costs by pushing for regulations that
permit substitution of credit derivative contracts for
loan loss provisioning.
The Bank of England published a provisional
letter on credit derivatives in late 1996. British regulators
classify bank assets as trading book or loan book.
Capital charges for loan-book assets are larger, reflect
ing their lesser liquidity. The Bank of England judged
the credit derivative market to be insufficiently liquid
to permit the more favorable trading-book classification.
To the extent that regulatory capital requirements are
binding on these institutions, this view limits use of
credit derivatives.15
In the U.S., the Federal Reserve and the Office
of the Comptroller of the Currency (OCC) have taken
different paths. The OCC holds that the credit derivative
market is too new to take broad regulatory measures.
OCC regulators are concerned that moving too quickly
would adversely influence the innovation process.
They are conducting case-by-case evaluations of in
stitutions’ credit derivative positions, responding as
appropriate. Since these decisions involve proprietary

Econom ic Perspectives

information, the trend in these decisions is not apparent.
The OCC seems aware of the potential for increasing
the efficiency of risk transfers and views its caseby-case approach as supporting this emerging
market segment.
The Federal Reserve has published two guide
lines on credit derivatives. In addition, a Federal
Reserve economist is considering the potential for
these contracts to increase systemic risk (Duffee
and Zhou, 1998).
The first guideline published by the Fed was
a Supervisory and Regulation Letter (SR 96-17) re
leased in August 1996. This letter primarily covers
credit contracts held in the banking book, so its ap
plication pertains primarily to end users of these
contracts. It directs bank examiners to base capital
requirements for a credit contract on the credit exposure
of the reference asset. The letter makes an analogy
between the present treatment of letters of credit and
the Fed’s intended treatment of credit derivatives;
that is, ascertain the credit exposure of the underlying
credit, determine the proportion of that credit exposure
present in the credit contract, then apply the capital
charge for credit exposures to the product of these.
This treatment does not appear to recognize risk re
ductions obtained through holding a diversified
portfolio of credits. In addition, the letter identifies
counterparty default on the credit derivative as a
credit exposure and requires capital on this risk, noting
that this aspect will primarily affect dealers.
The second guideline published by the Fed was
a Supervision and Regulation Letter (SR 97-18) released
in June 1997. This letter provides guidance for ex
aminations of trading accounts. For trading account
positions, banks can use either the standard capital
charge or a capital charge based on risk levels from
an approved internal model. The letter categorizes
trading-book contracts as either open positions,
matched positions, or offsetting positions and identifies
the types of risk for each: counterparty credit risk,
market risk, and credit risk from the asset underlying
the derivative contract. Open positions have exposures
to all three risk types. Matched positions pose only
counterparty credit risk, the other two risk types being
offset. Offsetting positions, for example, positions
whose payouts match in some but not all states, are
similar but the latter two types of risk are mitigated
not eliminated.
The letter directs examiners to classify positions
according to this matrix and apply standard capital
charges. Capital charges for counterparty risk apply
the following rule: If the underlying reference credit
is an investment-grade asset, the equity capital

Federal Reserve Bank of Chicago

charge is used; if the reference credit is a speculativegrade asset, the commodity capital charge is applied.
This treatment does appear to permit consideration of
diversification. The relatively favorable treatment of
credit derivatives for trading book assets vis-a-vis
assets held in the loan book gives banks an incentive
to move assets from the banking book to the trading
book. The strength of this incentive is mitigated by
the somewhat less favorable accounting treatment for
assets held in the trading book.

Economic consequences ofcurrent regulatorypolicy
Excepting bank trading books, regulators have
placed significant restrictions on the use of credit
derivatives. Credit derivatives used to insure assets
held in banking books, that is, most loans, must rep
licate the loss experience of the loan to obtain re
ductions in regulatory capital requirements. This
restriction implies that banks incur the full adverse
selection premium as if they had sold the loan. In
addition, the bank can be required to hold capital
against any counterparty risk encountered should
the bank’s counterparty fail to perform. Thus, the
credit derivative strategy will generally be dominated
by a strategy of selling loans. Therefore, institutions
that have previously maintained inventories of loan
loss provisions will generally find these preferable
to credit derivatives.
The bank can use credit derivatives to hedge
credit risk in assets held in bank trading books. Thus,
credit derivatives can be adopted when the bank is
willing to move assets from the banking book to its
trading book. This change requires the bank to mark
these loans to market. Historically, banks have been
reluctant to mark loans to their market values. This
reluctance implies that capital relief is unlikely.
Duffee and Zhou (1998) make an argument similar
to that of Grossman (1988). The lack of transparency
in the pricing of OTC transfers of credit exposures
can result in inefficient risk-bearing decisions. Imagine
a series of contracts linked in the sense that default
on any one increases the odds of other defaults. Full
transparency insures that investors can accurately
assess the risk and return from investing in these con
tracts. Less than full transparency implies that some
investors may underestimate risks so that capital
costs for firms creating additional contracts are too
low. This situation can result in excessive contracting
activity. If contracts begin to fail and loss experience
reveals the extent of oversupply, the market value of
outstanding contracts declines. If these failures are
seen as systemic, they could lead to social costs in
the form of government-sponsored bailouts.

The problem can be solved if contract transparency
is increased. However, making credit risk completely
transparent requires revelation of proprietary infor
mation. The Fed solves this problem by relying on its
bank supervisory functions to control the extent of
this risk. Absent a change in this policy, Fed policy
toward credit derivatives is likely to be determined by
its bank supervision concerns rather than by concerns
over transparency.
Exchange-traded contracts,16 on the other hand,
can improve the transparency of credit derivatives, but
the contracts must be written on observable bench
marks such as numbers of bankruptcies or bond prices.
As pointed out earlier, the use of benchmarks for
credit exposure involves basis risk.

Conclusion
1 have shown that under certain circumstances,
credit derivatives replicate the reduction in credit risk
accomplished by loan loss provisions. Using a oneperiod insurance contract to illustrate the functions
of a credit derivative, 1 compared the costs of credit
derivative contracts and loan loss provisions. When
the loan-provision amount is greater than the cost of
the credit derivative, the bank can increase its loans.
When the additional income from loans exceeds the
risk-adjusted opportunity cost of the loan provisioning,
the bank will find that credit derivatives dominate
loan loss provisions.
1 then priced the insurance contract using the
binomial model of Cox, Ross, and Rubinstein (1979).
This price represents a lower bound for the insurance
contract. Credit insurers will require compensation
for any adverse selection. Smith and Warner (1979)

explain the existence ofjoint benefits from contracts
structured to mitigate contracting problems. One
solution to this adverse selection problem is the speci
fication of drivers for contract cash flows determined
outside the bank. Use of an externally determined
driver will generally be less well correlated to the loss
experience of any single institution. This creates a
tradeoff between the adverse selection premium and
the cost incurred when the credit derivative fails to
cover the loss experience, that is, basis risk.
A contribution of this article is the identification
of two problems faced by the emerging credit derivative
contract market. The first is the reluctance of bank
regulators to permit relief from regulatory capital require
ments. The second is that contracts that successfully
avoid adverse selection problems are likely to have
broader appeal. These will generally be contracts
whose payouts are determined by performance indexes
mimicking the loss experience of many institutions.
It follows that liquidity will be greatest for contracts
based on external drivers, further increasing their
cost effectiveness over other forms of credit deriv
ative contracts.
I have shown how credit derivatives can be used
to lower the capital costs of banks, in particular, their
costs for holding regulatory capital. I have also shown
that credit derivatives can replicate the cash flows pro
vided by provisioning for loan losses. When this insur
ance function is accomplished at low cost, the bank
can increase its lending activities. Thus, outlays
made for credit derivatives can dominate the returns
offered by the safe-asset holdings generally used for
loss provisioning purposes.

NOTES
'KMV are the initials of the three founding partners of the
KMV Corporation, Steve Kealhofer, John Andrew McQuown,
and Oldrich Vasicek. Their method is described in McQuown
(1993).

6This concern is not without merit. Hartmann (1996) points
out that credit derivatives offer a speedier route for increasing
credit risk exposure. Banks may be tempted to use this route to
gamble for resurrection when capital levels are low.

2Both an overview and a technical description of CreditMetrics
are available on the Internet at www.riskmetrics.com/cm/
index.html.

7Certain accounting and tax benefits can also be derived by
retaining title to the underlying assets.

3For detailed coverage of this product, see the Internet site at
www. csfp. csh. com/csfpfod/html/csfp_ 10 .htm.
4This article covers the use of credit derivatives by financial
institutions. Frost (1997) describes corporate use of these
contracts.

5For a thorough description of the DPC structure, see Remolona,
Bassett, and Geoum (1996).

8Implicitly, the covariation between the interest rate and default
probability is also presumed to be zero.
9This is a more restrictive policy than the accountant’s use of
this term. A later section further develops the idea of loan loss
provisioning.

10This case can also be made by pointing out that the bank can
now choose between the linear combinations of default-free
investments earning r and risky loans earning r . The bank will
generally value this expansion of its opportunity set.

Economic Perspectives

“This view raises the concern that financial institutions pro
hibited from engaging in insurance activities may be prohibited
from participating in credit derivatives.

12An example of the Diamond intuition is the following. A bank
is constrained from accepting new loans because it is at its total
allowable level of risk. Were the bank able to increase its lending,
a portion of its present risk level could be eliminated though
diversification. A derivative can be used to reduce its exposure
to undiversifiable risks, allowing the bank to then increase
lending and lessen risk through diversification.
“When the up state pays zero, this point can be understood
through the insurance pricing equation above. Since both the
provisioning outlay and the credit derivative are discounted at r,
this interest rate impact is the same for both alternatives.
However, the down state payoff is also weighted by a term that
includes u - r in the denominator. As r rises, the weight declines
increasing the effect of an interest rate change on the credit
derivative.

14Net worth is the residual of assets after subtracting the pay
ments owed to all holders of nonequity claims; that is, depositors
and owners of debt. For purposes of this discussion net worth
can be construed as the value of the equity claims on a publicly
owned institution.

“The Financial Services Authority (FSA) has taken over
supervisory responsibility for UK banks. Releases by the FSA
appear to conform with the earlier policy defined by the
Bank of England. The releases are Board Notice 482 and
Board Notice 414,

“For example, the Chicago Mercantile Exchange recently
announced a futures contract on personal bankruptcies.

REFERENCES

British Bankers’Association, 1996, The BBA Credit
Derivatives Report, available (for purchase) on the
Internet at www.bba.org.uk/pubslist2.htm, November.

McQuown, J. A., 1993, “Market vs. accounting-based

measures of default risk,” KMV Corporation, proprietary
paper, available on the Internet at www.kmv.com/
insider/pdf.html, September.

Cox, John, Stephen Ross, and Mark Rubinstein,

1979, “Option pricing: a simplified approach,” Journal
ofFinancial Economics, Vol. 7, No. 3, September,
pp. 227-263.
Diamond, Douglas W., 1984, “Financial intermedia
tion and delegated monitoring,” Review of Economic
Studies, Vol. 51, No. 3, July, pp. 393-414.
Duffee, Gregg R., and Chunsheng Zhou, 1998,

“Credit derivatives in banking: Useful tools for manag
ing risk?,” Board of Governors of the Federal Reserve
System, Washington DC, working paper.

Remolona, Eli M., William Bassett, and In Sun
Geoum, 1996, “Risk management by structured deriv

ative product companies,” Economic Policy Review,
Federal Re serve B ank of New York, April, pp. 17-3 4.
Smith, Clifford W., Jr., and Jeremy B. Warner, 1979,
“On financial contracting: An analysis of bond
covenants,” Journal of Financial Economics, Vol. 7,
No. 2, June, pp. 117-161.
Smithson, Charles, 1997, Presentation given in June.

Watterson, Paul N., Jr., and Conrad G. Bahlke, 1997,
Frost, Joyce, 1997, “Corporate uses for credit deriva

tives,” International Treasurer, March.

“Credit derivatives 1997: Recent legal and regulatory
developments,” Futures dr Derivatives Taw Report,
March.

Grossman, Sanford J., 1988, “Insurance seen and
unseen: The impact on markets,” The Journal of
Portfolio Management, Summer.

Kwan, Simon, 1997, “Recent developments in loan loss

provisioning at U.S. commercial banks,” Economic
Tetter, Federal Reserve Bank of San Francisco, July.

Federal Reserve Bank of Chicago

Assessing the condition of Japanese banks:
How informative are accounting earnings?

Hesna Genay

Introduction and summary
There is little doubt about the current weak condition
of Japanese banks. Although they were never as
profitable as European or U.S. banks, Japanese banks
grew rapidly in the 1980s, buoyed by a strong domestic
economy and rapidly increasing asset prices. In 1980,
only one Japanese bank made the list of the ten
largest banks in the world, compiled by The Banker
magazine. By 1990, the four largest banks, and six
of the top ten, were Japanese. Moreover, the rapid
growth of Japanese banks was not confined to domes
tic markets. According to statistics compiled by the
Bank for International Settlements (BIS), the share of
Japanese loans in total international claims outstanding
was less than 20 percent in the early 1980s. By the end
of the decade, Japanese banks accounted for over
one-third of international bank assets (BIS, 1998). By
1990, the size and rapid expansion of Japanese banks
had earned them the moniker “mighty giants” of Japan.
The tide that carried Japanese banks to the top
ranks of international banks transformed into a series
of tsunami in the 1990s. The first signs of trouble
emerged with sharp declines in Japanese stock and
land prices. As a result, Japanese banks, which had
extensive equity holdings and loans collateralized by
real estate, saw significant declines in the value of
their assets and capital positions. The collapse of
U.S. commercial real estate prices and the 1990-91
recession in the U.S. put further pressure on Japanese
banks, which had invested heavily in this market. The
response of the Japanese banks was to turn to new
markets, the then rapidly growing South East Asian
economies. The current Asian crisis and the unresolved
asset quality problems in Japan have escalated
the amount of problem assets at Japanese banks to
dangerous levels.
Today, even the best performing Japanese banks
are facing liquidity pressures and some are struggling
to stay afloat. As of October 1998, the official amount

of nonperforming loans at Japanese banks was $600
billion, while some private analysts put the amount of
bad loans at over $1 trillion, representing roughly 20
percent of total loans outstanding. Moreover, accord
ing to some analysts, the Japanese banking system
has a shortfall of¥8 trillion in real net worth, even
after the injection of ¥10 trillion to ¥25 trillion in public
funds that is expected as a result of the ¥60 trillion
rescue plan passed in October 1998 by the Japanese
parliament. The impact of the crisis on the Japanese
economy and other financial markets is significant—
low rates of corporate investment and curtailed lending
are, at least partially, the result of problems in banking.1
Of course, other countries have also faced financial
crises.2 Among the more notable was the thrift and
banking crisis in the U.S. in the late 1980s and early
1990s, which resulted in the closure of 1,142 savings
and loan (S&L) institutions and 1,395 commercial
banks (Lindgren, Garcia, and Saal, 1996).3
Because banks are a source of funds for firms,
have an important role in the transmission of mone
tary policy, and are an integral part of the payments
system, the social costs of bank failures may be great
er than those of other types of businesses. Previous
studies on the determinants of bank profitability and
the likelihood of bank survival have shown that certain
bank characteristics are important in determining future
bank performance.4 Following this literature, I examine

Hesna Genay is an economist at the Federal Reserve
Bank of Chicago. The author would like to thank Elijah
Brewer III, George G. Kaufman, David Marshall, and
seminar participants at the Federal Reserve Bank of
Chicago for their valuable comments. The author would
also like to thank Mark Kawa and Evelyn Espina of the
Bank’s Supervision and Regulation Department for the
background information and data they provided, and
Scott Briggs, Thong Nguyen, and Praveen Chennareddy
for their diligent research assistance.

Economic Perspectives

the relationship between the performance of Japanese
banks in 1991-97 and their characteristics. In particular,
I focus on three questions.
One, how does the accounting performance of
Japanese banks relate to their financial characteristics?
Although the banking crisis in Japan is well recognized,
the precise financial condition of the banks and the
amount of, and losses from, their nonperforming loans
are uncertain. Differences between the disclosure,
accounting, and regulatory rules in Japan and other
industrial economies make it difficult to assess the exact
condition of Japanese banks and compare them with
other international banks. Furthermore, some analysts
interpret recent Ministry of Finance (MoF) actions
(such as allowing banks to value their security hold
ings at cost to avoid reporting valuation losses) as
attempts to mask the true condition of the banks; as a
result, they consider the reported results of Japanese
banks to be of little or no value. If the patterns between
bank performance and characteristics established in
previous studies are also evident in the Japanese
banking system, then even if the reported numbers
are not accurate, they would still provide usefiil signals
of bank performance.
Two, how does the stock market performance
of banks relate to their financial characteristics? In
particular, are the patterns between stock returns,
which are less subject to potential maneuvering by
banks, and financial characteristics consistent with
those observed in the accounting returns? If Japanese
accounting, disclosure, and regulatory practices
obscure the true performance of Japanese banks, then
the relationship between accounting earnings and
bank characteristics might not be consistent with that
observed in other countries. However, if market partici
pants are aware of these practices and their impact on
the condition of the banks, then market-based measures
of bank performance, such as stock returns, would
be little affected by these practices. As a result, any
inconsistency we might observe with accounting re
turns would not be evident in stock returns.
Three, how are the stock market and accounting
returns of banks related? Are the stock returns corre
lated with the accounting returns, or do shareholders
dismiss the accounting results as meaningless? If
accounting and disclosure practices of Japanese
banks obscure their condition to such an extent that
there is no additional information in their reported
results, then there would be no significant relationship
between accounting and stock returns.
Throughout the analysis, I explore potential differ
ences in these relationships among different types of
Japanese banks and over time. For the most part, press
reports and other analyses of Japanese banks focus on

Federal Reserve Bank of Chicago

the major banks (city, trust, and long-tern credit banks),
which account for more than 70 percent of Japanese
banking assets; however, their activities and character
istics differ significantly from those of regional banks.
Furthermore, the activities of banks, underlying eco
nomic conditions, and regulatory practices have
changed over time. These differences in bank charac
teristics and changes in the environment can potentially
influence the relationships I examine.
The results using accounting measures of per
formance indicate that some measures of asset quality
are significant determinants of Japanese banks’ earn
ings; and the relationships I document are consistent
with the results of previous studies. However, the
accounting returns of Japanese banks exhibit some
unexpected correlations with the market index, increases
in the number of business bankruptcies, and bank
capital. For instance, bank profitability, measured by
return on equity (ROE), is negatively correlated with
returns on the market index, indicating that banks are
less profitable when the stock market is performing
well. Further analysis shows that this and other puzzling
results with accounting earnings might be the result
of banks’ loan loss provisioning practices. In particular,
Japanese banks appear to increase their loan loss
provisions when their core profits and stock market
returns are high.
The results with banks’ stock returns show that
such income-smoothing does not affect their market
performance. Specifically, when performance is mea
sured by market returns, the puzzling results observed
with accounting returns disappear and we observe
correlations with the market index and the number of
bankruptcies consistent with expectations.
Despite the potential problems with the reported
earnings of Japanese banks, my results suggest that
accounting returns provided market participants with
useful information on banks’ condition in 1991-94:
Accounting and stock market returns are positively
and significantly correlated during this period. However,
the results also show that this relationship breaks down
in 1995-97, implying that the usefulness of reported
earnings has deteriorated in recent years.
As indicators of bank performance and character
istics, I use measures used by regulators and market
participants to assess the financial condition of banks.
My results suggest that Japanese accounting, dis
closure, and regulatory practices might have driven a
wedge between banks’ accounting and stock returns in
recent years. To the extent that such practices make
it more difficult to assess the condition of banks,
they introduce additional uncertainty to the market,
potentially increasing the risk premium required by
investors. The “Japanese premium”—the difference

1 3

between the interest rates paid by Japanese and other
international banks in the interbank markets—might
be considered a manifestation of this uncertainty.
Regulatory forbearance that allows economically
insolvent institutions to continue operations and
extends implicit or explicit guarantees to uninsured bank
claimants transfers wealth from deposit insurance
agencies, and hence taxpayers, to the shareholders
and debtors of insured institutions. The results in
previous theoretical and empirical studies indicate that
as a bank nears insolvency, more of its value is derived
from the value of subsidies and forbearance and the
correlation between stock market returns and the value
of the underlying assets declines. According to
Brickley and James (1986) and others, a bank has (in
addition to its tangible assets) a valuable intangible as
set in the form of access to underpriced, fixed-premium
deposit insurance and government forbearance
programs that modify insolvency rules. The capital
ized value of this intangible asset is embedded in the
bank’s stock market valuation, but is not reflected in
accounting values. When most of the market value of
an insured bank is in the form of this intangible asset,
movements in common stock returns need not be cor
related with movements in the value of the underlying
assets. In recent years, Japanese regulators have
delayed recognition of losses at banks and have
been reluctant to take strict actions against troubled
or insolvent institutions. Such regulatory forbearance
might account for the lack of correlation between
accounting and market returns of Japanese banks in
1995-97 when the deterioration in the banks’ financial
condition accelerated significantly. More recently, the
MoF has taken a number of steps to shore up banks’
reported capital base through accounting changes
and injection of government funds and has extended
government guarantees to all bank creditors through
the end of March 2001. These actions evoke recollec
tions of the initial response of regulators to the S&L
crisis in the U S.5 Experience with that crisis tell us that
regulatory forbearance can be a leaking lifeboat that
imposes significant costs on the economy and healthy
financial institutions, instead of the intended lifeline to
pull troubled firms to safety.6 If the financial revitaliza
tion laws passed by the Japanese parliament in October
1998 put an end to regulatory forbearance and allow
orderly resolution of insolvent institutions, they might
minimize the fiiture adverse impact of the banking crisis
on the economy.

Overview of Japanese banking7
Until the 1980s, functional segmentation, exten
sive regulations, restricted competition, government
intervention, and isolation from international markets

were the defining characteristics of Japanese financial
markets. The Japanese banking system underwent a
series of reforms in the late 1970s and 1980s (outlined
in appendix 1); however, the current system retains
some of its traditional characteristics.
To a certain extent, the markets are still segmented
across banking functions. Until the passage of the 1992
Financial System Reform Law, different institutions
conducted commercial, trust, and investment banking.
Similarly, until recently, different banks provided
short-term and long-term business loans. City and
regional banks traditionally provided short-term
financing to companies and were restricted to issuing
short-term liabilities. City banks traditionally have
focused on providing financing to large corporations
and have relied on large corporate deposits and Bank
of Japan credit for their funding. City banks were also
among the first Japanese banks to expand overseas.
The traditional business of regional banks, on the other
hand, has been the provision of short-term loans to
small- and medium-sized companies. Through their
branch network in their home prefecture and close
community ties, regional banks have relied primarily
on deposits from their loan customers and individuals
for funding.
Long-term business loans are provided by the
long-term credit and trust banks. Until recently, only
these institutions were allowed to issue long-term
liabilities. On the asset side of the balance sheet,
long-term credit banks provided commercial loans,
while trust banks focused on trust loans. Regulations
restricted long-term credit banks to issuing deposit
liabilities only to their borrowers and restricted trust
banks to raising fiinds through loan and money trusts.
However, over time, deregulation and increased com
petition among financial institutions have blurred the
lines separating the businesses of Japanese banks.
The regulations and laws governing banking
operations are formulated, implemented, and enforced
by the MoF. Until April 1998, when a new, independent
Financial Supervisory Agency (FSA) was established,
the MoF was the primary regulator of banks.8 Although
the MoF has the legal authority to license banks, en
force laws, and administer penalties for violations of
laws and regulations, it relies primarily on administrative
guidance for enforcement. Because one of the fiinctions
of the Bank of Japan is to ensure the safety and sound
ness of the financial system, it also has regulatory and
supervisory purview over banks, albeit to a lesser extent
than the MoF. Until the establishment of the FSA, both
institutions conducted examinations of banks.
Other government institutions in the Japanese
banking system include the Deposit Insurance Cor
poration, which insures bank deposits and collects

Econom ic Perspectives

insurance premiums, and the Resolution and Collection
Bank, which was established in 1995 to take over the
assets of failed institutions.
Despite the deregulation of banking activities in
recent years, Japanese banks have characteristics that
reflect their traditional roles. Some of these characteris
tics are evident in table 1, which shows the aggregate
balance sheets of four types of Japanese banks as of
the fiscal year ending March 31,1997.9
For banks that have traditionally provided long
term financing (long-term credit and trust banks), loans
excluding loan loss reserves (gross loans) represent
approximately 65 percent of total assets. Gross loans
account for approximately 72 percent of the assets of
city and regional banks that have traditionally provided

short-term financing. However, despite the greater
concentration of assets in loans, city and regional
banks have smaller loan loss reserves (both as a per
centage of assets and of loans) than long-term credit
and trust banks. The differences in loan loss reserves
might reflect differences in the composition of loan
portfolios of these institutions. For instance, on March
31, 1997, the credit exposure of the three long-term
credit banks to the riskier real estate, construction,
and finance sectors was 44.43 percent of their domestic
loan portfolio; loans to these three sectors represented
27.14 percent of the domestic loans at city banks.
The four types of banks invest roughly the same
fraction of their assets in securities. However, major
banks invest more in the equity of other companies,

TABLE 1

Balance sheets of U.S. and Japanese banks, March 31,1997

(percent unless indicated)
U.S. banks

Japanese banks

1.90

0.76

0.57

94.35
72.44
1.64
70.81
14.16
6.61
9.39

91.28
65.53
2.08
63.45
21.18
7.36
6.65

0.54

Large8

All

1.67

1.57

6.60

7.09

96.73
65.40
2.21
63.18
19.67
7.83
13.88

96.59
72.64
1.11
71.53
19.11
2.70
5.95

94.97
71.02
1.59
69.43
16.93
5.67
8.60

87.13
59.74
1.15
58.58
17.52
0.48
11.03

86.13
59.66
1.21
58.45
14.77
0.47
12.91

0.31

0.54

1.06

0.67

1.40

1.30

96.65
89.35
21.39
7.30

96.96
33.55
2.35
63.41

97.56
37.65
2.92
59.91

95.92
93.24
22.93
2.68

96.57
79.15
17.88
17.42

91.60
68.82
12.29
22.78

92.01
64.73
12.10
27.28

3.35

3.04

2.44

4.08

3.43

8.41

7.99

100.00

117

9,451

404

3,218.00

654.87

723.31

1,917.35

6,513.54

4,641.73

3,653.53

0.78

0.16

0.17

0.46

1.57

0.59

0.46

49.40

10.05

11.10

29.44

100.00

78.71

2.26

3.17

3.38

1.52

2.24

1.93

2.03

Earning assets
Gross loans
Loan loss reserves
Net loans
Security holdings
Equity investments
Other earning assets

Fixed assets
Total liabilities
Total deposits
Demand deposits
Other liabilities

Equity capital

Number of banks6

Total assets ($ billions)11

Long-term credit

All

Cash

Total assets or total
liabilities + capital6

City

Total assets/nominal GDP
Total banking assets, %
Loans loss reserves/
total loans, %

Regional

Trust

aLarge banks in the U.S. are defined as those with more than $1 billion in total assets as of March 31, 1997.
bBy Japanese accounting rules, loan loss reserves are recorded as a liability rather than as a contra-account for loans, which is the
practice in the U.S. To make the Japanese and U.S. figures comparable, the Japanese numbers were recalculated per U.S. practices.
cThe Japanese sample includes all banks for which there were 1997 data in Fitch-IBCA's Bankscope database and does not exclude
any observations. The differences in the number of banks in this table and in the tables that follow arise from exclusion of extreme
values and the requirement (by the nature of the analysis) that the banks in the following analysis have more than one year of data.
dYen amounts are translated to dollars at ¥123.72/$, the rate that was in effect on March 31,1997.
Note: Columns may not total to 100 percent due to rounding errors or missing categories.
Sources: Author's calculations based upon data from Federal Deposit Insurance Corporation, FDIC Statistics on Banking, available on
the Internet atwww.fdic.gov/databank/sob, 1998, and FitchIBCA, Bankscope, CD-ROM, 1998.

Federal Reserve Bank of Chicago

whereas regional banks invest more in Japanese public
bonds. Equity investments account for less than 3
percent of regional banks’ assets, but they represent
approximately 7 percent to 8 percent of major banks’
assets. The relatively greater investment in equity
securities reflects the major banks’ role in the indus
trial groups, the keiretsu, as major stockholders of
group companies.10
Japanese banks also differ in how they fund their
assets. Compared with major banks, regional banks
fund a greater percentage of their assets with equity
capital. Moreover, long-term credit and trust banks rely
less on deposits (less than 40 percent of funding) than
city and regional banks do (around 90 percent), reflect
ing the restrictions placed on deposit-taking at these
long-term finance institutions.
Not only do various Japanese banks have different
characteristics, but they also differ from U.S. banks in
terms of their activities and characteristics. Table 1 also
shows the aggregate balance sheets of all commercial
banks in the U.S. and the balance sheets of U.S. banks
with more than $ 1 billion in assets. Although there are
more than 9,000 banks in the U.S., bank assets in the
U.S. total to $4.6 trillion, compared with $6.5 trillion in
assets of 117 Japanese banks. Furthermore, bank assets
represent more than 1.5 times the Japanese nominal
gross domestic product (GDP), compared with 60 per
cent of nominal GDP in the U.S., reflecting the greater
role of banks in the Japanese economy.
Japanese and U.S. banks also differ in the extent
of leverage and composition of assets. Japanese banks
are more than twice as leveraged as U.S. banks. While
equity capital funds approximately 8 percent of U.S.
bank assets, it funds less than 4 percent of Japanese
assets. Japanese banks also invest more of their assets
in loans than U.S. banks. Loans excluding loan loss
reserves account for nearly 70 percent of Japanese
bank assets, but less than 60 percent of U.S. bank
assets. Loan loss reserves, as a fraction of both total
assets and gross loans, are higher at Japanese banks,
reflecting the differences in the conditions of the two
banking markets. However, as figure 1 shows, Japanese
banks’ loan loss ratios surpassed those of U.S. banks
only in 1997. In the early 1990s, U.S. banks’ loan loss
reserves covered 2.5 percent of their loans, compared
with less than 1 percent coverage for Japanese banks.
Japanese banks began to reserve for possible loan
losses aggressively only in 1996.
While the total amount of securities investment
is similar for Japanese and U.S. banks (approximately
17 percent of total assets), Japanese banks have
significantly more equity investments (nearly 6 percent
of assets) than U.S. banks (less than 0.5 percent of

assets), which are generally prohibited from making
such investments.
In addition, Japanese banks rely on deposits as
a source of funds more than U.S. banks. Deposits
fund nearly 80 percent of Japanese bank assets, but
less than 69 percent of the total assets of U.S. banks.
Other liabilities (such as fed funds purchases and
other nondeposit liabilities) account for 27 percent
of the assets of large U.S. banks, but only about 17
percent of the total assets of Japanese banks.
Figure 2 shows the performance of Japanese banks
relative to U.S. banks over the 1990-97 period. In
1990, when the U.S. was approaching the end of its
banking crisis, U.S. and Japanese banks had similar
ROEs and operating profits (figure 2, panels A and B).
However, when Japanese and U.S. banks are com
pared in terms of narrower performance measures,
such as operating profits before loan loss provisions
and interest margins, U.S. banks were more profitable
than Japanese banks even in 1990 (figure 2, panels
C and D). Hence, it was the higher level of loan loss
provisions at U.S. banks that made their performance
in 1990 comparable with that of Japanese banks.
Since 1990, the performance of Japanese and
U.S. banks has diverged significantly. During the
1990-97 period, U.S. banks improved their performance
by most measures, while Japanese banks’ performance
deteriorated.11 By 1997, Japanese banks were reporting
negative ROEs, while U.S. banks were enjoying record
levels of profitability. The differences are all the more
remarkable when performance is measured by return
on assets (ROA). During 1990-97, the average ROA
for U.S. banks was 0.95 percent, compared with 0.04
percent for Japanese banks. (The relative performance
of Japanese banks is poor even if they are put on a more
equal footing with U.S. banks in terms of underlying

Economic Perspectives

FIGURE 2

Performance of U.S. and Japanese banks
B. Operating profits/equity

C. Operating profits before LLP/equity

D. Interest margins

Notes: Panel A—return on equity is the ratio of net income to equity. Net income is the sum of operating profits, gains or losses on
equity, and special items minus income taxes. Panel B—operating profits is the sum of net interest revenue and net income from
other operations minus loan loss provisions. Panel C—operating profits before provisions is the sum of net interest revenue and net
income from other operations. Panel D—interest margins are defined as the ratio of net interest revenue to total earning assets.

Sources: Author's calculations based upon data from Federal Deposit Insurance Corporation, FDIC Statistics on Banking, available
on the lnternetatwww.fdic.gov/databank/sob, 1998, and FitchIBCA, Bankscope, CD-ROM, 1998.

economic conditions. For example, in 1987-91, when
the U.S. was in the midst ofa major banking crisis, U.S.
banks averaged 7.4 percent ROE, versus 0.3 percent
for Japanese banks in 1991-97.12)
The stock returns of Japanese banks reflected
their poor performance in 1990-97. As figure 3 shows,
Japanese banks had negative stock returns in five of
the eight years and underperformed the market in
seven of the eight years. Figure 3 also shows the
adverse impact of declining stock prices on the value
of Japanese banks, which hold significant amounts of
equity in other firms.
Table 1 and figure 2 show the differences in the
characteristics of different Japanese banks and the
poor performance of Japanese banks relative to U.S.
banks. How do the characteristics of Japanese banks
relate to their performance? Are the relationships be
tween the performance and characteristics of Japanese
banks similar to those observed in the U.S.? Next, I
examine these issues in more detail.

Federal Reserve Bank of Chicago

Performance and financial characteristics
A number of studies have examined the perfor
mance of banks and related it to bank characteristics
and activities. Because solvency of banking institutions
is of particular importance to the stability of financial
systems and because there were a large number of
failures among banks and S&Ls in the U.S. during
the 1980s, several studies have focused on factors
that determine the profitability and solvency of
depository institutions.13
Following this literature, I examine the ROE and
the stock market performance of Japanese banks in
1991-97.141 relate these performance measures to
bank characteristics that were found to be particularly
important determinants of bank performance in previous
studies: asset quality, capital ratio, liquidity, operational
efficiency, and size.
The relationship between asset quality and bank
earnings is closely related to the condition of the
overall economy. Banks that invest in riskier assets

result in Japan, that is, a negative relationship between
measures of asset quality and realized performance
of Japanese banks in 1991-97, when the Japanese
economy was subject to adverse shocks. I measure
asset quality and credit risk by the following variables:
the ratio of equity investments to total assets, the
ratio of loan loss provisions to loans, the ratio of net
loans to total assets, the ratio of domestic loans to
total loans, and the growth rate of assets (see box 1
for variable definitions).
The ratio of equity investments to total assets
measures the banks’ exposure to the performance of
other firms through their equity investments. As general
economic conditions deteriorate, the performance of
banks with a relatively high fraction of their assets
invested in the equity of other firms should be worse
than that of banks with lower equity exposure. Further
more, because equity securities are generally more
risky than debt securities, banks with more equity
investments may have lower realized profits when
stock prices decline. On the other hand, if equity in
vestments provide banks with more opportunities
for diversification, then banks with high fractions of
assets invested in equities would perform better than
other banks.
The ratio of loan loss provisions to loans can be
positively or negatively correlated with performance.

FIGURE 3

Japanese stock market performance
percent
30 r

1990

’91

’92

’93

’94

’95

’96

’97

Notes: Returns on banks stocks are holding period returns and are
calculated as the change in price from the end of the previous fiscal
year plus dividends paid during the year divided by the price at the end
of the previous fiscal year. Returns in the overall stock market are
calculated as percentage changes in the Tokyo Stock Exchange TOPIX
index over a fiscal year.
Sources: Author's calculations based upon data from the Tokyo Stock
Exchange, database, 1998, and Bloomberg Financial, database, 1998.

are likely to have higher expected profits. However,
higher asset risk implies lower realized profits when
the economy is experiencing a series of negative
shocks. Previous studies found that depository in
stitutions in the U S. that invested in riskier, or lower
quality, assets performed worse than others during
the 1980s and early 1990s. One would expect a similar

BOX 1

Definitions of variables

Annual stock returns—annual holding period

Loan loss provisions/loans—Loan loss provi

returns calculated as the change in stock price
plus dividends paid in the current period over
the previous period’s stock price.
BIS capital ratio—total risk-weighted capitalasset ratio as defined by the BIS.
Business bankruptcies—annual change in the
number of business bankruptcies, in percent.
Domestic loans/total loans—Domestic loans
divided by total (gross) loans.
Equity investments/TA—equity investments at
book value divided bv total assets, in percent.
Gross loans/TA—loans before loan loss reserves
divided by total assets, in percent.
Growth of TA—annual grow th rate of total assets,
in percent.
Interest margin—net interest revenue divided by
earning assets (loans plus investments), in percent.
Liquidity—demand deposits divided by bank
deposits plus cash plus securities in the trading
account; ratio of short-term liabilities to short
term assets, an inverse measure of liquidity.

sions (transfers to reserves, loan charge-offs,
loss on sale of loans to CCPC, write-off/down
of sovereign risk, loss shouldered for custom
ers, transfer to reserve for other credit losses,
write-down of other assets) divided by banking
loans (excludes trust loans).
Market return—annual change in the Tokyo
Stock Exchange TOPIX index, in percent.
Net loans/TA—net loans divided by total assets,
in percent.
Overhead ratio—personnel and noninterest ex
penses divided by earning assets (loans plus in
vestments), in percent.
ROE—net income divided by total book-value
capital. Net income includes operating profits,
gains/losses on sale of equity investments,
valuation losses on equity investments, special
items, and income taxes.
SIZE—total assets in logarithms.
TA—total assets in trillions of yen.

Economic Perspectives

If banks with riskier assets provision more than
other banks, then loan loss provisions measure credit
risk, and are likely to be negatively correlated with
realized profits. On the other hand, if banks that perform
better, or banks with more conservative management,
provision more for loan losses, then one would expect
a positive relationship between loan loss provisions
and performance.15 Empirical evidence onU.S.
banks shows that loan loss provisions and loan
loss reserves are negatively correlated with future
bank performance.16
The ratio of net loans to total assets measures
the banks’ credit risk, and the ratio of domestic loans
to total loans measures their domestic exposure. During
the sample period, loan quality, particularly the quality
of loans made to Japanese borrowers, was one of
the largest sources of risk to bank profitability. Con
sequently, one would expect banks with higher ratios
of loans to total assets and banks with more domestic
loans in their portfolio to have poorer performance
than other banks.
I also measure asset quality with the annual
growth rate of assets. During the U.S. thrift crisis,
some institutions tried to grow out of their problems
by expanding rapidly. Furthermore, additions to assets
at fast-growing institutions may increasingly involve
riskier assets. As a result, one might observe a negative
relationship between asset growth and realized profits.
On the other hand, if regulators are providing suffi
cient discipline, they may restrain the growth of in
stitutions that are in financial trouble and allow only
strong-performing banks to expand. Alternatively,
banks that grow relatively more may previously have
had good performance and/or expect to have good per
formance in the fiiture. In that case, one would observe
a positive relationship between growth and profitability.
In theory, performance can be positively or neg
atively related to capital ratios.17 For instance, in perfect
and competitive capital markets, higher capital ratios
would reduce risk and expected return on equity
(but would not change the weighted average cost of
funds). Moreover, because interest payments are tax
deductible, relying more on equity and less on debt
reduces after-tax earnings, generating a negative rela
tionship between earnings and capital. However, other
factors may lead to a positive relationship between
the capital and earnings of banks. Because banks retain
a portion of their earnings, over time more profitable
firms would have higher retained earnings, hence more
capital, than less profitable firms. Furthermore, equity
capital provides a cushion against losses, lowering
bankruptcy costs. In imperfect capital markets, banks
with more capital and lower bankruptcy costs are likely

Federal Reserve Bank of Chicago

to have lower interest costs and higher profitability
than other banks. In addition, when deposit insurance
is present and regulators have the authority to close
insolvent institutions, banks with profitable investment
opportunities have an incentive to be well capitalized
(Buser, Chen, and Kane, 1981; Keeley, 1990; and
Demsetz, Saidenberg, and Strahan, 1996). All these
factors point to a positive relationship between bank
performance and capital-asset ratios. Empirical evidence
indicates that banks with higher capital-asset ratios
are indeed more profitable and less likely to fail than
more leveraged banks. In this article, I measure the
capital position of Japanese banks by the ratio of
capital to risk-weighted assets, as defined by the
BIS capital accord.
Profitability is also related to liquidity. More liquid
banks are better able to meet adverse shocks and are
likely to face lower cost of funds in imperfect capital
markets, increasing their profitability. On the other hand,
liquid assets have lower expected returns than illiquid
assets, so banks with more liquid assets might have
lower expected earnings. In addition, banks choose
the level of liquidity of their assets. Therefore, if a bank
expects to face adverse shocks in the future, it may
choose to hold more liquid assets to cushion itself
against such shocks. In that case, one would observe
a negative relationship between profitability and liquidity,
since banks that expect lower profits would increase
their liquidity. In short, the relationship between
liquidity and profitability is ambiguous in theory and
is determined by the data. Empirical evidence points to a
positive relationship between liquidity and performance
of banks in the U.S. I measure liquidity by the ratio
of short-term liabilities to short-term assets, whereby
banks with higher ratios are less liquid than others.18
Operational efficiency, measured by the overhead
ratio, is also likely to be a key determinant of bank
profitability. To the extent that banks with high over
head ratios are less efficient, one would expect these
banks to perform worse than banks with lower over
head expenses. However, the overhead ratio is an
imperfect measure of efficiency and may also reflect
differences in banks’ product mix. For instance, nontraditional bank businesses may generate greater prof
its, but require more overhead expenses than traditional
banking. In that case, one would observe a positive
relationship between profitability and overhead ratios.
In general, previous studies have found that banks
with high overhead expenses perform worse than
other banks.
I also include size, measured by total assets, as a
control variable. Previous studies found that large
banks perform better than small banks.

1 9

In addition to these bank characteristics, I explore
the relationship between bank performance and
measures of aggregate economic activity. In particular,
I focus on stock market returns and the number of
business bankruptcies.19 As economic conditions
deteriorate, the number of bankruptcies increases. As
creditors, banks are directly affected by bankruptcies.
Hence, one would expect an increase in the number
of bankruptcies to be associated with higher loan
defaults and lower bank profits.
As noted above, Japanese banks have significant
investments in the equity of other firms. Therefore,
returns on the overall stock market affect the perfor
mance of banks, not only as an indicator of aggregate
economic conditions, but also through their impact
on the valuation of banks’ investments. As a result,
one would expect bank performance to be positively
correlated with returns in the stock market. Clearly,
banks with a relatively high fraction of their assets in
equity securities should benefit more from stock price
increases than other banks. To explore this relationship,
I interact the return on the market index with the ratio
of equity investments to total assets. If an increase
in the market index has a greater positive impact on
the performance of banks with more equity investments,
then the coefficient on the interaction term would
be positive.
My analysis is based on accounting results for
city, trust, long-term credit, and regional banks in 1991—
97 from FitchlBCA’s (1998) Bankscope database. My
initial analysis showed some extreme values of ROA,
ROE, and growth rate of assets, which were attributable
to mergers or insolvency. To avoid influencing the
results by including these extreme values, I deleted
observations in the top and bottom 1 percentile of
the distribution of ROA, ROE, and the growth rate of
total assets.20 The final sample contains 555 obser
vations for 88 banks.
The data also include daily stock prices of city,
trust, and regional banks for 1991-97 If om Bloomberg
(1997).21 Annual holding-period returns are constructed
using daily stock prices and dividend payments as
reported in the various editions of the Japan Company
Handbook ('ToyoKeizai, Inc., 1991-98).
The top panel of table 2 shows the mean values,
the standard deviations, and the minimum and maximum
values of the variables for the entire sample. The
average reported earnings and stock returns reflect
the poor performance of Japanese banks during this
period. Despite the exclusion of extreme values from
the sample, profitability varies greatly across banks
and over time. For instance, ROE ranges from -49.21
percent to 9.40 percent, indicating that while some
banks performed very poorly, others reported large,

positive profits. Similarly, banks differed in the
amount of their loan loss provisions. Although the
mean value for provisions was 0.59 percent, some
banks had no loan loss provisions, while others had
provisions as high as 9.61 percent of loans. There
are also differences in the asset composition and
operational efficiency of banks. For instance, net
loans ranged from 45.14 percent to 82.07 percent of
total assets, while equity investments ranged from
0 percent to 9.32 percent of total assets. In summary,
the sample statistics suggest that differences in banks’
characteristics across institutions and over time
might be significant.
The statistics in the bottom panel of table 2, the
mean values for different bank types and different
time periods, present further evidence of differences
in bank characteristics. Major Japanese banks differ
significantly from regional banks and characteristics
of Japanese banks changed significantly in the latter
part of the sample period. In particular, major banks
performed significantly more poorly than regional
banks in 1991-97. The average ROE for major banks
during this period was -0.40 percent, compared with
3.17 percent for regional banks. Other variables also
show significant differences in the characteristics of
major and regional banks, which were foreshadowed
by the statistics presented in table 1. Namely, major
banks invest less in loans but more in equities than
regional banks. Furthermore, major banks are more
liquid and have lower interest margins and overhead
expenses than regional banks. Regional banks also
provision less for possible loan losses. There are,
however, no significant differences in the capital ratios
of major and regional banks.
The last two columns in the bottom panel of
table 2 show the mean values of the variables in
1991-94 and 1995-97, respectively. These statistics
indicate that bank characteristics changed signifi
cantly over time. While there was no significant
difference in bank stock returns in the two periods,
ROEs were significantly lower in the later part of
the sample period.
Over time, Japanese banks also increased their
percentage of assets invested in loans and equity
securities. The increase in the ratio of domestic to
total loans reflects the aggregate decline in the
banks’ international loans. Furthermore, liquidity
of Japanese banks declined significantly in 1995-97,
which may reflect the higher costs of liquidity for
Japanese banks in interbank markets. Lastly, banks
raised their capital ratios and their provisioning for
loan losses in 1995-97. Below, I explore the relation
ship between bank characteristics and performance
more systematically.

Econom ic Perspectives

TABLE 2

Summary statistics

(percent unless indicated)
All banks
Mean
1991-97

Standard deviation
1991-97

Min
1991-97

Max
1991-97

Interest margin

1.58

0.59

-0.19

2.53

ROE

2.34

7.45

-49.21

9.40

-7.30

16.34

-50.01

58.17

TA, ¥ trillion

9.63

14.85

0.96

80.84

Growth of TA

0.33

12.58

-56.21

123.81

Annual stock return3

3.12

1.83

0.00

9.32

Net loans/TA

67.04

6.55

45.14

82.07

Domestic loans/total loans

Equity investments/TA

92.15

16.82

26.51

100.00

Liquidity

2.06

1.52

0.12

15.28

Overhead ratio

1.29

0.34

0.33

1.98

Loan loss provisions/loans

0.59

0.92

0.00

9.61

BIS capital ratio

9.26

0.81

7.28

13.61

Mean values for:
Major banks
1991-97
Interest margin

Regional banks
1991-97

All banks
1991-94

0.70

1.85**

1.47

ROE

-0.40

3.17**

4.41

Annual stock returns3

-9.26

TA, ¥ trillion

31.17

3.17**

Growth of TA

-0.31

0.52

-6.85

-8.32

All banks
1995-97
-j

-0.25* *

-5.96

9.95

9.24

-0.10

0.86

5.97

2.27* *

2.93

3.36**

Net loans /TA

64.04

67.94**

65.77

68.64**

Domestic loans /total loans

93.43

Equity investments /TA

68.14

99.34**

91.13

Liquidity

0.87

2.41 * *

1.58

2.66**

Overhead ratio

0.80

1.43**

1.26

1.32

Loan loss provisions/loans
BIS capital ratio

1.29
9.18

0.38**
9.28

0.24
9.14

1.02**
9.41 **

N (N= 555)
Number of banks

128
20

427
68

309
87

246
85

** indicates differences in means across bank types or subperic ds that are significant at the 1 percent level.
Excludes stock returns for long-term credit b anks.
Note: For variable definitions, see box 1.
Source: Author's calculations from FitchIBCA, Bankscope, CD-RO VI, 1998.

Determinants of accounting performance
Table 3 shows the parameter estimates when
banks’ ROEs are regressed on their characteristics.22
Appendix 2 describes my methodology in more detail.
Banks’ performance over a given period is related to
their characteristics at the beginning of the period;
hence, the results show the predictive power of current
bank characteristics for future performance.

Federal Reserve Bank of Chicago

Are the patterns observed in the Japanese banks’
accounting earnings and characteristics consistent
with those in other countries? The results in table 3
indicate the answer to this question is mixed.
For some characteristics, the relationship with
earnings is consistent with patterns observed in the
U.S. In particular, loan loss provisions and the ratio
of net loans to total assets are negatively correlated

TABLE 3

Returns on equity and bank characteristics
All banks
1991-97

Major banks
1991-97

Regional banks
1991-97

All banks
1991-94

Intercept

0.257

20.384

8.61 6

10.762***

Size

0.705*

0.1 58

0.064

-0.011

Growth of TA

0.000

0.003

0.000

Net loans/TA

-0.1 23* *

-0.122**

-0.028

1 .323

-0.080***

0.933

4.427

1 .850

1.641***

Overhead ratio

1 .448

0.559

1 .441

0.099

Liquidity

-0.300

-1 .444

-0.053

Loan loss provisions/loans

-0.91 4***

-0.388

-2.459***

BIS capital ratio

-0.1 99

-0.1 20

Equity investments/TA

-1 .878***

-4.417***

0.133***

1.632***

-1.690***
0.023

Market return x equity
investments/TA

-0.097***

-0.341 ***

Business bankruptcies

-0.039

Market return

R2
N

0.39
555

0.049
0.61
1 28

-30.066
0.401 *

Domestic loans/total loans

-0.1 81

9.090*
3.311

0.220**

0.220

All banks
1995-97

-0.409

0.31 6

1 .076

-0.1 62

0.839

-0.377***
-0.051 ***

-2.565***
0.290*

-0.052***

-0.014***

_O.114***

-0.025

-0.064***

-0.608

0.24
427

0.46
309

-0.157***
-0.464***
-0.092***
-1 464***
Equity investments/TA
-1 .21 1 *
-1 .431 ***
Parameter estimates when the market return and the ratio of equity investments to total assets
are included in regressions without an interaction term.
Market return

0.49
246

-0.090***

0.020

-0.286***

-2.664***

*, **, and *** indicate significance at the 10 percent, 5 percent, and 1 percent levels, respectively.
Notes: For variable definitions, see box 1. Parameter estimates that are in bold
indicate that there are significant differences across bank types or subperiods at the 5 percent level.
Source: Author's calculations from FitchIBCA, Bankscope, CD-ROM, 1998.

with earnings, indicating that banks with higher credit
risk performed worse than others. These measures of
asset quality are particularly strong determinants of
performance for regional banks, but are less informative
for major banks. Recall that, compared with regional
banks, major banks hold a smaller fraction of their
assets in loans; thus, these banks’ performance may
be more sensitive to fluctuations in other sources
of income, such as fee income and earnings from
security portfolios.
Banks with greater investments in equity securities
performed worse than others. This result shows that
when economic conditions were deteriorating, the
equity investments of Japanese banks exposed them
to greater risk and reduced their earnings. As shown in
the last row of columns 4 and 5 of table 3, the negative
impact was significantly worse in the 1995-97 period.
The relationship between profitability and other
bank characteristics is statistically weaker. Profitability
is significantly correlated with liquidity, size, and
growth rate of assets in only some specifications.
Furthermore, in contrast to the positive significant
relationship observed between bank earnings and

capital in other studies (for example, Berger, 1995, and
Demirguc-Kunt and Huizinga, 1997), Japanese banks’
earnings are not significantly related to their capital
ratios. At first glance, this result suggests that BIS
capital ratios have no impact on Japanese banks’
earnings. However, this conclusion is at odds with
anecdotal evidence which indicates that capital
management was of particular importance to Japanese
banks during this period. For instance, between 1992
and 1995, Japanese banks sold ¥2.7 trillion of subor
dinated debt to meet BIS capital requirements and
some major banks issued convertible securities to
raise capital. Furthermore, comments by MoF officials
and analysts suggest that the retrenchment of Japanese
banks from international lending is at least partially
motivated by their need to increase capital ratios. It is
unlikely that significant efforts by Japanese banks to
manage their capital positions had no impact on their
earnings.23 If capital management was important for
Japanese banks during the sample period, then the
impact of capital ratios on bank earnings would not
be measured accurately by the current analysis which
treats capital ratios as exogenous variables that are

Economic Perspectives

not influenced by bank characteristics. One would
need to take into account the factors that affect
banks’ capital management decisions before examining
the impact of capital on earnings. This type of analysis
is beyond the scope of this article.
Some of the relationships shown in table 3 are
inconsistent with our expectations and patterns ob
served in the U.S. Specifically, higher returns in the
Tokyo Stock Exchange, which imply more favorable
economic conditions, are associated with poor bank
performance. In addition, the coefficient estimates for
the interaction term between stock returns and equity
investments indicate that the negative correlation
between stock returns and earnings is stronger for
banks with more equity investments, particularly for
major banks and in the 1995-97 period. These results
are in direct contrast to our expectations. Further
analysis, however, revealed that the result was evident
only for measures of performance that include loan loss
provisions. There is a positive correlation between
pre-provision profits and stock returns. These results
suggest that Japanese banks provision more when
economic conditions are good.24 The correlations
between loan loss provisions in the current period and
other bank characteristics and economic conditions,
shown in table 4, point to a similar conclusion. Specifi
cally, banks provision more when they have higher
core earnings (operating profits before loan loss
provisions) and when the stock market performs well.
Furthermore, banks with higher equity investments
provision more than other banks. These correlations,

and the results with other performance measures, are
consistent with analysts’ assessment of the income
smoothing behavior of Japanese banks. The results
are also consistent with Moody’s (1997, 1998) reports
that to maintain their capital positions in recent years,
Japanese banks have sold their equity securities to
offset credit expenses.25
In addition, loan loss provisions are positively
correlated with the fraction of assets invested in
loans, indicating that banks with higher credit risk
provision more. However, there is a strong negative
correlation between loan loss provisions and the in
crease in the number of business bankruptcies in the
current period. This result is puzzling and gives further
evidence that Japanese banks’ provisioning practices
do not conform with conventional wisdom.
Lastly, the well-known credit quality problems
associated with Japanese borrowers in the 1990s
suggest a negative relationship between ROE and
the fraction of total loans allocated to domestic borrow
ers. The results in table 3 indicate that, in contrast to
our expectations, domestic loans were associated with
higher profitability in 1991-94. However, during 1995—
97 this relationship loses statistical significance.

Determinants of stock market performance
The consistency of the results in the previous
section with those in other banking studies was mixed.
Some of these results might be due to efforts by
Japanese banks to manage their regulatory capital
and to fund their credit expenses through sale of

TABLE 4

Correlations of loan loss provisions with other bank characteristics and economic indicators
All banks
1991-97

Major banks
1991- 97

Total assets

0.25***

-0.21 * *

Gross loans/TA

0.11**

Domestic loans/total loans

-0.36***

0.1 3
-0.07

Regional banks
1991- 97

All banks
1991-94

0.17* **

0.33***

0.34* **

0.09

0.09*

_0.40***

All banks
1995-97
0.33***
-0.03
-0.63***

BIS capital ratio

0.05

0.25* **

Equity investments/TA

0.60***

0.64* **

0.40* **

0.58***

0.67***

Loan loss provisions/loans
in the previous period

0.49***

0.36* **

0.45* **

0.10*

0.47***

Operating profits before
LLP/equity

0.49***

0.57* **

0.44* **

0.31 ***

0.47* **
-0.38* **

0.36* **

0.39***

0.30***

-0.29* **

-0.33***

0.31 ***

Market return

Business bankruptcies
N

Number of banks

_0.26***

-0.01

0.26***

-0.02

-0.1 0

0.46***

555

1 28

427

309

246

*, **, and *** indicate significance at the 10 percent, 5 percent, and 1 percent levels, respectively.
Note: For variable definitions, see box 1.
Source: Author's calculations from FitchlBCA, Bankscope, CD-ROM, 1998.

Federal Reserve Bank of Chicago

securities during favorable market conditions. If such
actions are transparent to investors, and analysts’
reports and anecdotal evidence suggest that they are,
then the puzzling results between bank characteristics
and market measures of performance would not exist.
Market participants would dismiss the reported numbers
as irrelevant and rely on other indicators of banks’
condition (for instance, analysts’ reports). In that case,
market measure of performance, such as bank stock
returns, would not be related to accounting profits;
and the relationship observed between stock returns
and bank characteristics would differ significantly from
the relationship observed with accounting profits.26
To explore this issue further, I relate the stock
returns of banks to the bank characteristics used
above. The results, reported in table 5, suggest that
while those accounting relationships that were consis
tent with our expectations are also evident for stock
returns, the puzzling results with accounting earnings
disappear when performance is measured by stock
market returns. Specifically, banks with more loans
and equity investments and banks with higher loan

loss provisions have lower stock returns. Furthermore,
size and profitability are positively correlated, partic
ularly for major banks. These results are consistent
with the results in the previous section and with the
results of other banking studies.
However, in contrast to the results in table 3,
the results in table 5 indicate that stock returns are
positively correlated with the market index. This
result implies that market participants perceive the
negative correlation of the market index with reported
earnings as an accounting artifact and see a positive
impact from an increase in the index on banks’ future
cash flows.
Another difference between reported accounting
profits and stock returns is their relationship with the
change in the number of bankruptcies. ROE is only
weakly correlated with bankruptcies (the only sig
nificant correlation is in 1991-94) and the results in
table 4 show a puzzling negative correlation between
loan loss provisions and bankruptcies. In contrast to
these relationships with accounting results and con
sistent with expectations, there is a strong negative

TABLE 5

Stock returns and bank characteristics

All banks3
1991-97
Intercept

Size

-1 5.222

1 .81 8*

Major banks3
1991-97
148.245*

7.216**

Growth of TA

-0.001

-0.023

Net loans/TA

-0.212*

Regional banks
1991-97

All banks3
1991-94

All banks3
1995-97

-1 3.818

-22.373

-0.133

1 .808

0.071

-0.001

-0.131

-0.394* *

4.704

-0.334

-0.089

-0.084

2.360

5.465

30.955

2.941

13.714

Overhead ratio

-4.119

-0.081

-4.894

-6.256

-2.996

Liquidity

-0.365

-3.683

-0.434

-0.797

0.672

Loan loss provisions/loans

-3.343***

-2.497***

-4.163***

0.073

-3.090***

BIS capital ratio

-0.325

Equity investments/TA

-1 .526* *

Domestic loans/total loans

Market return

Market return x equity
investments/TA
Business bankruptcies

R2
N

Market return

0.334***
0.071***
-0.280***

0.48
472

0.534***

2.182
-4.061 ***

2.196***

0.335
-1 .914**

0.697***

-0.204**

-0.104**

-0.159

-0.226***

0.73

0.986***

0.42

384

0.460***

-1 .803***
Equity investments/TA
-2.460**
-1 .385
Parameter estimates when the market return and the ratio of equity investments
to total assets are included in regressions without an interaction term.

1 .052

0.362

-0.1 05

-1 .566*
0.485**

0.488*

0.067***

0.081 *

-0.1 04

-1 .041

0.56

0.43
267

205

0.687***
-2.090**

0.659**

-0.059

*, **, and *** indicate significance at the 10 percent, 5 percent, and 1 percent levels, respectively.
aExcludes stock returns for long-term credit banks.
Notes: For variable definitions, see box 1. Parameter estimates that are in bold indicate that there are significant differences across
bank types or subperiods at the 5 percent level.
Source: Author's calculations from FitchIBCA, Bankscope, CD-ROM, 1998, and Bloomberg Financial, database, 1998.

Economic Perspectives

relationship between stock returns of banks and
increases in the number of bankruptcies. These re
sults suggest that although banks’ reported account
ing earnings exhibit no strong association with
bankruptcies, shareholders take into account the
adverse impact of bankruptcies on banks’ asset
quality and earnings.

Accounting and stock returns
Our results up to this point indicate that reported
earnings and stock returns of Japanese banks are
related to size and measures of asset quality in similar
ways. However, the results also point to some dif
ferences in the behavior of accounting and market
measures of performance. Given the differences in
the two measures of performance, how do they relate
to each other?
To answer this question, I examine the relation
ship between stock market and accounting returns
directly. I first estimate the market model for Japanese
bank stocks by regressing individual bank returns on
the market index. I then modify the market model by
including the return on equity as an additional explan
atory variable. If shareholders dismiss accounting
earnings as uninformative, one would expect the
coefficient on ROE to be insignificant.
First, as reported in the top panel of table 6, for
the entire sample, the coefficient on the market index
is positive. The coefficient is less than one, indicating
that when the overall stock market increases by 1 per
cent, bank stock returns increase by less than 1 per
cent. However, there are significant differences in how
the stock returns of major and regional banks move
with the market. A 1 percent increase in the market

index moves the stock returns of major banks by
more than 1 percent and those of regional banks by
less than 1 percent. Major banks own significantly
greater amounts of equity securities than regional
banks. Thus, a movement in the stock market affects
not only income from their operations, but also the
value of their equity investments, magnifying the
impact of changes in the market index.
Second, when the market model is augmented
with ROE, the coefficient on ROE is positive and
statistically significant. In addition, the fraction of the
variance in bank stock returns explained by the model,
R2, increases in most specifications when accounting
returns are included as explanatory variables. There
fore, for all the potential biases in the reported results
of Japanese banks, shareholders do not dismiss ac
counting earnings as meaningless.
However, the correlation between banks’ stock
market and accounting returns has decreased over time;
accounting returns are not significantly correlated
with stock returns in the 1995-97 period. Higher report
ed earnings in later years did not translate into higher
returns in the stock market as they had in the earlier
part of the 1990s, which implies that accounting profits
became less informative over time. This result suggests
that measures taken by banks to shore up their re
ported earnings and capital are not seen by market
participants as significant determinants of banks’
market performance and, instead, drive a wedge be
tween the banks’ accounting and market returns,
disconnecting the two measures of performance.
These results are consistent with the results of
other studies of U.S. banking showing that regulatory
forbearance decreases the correlation between the

TABLE 6

Stock market and accounting performance of Japanese banks

Intercept
Market return

All banks3
1991-94

Major banks3
1991-97

-4.393**
0.614**

-3.278**

-4.478**

-3.442**

-4.676**

1.021 **

0.530**

0.757**

0.515**

0.58

0.36

0.41

0.42

Intercept
Market return

Regional banks
1991-97

All banks3
1991-97

0.40
-4.81 9* *
0.643**

-2.246

-5.346**

1.208**

0.557**

ROE

0.217**

0.499**

0.313**

0.40

0.61

0.38

-9.944**
0.818**
1.150**
0.43

All banks3
1995-97

-4.643**
0.538**
0.128
0.42

* * indicates significance at the 5 percent level.
aExcludes stock returns for long-term credit banks.
Notes: For variable definitions, see box 1. Parameter estimates that are in bold indicate that there are significant differences in the
parameter estimates at the 5 percent level across bank types or subperiods.
Source: Author's calculations from FitchIBCA, Bankscope, CD-ROM, 1998, and Bloomberg Financial, database, 1998.

Federal Reserve Bank of Chicago

market value of equity and the value of net assets
in place.27 The market value of a bank’s equity is the
sum of the value of its net assets in place and the value
of deposit insurance subsidies and regulatory for
bearance. Accounting profits, on the other hand,
only reflect earnings from assets in place. As a bank
nears economic insolvency, the value of regulatory
subsidies and forbearance increases and shareholders
derive more of their value from subsidies rather than
assets in place. Except for the unlikely situations where
the value of assets in place is perfectly correlated
with changes in the value of subsidies, an increase
in the value of subsidies and regulatory forbearance
leads to a decline in the correlation between market
returns and value of assets in place.
The reluctance of Japanese regulators to force
recognition of loan losses and to impose penalties on
the shareholders of failing institutions is undoubtedly
valuable to banks’ shareholders. As the condition of
the banks deteriorated significantly in the later part
of the 1990s, the value of subsidies and forbearance
to shareholders might have increased significantly,
potentially accounting for the lack of correlation
between market and accounting returns of Japanese
banks in 1995-97.

Conclusion
Economic malaise, ever-increasing problem
loans, high credit expenses to provision for problem
loans, and low core profitability have taken their toll
on Japanese banks. In this article, I have examined
the performance of Japanese banks in recent years
and related it to variables used by regulators and
analysts to assess the condition of banks. The results
show significant differences between the performance
and characteristics not only of Japanese and U.S.
banks, but also of different Japanese banks and over
time. The results also show that although most
measures of bank asset quality are correlated with
accounting returns in line with expectations and the
results of other banking studies, other variables that
were found to be important determinants of bank
performance in the U.S. and elsewhere are not signifi
cantly related to the performance of Japanese banks.
Moreover, accounting profits are correlated with other
bank characteristics and economic variables in puzzling
ways. Additional evidence suggests that these puzzling
or inconsistent results may be due to income smooth
ing by banks. Specifically, Japanese banks appear
to increase their loan loss provisions when their core
earnings and the returns on the market are high.
However, such actions do not appear to affect the

stock returns of banks; the returns are positively cor
related with the return on the market index. My analysis
shows that although there might be problems with
the reported earnings of Japanese banks, accounting
returns still provided useful information to market
participants regarding the values of bank shares in
the early 1990s. However, the significance of this in
formation has decreased in recent years.
These results may reflect an increase in the value
of regulatory forbearance to bank shareholders. Since
the end of the period analyzed in this article, the MoF
has introduced a number of measures that indicate
an increase in regulatory forbearance. As outlined in
appendices 1 and 3, accounting changes have enabled
banks to increase their regulatory capital, government
purchases of banks’ preferred stock and subordinated
loans have injected capital to institutions experiencing
financial difficulties, and government guarantees
have been extended to all bank creditors through the
end of fiscal year 2001. Regulators typically forbear to
give ailing institutions time to recover. However, ex
perience tell us that forbearance imposes significant
costs on the economy by transferring wealth from
the deposit agencies, and hence taxpayers, to bank
shareholders and by increasing the cost of resolving
insolvent institutions. To the extent that the recent
financial revitalization laws resolve the insolvent in
stitutions and encourage solvent banks to deal with
their problems in a timely manner, they should greatly
improve the health of the Japanese financial system.

Econom ic Perspectives

APPENDIX 1

Developments in Japanese financial markets
Between 1952 and 1973, the Japanese economy exhibit
ed remarkable strength, averaging 10 percent real growth
per year. Financial institutions in general, and banks in
particular, were instrumental in achieving this strong
performance. By acting as conduits of funds from the
household sector to the corporate sector, they financed
exports and business investment that fueled the
economy. The goal of financial regulations during this
period was to provide a stable financial environment
conducive to growth. Regulation of interest rates kept
the cost of funds low for banks and corporations. The
positive slope of the yield curve ensured profits for
banks engaged in maturity transformation and fostered
a culture in which banking profits increased with size.
Segmentation of the financial markets across functions,
restrictions on portfolio activities of banks, and controls
over foreign exchange transactions and international
capital flows provided a stable system with restricted
competition. The collateralization requirements on all
debt issued and other restrictions on security issuance
meant that banks were the primary source of external
funds for corporations.
The policies of the high-growth period became
unsustainable, however, after the collapse of the Bretton
Woods system of fixed exchange rates and the first
oil crisis in 1973. As the government deficit ballooned,
it became harder for banks to absorb government
bonds without a secondary market for these securities.
The development of secondary markets in government
bonds enabled investors to circumvent interest rate
regulations on deposits, while soaring inflation in
creased the cost of these regulations. Increased inter
national trade and globalization of financial markets
provided further impetus for change and Japanese
financial markets underwent a series of reforms. By the
mid- to late 1980s, Japanese financial markets were
substantially liberalized. Regulations on interest rates
were gradually reduced, more financial instruments
for savings were allowed, restrictions on security issues
were relaxed, and portfolio activities of banks and
other financial institutions were expanded. For example,
the Foreign Exchange Law of 1980 allowed Japanese
corporations to finance their operations with foreign
currency denominated loans. This gave Japanese
businesses an alternative source of funding, which
increased the competitive pressures on Japanese banks.
At the same time, however, the law allowed Japanese
banks to borrow and lend freely in foreign currencies,
giving them entry into new markets. Japanese banks
took full advantage of this opportunity to expand

Federal Reserve Bank of Chicago

their international operations, including those in the
U.S. By 1990, Japanese banks had become the largest
foreign lenders to U.S. companies and financed most
of the record levels of Japanese direct investment in
the U.S. commercial real estate. In the meantime, soaring
stock and land prices in Japan during the second half
of the 1980s boosted banks’ unrealized gains on equity
holdings and enabled them to increase loans collateral
ized by property. By some private estimates, 50 percent
to 70 percent of new lending by Japanese banks in
1985-90 was collateralized by real estate.
With the collapse first of stock prices, then of
land values in the early 1990s, the first cracks in the
system appeared. Because a portion (up to 45 percent)
of unrealized gains on banks’ security holdings counts
as tier two capital under the BIS rules, the decline in
stock prices put significant pressure on banks’ capital
ratios. As early as mid-1991, press reports pointed to
difficulties faced by Japanese banks in meeting BIS
capital requirements. Regulators responded by allowing
banks to issue subordinated and perpetual debt. In
addition, banks sold loans and shifted lending from
low-margin markets (such as European and U.S. lend
ing) to higher margin segments (such as corporate
lending and leasing in Southeast Asia).
Early in the decade, declines in land prices were
welcomed by regulators. In fact, the MoF restricted the
growth of real estate lending in 1990 to discourage
speculative land deals. Anecdotal evidence suggests
that banks responded by shifting their real estate
lending to affiliates. Sharp declines in land prices
throughout the 1990s, however, reduced the value
of the collateral on loans and led to a significant
deterioration of asset quality at nonbank affiliates
of banks, such as housing loan companies. In 1993,
parent banks and other creditors restructured their
loans to the housing loan companies (the jusen), in
order to provide liquidity to these firms. Additional
declines in land prices, however, deteriorated the
condition ofthe jusen further. In December 1995, the
government announced the liquidation of seven
housing loan companies and the Housing Loan
Administration was established in July 1996 to takeover
the assets of the failed jusen.
The decline in land prices also had significant
adverse effects on the quality of loans at the banks.
In January 1993, the Cooperative Credit Purchasing
Company (CCPC) was established to purchase bad
loans and collateral backing such loans from the banks.
However, because the CCPC was funded by the
banks themselves, the plan was met with skepticism
by analysts from the outset. To date, the operations

of the CCPC have not stemmed the deterioration in
asset quality or have brought a decisive resolution
to the problem. The deterioration in the condition of
Japanese financial institutions in the 1990s and the
regulators’ response to the problem were evidenced
by the failure of several nonbanks and assisted mergers
of insolvent small banks with stronger banks in the
1991-95 period. The details of the assisted mergers
indicate that there were no losses to depositors and
very little penalties imposed on shareholders of failed
banks (Cargill, Hutchison, and Ito, 1997).
In November 1997, for the first time since World
War 11 a major Japanese bank, Hokkaido Takushoku
Bank, failed. Today banks continue to face continual
downgrading of their ratings, severe liquidity pressures,
higher funding costs in interbank markets, and de
clines in their stock prices. Regulators have responded
by guaranteeing all deposits, including interbank
deposits, through 2001 and giving unofficial guar
antees on other bank liabilities. In January 1998, the
government announced a ¥30 trillion program that in
creased the funds available to the Deposit Insurance
Corporation and injected ¥1.8 trillion of funds to shore
up banks’ capital base. In April 1998, prompt corrective
action regulations were implemented that required

fuller disclosure of nonperforming loans and more
adequate provisioning for problem loans. However,
the implementation of some of the prompt corrective
action regulations have been delayed, and in 1998 the
MoF implemented certain accounting changes aimed
at increasing regulatory capital of banks (see appendix
3). In June 1998, the Financial Supervisory Agency
took over the supervision of banks from the MoF.
Also in June 1998, the government announced the
“total plan,” designed to resolve the crisis. A modified
version of this plan became law in October 1998. The
¥60 trillion bail-out package involves the injection of
public money into banks on a voluntary basis to in
crease their capital base, as well as the nationalization
of insolvent banks. On the first day the law came into
force, nationalization of the Long-Term Credit Bank
of Japan (LTCB), which had been rumored to be in
solvent for a number of months, was announced. The
initial announcements indicated that all deposits, de
bentures, derivative contracts, interbank deals, and
subordinated debt of the bank would be honored.
The plan also called for the Deposit Insurance Corpora
tion to purchase the shares of the LTCB, which last
traded at ¥2.

APPENDIX 2

Relationship between accounting profits,
stock returns, and financial characteristics
of Japanese banks
In the first part of the analysis, I relate the account
ing profits of Japanese banks to a set of variables
that describe the banks’ characteristics and a set of
variables that measure aggregate economic activity.
Specifically, I estimate the following equation, using
ordinary least squares (OLS), and report the results
in table 3:
1)

ROE.t = (X + 3 Rmt + 0 A.,, + yZ +

K,, +

e.„

where ROE.tK return on equity for bank Z in fiscal
year Z;
is the return in the Tokyo Stock Exchange
in period Z, as measured by changed in the TOPIX
index; X,, is a vector of characteristics of bank Z,
calculated using information from fiscal year Z - 1;
is a set of measures for aggregate economic activity
in fiscal year Z; Kf l is Rmt multiplied by bank Z’s ratio
of equity investments to total assets in fiscal year Z - 1;
and e. t is an error term.
Similarly, banks’ stock returns are correlated
to their characteristics by estimating the following
equation using OLS:

2)7

/L-of-RA?
+0SA,
+y1 s'Z,t + (t)• sL,
, +et>st„y
i,t
"
m,t
i,t-l
j#g_i

where A ts is the stock market return of bank Z in period
Z and the S superscript indicates that parameter
estimates are for stock returns. The results from the
estimation of equation 2 are reported in table 5.
The interaction terms in equations 1 and 2 make
it difficult to determine the correlation between profit
ability and the ratio of equity investments to total
assets. To simplify the presentation of the results,
I reestimated equations 1 and 2 without the interaction
terms. The coefficient estimates for TOPIX and equity
investments from the “simplified” regressions are
reported as the last two rows in tables 3 and 5.
Lastly, in table 6, I report the results from the
OLS estimation of the following traditional and
“augmented” market models:

Ri,ts =a + "1
3,Rm,t,+ rU i,t,
Ri,t
s=u' + “
P,'
Rm,t +8ROEi,t +r\1 z,ft’,
1
where p,t and r|.f are error terms.

Economic Perspectives

APPENDIX 3

Differences in Japanese and U.S.
accounting practices
Some of the significant differences in the disclosure
and accounting rules in Japan and the U.S. are sum
marized below.
Nonperforming loans: In the U.S., loans that are
past due more than 90 days plus nonaccrual loans are
considered nonperforming. In Japan, the definition of
nonperforming loans has changed in recent years to
become more inclusive and more in line with U.S.
standards. Previously, only loans to bankrupt com
panies and loans past due more than 180 days were
considered nonperforming. However, since March 31,
1996, nonperforming loans have also included loans
to assisted companies and loans restructured to have
an interest rate below the official discount rate. On
March 31, 1998, the definition was expanded to include
loans past due more than 90 days and all restructured
loans. Despite these changes, however, loans with
partial interest payments, loans sold to the Cooperative
Credit Purchasing Company, nonperforming loans of
subsidiaries, and other loans for which the bank may
ultimately be held responsible are excluded from the
definition of nonperforming loans.
Also effective April 1, 1998, each bank is required
to self-assess its asset quality, dividing its credit ex
posures into the following four categories:1 category
I—exposures with no credit concerns are classified;
category II—“credit exposures on which each bank
has judged adequate risk management on an exposureby-exposure basis will be needed,” but where the clas
sification standard “varies significantly depending
on their respective management practices,” (Japan,
Ministry of Finance, 1998); category III—exposures
on which the banks have serious concerns and are
likely to incur losses, but cannot determine the timing
and amount of such losses; and category IV—credit
exposures that are noncollectible or of no value.2 On
January 12, 1998, the Ministry ofFinance (Japan, MoF,
1998) announced that 12.3 percent of Japanese banks’
total loans are classified in categories II through IV.
The bulk of the classified assets, 10.4 percent of total
loans, are in category II.
Loan loss provisions and reserves: U.S. account
ing rules require banks to maintain an allowance for
loan losses based on probability of collection and
expected future cash flows. Additional provisions are
made through periodic charges to operating expenses
and, thus, are fully tax-deductible. Loan loss reserves
are treated as a contra account on the assets side
of the balance sheet and, therefore, are deducted from

Federal Reserve Bank of Chicago

gross loans and total assets. Until April 1, 1998,
Japanese banks maintained three types of loan reserves.
General reserves for loan losses were maintained at
the maximum tax deductible level of0.3 percent oftotal
loans outstanding. The portion of loans determined
to be irrecoverable was reserved under specific re
serves, of which only 50 percent is tax deductible.
Banks could provision more than the tax deductible
amount with approval from the MoF. Analysts point
out that because loan loss reserves received a less
favorable tax treatment in Japan and because banks
were not required to increase provisions when the
present value of the loan declined below its face value,
Japanese banks did not fully provision for possible
loan losses. Some of these concerns were addressed
by the implementation of prompt corrective action
(PCA) regulations, effective April 1, 1998. Under the
PC A regulations, Japanese banks are expected to make
adequate provisions based on their self-assessment
of problem loans as outlined above. Lastly, most
banks maintain specific foreign loan reserves equal to
35 percent of loans to specific countries where transfer
risk maybe material. However, only 1 percent of the
outstanding loan amount is tax deductible. Reserves
are classified as liabilities and total loans and total
assets are reported gross of reserve amounts. Further
more, unlike U.S. banks, which can establish a loss
contingency reserve only when an event is probable
and the amount of losses can be established, Japanese
banks are allowed to establish discretionary reserve
accounts; transfers to and from such reserves might
allow Japanese banks to smooth their reported income.3
Charge-offpolicy: Under U.S. accounting prac
tices, once the extent and timing of losses arising from
a loan can be determined, expected losses are recog
nized through loan charge-offs. In Japan, loans are
charged off only when the debtor is in bankruptcy
and there is no hope of recovery, and banks need a
special MoF ruling to take loans off their books.
Valuation of securities: In the U.S., banks’
security holdings are classified under three separate
categories and methods of valuation. Japanese banks
classify their security investments as either for trading
or investment purposes; however, the classification
does not affect the valuation method. Listed securities
are valued at either the lower-of-cost-or-market
(LOCOM) value or at historical cost. Under the LOCOM
method, market value increases above cost are not
recognized and unrealized losses are recognized under
valuation reserves. Unlisted securities are generally
valued at cost; if the condition of the security issuer

deteriorates significantly, then the securities valuation
is reduced accordingly. The difference between the
market and book value of security holdings is referred
to as “latent revaluation reserves,” or more commonly
as “hidden reserves.”
BIS capital requirements: Similar to banks in
other countries, Japanese banks with international
operations are required to achieve a minimum total
capital ratio of 8 percent, based on standards issued
by the BIS.4 Within certain guidelines, regulators in
individual countries are allowed to determine what
constitutes capital. Consequently, there are differences
across countries in how banks can satisfy the capital
adequacy requirements. For instance, under U.S. reg
ulations, unrealized gains on securities do not count as
capital, but Japanese banks can use up to 45 percent
of hidden reserves as tier two capital. Low profitability,
high credit expenses for problem loans, and unfavor
able conditions in capital markets have put Japanese
banks’ capital position under pressure. In order to
provide some relief to banks, the MoF recently intro
duced certain measures. For example, since January
1998, Japanese banks have been allowed to value
securities at cost and avoid reported valuation losses;
however, if a bank chooses this valuation method, it
cannot use any portion of its unrealized gains as tier II
capital for BIS capital requirements. International
accounting standards generally do not allow higher-of-cost-or-market valuation for securities, which
in effect the MoF rule does. Again, effective January
1998, banks can value real estate at market values, and
45 percent of the valuation reserves count as tier two
capital. Most of the major countries, with the exception
of Germany and the U.S., also allow such valuation

reserves to count toward regulatory capital. In addition,
in March 1998, under its stabilization program, the
government purchased ¥1.8 trillion of banks’ preferred
stock and subordinated debt. All three measures have
increased Japanese banks’ regulatory capital base.
In addition, starting this year, if the maturities and the
other contractual features of loans and deposits from
the same customer meet certain requirements, banks are
allowed to net loan assets with the deposits of the
same customer. As a result, the risk-weighted assets of
banks are reduced, increasing their BIS capital ratios.
These categories are for disclosure purposes; for internal
purposes, Japanese banks typically classify their assets into
five categories: pass, special mention, substandard, doubtful,
and bankrupt.

2The classified exposures include off-balance-sheet guarantees
as well as loans, and the reserved and collateralized portion of
each exposure is classified in category I, independent of the
borrower’s financial condition. Because of these and other details
of the classification standards, the classified assets of a bank
cannot be linked directly to its disclosed nonperforming loans.
3In addition to these reserves for possible loan losses, Japanese
banks maintain reserves for expected losses on trading account
securities, government bonds, futures, and securities transactions.
4Although the BIS capital adequacy requirements were established
only for banks with international operations, regulators in the
U.S. require all banks to maintain the minimum BIS capital ratios.
However, Japanese banks with only domestic operations are
exempt from the BIS requirements. In recent years, Japanese
banks that experienced difficulties meeting the BIS requirements
have sold their international operations and, thus, are subject
only to the 4 percent capital requirement placed on banks with
no international presence. For instance, on March 31, 1998,
the MoF announced that the number of “internationally
operating banks” declined from 80 institutions to 45 institutions
and the number of “domestically operating banks,” which are
subject to the 4 percent capital requirement, increased from 67
institutions to 102 institutions.

NOTES
Tor more evidence on the economic impact of declining asset
prices and bank health, see Gibson (1995 and 1996); Kang and
Stulz (1997); Peek and Rosengren (1997); and Kaufman (1998)
Tor instance, a recent study notes that in 1980-96, over 130
countries experienced serious banking problems (Lindgren,
Garcia, and Saal, 1996).
Tor an overview of the S&L and banking crisis in the U.S.,
the resulting regulatory changes, and an assessment of the
regulatory reform, see Benston and Kaufman (1998) and
references therein.

Tor a concise review of the literature, see chapter three of
Lindgren, Garcia, and Saal (1996) and references therein.
Tor instance, just as the MoF allows Japanese banks to avoid
reporting valuation losses on security portfolios, in the 1980s
the Federal Home Loan Bank Board allowed S&Ls to defer

recognition of losses on asset sales. For details of the regulatory
accounting practices allowed by S&L regulators, see Benston
and Kaufman (1990), Barth (1991), and Ashley, Brewer, and
Vincent (1998).

6The cost of regulatory forbearance in the U.S. has been studied
by Eisenbeis and Horvitz (1994), Brinkmann, Horvitz, and
Huang (1996), Kane and Yu (1996), and others.
Tor a more detailed description of the Japanese financial markets
and regulatory developments, see Suzuki (1987), Cargill and
Royama (1988), Tatewaki (1990), Frankel and Morgan
(1992), Cargill, Hutchison, and Ito (1997), and Craig (1998).

8In addition to its regulatory function in the banking industry, the
MoF has other, broader responsibilities, such as regulation of
other financial institutions, setting fiscal policy, collecting taxes
and custom duties, drawing and allocating the government
budget, floating government bonds, and overseeing foreign
exchange transactions.

Economic Perspectives

9The banks in table 1 do not represent all banks in Japan, only
the largest ones. Second tier regional banks and institutions
that specialize in financing of small businesses and agriculture
are not included.

18I also used other measures of liquidity and capital (such as
book value of capital to total assets and BIS tier one capital ratio).
The results with these alternative measures were qualitatively
similar to those reported in the article.

10Keiretsu are one of the most distinguishing features of Japanese
organizational structure. Keiretsu are groups of companies
that maintain long-term relationships with each other through
cross shareholdings and customer-supplier relationships.
Financial institutions (typically a city bank, a trust bank, and
insurance companies) form the nexus of keiretsu and provide
debt and equity financing to group firms. Previous studies found
that keiretsu firms differ from other Japanese firms in significant
ways. (For a description of keiretsu relationships, see Nakatani
(1984), Sheard (1989), Genay (1991), Aoki and Patrick (1994),
and the references therein.) For instance, keiretsu firms recover
from financial distress faster than other Japanese firms (Hoshi,
Kashyap, and Scharfstein, 1990), and they may be less cash con
strained in their investments (Hoshi, Kashyap, and Scharfstein,
1991; and Hall and Weinstein, 1997). In addition, corporate
governance practices appear to be different in keiretsu: Banks
play a more central role in the governance of keiretsu firms
through their board representation (Kaplan and Minton, 1994),
and the shareholders of financial institutions in the keiretsu
respond differently to risk from the shareholders of other
financial firms (Genay, 1993). However, there is also evidence
that keiretsu relationships involve significant costs (Gibson,
1996; Kang and Stulz, 1997; and Weinstein and Yafeh, 1998).
Although anecdotal evidence suggests keiretsu relationships
are weakening, these groups and their financial institutions
continue to be major players in the Japanese economy.

19In the following analysis, I also used other measures of eco
nomic activity, such as the change in the yen-dollar exchange
rate, changes in short-term and long-term interest rates, chang
es in term structure, and dummy variables for years. The results
with respect to bank characteristics were similar to those reported
in the article. The results also indicated that Japanese banks face
some interest rate and foreign exchange risk. In particular,
depreciation of the yen is associated with lower bank earnings
and stock returns. Changes in the term structure are also negative
ly correlated with bank earnings. Specifically, increases in the
short-term gensaki rate (the three-month, riskless rate) are
associated with higher bank earnings, whereas increases in the
long-term (ten-year bell-wether bond) rates are negatively cor
related with bank earnings. Monthly stock returns of banks,
when significant, exhibit a similar relationship with changes
in the short- and long-term interest rates. However, there are
significant differences in the interest rate sensitivity of Japanese
banks in the pre- and post-1995 periods and across bank types.
The evidence with respect to long-term interest rates is
consistent with the results reported in Broussard, Kim, and
Limpaphayom (1998), which looks at the sensitivity of
Japanese banks in the 1975-94 period.

nThe slight decline in interest margins at U.S. banks during this
period reflects aggressive price competition in U.S. business lending
markets. Hence, the relatively greater profitability of U.S. banks
during 1990-97 is due mostly to higher fee and other income.

12Similarly, according to statistics reported by Demirgu2-Kunt
and Huizinga (1997), Japanese banks earned, on average, 0.10 per
cent return on assets (ROA) in 1988-95. Over the same period,
banks in the rest of the G7 countries earned 0.53 percent ROA.
13For some examples of this literature and other banking studies
that form the basis of the following discussion, see Brewer and
Garcia (1987), Berger, King, and O’Brien (1991), Kuester and
O’Brien (1991), Thomson (1992), Cole (1993), Berger (1995),
Brewer, et al. (1997), Demirgu9-Kunt and Huizinga (1997),
and references therein.

14Lack of sufficient numbers of Japanese bank failures pre
cludes me from analyzing the determinants of the solvency of
Japanese banks.

15I relate current bank performance to characteristics measured
at the end of the previous period. Therefore, although ROE is
negatively correlated with loan loss provisions in the current
period by definition, there might be a positive relationship
between current ROE and previous loan loss provisioning.
16There is also evidence that nonperforming loans reported by
U.S. banks are important predictors of future bank performance
and are significantly related to stock market value of banks’ equity.
For Japanese banks, definition of what constitutes a nonper
forming loan is less inclusive and has changed several times in
recent years (see appendix 3); as a result, it is more difficult to
measure the impact of nonperforming loans on Japanese bank
performance.

17For a detailed discussion of the relationship between earnings
and capital, see Berger (1995) and Brewer et al. (1997).

Federal Reserve Bank of Chicago

20Excluding these observations does not qualitatively affect
the results presented here.

21 Stock prices for the three long-term credit banks were un
available; hence these banks are excluded from the analysis of
stock returns reported in table 6.

22The following results on accounting profits remain qualita
tively the same if one uses ROA, rather than ROE.
23There is some evidence, for example, that the cost of issuing
convertible bonds was significant for Mitsubishi Bank (Ammer
and Gibson, 1996).
24The results with other measures of accounting profitability
are available from the author upon request.
25The statistics in table 2 indicate that Japanese banks have in
creased their equity investments in recent years. Although this
might appear inconsistent with anecdotal evidence on equity
sales, it is consistent with other anecdotal evidence that suggests
that banks repurchased their equity stakes in other companies
to maintain long-term relationships. Japanese banks accumulated
their equity stakes over a long period, beginning at the end of
World War II. Consequently, it is very likely that banks repur
chased these shares at higher prices than they originally paid. In
that case, the ratio of equity investments to total assets in table
2, reported as the lower of cost or market value, would increase.
26The largest shareholders of banks are other financial institu
tions and, for keiretsu banks, nonfinancial firms in the group
(Genay, 1993). To the extent that these shareholders are better
informed about the banks than other market participants, they
would be less likely to be misled by the reported numbers. If
the top shareholders trade on their information, or signal this
information to the market in other ways, the correlations of
stock returns with bank characteristics would reflect the market’s
information and would differ from those observed with ac
counting earnings.
27For example, see Brickley and James (1986), Kane (1985,
1986), Pyle (1986), Thomson (1987a and 1987b), and Unal
and Kane (1990).

REFERENCES

Ammer J., and M. S. Gibson, 1996, “Regulation and

the cost of capital in Japan: A case study,” Board of
Governors of the Federal Reserve System, Interna
tional Finance Discussion Paper Series, No. 556.

returns of financial institutions,” Journal of Financial
Economics, Vol. 16, No. 3, July, pp. 345-371.
Brinkman, E. J., P. M. Horvitz, and Y. Huang, 1996,

Aoki, M.,andH. Patrick (eds), 1994, The Japanese

“Forbearance: An empirical analysis,” Journal of
Financial Services Research, March, pp. 27-42.

Main Bank System, Oxford, UK: Oxford University
Press.

Broussard, J. P., K, A. Kim, and P. Limpaphayom,

Ashley, L. K., E. Brewer 111, and N. E. Vincent, 1998,

1998, “Bank stock returns, interest rate changes, and
the regulatory environment: New insights from Japan,”
Rutgers University, working paper.

“Access to FHLBank advances and the performance
of thrift institutions,” Economic Perspectives, Federal
Reserve Bank of Chicago, Second Quarter, pp. 33-52.
Bank for International Settlements (BIS), 1998,

International Banking and Financial Market
Developments, August.

Buser, S. A., A. H. Chen, and E. J. Kane, 1981, “Federal

deposit insurance, regulatory policy, and optimal
bank capital,” Journal ofFinance, Vol. 35, March,
pp. 51-60.
Cargill, T. F., M. M. Hutchison, and T. Ito, 1997,

Barth, J. R., 1991, The Great Savings and Loan De

bacle, Washington, DC: American Enterprise Institute.
Benston, G. K, and G. G. Kaufman, 1998, “Deposit

insurance reform in the FDIC Improvement Act: The
experience to date,” Economic Perspectives, Federal
Reserve Bank of Chicago, Second Quarter, pp. 2-20.
_________ , 1990, “Understanding the savings and
loan debacle,” The Public Interest, Spring, pp. 79-95.
Berger, A. N., 1995, “The relationship between capital

The Political Economy ofJapanese Monetary Policy,
Cambridge, MA: The MIT Press.
Cargill, T. F., and S. Royama, 1988, The transition of
finance in Japan and the United States: A comparative
perspective, Stanford, CA: Hoover Institution Press.
Cole, R. A., 1993, “When are thrift institutions closed?
An agency-theoretic model,” Journal of Financial
Services Research, Vol. 7, No. 4, pp. 283-307.

and earnings in banking,” Journal ofMoney, Credit,
and Banking, Vol. 27, No. 2, May, pp. 432—456.

Craig, V. V., 1998, “Japanese banking: Atime of crisis,”
Banking Review, Federal Deposit Insurance Corpo
ration, Vol. 11, No. 2, July.

Berger, A. N., K. K. King, J. M. O’Brien, 1991, “The

Deniirgiic-Kunt, A., and H. Huizinga, 1997, “Deter

limitations of market value accounting and a more re
alistic alternative,” Journal of Banking and Finance,
Vol. 15, No. 4/5, September, pp. 753-783.

minants of commercial bank interest margins and
profitability: some international evidence,” World
Bank, working paper.

Brewer, E., Ill, and G. G. Garcia, 1987, “Adiscriminant

Demsetz, R. S., M. R. Saidenberg, and P. E. Strahan,

analysis of savings and loan accounting profits,” in
Advances in financial planning andforecasting, Vol.
2, A Research Annual, Cheng F. Lee (ed.), Greenwich,
CT and London: JAI Press, pp. 205-244.

1996, “Banks with something to lose: The disciplinary
role of franchise value,” Economic Policy Review,
Federal Reserve Bank of New York, Vol. 2, No. 2,
October, pp. 1-14.

Brewer, E., Ill, H. Genay, W. E. Jackson 111, P. R.
Worthington, 1997, “ATrojan Horse or the Golden

Eisenbeis, R. A., and P. M. Horvitz, 1994, “The role

Fleece? Small business investment companies and
government guarantees,” Federal Reserve Bank of
Chicago, working paper series, Issues in Financial
Regulation, No. WP-97-22, December.

of forbearance and its cost in handling troubled and
failed depository institutions,” in Reforming Finan
cial Institutions and Markets in the United States,
George G. Kaufman (ed.), Boston: Kluwer Academic,
pp. 49-68.

Brickley, J. A., and C. M. James, 1986, “Access to

FitchIBCA, 199%, BankScope, CD-ROM.

deposit insurance, insolvency rules, and the stock

Economic Perspectives

Frankel, A. B., and I*. B. Morgan, 1992, “Deregulation

Kang, J. K, and R. M. Stulz, 1997, “Is bank centered

and competition in Japanese banking,” Federal
Reserve Bulletin, Board of Governors of the Federal
Reserve System, August, pp. 579-593.

corporate governance worth it? A cross-sectional
analysis of the performance of Japanese firms during
the asset price deflation,” Ohio State University,
working paper, No. 97-6.

Genay, H., 1993, “The ownership structure of Japanese
financial institutions,” Federal Reserve Bank of
Chicago, working paper, No. WP-93-19.

__________ , 1991, “Japan’s corporate groups,”
Economic Perspectives, Federal Reserve Bank of
Chicago, January/February, Vol. 15, No. 1, pp. 20-30.
Gibson, M. S., 1996, “More evidence on the link
between bank health and investment in Japan,”
Board of Governors of the Federal Reserve System,
International Finance Discussion Papers Series,
No. 549, May.

________ ,1995, “Can bank health affect investment?
Evidence from Japan,” Journal ofBusiness, Vol. 68,
July, pp. 281-308.
Hall, B. J., and D. E. Weinstein, 1997, “The myth of
the patient Japanese: Corporate myopia and financial
distress in Japan and the U.S.,” Proceedings of the
Thirty-Third Conference on Bank Structure and Com
petition, Federal Reserve Bank of Chicago, pp. 544-556.

Kaplan, S. N., and B. A. Minton, 1994, “Appointments

of outsiders to Japanese boards: Determinants and
implications for managers,” Journal of Financial
Economics, Vol. 36, pp. 225-258.
Kaufman, G. G., 1998, “Preventing banking crises in
the future: Lessons from past mistakes,” Review of
Monetary and Financial Studies, November.

Keeley, M. C., 1990, “Deposit insurance, risk, and

market power in banking,” American Economic
Review, Vol. 80, No. 5, December, pp. 1183-1200.
Kuester, K.A. and J. M. O’Brien, 1991, “Bank equity
values, bank risk, and the implied market values of
banks’ assets and liabilities,” Board of Governors of
the Federal Reserve System, Finance and Economics
Discussion Series, No. 7.
Lindgren, C.J., G Garcia, and M. L Saal, 1996, Bank
Soundness and Macroeconomic Policy, Washington,
DC: International Monetary Fund.

Hoshi, T., A. Kashyap, and D. Scharfstein, 1991,

Moody’s Investors Services, 1998, Banking System

“Corporate structure, liquidity, and investment: Evi
dence from Japanese industrial groups,” Quarterly
Journal ofEconomics, Vol. 106, May, pp. 33-60.

Outlook: Japan, Global Credit Research, July.

_________ , 1990, “The role of banks in reducing
the costs of financial distress in Japan,” Journal of
Financial Economics, Vol. 27, pp. 67-88.
Japan, Ministry of Finance, 1998, On the Banks’ SelfAssessment Result ofAsset Quality, available on the
Internet at http://www.mof.go.jp/english/elsftl.htm,
January 12.

Kane, E. J., 1986, “Appearance and reality in deposit

insurance: The case for reform,” Journal of Banking
and Finance, Vol. 10, pp. 175-188.

_________ , 1997, Major Japanese Banks ’ Financial
Flexibility Declines, Global Credit Research, December.
Nakatani, L, 1984, “The economic role of financial

corporate grouping,” in The Economic Analysis of
the Japanese Firm, M. Aoki (cd J, Amsterdam: NorthHolland, pp. 227-258.
Peek, J., and E. S. Rosengren, 1997, “Collateral damage:
Effects of the Japanese real estate collapse on credit
availability and real activity in the United States,”
Federal Reserve Bank of Boston, working paper,
No. 97-5.

Pyle, D. H., 1986, “Capital regulation and deposit

_________ , 1985, The Gathering Crisis in Federal
Deposit Insurance, Cambridge, MA: MIT Press.

insurance,” Journal of Banking and Finance, Vol. 10,
pp. 189-201.

Kane, E. J., and M. Yu, 1996, “Opportunity cost of

Sheard, P., 1989, “The main bank system and corporate
monitoring and control in Japan,” Journal ofEconomic
Behavior and Organization, Vol. 11, No. 3, May,
pp. 399-422.

capital forbearance during the final years of the FSLIC
mess,” Quarterly Review of Economics and Finance,
Fall, pp. 271-290.

Federal Reserve Bank of Chicago

Suzuki, Y. (ed), 1987, The Japanese Financial System,
Oxford, UK: Oxford University Press.

Toyo Keizai, Inc., 1991-98, Japan Company Handbook,
First Section, 1991-1998, Tokyo.

Tatewaki, K., 1990, Banking and Finance in Japan:
An Introduction io the Tokyo Market, London:
Routledge.

Unal, H., and E. J. Kane, 1990, “Modeling structural
and temporal variation in the market’s valuation of
banking firms,” Journal ofFinance, Vol. 45, No. 1,
March, pp. 113-136.

Thomson, J. B., 1992, “Modeling the bank regulator’s

closure option: A two-step logit regression approach,”
Journal ofFinancial Services Research, Vol.6, pp. 5-23.
________ , 1987a, “The use of market information in
pricing deposit insurance,” Journal ofMoney, Credit,
and Banking, Vol. 19, No. 4, November, pp. 528-537.

Weinstein, D. E., and Y. Yafeh, 1998, “On the costs

of a bank-centered financial system: Evidence from
the changing main bank relations in Japan,” Journal
ofFinance, Vol. 53, No. 2, April, pp.635-672.

_________ , 1987b, “FSLIC forbearances to stock
holders and the value of savings and loan shares,”
Economic Review, Federal Reserve Bank of Cleveland,
Quarter 3, pp. 26-35.

Economic Perspectives

Foreign growth, the dollar, and
regional economies, 1970-97

Jack L. Hervey and William A. Strauss

Introduction and summary
Midwestern manufacturing industries have under
gone a substantial transformation during the past
two decades that has positively influenced the region’s
economic growth. Extensive industrial restructuring
and technological innovation (both of which contrib
ute to increased productivity) have contributed to
this transformation. In addition, the region’s econom
ic growth, as well as that of other U.S. regions, is of
ten associated with economic developments external
to the domestic market, such as expansion in foreign
market demand, and favorable movements in the dol
lar exchange rate (that is, a dollar depreciation) during
the 1970s and from the mid-1980s to the mid-1990s.
Since the mid-1980s, international markets have
received considerable attention as drivers of growth
in manufacturing for the Midwest and other U.S.
regions. In particular, the coincidence of the recovery
ofthe Midwest economy,1 expanding foreign markets,
and the U.S. dollar’s depreciation in foreign exchange
markets since 1985 has led some observers to forge a
link between the recovery and foreign growth and the
dollar’s depreciation. From the late 1980s well into the
1990s, an association between the nation’s industrial
recovery and expansion, especially in the Midwest,
and the dollar’s depreciation was a common topic of
discussion, especially in the popular press. Examples
of this view are expressed by Koretz (1988) and
Prowse(1995).
In this article, we examine the impact on U.S.
regional economies of exchange rate change and for
eign demand growth. We address the following ques
tions: Are there differences in the exchange rates that
regions face? Did depreciation in the dollar exchange
rate measurably influence economic growth in the
Midwest and other regions in 1970-97? Does growth
abroad faced by the regional economies differ by
region? And to what degree did foreign economic

Federal Reserve Bank of Chicago

activity influence U.S. and regional economies in
1970-97? It is clear that different regions have differ
ent industrial structures. We expect these differences
to reflect regions’ trading partners and the industrial
makeup of their export basket. Therefore one would
expect these differences to be reflected in the foreign
exchange rates and foreign demand faced by different
regions.
We construct region-specific indexes for exchange
rates and foreign economic growth. We then examine
trends in these indexes. Next, we incorporate the two
region-specific index measures into a regression anal
ysis that addresses their impact on economic growth
in U.S. regions in 1970-97.2
The dollar experienced substantial variability in
its foreign exchange value during the period under
review; and the expected regional differences appear
prominently in our region-specific exchange rate in
dexes. The period was characterized by dollar depre
ciation from mid-1971 to mid-1980 and again from early
1985 to mid-1995. The bulk ofthe latter movement oc
curred from 1985 to 1988, although the dollar contin
ued to depreciate relative to several major currencies
into the mid-1990s. However, our analysis suggests
that Midwest manufacturing goods exporters, in the
aggregate, faced an appreciating dollar exchange rate
in 1988-96, rather than a continuation ofthe depreci
ating trend reflected in movements of the dollar rela
tive to several major currencies. The Midwest-specific

Jack L. Hervey is a senior economist and William A.
Strauss is a senior economist and economic adviser at
the Federal Reserve Bank of Chicago. The authors wish
to acknowledge the helpful contributions of Charles
Evans, David Marshall, Michael Kouparitsas, Keith
Phillips, and William Testa. Valuable research assistance
was provided by Timothy McKenna and Sharon Paulus.

exchange rate indexes reflect the heavy concentration
of the region’s export trade to other North American
markets, where the dollar was relatively strong. Ac
cording to this index, manufactured goods export
industries faced a real aggregate dollar exchange rate
that was higher (that is, had appreciated) at the end
of 1996 than in 1988 or even in 1970, the last full year
before the 1971 dollar devaluation and the subsequent
floating of the dollar.3 The appreciation of the dollar
exchange rate index observed for the Midwest, how
ever, was not characteristic for all regions of the
country. For example, the Far West indexes reflect the
relatively greater importance ofthe Pacific Rim markets,
and, consequently, show a weaker region-specific
dollar than the Midwest indexes. Far West manufac
turing industries experienced a marked dollar depreci
ation through the mid- 1990s.
The measurable economic impact of change in
the dollar exchange rate on overall regional economic
activity is less clear. Our statistical analysis examines
the relationship between regions’ output growth
(gross regional product—GRP) and the aggregate
dollar exchange rates they face. It suggests that while
region-specific exchange rates may exhibit different
trends for different regions of the country, variation
in these region-specific dollar measures may not be
an important factor explaining economic activity in
manufacturing industries. The ability of these region
al indexes to explain change in the gross measures of
regional economic activity is weak. On the other hand,
these foreign market indicators are significant with
regard to explaining change in total U.S. growth,
although the direction of the impact is apparently
through the import sector rather than the export sec
tor By reducing the dollar cost of imported factors of
production, such as raw materials and components,
an appreciating dollar may contribute to additional
domestic value-added output.
We also examine the relationship between changes
in regions’ output and their region-specific foreign
demand, that is, average economic growth in those
markets to which specific regions export goods. The
intuition is that the stronger the economic growth in
a region’s foreign markets, the greater the region’s
growth in exports to those markets will be and, in turn,
the greater the impact on the region’s economic growth.
Our results suggest that positive growth in a region’s
foreign markets tends to exert a positive impact on a
region’s manufacturing activity. However, the statisti
cal significance of the link is weak. Region-specific
foreign growth rates vary substantially. The concen
tration of the Midwest’s foreign markets in the Ameri
cas results in that region showing a substantially

lower rate of foreign income growth than most other
U.S. regions. In contrast, the Far West’s strong con
centration in Pacific Rim markets is reflected in rela
tively stronger foreign market growth during the
1970-97 period. For the U.S. as a whole, foreign
demand is a positive and highly significant contribu
tor to growth.
We include within our statistical analysis a oneperiod lag measure of U.S. economic activity for each
of the regions, which turns in positive and highly sig
nificant results. This strongly supports the contention
that the U.S. economy is still the primary factor influ
encing regional economic growth; this is especially
true for the Midwest. This result supports recent work
on regional input-output analysis.

The international economy and the U.S.
International markets have become increasingly
important to the U.S. economy during the past three
decades. Since 1960 the constant dollar volume of
U.S. goods exports has increased about eightfold.
Foreign demand for U.S. goods has also increased
relative to the total volume of U.S. goods production.
This is reflected in a substantial increase over time in
the proportion of domestic goods production enter
ing export markets. In 1960, for example, the real value
of U.S. goods shipped to foreign markets accounted
for about 8.5 percent of domestic goods output. By
1970, the export share of domestic goods output had
increased to about 11 percent and by 1995 it had
reached about 24 percent.4 At the same time that ex
ports were becoming a more important component of
the U.S. economy, there was also a redistribution of
output and exports among U.S. regions.5 While the
dollar value of Midwest exports of manufactured goods
increased substantially during this period, the Mid
west’s share of U.S. manufactured goods exports
actually declined, from over 30 percent in the early
1970s to a little over 20 percent in the early 1990s.
The postwar period also saw a change in the
world’s industrial and trade regime. Seven rounds of
multinational trade negotiations, beginning in the late
1940s with the establishment of the General Agreement
on Tariffs and Trade (GATT), contributed to an in
crease in world trade flows. Foreign countries and
industries recovering from the devastation of World
War II seized the opportunity created by increasingly
open markets. Rebuilt and relatively more efficient
manufacturing infrastructure in Europe and Japan
increased competitive pressure on the older, less effi
cient U.S. manufacturing industries.
The early postwar period also saw the U.S. dollar
emerge as the exchange rate standard in world trade.

Econom ic Perspectives

The dollar exchange rate standard, which grew out of
the 1944 Bretton Woods Agreement, began to break
down in 1971 when stress on the fixed exchange rate
regime forced the first of two dollar devaluations.
Eventual abandonment of the fixed exchange rates
regime came with the adoption of a floating dollar in
March 1973. Subsequent depreciation of the dollar
during the remainder of the decade helped keep U.S.
goods prices competitive in world markets and U.S.
exports continued to expand, increasing 214 percent
between 1973 and 1980. However, the value of goods
imports grew 254 percent over the same period. Thus,
even with the dollar’s depreciation during the 1970s,
foreign competition continued to increase.
During the first half of the 1980s, a period of his
torically high U.S. interest rates, foreign exchange
markets abruptly turned around and the dollar appre
ciated sharply through February 1985. This in combi
nation with the worldwide recession of the early 1980s
contributed to a deterioration in the price competitive
position ofU.S. goods in world markets (that is, the
foreign currency cost ofU.S. goods rose due to the
exchange rate effect). As a result, the value ofU.S.
goods exports declined 10 percent between 1981 and
1983 and remained below 1981 levels until 1987. On
the other hand, import growth slowed, but increased
in value by 55 percent from 1981 to 1987.
The increased intensity of international competi
tion contributed to turning the Midwest, the nation’s
manufacturing heartland of earlier decades, into the
Rust Belt during the late 1970s and the first half of
the 1980s. An economic recovery in the Midwest that
began during the second half of the 1980s coincided
with a realignment and restructuring of manufacturing
industries and a resumption in the rapid growth in
export markets. This confluence of developments
spawned the view, noted earlier, that the resurgence
in manufacturing was largely attributable to the sharp
depreciation ofthe dollar during 1985-88 and the more
gradual dollar depreciation through the mid-1990s. In
addition, however, one can not ignore the positive
impact of economic expansion in foreign markets and
the emergence of rapidly growing markets in Asia and
Latin America.
In short, exchange rate change and expansion in
international markets are widely held to have become
a more important influence on the U.S. and its region
al economies during the past three decades.

Exchange rates and economic growth
We examine two factors that influence U.S. inter
national trade, with reference to U.S. regional econo
mies. How do 1) exchange rate changes and 2)
changes in foreign demand influence these regional

Federal Reserve Bank of Chicago

economies? To answer this question, we construct
two measures—a region-specific exchange rate index
and a region-specific foreign economic growth index.
Our aim is to identify whether different regions
ofthe U.S. face different exchange rates; whether
there are differences across U.S. regions in the aver
age economic growth (foreign demand) they face in
their export markets; and whether the region-specific
measures of exchange rate change and foreign eco
nomic growth contribute to explaining changes in re
gions’ economic activity.

Why a regional exchange rate index?
At any time there is only one exchange rate for
the U.S. dollar vis-a-vis any other currency. We sug
gest, however, that different U.S. regions, by virtue
of their different industrial makeup and the foreign
markets in which their industries are active, face dif
ferent sets of exchange rates. Thus, a general obser
vation that the dollar is depreciating or appreciating
may have different implications from one region to
another. We construct a set of aggregate exportweighted dollar exchange rate indexes for selected
geographic regions. We identify broad manufacturing
industry classifications within each region. We focus
on identifying exchange rate variations and the dif
ferences in the composition of export markets that
influence selected U.S. regions. This is an area of
research only beginning to receive attention in the
literature; see Clark, Sawyer, and Sprinkle (1997 and
1999),6 Cronovich and Gazel (1998), and Hervey and
Strauss (1996 and 1998).
We can identify potential differences in exchange
rates faced by different regions of the country by
looking at regional trade patterns. Figure 1 shows
manufactured durable goods exports to major world
markets for the U.S. as a whole and eight regions.7
For example, the Midwest sends (1993-94 average)
nearly 60 percent of its manufactured durable goods
exports to markets in which the dollar has been histori
cally strong, that is, other North American markets
(46 percent to Canada and nearly 13 percent to Mexico),
while only 15 percent of its exports go to European
markets and only 4 percent to Japan. On the other
hand, durable goods industries in the Far West ship a
substantially higher proportion of their exports to
markets in which the dollar has depreciated: 17 per
cent to Japan and 25 percent to Europe.
Clearly, markets in which the dollar has appreci
ated in recent years (Canada and Mexico) have been
more important to durable goods manufacturers in
the Midwest than in the U.S. overall. The magnitude
and structure ofU.S. and Canada/Mexico trade
(cross-border trade) is of some concern with regard to

FIGURE 1

Share of durable goods exports by region or country of destination, 1993-94

aExcluding Mexico.
bExcluding Japan.
Source: Authors' calculations from Massachusetts Institute for Social and Economic Research, 1993 and 1994,
State ofexporter location, series 2, CD-ROM.

Economic Perspectives

the exchange rate and foreign demand growth issues
we examine in this article. The degree of integration
of these markets, especially in the production pro
cesses in the automotive and electronics industries,
may substantially reduce the influence of exchange
rate change on cross-border trade. Intra-firm trade,
while a part of the export-import statistics, may not
truly reflect a market exchange. Although there is a
large volume of trade, we cannot say how much of it
faces an exchange rate transaction (see box 1).

Construction of the regional dollar indexes
We identify nine regions: the U.S. and eight aggre
gations of states that correspond to the U.S. Depart
ment of Commerce, Bureau of Economic Analysis
(BEA), geographical breakdown of the United States.8
Our primary focus is on the BEA s Great Lakes (Mid
west) region (Illinois, Indiana, Michigan, Ohio, and
Wisconsin). Details of our calculations are in techni
cal appendix 1.
Abroad range of regional research examining
exchange rates and/or international trade effects has
focused on specific state effects, including, for exam
ple, Branson and Love (1986), Carlino (1990), Cough
lin and Pollard (1998), Hayward and Erickson (1995),
and a work by Cronovich and Gazel (1998) that exam
ines the impact of exchange rate and foreign income

change on state-defined measures of economic activ
ity, such as employment or exports. We do not report
individual state indexes here because of distortions
in gross export data that are exaggerated when using
state-level data. These distortions arise because the
complete manufacture of a product may not take
place within one state. More likely, the manufacture
of an intermediate component may be carried out in
one state, then shipped to another state or several
more states for further processing. State export by
destination data based on the value added in manu
facturing by state are not currently available. Future
work building on the regional input-output literature,
such as Israilevich, Hewings, Sonis, and Schindler
(1997), may provide these data.
We use regional aggregations of states’ exports
by destination from the U.S. Bureau of the Census,
State of export—Location of exporter series.9 These
data are from the shipper’s export declaration for the
state of location of the exporter, which means that the
value of intermediate goods eventually exported may
not be attributed to the appropriate state. The aggre
gation of states into BEA regions should reduce,
although it will not fully eliminate, this mismeasure
ment. The state/region and industry weights in the
indexes are based on U.S. exports by country of

BOX 1

Ro und-trip-trade: Canad a and Mexic o

Canada is an important market for Midwest man
ufacturing and, as such, a critical element in
the Midw est dollar index. Mexico is similarly
important to the Southwest. The unique relation
ships the U.S. has with Canada and Mexico raise
concern about the interpretation of the regional
indexes.
In certain industries, there is a high level of
integration of production facilities across the
borders (in particular, the automotive industry
across both borders and the maquiladoras indus
tries along the U.S./Mexico border). The ques
tion arises whether the effect of a change in the
Canadian dollar/U.S. dollar or Mexican peso/U.S.
dollar exchange rate is the same for an integrated
firm (with cross-border intra-firm transactions,
or round-trip trade) as for unrelated firms (with
cross-border inter-firm transactions). Are these
transactions booked in U.S. dollars or do exchange
rates make a difference?
There appear to be no simple answ ers to
these questions. Conversations with individuals
in the auto industry suggest that even for such a

Federal Reserve Bank of Chicago

cross-border integrated industry, exchange rate
change does make a difference, but in the long
er-term decisions such as plant investment and
location. In that case, the existence of an inte
grated market across borders might not bias the
impact of exchange rate changes on the regional
indexes viewed in a long-term context. In other
integrated industries, transactions are denomi
nated in dollars and the exchange rate transla
tion occurs only if the final product enters a
third country foreign market.
For the maquiladoras industries on the Mexico/U.S. border, most cross-border transactions
are denominated in U.S. dollars. Thus, change in
the peso/dollar exchange rate does not have a
direct effect on these transactions. Nonetheless,
a peso devaluation, for example, will influence
the local value-added portion, that is reduce in
terms of dollars (through cheaper labor and com
ponents), the dollar value of the transaction if
the final product is shipped back (imported) to
the U.S. So, even in this case exchange rate
change counts to some degree.

destination by industry for 1993 and 1994.10 The Cen
sus location data are adjusted for exports unallocated
by state and industry classification and are made
available by the Massachusetts Institute for Social
and Economic Research (MISER).11
To make this project a more manageable size, we
grouped the 20 two-digit manufacturing Standard In
dustrial Classifications (SICs) into three broad class
es—total manufactured goods, durable manufactured
goods, and nondurable manufactured goods.12 Our
region and industry breakdown results in 27 exchange
rate indexes. The full range of regions and industries
would have resulted in 180 indexes. However, the size
restriction imposes a cost. Further breakdown of the
industries might provide more information on the im
pact of round-trip trade with Canada and Mexico—
essentially intra-firm transactions where, at least in
the short-term, intermediate goods traverse the border
without entering the price/market system.
To provide a known index for comparison at the
national level, our regional indexes incorporate the
currencies of the same 44 countries as the J. P. Morgan
(JPM) real effective exchange rate indexes. These
countries account for more than 90 percent of U.S.
goods exports.
The use of export-only weights is an unusual
methodology in the construction of an aggregate
exchange rate index. Aggregate exchange rate indexes
typically use a weighting mechanism based on bilateral
trade weights (as in the JPM noted above), for exam
ple, U.S. exports plus imports by country of destination
or source, or multilateral trade weights, for example,
total world trade (that is, total exports plus total im
ports) for each country. Due to the lack of the lack of
available import data by state, we are constrained to
constructing a regional index based on export weights.
A multilateral trade weighting scheme (for example, as
used in the Federal Reserve Board’s nominal tradeweighted index) would be better able to account for the
third-country effects of exchange rate changes. Howev
er, multilateral weights are not applicable to U.S. regional
indexes because the same country weight (that is, its
share of world trade) would apply to each region.13
The use of export-only weights requires that con
clusions be carelully stated. These indexes relate only
to an aggregation of exchange rates that exporters
face directly. There are two areas of inquiry of interest
with respect to exchange rates: 1) the impact of ex
change rate change on the regional economy and
regional manufacturers through their export markets,
and 2) the impact of exchange rate change on the
regional economy and manufacturers as influenced
by imports. At this stage, we can only address the
exporter side.14

A final issue of concern in the construction of
our regional indexes relates to the 1993-94 period we
use for the export and industry weights. It is well
established in the literature that there were marked
changes in U.S. trade patterns from 1970 through the
late 1980s. Hickok and Orr (1989-90), Hervey (1990),
and Hickok (1991) document substantial changes in
the foreign market shares and industrial composition
of U.S. exports during that period. The selection of
the fixed period (1993-94) base for trade weights rais
es a question about the potential bias in the indexes
as they move away from the base period. Hervey and
Strauss (1987) suggest that export weights that use
a moving average (for example, a 12-quarter moving
average) to account for change over time in the com
position of trade by destination would be a consider
able improvement over the fixed-period weights. In
addition, Coughlin and Pollard (1998) make a case for
the use of chained weights in the construction of
aggregate exchange rate indexes to lessen the wellknown problems associated with the fixed-period
base of the Laspeyres-type index used in most aggre
gate exchange rate indexes, including the ones in this
study. Acknowledging these shortcomings, we are con
strained to use fixed-year weights here because of the
limited availability of consistent historical state-export
data prior to 1993.
An exchange rate is a price of one currency in
terms of another. But it is not the only relevant price.
Rates of change of within-country prices across
countries are also of interest, especially when coun
tries experience marked differences in inflation. A
change in the exchange rates tells only half the story.
The focus of exchange rate adjustment should be on
real exchange rates. The preferable internal price
series for this exercise is one that relates specifically
to the goods traded. However, price series with such
detail are not available for the spectrum of countries
and industry groups included in our regional indexes.
We use producer price series provided by J. P. Morgan.15
The exchange rate series for countries in the in
dexes are monthly averages from the International
Monetary Fund, International Financial Statistics
series, except for Taiwan and Hong Kong, which are
from the Federal Reserve Board series.

Regional exchange rate indexes
The regional export-weighted dollar indexes in
some cases contradict the common perception that
the dollar continued to depreciate over much of the
period 1970-97, particularly in 1988-96. Figure 2 plots
the region-specific dollar and the U.S. dollar for the
periods 1970-97 and 1988-96. In the aggregate, the

Econom ic Perspectives

Midwest index (figure 2, panels E and F) shows an
appreciating trend for both periods, particularly for
1974-97.16
For 1988-96, divergent trends from that recorded
for the U.S. appear primarily in the indexes for the

Federal Reserve Bank of Chicago

Southwest, the Mountain states, and to some degree
the Mideast and Far West. While the differences are
not large, the deviations of the regional indexes from
the U.S. index are either consistently positive (espe
cially for the Midwest, Southeast, and Southwest) or

negative (in the case of the Mideast, Far West, and
Mountain states) (see figure 3).
Figure 4 provides an interesting perspective on
the Midwest’s dollar index relative to those ofthe
other regions. In most cases the Midwest dollar index

for manufactured goods deviates substantially from
the indexes of the other regions and is higher than
those ofthe other region-specific indexes, peaking in
1995 at 20 percent versus the Mountain states, 17
percent versus the Far West, 15 percent versus the

Economic Perspectives

Mideast, and about 12 percent versus New England.17
On the face of it, this suggests that Midwest export
ers of manufactured goods are facing a substantial
real exchange rate appreciation (price disadvantage)
in their foreign markets, relative to other regions.

Federal Reserve Bank of Chicago

The industrial composition of exports also appears
to influence the regional exchange rate pattern. The
U.S. export-weighted real indexes for durables and
nondurables are virtually identical. Figure 5 plots the
percentage deviation in the regions’ durables and

nondurables indexes versus the U.S. indexes. For
example, in 1995 the Midwest’s exchange rate index
for durable manufactures was more than 9 percent
higher than the comparable U.S. index. However, the

Midwest’s nondurables index was only 2 percent
higher than the comparable U.S. index. Generally, the
region-specific nondurables indexes are less volatile
and follow the national index more closely than do

Economic Perspectives

the durable goods indexes. The Southwest is the
only region to show a large positive deviation from
the U.S. for nondurables (it shows a similar deviation
for durables). We suspect that this atypical result for

Federal Reserve Bank of Chicago

nondurables may be related to the high concentration
of maquiladoras industries (mostly U.S. industries
on the Mexican side of the border that produce for
the U.S. market; see box 1) along the U.S./Mexico

border and the large volume of crossborder trade,
especially in the textile and apparel industries.

Foreign demandfaced by regional markets
In an open economy, foreign demand contributes
to overall demand on the economy’s output. We would
expect economic expansion or contraction in foreign
markets to influence the growth in exports to those
markets. Thus, we would expect economic activity in
U.S. regions that export to be positively influenced
by growth abroad. Furthermore, because economic
growth is not uniform across countries and regions
do not export uniformly across foreign markets, we
would expect economic growth abroad to influence
U.S. regional economies differently.
To measure foreign economic growth that is
unique to the markets served by individual regions, we
use an average foreign gross domestic product (GDP)
growth (in real terms) that is exchange rate neutral
(measured in terms of an individual country’s home
currency). Details of our calculations are in technical
appendix 2. We construct nine “region-specific” for
eign GDP growth rate series—for eight regions and
the U.S. Cronovich and Gazel (1998) take a similar
approach to measuring export-weighted foreign mar
ket growth for individual states.
We weight GDP growth rates for 20 major export
markets for each region based on the market share of
that region’s manufactured goods exports to each
country (in 1993-94). Thus, the more important a spe
cific country is for a region’s exports, the larger is the
weight placed on that country’s GDP growth; and the
countries included may vary across regions. We use
the major markets in the MISER location of exporter
series for 1993-94. Individual country GDP growth
rates are from the International Monetary Fund, the
United Nations, and individual country sources.
Figure 6 presents the region-specific foreign GDP
growth rates relative to the U.S.-specific rate. (A pos
itive number indicates that growth in a region’s foreign
markets is larger than growth in foreign markets for
the U.S.) The data suggest that, for the most part,
aggregate regional foreign market growth tends to
deviate substantially from the U.S. average. Certain
regions appear to experience consistently faster or
slower growth in their foreign markets than does the
U. S. In particular, foreign demand growth in Midwest
markets (heavily concentrated in the Americas) was
well below the U.S. average throughout the 1970-97
period. On the other hand, the Far West markets,
heavily concentrated in the Pacific Rim, recorded
consistently higher growth. The Southwest shows a
highly variable growth pattern with broad swings in the
early and late 1980s and mid-1990s; these are probably

largely attributable to the sharp deterioration in economic
activity in Latin America and Mexico in the 1980s (credit
crisis) and Mexico in the mid-1990s (depreciation ofthe
peso and subsequent austerity measures).

Regional economic activity: Regional exchange
rates andforeign economic growth
Our region-specific dollar exchange rates and
region-specific foreign economic growth measures
suggest that different regions of the country face
different mixes of exchange rate and foreign growth
change and may be influenced differently by devel
opments in international markets. Do these regional
measures of international exposure measurably influ
ence economic growth within these regions? Studies
such as Cronovich and Gazel (1998) indicate that
region-specific exchange rate and foreign demand
growth influence regional export markets. But are the
effects large enough to significantly influence overall
regional economic growth?
To explore this further, we use four regression
models to explain variation in regions’ real output or
gross regional product (GRP). We use four dependent
variables for output—total GRP, GRP attributable to
manufacturing industries, GRP attributable to durable
manufacturing industries, and GRP attributable to non
durable manufacturing industries (the last three indus
try designations parallel the definition of the regional
exchange rate indexes). The analysis views the two
constructed international market variables—regionspecific exchange rate indexes and region-specific
foreign GDP growth—as externally generated shocks
to the regional economies. One may quibble with this
assumption, as clearly the relationship between the
growth ofU.S. and foreign economies is not entirely
independent and U.S. foreign exchange rates are not
independent ofU.S. domestic monetary and econom
ic policy. It is also true that within the U.S., develop
ments in regional growth are not independent of
developments in other regions or of the U.S. economy.
To isolate these domestic influences on regional
growth, we include two measures of recent domestic
economic activity—previous period U.S. GDP growth
and previous period own-region economic growth.
We also include a measure of oil prices as an indepen
dent, external shock variable. Other researchers, for
example, Davis, Loungani, and Mahidhara (1997), have
found this to be meaningful in explaining regional
economic growth. (One might question the external
or supply shock nature of this variable. During the
1970s and early 1980s this nature of oil price determi
nation was reasonably clear, but it has been less clear
since them). Technical appendix 3 provides a more de
tailed discussion of the makeup of the model.

Econom ic Perspectives

FIGURE 6

Difference in foreign GDP growth (region less U.S.)
A. New England

B. Mideast

C. Midwest

D. Plains
percentage points

percentage points

E. Southeast

F. Southwest

percentage points

_2 I i i i i i i_i i i i i i i i i_ i i i i i i i i i i i i
1972
'77
'82
'87
'92
'97
G. Mountain

What results do we expect from this analysis?
We expect the lagged measures of domestic output
growth (U.S. and regional) to be positively related to
current output growth. We normally expect oil price
change to be related negatively to output. If oil prices

Federal Reserve Bank of Chicago

H. Far West

increase, production that is energy intensive or
product that is energy intensive in its use becomes
less competitive and output declines, pending a
redistribution in resources and output. However, for
a region in which energy production is an important

TABLE 1

Regression results for regional output equations
T0t£ I GRP

1. United States
Region-specificforeignGDP
Priceofcrudeoil byindustry
Real GDP, U.S. by industry
RealGRP byindustry
Exchange rate index by industry
Exchange rate index by industry.
one-period lag
Exchange rate index by industry.
two-period lag
Sign/significanceofsumof
exchange rate variables
R2
2. New England
Region-specificforeignGDP
Priceofcrudeoil by industry
Real GDP, U.S. by industry
RealGRP byindustry
Exchange rate index by industry
Exchange rate index by industry.
one-period lag
Exchange rate index by industry.
two-period lag
Sign/significanceofsumof
exchange rate variables
R2
3. Mideast
Region-specificforeignGDP
Priceofcrudeoil by industry
Real GDP, U.S. by industry
RealGRP byindustry
Exchange rate index by industry
Exchange rate index by industry.
one-period lag
Exchange rate index by industry.
two-period lag
Sign/significanceofsumof
exchange rate variables
R2
4. Midwest
Region-specificforeignGDP
Priceofcrudeoil by industry
Real GDP, U.S. by industry
RealGRP byindustry
Exchange rate index by industry
Exchange rate index by industry.
one-period lag
Exchange rate index by industry.
two-period lag
Sign/significanceofsumof
exchange rate variables
R2
5. Plains states
Region-specificforeignGDP
Priceofcrudeoil by industry
ReaIGDP.U.S. byindustry
RealGRP byindustry
Exchange rate index by industry
Exchange rate index by industry.
one-period lag
Exchange rate index by industry.
two-period lag
Sign/significanceofsumof
exchange rate variables
R2

1.3701s (0.2318)
-0.0271s (0.0143)
0.0330 (0.1237)
NA
0.1684s (0.0835)

Manufacturing GRP

Durable GRP

Nondurable GRP

2.0616s (0.5392)
-0.0668s (0.0334)
0.0440 (0.1517)
NA
0.2026 (0.1937)

2.7035s (0.7246)
-0.0610 (0.0453)
0.1118 (0.1555)
NA
0.2478 (0.2625)

1,2807b (0.4734)
-0.0705b (0.0283)
-0.1926 (0.1658)
NA
0.1252 (0.1659)
0.1073 (0.1483)

0.0838 (0.0771)

0.0461

(0.1855)

-0.0318 (0.2537)

0.0859 (0.0768)

0.2316 (0.1813)

0.3796 (0.2421)

0.69
0.0026
-0.3369
0.8534s
0.2776s
0.0221

(0.4841)
(0.2790)
(0.2851)
(0.1016)
(0.0886)

0.55
0.8844
-0.3519
0.4744b
0.2972b
0.1895

0.9461
-0.5662
0.4063s
0.3769b
0.1497

(0.1541)
+b

0.49

0.51

(0.6093)
(0.4367)
(0.1922)
(0.1152)
(0.1310)

0.1321

(0.8463)
(0.5155)
(0.1984)
(0.1389)
(0.1820)

1.0889b
-0.2970
0.3549s
0.0270
0.2504b

(0.4559)
(0.5276)
(0.1843)
(0.1363)
(0.1191)

0.0235 (0.0783)

-0.0255 (0.1198)

-0.0703 (0.1739)

0.1699 (0.1110)

0.1083 (0.0784)

0.1655 (0.1342)

0.2199 (0.1920)

0.0982 (0.1220)

0.82
-0.3516
-0.0737
0.9262s
0.1208
-0.0745

0.75
(0.2589)
(0.0493)
(0.1602)
(0.0866)
(0.0594)

-0.1828
0.1051
0.9049s
-0.0267
-0.0721

(0.2899)
(0.0840)
(0.0972)
(0.0691)
(0.0805)

0.69

0.68
-0.4906
-0.3048b
1.0667s
-0.0094
-0.2360b

(0.4170)
(0.1317)
(0.1046)
(0.0706)
(0.1138)

-0.0200
-0.0410
-0.8095s
0.0126
0.0777

(0.2438)
(0.0728)
(0.1078)
(0.0859)
(0.0762)

0.0177 (0.0491)

-0.0549 (0.0739)

-0.1842 (0.1095)

0.0905 (0.0653)

-0.0012 (0.0491)

0.0810 (0.0759)

0.0764 (0.1093)

0.0517 (0.0734)

0.87
0.1290
-0.1266
1.1769s
0.0460
0.0134

0.90
(0.3733)
(0.1152)
(0.2583)
(0.1073)
(0.1164)

0.1229
-0.0130
1.4066s
0.0003
0.1411

(0.4719)
(0.1205)
(0.1826)
(0.0829)
(0.1649)

0.2111
-0.0474
1.4053s
-0.0656
0.2538

0.89

0.91

(0.5822)
(0.1164)
(0.1692)
(0.0817)
(0.2110)

-0.0841
-0.0840
1.0179s
0.0470
0.0330

(0.2715)
(0.1107)
(0.1412)
(0.0954)
(0.0959)

0.0343 (0.1078)

0.1907 (0.1580)

0.2705 (0.2013)

0.0916 (0.0916)

-0.0426 (0.1024)

-0.2000 (0.1589)

-0.3019 (0.2013)

-0.0652 (0.0978)

0.79
0.5244
-0.0266
0.7803b
-0.0311
0.1112

0.87
(0.5027)
(0.2854)
(0.2981)
(0.1255)
(0.1038)

0.0779
-0.0803
-0.9214s
0.0343
-0.0276

(0.3307)
(0.2368)
(0.1069)
(0.0638)
(0.0795)

0.86

0.87
-0.0328
0.0743
1.0802s
-0.0817
-0.1158

(0.6044)
(0.4214)
(0.1477)
(0.0807)
(0.1584)

-0.1105
-0.1620
0.8721s
-0.0305
0.0362

(0.3983)
(0.3419)
(0.1811)
(0.1437)
(0.1025)

-0.0914 (0.0930)

0.0067 (0.0789)

-0.0940 (0.1580)

0.0790 (0.0974)

0.0065 (0.0945)

0.0134 (0.0824)

0.1262 (0.1612)

-0.1192 (0.1066)

0.92

0.88

0.67

0.73

Economic Perspectives

TABLE 1 (CONT.)

Regression results for regional output equations
Total GRP
6. Southeast
Region-specific foreign GDP
Priceofcrudeoil byindustry
Real GDP, U.S. by industry
RealGRP byindustry
Exchange rate index by industry
Exchange rate index by industry.
one-period lag
Exchange rate index by industry.
two-period lag
Sign/significanceofsumof
exchange rate variables
R2
7. Southwest
Region-specificforeignGDP
Priceofcrudeoil by industry
Real GDP, U.S. by industry
RealGRP byindustry
Exchange rate index by industry
Exchange rate index by industry.
one-period lag
Exchange rate index by industry.
two-period lag
Sign/significanceofsumof
exchange rate variables
R2
8. Mountain states
Region-specificforeignGDP
Priceofcrudeoil by industry
RealGDP.U.S. byindustry
Real GRP by industry
Exchange rate index by industry
Exchange rate index by industry.
one-period lag
Exchange rate index by industry.
two-period lag
Sign/significanceofsumof
exchange rate variables
R2
9. Far West
Region-specificforeignGDP
Priceofcrudeoil by industry
Real GDP, U.S. by industry
RealGRP byindustry
Exchange rate index by industry
Exchange rate index by industry.
one-period lag
Exchange rate index by industry.
two-period lag
Sign/significanceofsumof
exchange rate variables
R2

0.3288
-0.0223
0.7569'
0.0253
-0.0254

Manufacturing GRP

(0.2275)
(0.0470)
(0.1349)
(0.0720)
(0.0551)

0.4735b
-0.0839s
0.7792'
-0.1369'
0.0215

(0.1853)
(0.0452)
(0.0598)
(0.0433)
(0.0515)

Durable GRP

0.6775'
-0.2474'
0.7987'
-0.1112'
0.1941'

Nondurable GRP

(0.2103)
(0.0689)
(0.0514)
(0.0364)
(0.0571)

0.4391
0.0156
0.8902'
-0.0877
-0.1082

(0.2986)
(0.0618)
(0.1217)
(0.0836)
(0.0875)

0.0052 (0.0454)

0.0578 (0.0463)

0.0336 (0.0497)

0.0643 (0.0854)

0.0608 (0.0449)

0.0897s (0.0475)

0.2413' (0.0523)

0.0046 (0.0845)

0.1457
0.5831'
0.7059'
0.1856
-0.0674

0.97

0.90
(0.3645)
(0.1953)
(0.2299)
(0.1681)
(0.1091)

0.97

-0.2328
0.7880s
1.2613'
-0.0903
-0.1298

(0.4978)
(0.4386)
(0.1862)
(0.1264)
(0.1511)

0.6150
1.0224
0.8633'
0.0192
-0.0999

0.88
(0.7541)
(0.6933)
(0.2163)
(0.1704)
(0.2479)

-1.2066s
0.8779
2.0035'
0.0826
-0.3202s

(0.6677)
(0.5392)
(0.2900)
(0.1278)
(0.1847)

-0.0696 (0.0778)

-0.3706b (0.1368)

-0.5186b (0.2365)

-0.1644 (0.1294)

0.0458 (0.0810)

-0.0992 (0.1225)

0.1177 (0.2119)

-0.2176 (0.1497)

0.77

0.57
0.2707
0.9133
0.6981b
0.3770b
0.0634

(0.5464)
(0.8612)
(0.3133)
(0.1482)
(0.0973)

-0.0451 (0.0892)
-0.0205 (0.0937)

-0.5019
1.5065
1.0037'
0.0719
-0.0029

0.0991
-0.0411
0.8021'
0.3436'
0.0305

0.0431

(0.1049)

(0.3017)
(0.0784)
(0.1790)
(0.1117)
(0.0625)

-0.0965 (0.0632)

-1.0768
0.1782
1.0236'
0.0218
-0.1459

(0.5131)
(0.2038)
(0.1636)
(0.1178)
(0.1224)

-0.2211s (0.1163)

(0.1267)

0.74

(0.5240)
(1.8408)
(0.2087)
(0.1499)
(0.1367)

-0.2408s (0.1175)

-0.0086 (0.1621)

0.1282 (0.1259)

0.6058
-0.1037
0.5373'
0.3733'
0.1187

0.60
(0.6292)
(0.2124)
(0.1495)
(0.1369)
(0.1523)

0.2708
0.1416
1.0726'
-0.0155
0.2322s

(0.4996)
(0.2751)
(0.2003)
(0.1137)
(0.1304)

-0.0823 (0.1540)

-0.4004' (0.1124)

0.0543 (0.1632)

0.1728 (0.1182)

0.4088
3.4602s
0.8708'
0.0150
0.2325

(0.7794)
(2.2401)
(0.1855)
(0.1261)
(0.1612)

0.71

0.6949
-0.1330
0.6172'
0.2524b
0.2096

0.1091

0.71

-0.1398 (0.1536)

0.78

0.0052 (0.0582)

0.77

(0.4991)
(1.5517)
(0.1545)
(0.1019)
(0.1070)

-0.1803s (0.0975)

0.47

0.68

0.69

0.76

Notes: Superscript a, b, and c indicate significance at the 10 percent, 5 percent, and 1 percent level, respectively.
Numbers in parentheses are standard errors. Price of crude oil is refiner's acquisition price relevant to given industry
in given region. Exchange rate indexes are regional export-weighted exchange rate indexes. NA indicates not applicable.
GDP is gross domestic product. GRP is gross regional product.

industry, higher energy prices might lead to higher
output. In our model, oil prices reflect the importance
of a region’s GRP (total, manufacturing, durable, and
nondurable) relative to the comparable U.S. measure.
Thus, the more oil-intensive an industry or region is

Federal Reserve Bank of Chicago

relative to the U.S., the greater the expected impact of
oil prices.
We expect foreign output to relate positively to
a U.S. region’s output. The greater the importance
of manufactured exports to a region’s output, the

greater the impact we expect GDP changes in its for
eign markets to have on that region.
Regional exchange rate measures in this study
are export-weighted. Looking at export industries
only, we expect exchange rate change to have a nega
tive impact on regional output. That is, an apprecia
tion in the dollar exchange rate increases the foreign
currency price of the U.S. region’s exports; the higher
price to foreign buyers of imports from the U.S. reduces
those imports (reduces U.S. exports) and contributes
to reduced total purchases of the U.S. region’s output.
The magnitude of that impact depends upon the elas
ticity of foreign demand for the goods exported by
the region. However, we know that imports also influ
ence domestic output, and they may do so in a posi
tive way. For example, regions/industries that import
large quantities of intermediate products will respond
positively to an appreciation of the dollar. An appre
ciating dollar means lower prices to U.S. producers for
imported components (the magnitude depending upon
the degree of price passthrough). Lower production
costs may lead to increased output. Again, the mag
nitude of this impact depends on price sensitivity at
the production and final sales levels. The stronger the
import effect, the more likely the exchange rate/output
relationship will be positive. Consequently, the sign
of the exchange rate variable, as it influences regional
output, is ambiguous.
Table 1 (pages 48 and 49) shows the results of our
regression analysis. In the U.S. (panel 1), the coeffi
cients for foreign demand are positive as expected
and highly significant at standard statistical levels.
This provides strong support for the view that eco
nomic growth in foreign markets is a positive factor
contributing to U.S. economic activity. The relative
size of these coefficients by industry also suggests
that the U.S. manufacturing sector is more sensitive to
changes in foreign demand than is total U.S. GDP.
(This is what we would expect given that manufactured
goods exports account for a substantially larger share
of exports than does manufactured goods output of
total GDP.) Within manufacturing, durable goods indus
tries are more sensitive to foreign demand change
than nondurable goods industries. The oil price variable
is negative, as expected, indicating that an increase
in oil prices tends to be a drag on economic activity.
The effect is most significant for the nondurable sec
tor, which includes petroleum. The export-weighted
exchange rate variable, which enters the regression
contemporaneously and with one- and two-period
lags, exhibits positive signs in all cases but is signifi
cant in only one of the 12 industry/lag relationships.
Taken together, however, the exchange rate variables

in the four regressions are statistically significant.
The positive sign lends some support to the potential
positive relationship noted earlier between exchange
rates/imports and domestic output.
The results of the regional equations are highly
variable. What stands out is the universally positive
and significant impact on regional economic activity
of the U.S. activity variable (table 1, panels 2 through
9, line 3). This result shows that the national economy
is the primary influence on regional economic activity,
as pointed out recently in the regional input-output
literature (see for example, Hewings, Schindler, and
Israilevich, 1998).
At the regional level, the impact of oil prices on
output generally shows the expected negative sign.
However, the impact of this variable on regional eco
nomic activity is only occasionally statistically signifi
cant. Manufacturing and durable manufacturing
industries in the Southeast (table 1, panel 6) show a
significant negative influence of oil price change on
output. The two regions that show positive (though
generally not statistically significant) oil price/output
relationships are, as one might expect, in the energy
producing regions—the Southwest (panel 7) and the
Mountain states (panel 8).
The variables that we are particularly interested
in, that is, region-specific foreign GDP growth and
regional export-weighted exchange rates, also give
mixed results.
The expected positive sign of region-specific for
eign GDP growth on regional economic activity is sup
ported in 21 of the 32 region/industry categories. In
most cases, however, even where the expected posi
tive relationship exists, the statistical significance is
weak. In the Southeast (panel 6, line 1), the manufac
turing and durable classifications record positive and
significant relationships. In New England (panel 2,
line 1), the foreign demand variable contributes to a
positive and statistically significant impact on output
in the nondurable manufacturing industry. The magni
tude of the coefficients, which may be interpreted as a
measure of the sensitivity of the region’s economic
growth to foreign economic growth (income elasticity)
is modest. In the Southeast, for example, a 1 percent
increase in foreign GDP growth would have about a
0.7 percent positive impact on GRP.
The regression results reflecting the impact of
changes in regional exchange rates on output also
show mixed results. Of the 32 region/industry equa
tions (excluding the U.S. equation), half of the sums
of the coefficients (table 1, line 8 of each panel) of the
exchange rate lag structure (table 1, lines 5, 6, and 7)
are negative and half positive. Coefficients on five of

Econom ic Perspectives

the positive sums (New England, the Mideast, and
the Southeast) are significantly different from zero;
three of the negative sums (Mideast and Southwest)
are significant. As mentioned earlier, a negative impact
of an appreciating dollar, for example, indicates that
the higher price of a region’s exports results in a reduc
tion in foreign purchases (reduced exports from the
region), which has a negative impact on the region’s
output. A positive impact from an appreciating dollar
may indicate that the lower dollar cost of a region’s
imports of production components ultimately leads to
an increase in GRP. We suggest that the difference
between these two outcomes may be due to variation
in the industrial composition across regions, which
we are unable to discern with the levels of industry
aggregation we use in this study.
As noted at the outset, we are particularly inter
ested in the economic revival of the Midwest during
the late 1980s. The importance of international markets
and exchange rate change in the post-1985 period
have been widely touted as influential in the region’s
recovery. Our results suggest that the Midwest econ
omy is critically dependent on the U.S. economy (table
1, panel 4, line 3). Indeed, based on the magnitude of
the coefficients, the Midwest economy is significantly
more sensitive to conditions in the U.S. economy than
are the other seven regions. This holds true for man
ufacturing industries overall and for durable manufac
turing. The international sector variables we identify,
region-specific foreign economic growth and regionspecific exchange rates, do not appear to provide a
significant additional explanation for Midwest eco
nomic activity, although the signs of the coefficients
are plausible.

Conclusion
Although regional economies are part of the U.S.
economy, regions differ substantially in their industrial
makeup and the extent of their involvement in interna
tional markets. While they face a common external
border and a common set of national exchange rates,
different regions and their industries may face a dif
ferent set of exchange rates and foreign demand con
ditions. We have examined these differences through
the construction of region-specific exchange rate and
foreign GDP growth indexes.
Our export-based exchange rate indexes indicate
that Midwest manufactured goods exporters, for
example, faced an appreciating dollar from the late
1980s until the mid-1990s related to the composition
of their foreign markets and their heavy concentration
in durable goods industries. This contradicts the

Federal Reserve Bank of Chicago

common perception, based on exchange rate trends
for major currencies, that the dollar was depreciating
during that period.
Our foreign GDP growth indexes also suggest
some variation in the rate of foreign economic growth/
demand faced by various U.S. regions. Because the
Midwest’s primary foreign markets are at the low end
of the growth spectrum, its region-specific foreign
growth has been the lowest of the eight BEA eco
nomic regions since 1980. With exports going mainly
to high-growth Asian economies during the 1970s
to mid-1990s, the Far West is at the high end of the
spectrum.
These region-specific measures of foreign market
influence do not appear to consistently contribute to
a statistically significant, measurable impact on total
regional economic activity. Forthe U.S., foreign GDP
growth does show a strong positive and significant
impact on economic activity. However, in only one
region, the Southeast, is the impact of foreign demand
growth strong enough to impose a positive and signif
icant impact on GDP growth in manufacturing and dura
ble manufacturing. The exchange rate measures show
a significant impact on U.S. GDP, but this is apparent
ly through the terms of trade effect on imports, which
promotes domestic output through the lower relative
cost of component imports. This pattern holds for sev
eral of the regions (New England, the Mideast, and the
Southeast) for selected manufacturing classifications.
In only one region, the Southwest, does the exchange
rate variable appear to negatively and significantly
influence the region’s overall economic activity.
The Midwest economy does not respond signifi
cantly to the foreign GDP or exchange rate variables,
given the statistical formulation we use here. Howev
er, our results indicate that the impact of the domestic
economy variable (the home market) is significantly
more important for the Midwest on an industry by
industry basis than for the other regions (though the
variable is significant for all regions). The only region/
industry equations with larger sensitivity measures
on an industry by industry basis are nondurables,
most likely petroleum-related, in the Southwest and
Far West.
Finally, the main implication of this study rein
forces recent work in regional input-output research.
While international markets are certainly important to
the U.S. economy, from a regional perspective the U.S.
economy is still the primary factor influencing eco
nomic growth. A healthy U. S. economy is first and fore
most in its influence on regional output; our results
indicate that this is especially true for the Midwest.

TECHNICAL APPENDIX 1

Regional export-weighted dollar
Calculation ofthe regional export-weighted dollar
takes the following form:

xRi,

PPIit 1

7=1 PP.o

111 US,t _

44 r

100

RGTWDk„

Wg'kjj

where
RGTWD = regional export-weighted dollar,

k = U.S. region with n states,
j= country (1 to 44),

z = U.S. manufacturing industry category
(SIC 20-39),
Z = time period; observations are monthly
January 1970 through December 1997.

(The indexes are available through July
1998, reflecting the widespread appreciation
ofthe U.S. dollar that began in 1996;
however, the focus of this article ends
with 1997.),
XR = exchange rate of country j with respect
to the U.S. dollar (foreign currency/U.S.
dollar),

PPI= Producer (wholesale) Price Index for
country j ortheU.S., 1990= 100, and
Wgt = share of U.S. exports of industry Z,
from region k, to country j.
(Weights are an average of 1993
and 1994 U.S. good exports.)
Note: The indexes are constructed with the base year
1990 = 100. For expository purposes the indexes are
rescaled to 1970 = 100.

TECHNICAL APPENDIX 2

Region-specific foreign GDP growth
Calculation ofthe region-specific foreign real GDP
growth rates takes the following form:
20

GDPfrskJ = X (/ V/20i.) x (GDPfJ t),
7=1

where
GDPfrs= export-weighted average of annual GDP
growth rate (In) of region Cs 20 major
foreign export markets for time Z (regionspecific GDP),

GDPf = annual real GDP growth rates (In) for
countryj for time Z,

Ano k = sum value of exports of manufactured
goods (annual) from region k to country
j; 20 major foreign markets (average
for 1993-94),

X= value of region k's manufacturing goods
exports (annual) to country j (average
for 1993-94),

k = U.S. regions one to eight, plus U.S. total,
and
j = country one to 20 major export markets
for region k.
Note: Period covered is 1970 through 1997. China is not in
cluded in the 20 major foreign markets (GDP growth rates
are not available prior to 1978).

TECHNICAL APPENDIX 3

Impact of exchange rates and foreign
demand growth, OLS model
The central question of this study is whether the in
teraction between U.S. regional economies and their
respective international markets shows differences
across regions with regard to the exchange rates they
face and economic growth in their foreign markets.
There appear to be differences across regions in both
the exchange rate aggregates and the foreign market
growth aggregates.

Do these region-specific measures of exchange
rates and foreign economic growth have a measur
able impact on the regions’ economic activity? We
address this question using an ordinary least squares
model that identifies three “shock” variables’ (regionspecific foreign GDP, region-specific exchange rates,
and oil prices) impact on the gross regional product
(GRP) ofthe eight U.S. Bureau of Economic Analysis
(BEA) regions, plus the U.S.

Economic Perspectives

We base GRP for a region on data from the BEA
“Gross state product by industry” series for 1977-96.
These data are available in nominal and real values for
total, all manufacturing industries, durable manufac
turing industries, and nondurable manufacturing indus
tries. We extended the nominal GRP series to 1970-76
and 1997 using BEA earnings data in the appropriate
industry class. We then deflated the estimated nomi
nal GRPs using one of several standard price indexes,
based on the strength of the correlation between the
standard indexes and the implicit deflator between nomi
nal and real GRP by region and industry for 1977-96.
Oil price (OIL) is defined as average refiners’ acqui
sition cost (domestic and foreign sources). Nominal
prices are deflated using CPI less energy. Real oil
prices enter the region and industry equations in the
following form: Where total G7?P for the region is the
dependent variable, OIL enters the equation as the yearto-year percent change (In) in its full price adjusted
value. Where G7?P for a region is defined by an indus
try category, for example, durable manufacturing, OIL
enters the equation as the year-to-year percent
change in the full price adjusted value multiplied by
the region’s GRP in durable manufacturing share of
U.S. durable manufacturing GRP. In this form, the
more important a region’s durable manufacturing is in
U.S. durable output, for example, the heavier the oil
price weight will be.
The regional export-weighted exchange rate is
defined in technical appendix 1. Region-specific for
eign GDP growth is defined in technical appendix 2.
In addition, to account for the influence of the domes
tic economy, we include one-period lags of U.S. GDP
and the GRP ofthe region in question.

The regression equation takes the following form
(all observations are annual)
G7^,,. = a+bSGRPk__USi l,^ + bOGRP^

+b3(GDPfrsk')+b4(OILkiliil')

+b^RGTWDkr i .__o } + b6(RGTWDkt_x . _x)
+b^RGTWDki_xJi_2\

where
GRP= gross regional product (real) as defined
above, percent change (In),
GDPfrs = region-specific foreign GDP (real) as
defined in technical appendix 2—enters
equation in the contemporaneous period,
percent change (In),
RGTWD = region-specific export-weighted dollar (real)
as defined in technical appendix 1—enters
equation contemporaneously and with
one-period and two-period lags, percent
change (In),
OIL = refiners’ acquisition price for oil (real)
defined above—enters equation with a
one-period lag,
k = regions one through nine (U.S. total and
eight BEA regions), and
/ = industry classification (all SICs, aggregat
ed all manufacturing SICs, aggregated
durable manufacturing SICs, and aggregat
ed nondurable manufacturing SICs).
Variables are in log changes, except as defined
above for oil.

NOTES
1This article grew out a research project conducted as part of
the Federal Reserve Bank of Chicago’s year-long study of the
Midwest economy. Summaries of the six Midwest Assessment
conferences and a project report are available on the Internet
at www.frbchi.org. Research papers are also available from the
Bank’s Public Information Department on request.
2Some would question the assertion that exchange rates and
foreign economic growth are externally determined variables
relative to the U.S. economy, given the interdependence of
the world’s economies. Certainly international interdepen
dence has increased during the past 20 to 30 years. However,
we would argue that the advent of floating exchange rates in
the early 1970s unlinked many foreign economies from the
U.S. economy and the dollar, in the sense that U.S. monetary
policy no longer determined world monetary policy.

3The broad-based appreciation of the U.S. dollar relative to
nearly all other currencies in 1997 through mid-1998 further
accentuates the apparent strength of the Midwest dollar,

Federal Reserve Bank of Chicago

relative to the earlier periods. It also dramatically affects
the exchange rates of those regions heavily influenced by the
Asian markets.

4These figures are based on U.S.D.C. National Income and
Product Account data, Survey of Current Business, tables 1.4
and 4.1 (selected issues). A more complete discussion of export
shares of output is in Hervey (1995).
Estimated from data in U.S. Department of Commerce,
Bureau of the Census, “Exports from manufacturing establish
ments,” Analytical Report Series, Annual Survey ofManufac
tures, selected issues 1983-91, and Origin ofExports from
Manufacturing Establishments, selected issues 1969-81.

6Clark, Sawyer, and Sprinkle (1997 and 1999) have found
“nontrivial differences” between a “Southern” export-weighted
dollar index and an index constructed for the rest of the U.S.
They have also found differences in similarly constructed
indexes of U.S. census regions versus a total U.S. index.

7The seven foreign markets defined in figure 1 account for 100
percent ofU.S. goods exports during 1993 and 1994. They
include the 44 countries used in this study, which accounted for
91.5 percent of U.S.goods exports and “all other” markets. The
groups are defined as: North America—Canada and Mexico;
Latin America—Argentina, Brazil, Chile, Colombia, Ecuador,
Peru, and Venezuela; Europe—Austria, Belgium, Denmark,
Finland, France, Germany, Greece, Ireland, Italy, Netherlands,
Norway, Portugal, Spain, Sweden, Switzerland, and the United
Kingdom; Japan; Southeast Asia—Australia, Hong Kong, India,
Indonesia, the Republic of Korea, Malaysia, New Zealand,
Pakistan, Philippines, Singapore, Taiwan, and Thailand; Africa—
Morocco, Nigeria, and South Africa; Other—Kuwait, Turkey,
Saudi Arabia, and all other.

8New England—Connecticut, Maine, Massachusetts, New
Hampshire, Rhode Island, and Vermont; Mideast—Delaware,
District of Columbia, Maryland, New Jersey, New York, and
Pennsylvania; Great Lakes—Illinois, Indiana, Michigan, Ohio,
and Wisconsin; Plains—Iowa, Kansas, Minnesota, Missouri,
Nebraska, North Dakota, and South Dakota; Southeast—
Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana,
Mississippi, North Carolina, South Carolina, Tennessee,
Virginia, and West Virginia; Southwest—Arizona, New Mexico,
Oklahoma, and Texas; Mountain—Colorado, Idaho, Montana,
Utah, and Wyoming; Far West—Alaska, California, Hawaii,
Nevada, Oregon, and Washington.
9U.S. Department of Commerce, Bureau of the Census, FT-900
Supplement, “Location of exporter” series.
10The “Location of exporter” series was first made available
(on a continuous basis) by the U.S. Department of Commerce
for 1993. We consider this data series to be superior to the
Department’s “Origin of exporter” series, available from
1987, which biases the valuation of exports by individual
states toward those states where the port of export is situated.

nThe state export data as reported by the Bureau of the Census
contain a substantial category of “unallocated” exports. The
Massachusetts Institute for Social and Economic Research,
“MISER state of exporter location data (series II),” adjusts
these data to account for the unallocated portion. The adjusted
data series are made available on a by-state-by-country of des
tination at the two-digit SIC classification. In 1994, these

adjustments accounted for nearly 7 percent of total manufac
tured exports.

12Durable goods: SIC-24, lumber and wood products; 25, furni
ture and fixtures; 32, stone, clay, and glass products; 33, primary
metal industries; 34, fabricated metal industries; 35, industrial
machinery and equipment; 36, electronic and other electric
equipment; 37, transportation equipment; 38, instruments and
related products; and 39, miscellaneous manufacturing industries.
Nondurable goods: SIC-20, food and kindred products; 21,
tobacco manufactures; 22, textile mill products; 23, apparel
and other textile products; 26, paper and allied products; 27,
printing and publishing; 28, chemicals and allied products; 29,
petroleum and coal; 30, rubber and miscellaneous plastics; and
31, leather and leather products.
13In sum, the third-country issue boils down to this: The dollar
may experience a real depreciation or appreciation relative to
a bilateral trading partner. That exchange rate change affects
the relative competitiveness not only ofU.S. goods versus the
bilateral partner, but also ofU.S. goods versus third-country
trading partners. The aggregate exchange rate construction
we use here does not allow us to address this issue.
14A scheme utilized by Hayward and Erickson (1995), who in
a somewhat different context sought to measure the size of
import competing industries by state by SIC, appears poten
tially useful in getting to the import competitiveness issue.
This work is being extended to include an aggregate bilateral
index that uses a modification of the Hayward-Erickson mea
sure for imports by region.

15These price indexes were provided by the Economic Research
Group of J. P. Morgan through December 1997. Based on
availability of data, some countries’ price indexes are versions
of their consumer price index.
16The monthly regional index series for the three industry cat
egories for each of the regions (currently January 1970 through
July 1998) are available from the authors on request.

17Percentage changes in the indexes are reported on a logarith
mic basis.

REFERENCES

Board of Governors of the Federal Reserve System,

1998, “Trade-weighted indexes (INTL, JRXWGK N.M),”
Washington, DC, proprietary database, various
releases, July.
Branson, W. H., and J. P. Love, 1987, “The real ex

change rate and employment in U.S. manufacturing:
State and regional results,” National Bureau of Eco
nomic Research, working paper, No. 2435, November.
________ , 1986, “Dollar appreciation and manufac
turing employment and output,” National Bureau of
Economic Research, working paper, No. 1972, July.
Carlino, G.A., 1990, “Should states fear the effects of a

Clark, D. P., W. C. Sawyer, and R. L. Sprinkle, 1999,
“Regional exchange rate indexes for the United
States,” Journal of Regional Science, forthcoming.

_________ , 1997, “The value of the ‘southern’
dollar,” Review ofRegional Studies, Vol. 27, No. 2,
pp. 185-193.
Coughlin, C. C., and P. S. Pollard, 1998, “Constructing

and using national and regional TWEXS: The case
for chaining,” paper presented at the Workshop on
Regional Economic Indicator Models held in June
1998, sponsored by the Federal Reserve Bank of St.
Louis in Braga, Portugal, revised in September.

changing dollar?,” Business Review, Federal Reserve
Bank of Philadelphia, September/October, pp. 3-12.

Economic Perspectives

Cronovich, R., and R. Gazel, 1998, “Do exchange
rates and foreign incomes matter for exports at the
state level?,” Journal of Regional Science, Vol. 38,
No. 4, pp. 639-657.

Hickok, S., and J. Orr, 1989-90, “Shifting patterns of

Davis, S. J., P. Loungani, and R. Mahidhara, 1997,

International Monetary Fund, 1970-97, International
Financial Statistics, Washington, DC, various issues.

“Regional labor fluctuations: Oil shocks, military
spending, and other driving forces,” Board of Gover
nors of the Federal Reserve System, International
Finance, discussion papers, No. 578, March.
Hargreaves, D., 1994, “Currency indices for emerging

markets,” Economic Research Note, Morgan Guaranty
Trust Company, Economic Research Group, January 7.

________ , 1993, “Effective exchange rates: OECD
currencies,” Economic Research Note, Morgan Guar
anty Trust Company, Economic Research Group,
December 30.
Hayward, D. J., and R. A. Erickson, 1995, “The North

American trade of U.S. states: A comparative analysis
ofindustrial shipments, 1983-91,” International
Regional Science Review, Vol. 18, No. l,pp. 1-31.
Hervey, J. L., 1990, “Changing U.S. trade patterns,”

Economic Perspectives, Federal Reserve Bank of
Chicago, March/April, Vol. 14, No. 2, pp. 2-12.
Hervey, J. L., and W. A. Strauss, 1998, “Aregional

export-weighted dollar: An examination of the region
al impact of exchange rate change,” International
Regional Science Review, Vol. 21, No. 1, pp. 1-20.
________ , 1996, “Aregional export-weighted dollar:
A different way of looking at exchange rate changes,”
paper presented at the Federal Reserve Bank of
Chicago conference, Assessing the Midwest Econo
my, Global Linkages to the Midwest Economy, work
ing paper, No. GL-2, September 18.
_________ , 1987, “The international value of the dol
lar: An inflation-adjusted index.” Economic Perspec
tives. Federal Reserve Bank of Chicago, January/
February, Vol. 11, No. l,pp. 17-28.
Hewings, G. J. D., G Schindler, and P. Israilevich,

1998, “Interstate trade among Midwest economies,”
Chicago EedLetter, Federal Reserve Bank of
www.jpmorgan.Chicago, May.
Hickok, S., 1991, “The shifting composition of U.S.

manufactured goods trade,” Quarterly Review,
Federal Reserve Bank of New York, Spring, Vol. 14,
No. 4, pp. 36^17.

Federal Reserve Bank of Chicago

U.S. trade with selected developing Asian economies,”
Quarterly Review, Federal Reserve Bank of New York,
Winter, Vol. 14, No. 4, pp. 36—47.

Israilevich, P., G. J. D. Hewings, M. Sonis, and G.
Schindler, 1997, “Forecasting structural changes with

a regional econometric input-output model,” Journal
ofRegional Sciences, Vol. 37, No. 4, pp. 565-590.
Koretz, G, 1988, “Will U.S. factories need another

exchange rate fix?,” Business Week, August 8, p. 18.
Little, J. S., 1989, “The dollar, structural change and
the New England miracle,” New England Economic
Review, Federal Reserve Bank of Boston, September/
October, pp. 47-57.
Massachusetts Institute for Social and Economic re
search (MISER), 1993-94, State ofExporter Location,

series II.
Morgan Guaranty Trust Company, Economic Re
search Group, 1998a, “Price indices for the 45 coun

tries used in the J. P. Morgan real effective exchange
rate indices,” July, unpublished data.

________ , 1998b, “OECD and emerging markets cur
rencies,” available on the Internet at www.jpmorgan.
com/MarketDatalnd/Forex/currlndex.html, January.
Prowse, M., 1995, “Midwest toasts dollar’s decline,”
Financial Times, March 23, p. 6.
United Nations, 1985/6-95, Statistical Yearbook,
New York, various issues.

U.S. Department of Commerce, Bureau of the Cen
sus, 1993-94, U.S. Merchandise Trade, FT-900 Sup

plement, selected issues.
________ , 1994-95, U.S. Merchandise Trade, FT-900,
selected issues.
________ , 1983-91, “Exports from manufacturing
establishments,” Analytical Report Series, Annual
Survey of Manufactures, selected issues.
________, 1969-81, Origin ofExportsfrom Manufac
turing Establishments, selected issues.
U.S. Department of Commerce, Bureau of Economic
Analysis, 1998, “Gross state product by industry,

1977-96,” available on the Internet at www.bea.doc.
gov/gsp/gsplist.htm.

The business cycle:
It’s still a puzzle

Lawrence J. Christiano and Terry J. Fitzgerald

Introduction and summary
Good fiscal and monetary policy requires a clear un
derstanding of the workings of the economy, espe
cially what drives the business cycle—the periodic
ups and downs in economic activity. Since at least
the late 1800s, a full swing from the start of an eco
nomic expansion to a recession and back to the start
of another expansion has generally taken between
two and eight years. Every citizen is keenly aware of
the state of the economy, whether it is in prosperity
or recession.
Everyone is so conscious of the business cycle
because most sectors of the economy move up and
down together.1 This phenomenon, referred to as
comovement, is a central part of the official definition
of the business cycle. The definition is set by the Na
tional Bureau of Economic Research (NBER), which
decides when recessions begin and end. Under the
NBER’s definition,
“... a recession is a [persistent] period of decline
in total output, income, employment, and trade, usu
ally lasting from six months to a year, and marked by
widespread contractions in many sectors of the
economy.”2
Even though comovement is a defining character
istic of the business cycle, in recent decades macro
economists have tended to focus on understanding
the persistence in the ups and downs of aggregate
economic activity. They have generally been less con
cerned with understanding the synchronized nature
of this pattern across sectors. In part, the omission
reflects the conceptual difficulties inherent in think
ing about an economy with many sectors.3 Standard
models of business cycles assume there is only one
good being produced and so they consider only one
economic sector. These models do not encourage
thinking about the comovement of economic activity
across many sectors. Since these models were first

introduced, in the late 1970s and early 1980s, the state
of macroeconomics has advanced rapidly. The con
ceptual and computational barriers to thinking about
multiple sectors are quickly falling away. As a result,
recent years have witnessed a renewed interest in
understanding comovement.
We have two objectives in this article. The first
is to document business cycle comovement. We
examine data on hours worked in a cross section of
economic sectors. We examine the business cycle
components of these data and show that the degree
of comovement is substantial. Our second objective
is to analyze explanations for this comovement. We
find that none is completely satisfactory. Still, this
is a growing area of research, and we are seeing
some progress.

Identifying comovement
To study comovement across sectors over the
business cycle, we need the following two things: a
measure of the level of economic activity in the vari
ous sectors of the economy; and a precise definition
of what we mean by the business cycle component of
the data. Below, we address these two issues. After
that, we present our results, characterizing the degree
of comovement in the data.

Lawrence J. Christiano is a professor of economics at
Northwestern University, a consultant to the Federal
Reserve Bank of Chicago, and a research associate at
the National Bureau of Economic Research. Terry J.
Fitzgerald is an economist at the Federal Reserve Bank
of Cleveland. The title of this article is modifiedfrom
Kocherlakota’s (1996) article. The findings for business
cycle analysis are similar to Kocherlakota’s for the
equity premium puzzle. The authors are gratefulfor
discussions with Michelle Alexopoulos, Stefania Albanesi,
Paul Gomme, Henry Siu, and Robert Vigfusson.

Economic Perspectives

The data
We measure economic activity in a given sector
by the number of hours worked in that sector. Table 1
lists the sectors we consider. The hours worked mea
sure that covers the most sectors is total private hours
worked.4 This measure covers all sectors of the econ
omy, except government and agriculture. It is broken
into hours worked in goods-producing industries and
in service-producing industries. Goods-producing in
dustries are further broken into mining, manufactur
ing, and construction. Similarly, service-producing

industries are broken into five subsectors. The sub
sectors of manufacturing, durable goods and nondu
rable goods, are broken into yet smaller sectors. The
data in the third column give an indication of the rela
tive magnitude of each subsector. In particular, any
given row reports the average number of people
employed in that sector, divided by the average num
ber of people employed in the sectoral aggregate to
which that sector belongs. Thus, for example, 58 per
cent of manufacturing employment is in the durable
goods sector and 42 percent is in the nondurable

TABLE 1

Properties of the business cycle components of hours worked
Variable
number

Hours worked variable

1
2
3
4
5
6
7
8
9
10
1 1
12
1 3
14
1 5
16
1 7
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33

Total private
Goods-producing industries
Mining
Construction
Manufacturing
Durable goods
Lumber and wood products
Furniture and fixtures
Stone, clay, and glass products
Primary metal industries
Fabricated metal products
Machinery, except electrical
Electrical and electronic equipment
Transportation equipment
Instruments and related products
Miscellaneous manufacturing
Nondurable goods
Food and kindred products
Tobacco manufactures
Textile mill products
Apparel and other textile products
Paper and allied products
Printing and publishing
Chemicals and allied products
Petroleum and coal products
Rubber and misc. plastics products
Leather and leather products
Service-producing industries
Transportation and public utilities
Wholesale trade
Retail trade
Finance, insurance, and real estate
Services

Relative
magnitude

Relative
volatility

1 .00
.33
.03
.1 7
.80
.58
.06
.04
.05
.09
.13
.19
.1 5
.1 7
.08
.04
.42
.21
.01
.1 1
.1 5
.09
.16
.13
.02
.09
.03
.67
.10
.10
.31
.10
.38

1 .00
3.91
5.46
6.75
3.92
6.90
10.18
8.14
4.98
9.89
7.21
11.10
8.75
7.83
5.03
3.23
1 .39
.16
1 .83
3.92
2.64
1 .97
.91
1 .01
2.02
7.82
2.71
.25
.87
.65
.36
.35
.19

Business cycle
comovement

.00
.99
.38
.88
.97
.97
.89
.94
.95
.86
.96
.93
.88
.89
.76
.90
.91
.50
.08
.76
.85
.85
.90
.80
.16
.89
.64
.93
.95
.87
.87
.48
.49

Notes: The column labeled "Relative magnitude" reports an indication of the relative magnitude of each sector. Any given row reports the average number of people
employed in that sectordivided by the average number of people employed in the sectoral aggregate to which that sector belongs, for example, 58 percent of
manufacturing employment is in the durable goods sector and 42 percent is in the nondurable goods sector. The column labeled "Relative volatility" reports the
variance of the business cycle component of the logarithm of hours worked in the indicated row variable divided by the varianceofthe business cycle component of
the logarithm oftotal private hours worked. The column labeled" Business cyclecomovement"iscalculated using the process described innote6ofthearticle.
Source: Authors'calculations from data of DRI Basic Economics database, 1964-96.

Federal Reserve Bank of Chicago

goods sector. Also, the largest goods-producing indus
try, by far, is manufacturing, which has 80 percent of
all goods-producing employees.
Next, we try to characterize how much business
cycle comovement there is across the economic sec
tors we consider. That is, if we limit ourselves to the
business cycle range of fluctuations in the data—
fluctuations that last between two and eight years—
to what extent do the data move up and down
together?5

Business cycle component ofthe data
A detailed discussion of our notion of the business
cycle component of the data is in technical appendix
1. Figure 1 illustrates the basic idea behind what we
do. The choppy line in panel A of figure 1 displays
total private hours worked. The reported data are the
logarithm of the raw data. The advantage of using the
logarithm of the data in this way is that the resulting
movements correspond to percent changes in the

underlying raw data. The deviations between the
actual data and the trend line in panel A of figure 1
are graphed in panel B. Those deviations contain the
rapidly varying, erratic component, inherited from the
choppy portion of the data that is evident in panel A.
The smooth curve in panel B is our measure of the
business cycle component of the total private hours
worked data. Specifically, that measure excludes both
the trend part of the data and the rapidly varying,
erratic component. It includes only the component of
the data that contains fluctuations in the range of two
to eight years. According to our approach, the econo
my is in recession when our business cycle measure is
negative and in prosperity when it is positive.
Figure 1 also compares our measure of the busi
ness cycle with the one produced by the NBER. The
start of each shaded area indicates the date when,
according to the NBER, the economy reached a busi
ness cycle peak. The end of each shaded area indicates
a business cycle trough. Note how total
private hours worked fall from peak to
trough and then generally grow from
trough to peak. An obvious difference in
the two business cycle measures is that
ours is a continuous variable, while the
NBER’s takes the form of peak and trough
dates. As a result, our measure not only
indicates when a recession occurs, but
also the intensity of the recession. Apart
from these differences, however, the two
measures appear reasonably consistent.
For example, note that near the trough
of every NBER recession, our measure
of the business cycle is always negative.
But the two measures do not always
agree. According to our measure, the
economy was in recession in 1967 and in
1987, while the NBER did not declare a
recession during those periods. In part,
this is because there must be several
months’ negative employment growth
before the NBER declares a recession.
However, our procedure only requires a
temporary slowdown.
Figure 1 provides informal evidence
in support of the facts we wish to docu
ment. As noted in the introduction, the
NBER must see a broad-based decline
before declaring a recession. Thus, the
NBER dates in figure 1 indicate periods
when many economic sectors showed
weakness. Since these dates roughly co
incide with periods of weakness in total

Economic Perspectives

private hours worked, this is consistent with the view
that most sectors move up and down together, at
least in the two- to eight-year frequency range. We
stress, however, that the NBER’s dating procedures
are informal. Our objective in this section is to provide
a formal, quantitative assessment of the degree of
comovement among economic sectors.
We computed a business cycle component for
each of the 33 series listed in table 1. As we anticipated,
we find that the business cycle components in most
of the series move together closely. This is true, de
spite a striking lack of uniformity in other respects.
For example, note how different the trends in figure 2
are. The first two columns report data for the goodsproducing industries and its major components. The
second two columns report the analogous data for
the service-producing industries. Generally, trend em
ployment is down in the goods-producing industries
and up in the service-producing industries. The lev
els of volatility in the business cycle components of
the various series are also very different. The fourth
column of table 1 reports the variance of the business
cycle component of a variable, divided by the variance
of aggregate hours worked. The relative variance of
hours worked in goods-producing industries is typi
cally quite high, substantially above 2, and it is quite
low for service-producing industries. That goodsproducing industries are volatile relative to the serviceproducing industries is well known.

Measuring business cycle comovement
Despite the very substantial differences in the
trends of the data series shown in figure 2, their move
ments over the business cycle are quite similar. Figure 3
illustrates the business cycle components of the same
variables used in figure 2. In each case, we computed
the business cycle component using exactly the same
method underlying the calculations in panel B of fig
ure 1. Each graph contains the business cycle compo
nent of the variable indicated and the business cycle
component for total private hours. This was taken
directly from panel B of figure 1.
In most of the series in figure 3, the data move
up and down closely with the business cycle compo
nent of total hours worked. There are some exceptions.
For example, the business cycle movements in mining
bear little resemblance to the business cycle move
ments in total hours worked. At the same time, mining
represents a very small part of the private economy
and employs only 3 percent of workers in the goodsproducing industry. Another exception is the finance,
insurance, and real estate (FIRE) industry, whose
business cycle component exhibits reasonably high

Federal Reserve Bank of Chicago

comovement with aggregate employment until the
1980s, after which this relationship breaks down.
To measure the degree of business cycle comove
ment between a given series and total hours worked,
we use a statistic that is like the square of the correla
tion between the business cycle components in the
two variables. Our statistic measures the fraction of
the variance in the series that can be accounted for
by the total hours worked data.6 If this number is, say,
98 percent, this means that 98 percent of the business
cycle variance in the variable can be accounted for by
the business cycle in aggregate hours worked. These
results are reported in the fifth column of table 1. As
expected, the results indicate that this measure of
comovement is relatively low, in the sense of being
below 0.50, for the mining, FIRE, and services sec
tors. Overall, however, the degree of comovement by
this measure is high.
Going one step further in the level of disaggrega
tion, we can get an idea about the comovement in the
components of durable and nondurable manufactur
ing. Figure 4, panel A displays the business cycle
movements in the components of durable manufac
turing sectors. Panel B does the same for nondurable
manufacturing. In each case, the data series graphed
at the top of the figure is the business cycle compo
nent of total hours worked. The series are presented
so as to allow one to focus exclusively on the degree
of comovement between them. Thus, we added a con
stant to each series to spread them out across the
figure and divided each series by its sample standard
deviation, so that the standard deviation of the
reported data is unity in each case.7 The number to
the right of each line in the figure identifies the data
series. Figure 4 also displays the NBER peak and
trough dates as a convenient benchmark.
Figure 4, panel A shows that the comovement
among sectors in durable manufacturing is very high.
With only one minor exception, the variables move
closely with each other and with aggregate employ
ment. The exception is that instruments and related
products, series 15, does not move closely with the
other variables during 1987, when the other business
cycle components are signaling a recession. However,
overall the degree of comovement is strikingly high.
Figure 4, panel B shows that the business cycle co
movement in the nondurable goods manufacturing
industries is lower than in the durable goods sector.
Two variables that do not comove closely with the
others at business cycle frequencies are tobacco man
ufactures, series 19, and petroleum and coal products,
series 25. Both these variables are rising in the first
and last NBER recession periods in our data. The

FIGURE 2

Hours worked in various sectors: Data and trends
A. Goods-producing industries

D. Mining

G. Service-producing industries

J. Retail trade

logarithm

B. Manufacturing, durable goods

E. Construction

H. Transportation and utilities

K. Finance, insurance, and real estate

logarithm

C. Manufacturing, nondurable goods

F. Wholesale trade

I. Services

logarithm

E co n o m ic P ers p ecti ves

Source: Authors' calculationsfromdataofDRI Basic Economics database, 1964-96.

F ed era l R eserv e B an k o f C h ic ag o

FIGURE 3

Business cycle component comparison: Total hours worked versus hours worked in various sectors
A. Goods-producing industries

B. Manufacturing, durable goods

D. Mining

G. Service-producing industries

J. Retail trade

logarithm

E. Construction

H. Transportation and utilities

K. Finance, insurance, and real estate

logarithm

C. Manufacturing, nondurable goods

logarithm

F. Wholesale trade

I. Services

logarithm

-0.06 Ll 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1967
’72
’77
’82
’87
’92
Note: The information displayed in the "total hours worked" line is "business cycle component," taken from figure 1, panel B.

Source: Authors' calculations from data of DRI Basic Economics database, 1964-96.

comovement statistic for these variables reported in
table 1 is very low, 0.08 for tobacco manufactures
and 0.16 for petroleum and coal products. The other
variables in nondurable manufacturing display stron
ger comovement, with comovement statistics of 0.50
or higher.
Up to now, the statistics we have used to charac
terize comovement emphasize association with aggre
gate hours worked. This nicely complements the
visual evidence in the graphs. However, there is a
pitfall to relying exclusively on statistics like this to
characterize comovement. A simple example illustrates
the point. Suppose there is a variable, yt, which is the
sum of two other variables, y]t and yy.
y,=y1,+y2,Suppose further thatv(; andv,; are uncorrelated. No
one would say there is comovement between these
variables. Still, each variable is strongly correlated
with the aggregate. To see this, take the simple case
where the variance ofy;, and v,, is c?2. Then, the corre
lation between j/., and y,is0. 71, forz = 1,2, despite the

absence of comovement between the variables.8 This
example exaggerates the point somewhat, since results
are less severe when there are more than two sectors.9
Still, this pitfall is of some concern.
With this in mind, we consider the correlation
between the business cycle components of all the
variables. A difficulty with this is that there are many
such correlations. For example, with three variables,
there are three possible correlations, with four there
are six, with five there are ten, and with n there are
n(n- l)/2. So, with» = 33, there are 528 possible cor
relations. It is a challenge to organize and present
this many correlations in a coherent way. We present
the mean and histogram of the correlations for different
subsectors in figure 5. The histograms display, on the
vertical axis, the fraction of correlations lying within a
given interval, whose midpoint is indicated on the
horizontal axis.10
Figure 5, panel A displays the correlations for the
finest levels of aggregation for which we have data.
This means hours worked in mining, construction, the
20 components of manufacturing, and the five compo
nents of the service-producing industries. Thus, we

FIGURE 4

Business cycle component of total hours worked and hours worked
in various manufacturing subsectors
A. Two-digit manufacturing durable
and total hours worked

B. Two-digit manufacturing nondurable
and total hours worked

Notes: The number on the right indicates the variable sector from table 1 that is being highlighted. Panel A shows:
1 —total private hours; 7—lumber and wood products; 8—furniture and fixtures; 9—stone, clay, and glass products;
10—primary metal industries; 11—fabricated metal products; 12—machinery, except electrical; 13—electrical and electronic
equipment; 14—transportation equipment; 15—instruments and related products; and 16—miscellaneous manufacturing.
Panel B shows: 1—total private hours; 18—food and kindred products; 19—tobacco manufactures; 20—textile mill products;
21 —apparel andothertextile products; 22—paper and allied products; 23—printing and publishing; 24—chemicals and allied
products; 25—petroleum and coal products; 26—rubber and miscellaneous plastics products; and 27—leather and leather products.
Each variable has been scaled by its sample standard deviation and a constant has been added in ordertospread outthedata
inthe panels. Shaded areas indicate recessions as determined bythe National Bureau of Economic Research.

Source: Authors' calculations from data of DRI Basic Economics database, 1964-96.

Economic Perspectives

have 27 data series, with 351 correlations between them.
Figure 5, panel A indicates that the mean of these corre
lations is 0.55. When we eliminate the three data series
that we already know do not display strong business
cycle comovement, the mean rises to 0.68. The histo
gram shows that there is a substantial fraction of high
correlations in these data. We infer that the data are
consistent with the impression from the preceding
statistics that there is considerable evidence of comove
ment. Figure 5, panel B presents the results for the
manufacturing durable sector. Consistent with our
previous findings, the degree of comovement in this
sector is very high, with a mean correlation of 0.82.
Figure 5, panels C and D show the results for the
nondurable manufacturing sector and the serviceproducing sectors, respectively. Again, the results are
consistent with the notion that there is less comove
ment in these sectors than in manufacturing durables.
Still, the degree of comovement is substantial, with
mean correlations in excess of 0.6 if we consider all sec
tors except tobacco and petroleum and coal products.

Federal Reserve Bank of Chicago

In conclusion, we find that there is substantial
business cycle comovement in the data. Only two rel
atively small sectors—tobacco manufactures and pe
troleum and coal products—exhibit little tendency to
move up and down with general business conditions
over the business cycle.

Explaining business cycle comovement
What is it that at times pulls most sectors of the
economy up, and at other times pushes them down
again? This is one of the central questions in busi
ness cycle analysis. Although economists have de
veloped a number of possible explanations, the
phenomenon remains a puzzle.
In a classic article devoted to this puzzle, Robert
E. Lucas, Jr., conjectures that the resolution must lie
in some sort of shock that hits all sectors of the econ
omy, a so-called aggregate shock (Lucas, 1981).
Many economists today would probably agree with
this conjecture. That is why, in practice, the search for
the ultimate cause of business cycles often focuses

on identifying an aggregate shock. However, re
search conducted since Lucas published his article
suggests that identifying the cause of business cycles
may not be so simple.
First, even if we do manage to identify a shock
that clearly affects the whole economy, it does not
necessarily follow that shock is responsible for the
business cycle. A shock might well be experienced by
all sectors of the economy, but they need not all re
spond in the same way. The business cycle shock, if
indeed there is only one, seems to lead to a synchro
nized response across sectors. Second, we now know
that the search for a single aggregate shock may itself
be off base. Following the work of Long and Plosser
(1983), we know that, at least in theory, disturbances
to individual industries, even if they are uncorrelated
across industries, could result in comovement.
Currently, there is no consensus among econo
mists as to what causes business cycles and, in partic
ular, their key feature, comovement. At the same time,
researchers are exploring a large range of possibilities.
Next, we provide a selective overview of this research.
A natural starting point is what is perhaps the
most thoroughly developed theory of business cycles,
the real business cycle theory associated with Kydland
and Prescott (1982), Long and Plosser (1983), and
Prescott (1986).11 We focus specifically on the stan
dard real business cycle model, developed in Hansen
(1985) and analyzed in Prescott (1986). Although that
model posits an aggregate shock, it is inconsistent
with business cycle comovement. We then explore
two sets of modifications to this model. The first can
be viewed as natural extensions of the model. The
second depart more significantly from the model’s
assumptions.

Standard real business cycle theory12
A key component of real business cycle theory
is a production technology. This is a relationship that
specifies how much output a firm can obtain from a
given amount of capital and labor resources. This
technology is subject to shocks. Sometimes a good
shock occurs and more output can be produced for a
given level of inputs. In this case, we say the technol
ogy has been shifted up. A good technology shock
might reflect the implementation of a more efficient
way to organize the work force, the acquisition of
more efficient manufacturing equipment, or perhaps
the discovery of a way to alter the firm’s product so
that it better meets customers’ needs. At other times,
a bad technology shock can shift a production tech
nology down. Abad shock might reflect bad weather,
a labor dispute, an accident in the workplace, a machine

breakdown, or a government policy that encourages
an inefficient way of organizing production. Accord
ing to real business cycle theory, business cycle
expansions reflect that shocks affecting firms are
mostly on the positive side, while recessions reflect
periods when most firms’ shocks are on the negative
side. Standard formulations abstract from the differ
ences between firms and simply assume they all have
the same production technology and are affected by
the same shock. Thus, real business cycle theory
proposes that the aggregate shock to which Lucas
refers is a productivity shock.13
The standard real business cycle model not only
assumes that all firms are affected by the same produc
tivity shock, but also that there is just one type of
good produced (and, therefore, one industry sector)
in the economy. At least at first glance, this model
does not seem useful for examining business cycle
comovement among many sectors. However, it has
recently been pointed out that this impression is mis
leading.14 In fact, one can use the model to examine
business cycle comovement. When we do so, we find
that its implications are strongly counterfactual. The
standard real business cycle model is at variance with
the observation of business cycle comovement, despite
the fact that it views the economy as being driven by
a single aggregate shock. Understanding why it is
incompatible with comovement is useful for gaining
insight into the various lines of inquiry researchers
have pursued.
The standard real business cycle model imagines
that households interact with firms in competitive mar
kets, in which they supply labor and physical capital
and demand goods for consumption and to add to
their stock of capital. Although there is only one type
of production technology in this model, we can rein
terpret the model to suggest that one type of firm pro
duces goods for consumption (the consumption goods
industry) and another type produces new investment
goods for maintaining or adding to the stock of capi
tal (the investment goods industry).
When a positive productivity shock hits, so that
the real business cycle model shifts into a boom, the
output of both consumption and investment goods
industries increases. However, there is a relatively
larger increase in the output of investment goods.
This reflects a combination of two features of the mod
el. First, a positive technology shock increases the ex
pected return to investment, raising the opportunity
cost of applying resources to the consumption sector
Second, the model assumes that households prefer
not to increase consumption substantially during
booms but to smooth consumption increases over a

Econom ic Perspectives

longer time horizon. The increase in the demand for
investment goods relative to consumption goods that
occurs in a boom implies, in the standard model, that
capital and labor resources are shifted out of the pro
duction of consumption goods and into the produc
tion of investment goods. The model does predict a
small rise in consumption in a boom. However, this
rise is driven by the favorable technology shock, which
is not fully offset by the flow of productive resources
out of that sector. Thus, the model implies that hours
worked in the consumption sector are countercyclical,
in contrast with our empirical findings in the previous
section. This is a feature of the model, despite its im
plication that total hours worked rise in a boom. That
is, the additional hours of work all flow into the invest
ment good sector. The standard real business cycle
model also implies that investment in capital for use
in the consumption sector is countercyclical. This too,
is counterfactual, according to the results reported in
Huffman and Wynne (1998).
So, this model is strongly at variance with comove
ment. Why is this so? The result may seem especially
surprising to those who expect an aggregate shock to
all sectors of the economy to produce comovement.
Intuitively, there are two related ways to under
stand the model’s implication that inputs are allocated
away from the sector that produces consumption
goods during a boom. One is that the model overstates
the value of leisure at that time. This inflates the cost
of allocating labor resources to the consumption sec
tor then. The other is that the model understates the
value of the output of the sector producing consump
tion goods in a boom. This undercuts the incentive
to allocate resources to that sector then.15

Natural extensions of the standard theory
Among the various extensions to the model that
economists have pursued,16 we focus on approaches
that stress 1) factors that prevent the rise in the mar
ginal utility of leisure in a boom and 2) factors that
prevent the decline in the value of the output of the
consumption sector in a boom. As in the discussion
above, the work we survey here assumes two market
sectors.17

Value ofleisure
One factor that can slow the decline of the mar
ginal utility of leisure when the economy moves into
a boom was explored in Benhabib, Rogerson, and
Wright (1991) and Einarsson and Marquis (1997). Each
of these papers points out that if there is a third use
of time, in addition to leisure and time spent working
in the market, and if that use of time declines during a

Federal Reserve Bank of Chicago

boom, the marginal utility of leisure need not increase
as market effort increases.18
Benhabib, Rogerson, and Wright (1991) suggest
that the third use of time may be working in the home.
For example, the amount of leisure time enjoyed by a
homemaker may not change significantly if the home
making job is exchanged for a market job. Consider
ations like this lead Benhabib, Rogerson, and Wright
to argue that work time can be reallocated from the
home to the market during a boom without substan
tially raising the marginal utility of leisure .19
Einarsson and Marquis (1997) suggest that the
third use of time may be time spent accumulating
human capital, such as going to school. In principle,
this is an appealing idea, since it is known that time
spent in educational pursuits goes down in business
cycle expansions. Their work is primarily theoretical,
however. A crucial issue one would have to address
in pursuing this explanation at the empirical level is
whether the time spent on education is sufficiently
countercyclical, in a quantitative sense, to have a sub
stantial effect in a suitably modified real business
cycle model. In assessing this, one would have to
confront a substantial measurement problem. In par
ticular, time spent in educational institutions is only
part of the time spent in education. Some of that time
is applied in the workplace, by diverting workers from
direct production. Our understanding is that there do
not exist reliable measures of this use of time.

Value ofthe output ofthe consumption sector
Several papers attempt to get at comovement by
reducing the decline in the value of output in the con
sumption sector during booms. For example, Baxter
(1996) adapts the standard real business cycle model
by assuming that the consumption of market goods
and the services of home durables are good substi
tutes. An example of two goods that substitute is
a movie watched in a theater (a market good) and a
movie watched on a home television set (a home
durable good).20
Under Baxter’s substitutability assumption, the
appropriate measure of household consumption is not
just market consumption, but consumption of market
goods plus the service flow on the stock of home
durables. If home durables consumption is sufficiently
large, then a given jump in the consumption of market
goods leads to a smaller percent drop in the marginal
utility of consumption. In the extreme case where the
stock of home durables is extremely large and accounts
for essentially all of consumption, then a rise in mar
ket consumption would produce essentially no drop
in the marginal utility of consumption.21 Although

Baxter shows that this mechanism does indeed pro
duce comovement in employment across consump
tion and investment sectors in her model, there is a
sense in which the comovement is not strong
enough. That is because investment in the capital
used in the two sectors is essentially uncorrelated.
As noted above, the data suggest that investment
across sectors comoves as well, in addition to output
and employment.
One can also view the home production ap
proach ofBenhabib, Rogerson, and Wright (1991) as
a strategy to generate comovement by reducing the
decline in the value of output in the consumption
sector during booms. In a boom, as labor is allocated
away from home-produced goods toward the produc
tion of market goods, the marginal utility of the mar
ket good does not fall much because the market and
home goods are assumed to be highly substitutable.22
This allows the value of output in the consumption
sector to rise sufficiently so that employment in that
sector is procyclical. A shortcoming of the analysis,
emphasized by Benhabib, Rogerson, and Wright
(1991), is that the high substitutability between home
and market goods needed for comovement of labor
inputs hurts on another dimension. It has the effect
that purchases of durables are countercyclical over
the cycle.23
Christiano and Fisher (1998) take another ap
proach. Following Boldrin, Christiano, and Fisher
(1995), they modify a standard real business cycle
model in two ways. First, they specify that it takes
time before labor can shift between economic sectors
in response to a shock. The reasons for this are not
modeled explicitly, but the assumption is motivated
with an informal reference to such factors as the
search and training costs which inhibit real world la
bor mobility between industry sectors. This assump
tion alone is not sufficient to guarantee comovement,
however. Without fiirther changes, their model pre
dicts that resources would be reallocated out of the
consumption sector and into the investment sector
as soon as labor becomes fully mobile, which they
specify to occur in three months’ time. As a result,
this version of the model is still inconsistent with the
evidence on business cycle comovement. Christiano
and Fisher therefore introduce a second modification,
by changing the specification of household prefer
ences over consumption. They assume that house
holds have a tendency to become accustomed to the
level of consumption they have enjoyed in the recent
past. This property of preferences is known as habit
persistence. A household with habit persistence pref
erences whose consumption has recently increased

is particularly unhappy if later it must return to its
previous level of consumption.24 Habit persistence
preferences have the implication that when consump
tion rises in a boom, the marginal value of continuing
to maintain consumption at a high level is increased.
Christiano and Fisher show that habit persistence
and limitations on the intersectoral mobility of labor
are sufficient to produce comovement in hours
worked and investment. To our knowledge, this is the
only quantitative model in the comovement literature
with this property.
The credibility of this result depends on the
credibility of the underlying assumptions. The as
sumption that there are limitations on the speed with
which productive resources can be transferred across
sectors seems uncontroversial, though the model cer
tainly takes an extreme stance. What does call for a
defense is the assumption of habit persistence prefer
ences. One defense is that these preferences help to
account for observations that otherwise seem puz
zling. For example, Boldrin, Christiano, and Fisher
(1995) show that, consistent with results in Constantinides (1990), habit persistence and limited intersec
toral mobility can account for the magnitude of the
observed average premium in the return on equity
over risk-free securities. The solution to this premium
has eluded many researchers.25 In addition, Christiano
and Fisher (1998) show that habit persistence can
help account for the so-called inverted leading indi
cator property of interest rates, that high interest
rates tend to forecast bad economic times. King and
Watson (1996) document that standard models have
difficulty accounting for this observation.26
A third approach toward understanding comove
ment was recently pursued by Hornstein and Praschnik
(1997). They observe that some of the output of the
sector that produces consumption goods (the non
durable goods sector) is also used as intermediate
goods in the production of investment goods. For
example, both households and investment-good pro
ducing firms make use of the services of the transpor
tation sector Hornstein and Praschnik (1997) modify
a real business cycle model to accommodate this fea
ture of the economy. The modification has the effect
of increasing the value of output in the consumption
sector in a boom. This increased value reflects the in
creased need for the output of the consumption good
sector during a boom for use in the investment good
sector.27 We refer to this demand channel going from
investment sector to the nondurable goods sector as
the intermediate goods channel.
There are two shortcomings of the Hornstein
and Praschnik (1997) analysis. First, the model is not

Econom ic Perspectives

17) nondurable good sectors is substantially procy
clical (that is, the comovement statistic is 0.45 or
higher), even though the strength of the intermediate
good channel (the magnitude of'/J varies from almost
zero in the case of food to nearly 0.25 in the case of
wholesale trade. Interestingly, although the comove
ment in mining is moderately weak in our data set, it
is one of the sectors in which the intermediate goods
channel is the strongest. Conversely, the comove
ment in apparel is strong, although this sector’s inter
mediate goods channel is almost nonexistent. Based
on these results, we suspect that the intermediate
goods channel to the investment sector plays at best
only a small role in accounting for comovement of
employment in nondurable goods.30 To further ex
plore the Hornstein and Praschnik hypothesis, one
would have to construct a version of their model with
a disaggregated nondurable goods sector and see if
it is consistent with comovement, in the sense of be
ing able to reproduce patterns like those in figure 6.31

consistent with the observed comovement in invest
ment across sectors. That is, the intersectoral linkag
es in the model are not strong enough to produce full
comovement. Second, data on subsectors of the non
durable goods sector cast doubt on the notion that
the intermediate good channel is the only reason
there is comovement. We studied the subsectors of
the nondurable goods sector and found that there is
considerable variation in the fraction of total output
sent to the investment goods industry. But, as docu
mented in the previous section, most of these sectors
nevertheless display strong business cycle comove
ment. Figure 6 is a scatter plot of the subsectors’ de
gree of comovement, drawn (with one exception) from
the fifth column of numbers in table 1, against the
strength of each sector’s intersectoral linkage with
the investment sector, I. The variable I is the fraction of the gross output of a sector which is allocated
to intermediate goods destined directly or indirectly
for the production of final investment goods.28 (Tech
nical appendix 3 has details of how we computed
this.) The numbers in parentheses in figure 6 indicate
the relative magnitude of the gross output of the sec
tor (gross output of the sector in 1987, divided by the
sum of the gross outputs across all 17 sectors). Hornstein and Praschnik’s concept of the nondurable
goods sector is broader than the one in table 1. They
also include agriculture; retail trade; wholesale trade;
transportation, communication, and utilities; servic
es; FIRE; and mining.29
Figure 6 shows that employment in most (13 of

Alternative approaches

Here, we summarize three other approaches that
may ultimately lead to a satisfactory explanation of
business cycle comovement—strategic complemen
tarities, information externalities, and efficiency
wage theory. The first two approaches emphasize the
importance in business decisions of expectations
about the future. They draw attention to the possibil
ity that general shifts in expectations may trigger
business cycle fluctuations. If so, these shifts in ex
pectations may well constitute the ag
gregate shock to which Lucas (1981)
FIGURE 6
refers. The third approach looks at effi
Business cycle comovement in nondurable goods sectors
ciency wage theory. Although promis
ing,
the ability of these three theories to
comovement
1.00
Transportation, communication,
quantitatively account for the comove
Printing (0.02)
■ and public utilities (0.13)
ment aspect of business cycles is yet to
■
" ■
■ Paper (0.02)
Rubber (0.02) ■ ■
be
fully explored.
Apparel(0.02) tracje(QQ7)
Chemicals(0.03)
Wholesale
0.75

Textiles
(0.01)

■ Leather products (0.00)
0.50

trade (0.07)

Food (0.06)
" ■

■ ■ Services (0.28)
I
Finance, insurance,
and real estate (0.21)

Mining (0.02)
a

0.25

Agriculture (0.04)
a

■ Petroleum (0.02)

Tobacco (0.01)

I
0.00

0.05

I
0.10

0.15

I
0.20

I
0.25

importance of intermediate goods channel to investment sector (/c)
Note: Plot of sector is calculated using the comovement statistic reported in
table 1 and as explained in note 6. The number in parentheses following the sector
name indicates the relative size of that sector.
Source: Authors'calculations from data of DRI Basic Economics database, 1964-96.

Federal Reserve Bank of Chicago

Strategic complementarities
Suppose there are two people, A
and B. Each has to decide on a level of
work effort: high or low. Suppose that
the net gain to A of exerting a high level
of effort is greater ifB exerts a high level
of effort and that B is in a similar posi
tion.
The situation is depicted in table 2.
I
Table 2 has four entries, one for
each possible combination of work ef
fort. In each entry, the first number indi
cates the net gain to A, and the second
number indicates the gain to B. Suppose
A exerts high effort. Then, ifB is putting

out high effort too, A receives 5. But, ifB exerts low
effort, then A receives -5. The situation is the same
for B. Table 2 captures the idea that the gain to either
person from exerting high effort is high only when
the other person exerts high effort. A situation like
this is said to be characterized by strategic comple
mentarity. What do we expect to happen? If the two
people could sit down and reach an agreement, they
would clearly both choose to exert high effort. But
what if they have difficulty coordinating in this way?
There are now two possibilities. One is that each ex
pects the other to exert low effort, in which case each
finds it optimal to exert low effort. This would put the
two people in the bottom right box, with a low payoff
going to each. They would stay there until they
found a way to communicate and reach an agreement
or until something happened to alter their expectation
about the other’s plans. Another possibility is that
each expects the other to exert high effort, in which
case it is in the private interest of each person to
exert high effort. This situation could persist for a while,
again, unless something happens to shift one person’s
expectations about what the other one will do.
What does this have to do with business cycles
and comovement? Possibly a lot. There are aspects
of business decisions that exhibit strategic comple
mentarity. For example, suppose a firm is considering
reopening a plant or starting a large capital invest
ment project. Suppose the project involves a sub
stantial outlay of funds, not just to hire more workers
but also to purchase materials and supplies from oth
er firms. The higher the sales the firm expects in the
future, the more inclined it will be to shift to a high
level of activity in this way. However, much of a
firm’s sales come from other firms. And those sales
are greater if other firms are themselves operating at a
high level of activity, for example, reopening plants or
undertaking new capital investment projects. So, firm
A has a greater incentive to shift to a high level of ac
tivity if it believes firm B plans to operate at a high
level of activity.
What do we expect in this situation? Coordina
tion in this setup is much more difficult than in the

two person example. There are millions of firms in the
economy and, even if it were technically feasible for
some firms to coordinate, the antitrust laws represent
another barrier. In light of these considerations, we
might well expect to find results similar to those in
the two person example. Thus, if firms were pessimis
tic about prospects for future sales, they would choose
to be inactive and their pessimistic expectations would
be fulfilled.32 Optimistic expectations would be selffulfilling in the same way. It is clear that in this setting,
expectations have the potential to act as an aggre
gate shock driving the business cycle. Of course,
that does not guarantee that they can necessarily
account for comovement.33 This is an important topic
of research.34

Information externalities
Another potential source of comovement is the
way information about the state of the economy is
transmitted to individual firms. Forecasts of the future
strength of the economy are a factor in individual
firms’ current investment decisions. If a firm observes
a series of construction projects being initiated by
other firms, it may infer that those other firms have
information that bodes well for the general economic
outlook. When the firm combines this inference with
its own information about the economic outlook, it
may decide to invest too. Other firms may follow for
similar reasons.
These considerations are logically distinct from
the strategic complementarities discussed above. There,
a firm is interested in the actions of other firms because
these actions have a direct impact on the firm’s prof
itability. Here, a firm is interested in the actions of other
firms because of the associated information externality.
The externality refers to the fact that a firm’s action
may reveal information it has on something of inter
est to other firms, such as the state of the economy.
It is a positive externality, unlike the more familiar
examples of externalities which tend to be negative.35
We present an example, taken from Banerjee
(1992), to illustrate the sort of things that can happen
when there are information externalities. When firms
look to what other firms are doing for guidance in de
ciding what they should do, this can lead to what
Banerjee (1992, p. 798) calls herd behavior, a situation
with “everyone doing what everyone else is doing,
even when their private information suggests doing
something quite different.” It hardly needs to be stat
ed that herd behavior sounds like comovement.
Here is the example. Suppose there are 100 peo
ple trying to decide between two restaurants, A and
B. Each person knows very little about the two

Economic Perspectives

restaurants, but thinks the odds favor A slightly. In ad
dition, each person receives a signal about the relative
quality of the two restaurants. For example, one per
son may read a review of the two restaurants in a
travel guide. The review is several years old, howev
er, and the signal may not be accurate. The signals re
ceived by each of the 100 persons are equally reliable.
Everyone knows this, but they do not know what sig
nal the others received. Now, suppose that 99 people
get a signal that suggests B is better than A, while
one person gets a signal that A is better than B. If all
information were known to everyone, they would rec
ognize that the preponderance of the evidence favors
restaurant B and all 100 people would go to B. How
ever, what actually happens is that the 99 people
whose signal indicates B is better ignore their signal
and flock to restaurant A, following the one person
who received the signal that A is better.
This result is not due, as one might suppose, to
an assumption that agents are irrational. On the con
trary, the example assumes agents are completely
rational. The result reflects that not all people make
their decisions at the same time. Some have to be
first, and as a result, the information they have has
disproportionate impact, since almost everyone else
is watching them. This timing assumption does not
seem unrealistic. In practice, the exact timing of firms’
decisions is not completely under their control.36
The example adopts an extreme version of the
assumption that the timing of a decision is out of the
agents’ control, specifying that someone must choose
first, then someone else must choose second after
observing the choice of the first, and so on. The per
son choosing first happens to be the one who receives
the signal that A is better than B. Since person l’s
suspicion that A is better is apparently confirmed by
the signal, this person rationally chooses A. The sec
ond person’s signal suggests that B is better. However,
person 2 knows that person l’s signal must have
favored A. Since the two signals are equally reliable,
they cancel in the mind of person 2. Since person 2
originally thought restaurant A was better, the ratio
nal thing for person 2 to do is to go to restaurant A.
Person 3 is in precisely the same position as 2, because
person 3 knows that, given person 1 went to A, per
son 2 would have gone to A no matter what signal
she received. That is, person 3’s observation that
person 2 went to A provides no information at all
about the relative quality of the two restaurants. Being
in the same position as 2, person 3 also chooses A
regardless of the signal received. In this way, all 99
people after the first ignore their own signal and go

Federal Reserve Bank of Chicago

to restaurant A. Although there is a lot of information
in the economy about the relative quality of the res
taurants, one person acts on a small piece of it, and
everyone else follows.
This example and others like it hold out some
hope that a fully developed business cycle model
incorporating information externalities might exhibit
the synchronization of behavior across economic
sectors that we observe over the business cycle. How
ever, the above example only illustrates how informa
tion externalities can lead rational people to ignore
information and synchronize on bad decisions. Syn
chronization of actions would have occurred anyway,
even if there had been no information problem and all
signals had been known to everyone. Another con
cern with this example is how heavily dependent it is
upon details of the environment. For example, the
outcome is very different if two people are required to
choose a restaurant first. In this case, the 99 people
who received the signal that B is better than A go to
B.37 38 Despite these considerations, we believe the
growing literature on information externalities may
eventually provide at least part of the explanation for
business cycle comovement.39

Efficiency wage theory
A third strategy for understanding comovement
is to make use of efficiency wage theory.
Efficiency wage theory: A sketch
Under this view of labor markets, the amount of
effort a worker makes (the worker’s efficiency) depends
on the wage that the worker is paid. Development
economists hypothesized that in economies at a very
early stage of development, a higher wage leads to
greater worker efficiency because it facilitates improve
ments in diet and health. Efficiency wage theory holds
that a higher wage also results in greater worker effi
ciency in a modern, developed economy, but for differ
ent reasons. Because employers cannot perfectly
monitor the amount and quality of work effort expended
by their employees, there is a temptation for workers
to shirk. Efficiency wage theory says that a high wage
rate is an effective way to combat this temptation.
The higher the wage, the more a worker has to lose
if caught and fired for poor job performance.
The simplest version of this idea was articulated
by Robert Solow,40 who theorized that the firm selects
a wage rate, the efficiency wage, which maximizes
worker effort per dollar paid. The firm is not willing
to pay more because the resulting increase in worker
effort would not be enough to warrant the extra cost.
The firm is also not willing to pay less, because the

FIGURE 7

Efficiency wage model
Marginal
productivity, wage

resulting fall in effort would exceed the fall in cost.41
In the Solow model, the amount of effort expended
per hour by a worker is a function only of the current
wage and, for example, does not depend on the general
state of business conditions. As a result, the efficiency
wage rate does not vary over the business cycle under
Solow’s efficiency wage theory.
The firm also has to decide how many workers to
hire. It hires workers up to the point at which the mar
ginal productivity of the last worker is just equal to the
efficiency wage.42 The downward sloping marginal
productivity of labor curve in figure 7 shows how the
marginal productivity of labor is lower at higher lev
els of employment. At the level of employment, L, the
marginal productivity of the last worker hired is equal
to the efficiency wage. Since the efficiency wage in
the Solow model is a constant, it follows that employ
ment over the business cycle is determined by the re
quirement that the marginal product of labor does not
change. The downward sloping curve in figure 7,
marginal productivity of labor', shows the marginal
productivity curve after it has been shifted up by a
positive technology shock.43 If the firm kept employ
ment fixed at L when technology shifted up, marginal
productivity would rise to W', a point far above the
efficiency wage. By increasing L to L', the firm keeps
marginal productivity unchanged despite the shift up
in technology.44
A notable feature of efficiency wage theory is
that labor supply plays no role in the determination
of the wage rate. The theory assumes that there are
more workers willing to work than the firm is willing

to hire at the efficiency wage. Still, un
employed workers cannot bid the wage
down below the level of the efficiency
wage. Firms are not interested in hiring
workers at such a low wage because
they fear it would not provide workers
with enough incentive to work hard. The
quantity of unemployed people is the
number who are willing to work at the ef
ficiency wage, minus the number that
firms want to hire. Note how the upward
sloping labor supply curve in figure 7 is
shifted to the right. At the efficiency
wage, Ls, workers would like to work, but
only! are hired, so unemployment is TA
- L. At the higher level of technology,
unemployment falls to Ls-L'.

Efficiency wage theory and business
cycle comovement
How might efficiency wage theory help account
for business cycle comovement? Suppose the busi
ness cycle is driven by an aggregate, real-businesscycle-type technology shock. As we explained earlier,
in the standard real business cycle model such a
shock does not lead to comovement in employment.
In that model, a positive shock leads to a transfer of
resources—labor and capital—away from the firms
producing consumption goods and toward the firms
producing investment goods. Now suppose the labor
market part of the real business cycle model is replaced
by efficiency wage theory, which implies that firms
vary the number of workers they employ to ensure
that the marginal product of labor remains constant
and equal to the efficiency wage rate. So, when a posi
tive technology shock shifts up the marginal produc
tivity of labor, employment must increase to maintain
equality between the marginal product of labor and
the efficiency wage.45
We indicated earlier that a positive real-businesscycle-type shock pushes up the production functions
and the marginal labor productivity curve of all firms.
According to efficiency wage theory, this results in
an increase in employment by all firms, as they seek
to maintain equality between marginal labor produc
tivity and the unchanging efficiency wage rate. This
is the intuition underlying the idea that efficiency
wage theory may help explain business cycle co
movement.46- 47
Have we now established that efficiency wages
are sufficient to account for comovement? Absolute
ly not. When we examine efficiency wage theory more
closely, we discover that it need not necessarily work

Economic Perspectives

as just outlined. The relationship between how hard a
worker works and the wage rate (the worker’s effort
function) is a function of the household’s attitude to
ward risk, the resources it has available if the worker
is caught shirking and fired, the probability of being
caught conditional on shirking, and the precise con
sequences of being fired for shirking. The household
effort function used in the analysis must integrate all
these factors in a logically coherent way. In addition,
it must be consistent with other household decisions,
such as how to split income between consumption
and saving. To build confidence in the idea that effi
ciency wage theory helps account for comovement,
we must integrate all these aspects of the household
into a coherent framework which also includes firms
and their decisions to see if it works.
To understand why it might not work, recall the
Solow model’s assumption that worker effort is a func
tion only of the wage rate. That is what led to our con
clusion that the efficiency wage is a fixed number,
unrelated to the state of the business cycle. But the
logic of the efficiency wage argument suggests that
the Solow assumption may not be consistent with ra
tional behavior by households. According to efficiency
wage theory, what motivates hard work is the fear of
losing a high-wage job. Of course, the cost of that
loss is not a function of the wage rate alone. It is also
a function of the amount of time the worker can expect
to be out of a job after being fired. This suggests that
the horizontal line in figure 7 shifts up in a boom, when
the duration of unemployment is low.48 However, if it
shifts up high enough, the comovement result could
disappear. This highlights the importance of integrating
efficiency wage theory into a logically coherent model,
before we conclude that it provides a solid foundation
for understanding business cycle comovement.
Important steps have been taken in this direc
tion, for example, Shapiro and Stiglitz (1984) andDanthine and Donaldson (1995).49 Recent work by Gomme
(1998) and Alexopoulos (1998) makes a significant

further contribution toward understanding the impli
cations of efficiency wage theory for business cy
cles. However, this work does not focus on the
implications for business cycle comovement. We ar
gue that doing so is a good idea.50

Conclusion
A key feature of the data is that, in a frequency
range of two to eight years, output, employment and
investment across a broad range of sectors move up
and down together. We have documented this phenom
enon—business cycle comovement—as it pertains to
employment. Our survey of possible explanations for
it is by no means exhaustive. Many other approach
es—those based on sticky prices and wages, countercylical markups, and credit market frictions—also
deserve consideration.51 Still, we have covered a wide
range of models, from straightforward modifications
to standard business cycle theory to theories that
suggest analogies between businesspeople and herds
of animals.
Many of the approaches we have surveyed are
in early stages of development, while some have been
developed to the point where their implications have
been quantified and compared with the data. Among
these, only one has been shown to be consistent with
the observed strong comovement in output, employ
ment, and investment across sectors of the economy—
the model presented in Christiano and Fisher (1998).
This model incorporates a specification of household
preferences, habit persistence, that is not currently
standard in the macroeconomics literature. We believe
that the success this model has in generating comove
ment warrants giving this specification of preferences
further consideration.
Because comovement is such a central feature
of business cycles and because we do not have a
generally agreed upon theory of comovement, we
conclude that the business cycle is still a puzzle.

TECHNICAL APPENDIX 1

Extracting the business cycle component
of a time series
In casual discussions of economic time series, we
often think of the data as being the sum of compo
nents that have different frequencies of oscillation:
the business cycle component lasting two to eight
years, components lasting shorter periods, etc. The
theory of the spectral analysis of time series makes
this intuition rigorous. It clarifies how one can think

Federal Reserve Bank of Chicago

of data as being composed of components that fluc
tuate at different frequencies. The method we use to
extract the business cycle component of economic
time series builds on this theory. For this reason, we
begin with a brief section which attempts to convey
the basic intuition of spectral analysis. The second
section uses this intuition to describe and motivate
our method for extracting the business cycle compo
nent of a time series.

Decomposing a time series into
frequency components
At the core of the spectral analysis of time series
is the view that the data can be thought of as the sum
of periodic functions. The purpose of this section is
to explain this. We begin by reviewing the basic peri
odic function used in spectral analysis, which is com
posed of a sine and a cosine function.
Consider the following cosine function of time, /:
cos (/co), 1 = 0, 1,2,....
A graph of this, with cos (/co) on the vertical axis
and / on the horizontal axis, exhibits the oscillations
between 1 and-1 familiar from high school trigonom
etry. Recall too, that the period of the cosine function
is 2n. That is, cos(v) = cos(v + 2nh), for/?= 1,2,....
Thus, after the argument of the cosine function in
creases by 2n, the function repeats itself in a periodic
fashion.
What is of interest here is the period of oscilla
tion of cos(/co), expressed in units of time. This is the
amount by which / must increase so that /co increases
by 2n. Thus, suppose /] is a given point in time. We
want to know what is the later point in time, /2 > /p
when the cosine function begins to repeat itself. This
is just /2 such that /2co-/1co = 2tt, or, < -/, = 2n/co.
Thus, the period of oscillation of cos(/co), in units of
time, is 2n/co. The parameter co is referred to as the fre
quency of oscillation.
The function, sin(/co), behaves similarly. It fluctu
ates between 1 and -1, and has a period of oscillation
of 2tc/co. Thus, the two functions have the same ampli
tude (magnitude of vertical variation) and period.
However, the sine function has a different phase than
cos(/co). For example, a graph of the two functions to
gether shows that one looks like the other, apart from
a horizontal shift. The phase difference between the
two functions is a measure of the magnitude of this
horizontal shift. Figure A1 displays sine and cosine
functions for/= 0, 1,..., 200. The period of oscilla
tion is 100, so that the frequency is co = 2n/100. Thus,
the figure displays the graphs of cos(/2n/100) and
sin(/27t/100).
We can now describe the central periodic function
in spectral analysis, namely the linear combination of a
sine and a cosine function
1) acos(/co) + 7>sin(/co),

where a and /) are parameters. This function obviously
has a period, in units of time, equal to 2n/co. But, its

amplitude and phase depend on the values of a and
/). If a and /) are both very small, the resulting function
has very small amplitude and if a and /) are both very
large it has a large amplitude. Also, as the size of a is
increased and the size of b is decreased, the phase of
the function shifts, as more weight is allocated to the
cosine and less to the sine.
It turns out that sums of periodic functions like
equation 1 look very much like actual data. Thus,
suppose we have a time series of data, xp t= 1,..., T.
To see that x, can be expressed as a sum of periodic
functions, suppose we specify 772 (suppose 7'is even)
such functions, each with a different frequency of
oscillation co. For concreteness, let
2tylT, forj = 1,
..., 7/2. To distinguish the parameters associated with
each of these functions, we denote them by a and b
fory = 1,..., 772. It should not be surprising that, in
general, a time series, xp ..., xT can be written as the
sum of these 772 functions

co =

2) xt = a j cos(/C0[) + />! sin(/C0i) + ...

+aT/2 cos(/cor/2) + Z>r/2 sin(/cor/2),
for / = 1,..., T. That is, we can always find values for
the 7'parameters, (a b:,j = 1,..., 7/2), so that the T
equations, equation 2 for / = 1,..., 7', are satisfied. To
see this, consider the following regression. Let the
explanatory variables be:
••• cos(cor/2)

sin(cor/2)

cos/cOj)

sin/Wj)

cos^cOj)

sin(2®i) ••• cos(2cor/2)

sin(2cor/2)

cos)!®!)

sin(7(0i) ••• cos(7cor/2)

sin(7cor/2)

X— .

Economic Perspectives

Let the 7’x 1 vector of‘independent variables,’L
and the T 1 vector of regression coefficients, (1 be

• M**

, p=

aT!1
_hTi2 _

Then, the regression is

Y=X$ + u.
Note, however, that since the number of explanatory
variables is T, the error term is exactly zero, and p is
computed from (T X)-1 X Y = X-1 Y.1 Thus a time se
ries of length T can be expressed exactly as the sum
of 772 simple processes like equation 1, each having a
different frequency of oscillation.
Unfortunately, taken literally, equation 2 is not a
very sensible way to think of the data. With T obser
vations, one has only to compute the as and the bs
and then the T + 1 st observation can be predicted ex
actly, No one believes that there is any way to use 7’
observations on any economic data series and predict
the next observation exactly. Imagine, for example,
that you could do this with the Dow Jones Industrial
Average. If you could, then after one minute of read
ing this appendix, you would have the information
needed to go out and become fabulously wealthy
Of course, one could instead suppose that the
data are a realization from an expression like equation
2, in which the number of periodic functions exceeds
the number of observations by, say, 10. In this case,
there is no longer the implication that one can per
fectly predict next period’s value ofx(. However, there
is the implication that after 20 more periods, the data
series will then become perfectly predictable. No one
would think this is a sensible way to view economic
data either. The theory of spectral analysis assumes,
sensibly, that no matter how many observations on x(
one accumulates, the data never become perfectly
predictable. That is, it in effect assumes that the num
ber of periodic functions in equation 2 is infinitely
large by comparison with the size of the available data
set. When this is so, equation 2 is written in the form
of an integral, as follows:2

3) x, =/“[alrajcosfmt)+fe(ra)sin(<B/)]i7(D,
where a(m) and Z>(©) are functions of co. In view of
these observations, it is perhaps not surprising that

Federal Reserve Bank of Chicago

any covariance stationary time series process, x, can
be expressed in the form of equation 3 (Koopmans,
1974).

Extracting the business cycle component
Equation 3 allows us to make precise the notion
of extracting the business cycle component of xf That
representation views the time series process, x(, as
the sum of components with periods of oscillation
2jt/co for co lying in the interval 0 to
In monthly
data, the business cycle corresponds to components
with period greater than 24 months and less than
96 months. In terms of frequencies of oscillation,
this corresponds to co belonging in the interval
co = 2ji / 96 to co = 2ji / 24. Thus, we seek the
business cycle component of x(, yt, such that

it.

y, = jff[a(ro)cos(<D/) + ^(®) sin(co/)]Jco.

It is well known that yt can be computed as a par
ticular centered moving average of observations on
the observed data, x(

+xm) + B2(xi_2+x(+2)+...,

where

_ sm(7m)-sin(jm)

There is a major practical stumbling block to using
equation 5 for extracting the business cycle compo
nent ofxf. It requires an infinite amount of data! Some
sort of approximation is needed, if one is to estimate
yt given only the available data, xp ..., x(.
An extensive analysis of this problem appears in
Christiano and Fitzgerald (1998), which also provides
a review of the related literature. We provide only the
briefest review of that discussion here, just enough
to enable us to describe exactly how we isolated the
business cycle component of the data.
We denote our approximation ofy(byj>, Here, we
focus on the approximations of the following form:
7) yt = Box, + % (x(_, + x<+1) + ...

+ BK(xt_K + xl+K).
That is, we approximate^ by a finite ordered, cen
tered, symmetric moving average. But, how should we
choose the weights? The natural way is to choose

them so that is as close toy, as possible, that is, so
that they solve
8)

Also.
T

ar/2 -

. min £(y,-y,)2.
Bj, /=0,l.. V

YcOS(0}T/Tt)xt IT
t=l

IT.

bT/2 ~
,t=i

The solution to this problem is a function of the
details of the time series representation ofx. For ex
ample, if we suppose that .v, is a random walk, that is,
.V, - xtis a process that is uncorrelated over time,
then the solution is:3

2To gain further intuition into the relationship between equa
tions 2 and 3, it is useful to recall the simplest definition of
an integral, the Riemann integral. Thus, let fly) be a function,
with domain y < y < y . Let y .,j = 1, ..., M be a set of numbers
that divide the domain into M equally spaced parts. That is,

Ti = NAW2
9) B.=B.,y=0,...,W-l

where
= (T — T) / M. Note that yM = y. The integral off
over its domain is written,

A =—[4-So + A + ^2 + •••
g/(y)rfy.

Suppose the data at hand are xp ..., xT, so that
the objects ofinterest are, vp ...,yT. We computedy36,
...,yT_36 as follows. Fory36 we applied equation 7 with
K= 35, fory37 we applied equation 7 with A'- 36, and
so on. For each y, that we computed, we used the
largest value of A? possible. Christiano and Fitzgerald
(1998) argue that this procedure for estimatingy, works
well in terms of optimizing equation 8, even if the true
time series representation of.v,is not a random walk.
They show that an even better approach uses an
asymmetric set of weights, so that the estimate of v,
for each Z uses all the available data on x,.
xNote that sin(Zcor/2) = 0 for all integers t. Since the right col
umn ofX is zero in this case, X is singular and so cannot be in
verted. In practice, the last column of X is replaced by a
column of ones, to accommodate a non-zero sample mean in
x. Under these conditions, the columns of A are orthogonal, so
that JU1!' is trivial to compute. In particular, for j = 1,..., T/2 - 1:

This is approximated by the sum of the areas of the
Affly) by rectangles:
M
I /(y7)AM.

7=1
The Riemann interpretation of the integral is that it is
the limit of the above sums, as M —> oo. The relationship be
tween the above finite sum and the integral resembles that
between equations 2 and 3 if we adopt y. = co. = iTij/T, AM = Iti/T,
M = T/2, fly) = a(co) cos(co/) + Z?(co.) sin(co.Z), a(cop = aTI 2k,
Z>(co.) = bll'IlK.

3Actually, the theory as we summarized it here technically does
not accommodate nonstationary processes like random walks.
Christiano and Fitzgerald (1998) discuss standard ways of ex
tending the theory to this case. Also, optimizing the mean
square error criterion, equation 7, requires a constant term in
equation 8. See Christiano and Fitzgerald (1998) for more de
tails.

? T

a,- = v X cos(<w7f)xr,
z=i
? T

TECHNICAL APPENDIX 2

Comovement and the elasticity
of substitution
The standard real business cycle model assumes that
the elasticity of substitution between capital and labor
in production is unity In the text, we discussed a
result due to Benhabib, Rogerson, and Wright (1991):
With this kind of production function and with utility
functions consistent with balanced growth, comove
ment in employment is impossible (see note 16). Here,
we show that comovement is a technical possibility
when the elasticity of substitution is different from
unity. However, we find that comovement does not

occur for plausible parameter values. These results
suggest that attempts to account for comovement by
adjusting the elasticity of substitution in production
in a standard real business cycle model are unlikely
to be successful.
We begin by describing a version of the standard
real business cycle model. We assume that house
holds have identical preferences of the following form:

Economic Perspectives

o, t|/ >

where
0 satisfy the various conditions required
for utility to be strictly concave. Also, Cf> 0 denotes
per capita consumption, and Lt denotes per capita
hours worked. We require 0 < Lt < 1. The resource
constraint is
c, + £,+1-(!-§)£, < (l-a)Z-,v + a£,

Here, 0 < § < 1 is the rate of depreciation on capi
tal and 0 < a < 1 is a parameter. The elasticity of substi
tution between capital and labor is v > 0. Also,
log(z,) = p log(z,_!) + e,, 0 < p < 1,
where er is a zero mean random variable, that is inde
pendently distributed over time. Finally, k>0 denotes
the beginning of period Z stock of capital, which is a
given quantity at time Z.
As noted in the body of the article, it is possible
to interpret this as a two sector model: one in which
consumption goods, cp and investment goods, kf+1 (1 - §) kp are produced in different sectors by differ
ent firms. It is assumed that both sectors use the same
production function, the one stated above, and that
capital and labor can move freely between the two
sectors, subject only to the obvious constraint that
the sum of capital and labor in the two sectors equals
kf and Lf, respectively. Thus, letting La, ka and L.f, k.t
denote the amount of labor and capital, respectively,
used in the consumption and investment sectors,
we require

t|Z

(/VO-SMcVO-H
were generated using our approximation to the
model’s solution. The 1,000 observations were then
used to compute the correlations, r'ic between La and
4 and the correlations, o' , between/ , and/ A
model exhibits comovement in employment and in
vestment if both
0. We found
0 for
each value of n using the benchmark parameterization.
We repeated these calculations several times,
each time perturbing one, and only one, of the param
eters in the benchmark parameterization. We consid
ered the following alternatives: CT = 5; p = 0.99; p = 0.0;
steady state hours equal to 0.1; steady state labor’s
share equal to 0.3; § = 0.05, § = 0.01; and 3 = 0.97,
3 = 0.995. The perturbations in o, p, 3, and § did not
produce a parameterization exhibiting comovement.
The reduction in labor’s share resulted in comovement
in employment, but not investment, for values ofv
between about 3 and 13. Lowering steady state hours
to 0.10 also resulted in comovement in employment
but not investment. Hence, we conclude that altering
the elasticity of factor substitution in production
does not improve the standard real business cycle
model’s ability to reproduce full comovement for rea
sonable parameter values.

p't, pjc >

As mentioned in note 16, the marginal product
of labor in the sector producing the consumption
good equals households’ marginal rate of substitution
between consumption and leisure.

c V —
— z v --------------M
(1-4-4
We drop the time subscripts to simplify the nota
tion. Rearranging this equation, we obtain

^ll-l

(1-4-44=4

Note first that when = 1, we reproduce the result
in note 16, which indicates that L and £ cannot move
in the same direction. When V 1, this reasoning no
longer holds. We can see intuitively that employment

Federal Reserve Bank of Chicago

in the two sectors might move together with
1. In
particular, consider the case
1. In this case, we
have found in many numerical examples that C/z falls
with a rise in z due to a positive shock in e. Continuity
suggests that this also happens when is a little above
unity. We conclude that if the resulting rise in (C/z)1 ''
is sufficiently large, then it is possible for both 4 and
4 to increase in response to a positive shock in e (see
note 16 for the sort of reasoning used here).
We approximated the solution to this model us
ing the undetermined coefficient method in Christiano
(1991). We assigned parameter values in the following
way. For our baseline parameterization, we set () = 0.99,
§ = 0.025, p = 0.95, ando = 1. We chose and a to
guarantee that, in the model’s steady state, labor’s
share of income is 0.64 and steady state hours
worked is 0.30. An empirical defense for the choice of
values for 3, §, p , labor’s share, and steady state hours
worked may be found in the real business cycle litera
ture. For the calculations reported below, we set o =
0.01, and drew 1,000 observations on 4 from a normal
distribution with mean zero and standard deviation, o .
In our first experiment, we considered values of
on a grid between 0.7 and 20. For each value ofv,
1,000
observations on /.ct,and£„
and
’
ir

p'c, pjc <

TECHNICAL APPENDIX 3

Analysis of the input-output tables

Solving this for Y, we get

Our analysis of the input-output tables is based on
the 1987 benchmark, 95 variable input-output table
for the U.S. economy. Our main objective here is to
define the fraction of a sector’s final output which is
used directly or indirectly in the production of final
investment goods. Let L denote the vector of gross
outputs for the production sectors of the economy.
Let A = [a ] be the matrix of input-output coefficients.
That is, a is the quantity of the zth industry’s output
used to produce one unit of they'th industry’s output.
Let If, C, G, O denote the vectors of gross private fixed
investment, personal consumption expenditures, gov
ernment (federal, state, and local) purchases, and
‘other’ for each sector. Here, ‘other’ is essentially ex
ports minus imports. Total output, Y, is broken down
into a part allocated to intermediate inputs, AY, and a
part allocated to final output, /' + C + G + Oas follows:

Y=Y,+Y
if
C + YG +YO’
Y=[I-A]'i,i=If,C,G,O.
For convenience, we report K, z = 7( C, G, O for
the 95 sectors of the U.S. economy which are includ
ed in the input-output table underlying the analysis
reported in figure 6. Table Al reports results for the
17 sectors of the nondurable goods industry, as de
fined in the Hornstein and Praschnik (1997) analysis.
That table reports the input-output table industry
numbers that make up the industries whose name is
in the middle column. Table A2 reports the numbers
for the other sectors. The sum of the numbers in a
row must be unity.

AY+P + C + G + O=Y.
TABLE A1

Results for consumption
l-O industry number

1-0 industry title

1+2+3+4

Agriculture, forestry, and fisheries

0.060

0.894

0.052

-0.006

5+6+7+8+9+10

Mining

0.207

0.893

0.181

-0.282

Food and kindred products

0.018

0.962

0.041

-0.021

Tobacco products

0.000

0.914

0.000

0.086

16+17

Textile mill products

0.185

0.995

0.072

-0.252

18+19

Apparel and othertextile products

0.037

1.284

0.041

-0.362

24+25

Paper and allied products

0.103

0.833

0.112

-0.047

26A+26B

Printing and publishing

0.058

0.795

0.121

0.026

27A+27B+28

Chemicalsandallied products

0.180

0.698

0.138

-0.016

Petroleum refining and related products

0.105

0.782

0.145

-0.032

Rubberand miscellaneous plastics products

0.246

0.761

0.127

-0.134

33+34

Footwear, leather, and leather products

0.031

2.154

0.037

-1.222

65A+...+68C

Transportation,communications, and utilities

0.107

0.740

0.123

0.029

69A

Wholesaletrade

0.232

0.589

0.098

0.082

69B

Retail trade

0.066

0.919

0.015

0.000

7OA+7OB+71A+71B

Finance, insurance, and real estate

0.061

0.877

0.043

0.020

72A+... + 77B

Services

0.076

0.879

0.044

0.002

Notes: Yj measures amount of gross output of industry in indicated row sent directly or indirectly to industry/.where/ = lf, C, G, 0.
Row numbers are scaled so they sum to unity.
Source: Authors' calculations based on data from U.S. Department of Commerce, Bureau of Economic Analysis, 1992,
Survey of Current Business, Volume 72, Number 4, April.

Economic Perspectives

TABLE A2

Results for nonconsumption
1-0 industry
number

l-O industry title

New construction

0.805

0.000

0.195

0.000

Maintenance and repair construction

0.180

0.574

0.243

0.002

Ordnance and accessories

0.011

0.051

0.838

0.100

20+21

Lumber and wood products

0.542

0.340

0.169

-0.051

22+23

Furniture and fixtures

0.477

0.585

0.067

-0.128

29A

Drugs

0.018

0.963

0.125

-0.107

29B

Cleaning and toilet preparations

0.018

0.949

0.033

0.000

Paintsand allied products

0.422

0.445

0.168

-0.036

Glass and glass products

0.202

0.798

0.116

-0.116

Stone and clay products

0.575

0.314

0.206

-0.095

Primary iron and steel manufacturing

0.515

0.501

0.207

-0.223

Primary nonferrous metals manufacturing

0.400

0.485

0.247

-0.132

Metal containers

0.057

0.891

0.064

-0.012

Heating, plumbing, andfabricated structural metal products

0.604

0.201

0.198

-0.002

Screw machine products and stampings

0.358

0.641

0.132

-0.131

Otherfabricated metal products

0.377

0.573

0.175

-0.124

Engines and turbines

0.377

0.362

0.246

0.015

44+45

Farm, construction, and mining machinery

0.731

0.151

0.097

0.021

Materials handling machineryand equipment

0.876

0.134

0.105

-0.115

Metalworking machineryand equipment

0.779

0.261

0.108

-0.147

Special industry machineryand equipment

0.962

0.154

0.028

-0.145

General industrial machineryand equipment

0.729

0.305

0.130

-0.164

Miscellaneous machinery, exceptelectrical

0.309

0.438

0.258

-0.006

Computerand office equipment

0.786

0.148

0.156

-0.090

Service industry machinery

0.636

0.289

0.120

-0.045

Electrical industrial equipmentand apparatus

0.639

0.308

0.168

-0.114

Household appliances

0.242

0.842

0.045

-0.129

Electric lighting and wiring equipment

0.471

0.447

0.185

-0.104

Audio, video, and communication equipment

0.626

0.564

0.206

-0.396

Electronic components and accessories

0.338

0.437

0.322

-0.097

Miscellaneouselectrical machineryand supplies

0.321

0.687

0.148

-0.156

59A

Motor vehicles (passenger cars and trucks)

0.478

0.776

0.051

-0.304

59B

Truck and bus bodies, trailers, and motor vehicles parts

0.437

0.746

0.080

-0.263

Aircraft and parts

0.134

0.049

0.546

0.270

Othertransportation equipment

0.145

0.543

0.336

-0.024

Scientific and controlling instruments

0.442

0.166

0.372

0.020

Ophthalmic and photographic equipment

0.347

0.590

0.228

-0.165

Miscellaneous manufacturing

0.175

1.121

0.071

-0.368

Federal government enterprises

0.079

0.814

0.104

0.003

State and local government enterprises

0.033

0.928

0.043

-0.003
0.000

Noncomparable imports

Scrap, used, and secondhand goods

0.000

-9.699

7.493

1.830

1.377

General government industry

0.000

1.000

0.000

Rest of the world adjustment to final uses

0.000

Household industry

0.000

1.000

0.000

Inventory valuation adjustment

0.000

1.000

Notes: ^.measures amount of gross output of industry in indicated row sent directly or indirectly to industry/, where / = lf, C, G, 0.
Row numbers are scaled so they sum to unity.
Source: Authors' calculations based on U.S. Department of Commerce, Bureau of Economic Analysis, 1992,
Survey of Current Business, Volume 72, Number 4, April.

Federal Reserve Bank of Chicago

NOTES
xSee Burns and Mitchell (1946, p. 3), Lucas (1981, p. 217),
and Sargent (1979, p. 215).
2This definition was taken from the NBER’s web address,
http: //www. nber. org/cycl e s. html.

3An important exception is Long and Plosser (1983), which
does allow for multiple sectors. Their model economy is
straightforward to analyze because they adopt several key
simplifying assumptions. For example, they assume the entire
stock of capital in each sector wears out within three months.
However, these assumptions make the model ill-suited for
quantitative, empirical analysis. It took many years before
economists undertook a systematic empirical analysis of ver
sions of the Long and Plosser model without the key simplifying
assumptions (see Horvath [1998a, b]).
4Employment data are taken from DRI Basic Economics
database. The hours worked data are indexes of aggregate weekly
hours of production or nonsupervisory workers on private
nonagricultural payrolls by industry. The data on numbers of
workers are workers on nonagricultural payrolls, by industry. All
data are monthly and seasonally adjusted and cover 1964:Q195:Q3.
5Other studies of this question include Baxter (1996), Cooper
and Haltiwanger (1990), Hornstein and Praschnik (1997),
Huffman and Wynne (1998), and Murphy, Shleifer, and
Vishny (1989).
6Our statistic is the regression R2 obtained by regressing the
business cycle component of that series on the business cycle
component of total hours worked, at lags 0, 1, and -1. We allow
next month’s employment and the previous month’s employ
ment to enter this relationship because we do not want our
measure of comovement to be low just because a variable may
be out of phase with total private hours worked by only one
month. If we did not include these lags, our regression R2 would
coincide exactly with the square correlation referred to in the
text. We construct our statistic as follows. Let y denote the
business cycle component of a given sector’s employment. Let
xt denote the corresponding measure of total hours worked. We
consider the regression ofy on x{, x , and x py = aftxz+ b^x^ +
o^x/+1 + e<5 where a. represents the estimated coefficients.
Then, the R2s reported in the table are var (pQx + ot x + a2x/+1)/
vtfr(y,).
7Table 1 shows the volatility in each of these data series.
8The correlation between y andy is corr(yt,ya) = Cov(y ,y^/
[Kzr(y)lYir(y )]*. But, Cov(yt, >>It) = a2 and Varfy^ = 2a2, Var(yjt)
= a2. Substituting these results into the formula, we get
corr(yeya) = lN2.-0.71.
’Suppose^ =yh + —
The logic of the previous note leads
to corrty ,y^) = \hln. With n = 33, this is 0.17, after rounding.
10The midpoints are -0.35, -0.25, -0.15, ..., 0.85, 0.95. In
each case, the interval has length 0.1 and extends 0.05 above
and below the midpoint.

“Real business cycle theory has evolved considerably in recent
years and now encompasses a wide variety of conceptions of
the economy. The definition proposed by Prescott (1991,
p. 3) reflects this: “Real business cycle theory is the applica
tion of general equilibrium theory to the quantitative analysis

of business cycle fluctuations.”
12This section and the next one draw heavily on work by
Christiano and Fisher (1998).
13Some might want to dismiss the notion of a technology shock
that affects all firms simultaneously as too preposterous to de
serve consideration. Such a person may find it more plausible
to think of technology shocks as things that are idiosyncratic
to individual firms. Most of the examples of technology shocks
given in the text certainly suggest this. This is the line that
Lucas took when he dismissed the idea that a technology shock
might be the aggregate shock needed to account for business
cycles. He argued that, although technology shocks are no
doubt important at the firm level, they could not be important
for economy-wide aggregate output: He expected that firms
affected by positive productivity shocks would be balanced by
firms experiencing negative shocks. Work of Shleifer (1986)
and Dupor (1998) suggests that the Lucas reason for dismissing
technology shocks as an important impulse to business cycles
may be premature. These researchers emphasize the distinc
tion between the time that a new technological idea arrives in
the firm, and the time the firm implements it. Consistent with
Lucas’s intuition, the exact timing of arrival of ideas may well
be idiosyncratic at the firm level. In this case, the economy
wide average rate of arrival of new ideas would be constant:
Firms discovering ideas for new products or labor-saving ways
to produce output would be balanced by firms experiencing no
progress or even regress. What Shleifer and Dupor emphasize,
however, is that it is not the arrival of new ideas per se that
shifts up production functions. Rather, it is the implementa
tion of the new ideas that does this. They point out that there
may well be plausible mechanisms in an economy which lead
firms to implement new, technology-shifting ideas at the same
time. These mechanisms involve “strategic complementari
ties,” which are discussed further below.
14See, for example, Benhabib, Rogerson, and Wright (1991).
15Formally, this is what we have in mind. A standard real busi
ness cycle model, with unit elasticity of substitution in produc
tion between capital and labor, implies that the value of the
output of the sector producing consumption goods, measured
in utility units, is proportional to the value of the labor used in
that sector, also measured in utility units. The value of the out
put of the consumption sector is the product of the total out
put of that sector, T, and the marginal utility of consumption,
u . The value of the labor used in the sector producing consump
tion goods is the product of the labor used in producing con
sumption goods, Lc, and the marginal utility of leisure, ur Thus,

aYuc = UjLc.
This is just a rearrangement of the usual static efficiency con
dition that specifies that the marginal product of labor in pro
ducing the output of the consumption sector, V.Y/L , must equal
the marginal rate of substitution between consumption and lei
sure, uju . Note that if the term on the left of the equality falls
(‘the value of the output of the sector producing consumption
goods falls’) and rises (‘the marginal utility of leisure rises’),
then/, must fall.
16The inability of the standard real business cycle model to pro
duce comovement is surprisingly robust. Standard specifications
of that model hold that the marginal rate of substitution be
tween consumption and leisure is \|/C/(1 - L -L$, where L is
employment in the consumption sector, L. is employment in
the investment good sector, and 1 - L - L. is leisure. Also, Vj/

Economic Perspectives

and £, are non-negative constants. In Hansen’s (1985) indivisi
ble labor model, £ = 0. In his divisible labor model, £ = 1. The
standard model assumes a Cobb-Douglas production function,
so that the marginal product of labor is proportional to aver
age labor productivity in the consumption good producing sec
tor. Equality between the marginal product of labor and the
marginal rate of substitution between consumption and leisure
implies:
C
(7__
=------ vi/C
-------Lc Q.-Lc-lrf'

Cancelling consumption on the two sides and rearranging, we get
-(1-ZC-Z,)5=ZC.
V
From this it is easy to see that if, for whatever reason, L. or L
moves, then the other variable must move in the opposite direc
tion. This demonstration summarizes a discussion in Benhabib,
Rogerson, and Wright (1991) and in Murphy, Shleifer, and
Vishny (1989). The result holds for the entire class of utility
functions identified by King, Plosser, and Rebelo (1988) as
being consistent with balanced growth. However, the same
cannot be said for the entire class of production functions con
sistent with balanced growth. In particular, the result does not
hold for production functions in which the elasticity of substitu
tion between capital and labor differs from unity. We demon
strate this in technical appendix 2. We also show, however, that
for plausibly parameterized versions of the standard real business
cycle model, departures from unit elasticity of substitution in
production do not help the model reproduce comovement.
17One paper that is often mentioned in the comovement litera
ture is Huffman and Wynne (1998). However, their focus is
primarily on comovement in investment and output. They
largely abstract from comovement in employment by making
assumptions that make labor in the consumption sector essen
tially constant. They specify that the elasticity of substitution
between labor and capital in the consumption sector is nearly
unity, and that — 0. The argument in note 16 explains why
their model has the implication that L is essentially constant.
18Suppose L is the third use of time. Then the equation in note
16 is modified as follows:

a
V

—(1-4-L -L y

=l .
c

Consider a car, for example. Ownership of a car makes it more
attractive to go out on long road trips that require purchasing
market goods like hotel and restaurant services. This suggests
that cars and market goods are complements. A moment’s fur
ther thought about this example suggests that most household
durables actually cannot be neatly labeled as either complements
or substitutes for market consumption. For example, an auto
mobile is also a substitute for market goods because it reduces
the need for market services like cab, train, and airplane rides.
Similarly, consider the biggest household durable of all, the
home. It substitutes for hotel and restaurant services and com
plements market goods such as party goods, telephone services,
and food. Thus, intuition is ultimately not a good guide to
assessing Baxter’s assumption about the substitutability of
durables and market goods. Ultimately, this must be assessed
through careful econometric work to determine whether, on
average, market goods and durables are more like substitutes or
complements.

21Consider the limiting case of perfect substitutability, so that
consumption is C + D, where C is market consumption and D is
the service flow from the stock of home durables. With log
utility, the marginal utility of market consumption is 1/(C +
D). Suppose D is fixed. Then a given jump in C reduces margin
al utility by less, the larger is D.
22Remarks in note 20 about Baxter’s work are obviously relevant
here too. Intuition is a very confusing guide, at best, regarding
the plausibility of Benhabib, Rogerson, and Wright’s assump
tion that the elasticity of substitution between home-produced
and market-produced goods is high. The parameter must be es
timated econometrically. This was done in Rupert, Rogerson,
and Wright (1995), who report, based on data from the Panel
Study on Income Dynamics, that the elasticity of substitution
indeed is high.

23Because the model predicts that consumption rises in a boom,
the high degree of substitutability between home and market
goods causes the marginal value of home goods to drop in a
boom. This in turn causes a drop in the value of home durables,
leading households to reduce their purchases of new durables.
This implication is strongly counterfactual, however, since
durables are in fact highly procyclical. Interestingly, Baxter’s
(1996) model seems to avoid this tension. In particular, her
model generates comovement between employment in the
consumption and investment industries and simultaneously
implies that durable goods purchases are procyclical.

Evidently, now it is possible for both L and L. to be procycli
cal, as long as L is sufficiently countercyclical.

24Boldrin, Christiano, and Fisher (1995) adapt the habit persis
tence specification of preferences proposed in Constantinides
(1990) and Sundaresan (1989).

19Closely related to this is their recommendation that econo
mists work with the following utility function in consumption
and leisure: u [c - \|/Q Z1+*/(l + \|/)] » where V|/, V|/o > 0 and u is a con
cave, increasing utility function. The marginal rate of substitu
tion between consumption and leisure with this utility function
is
Substituting this into the employment condition in note
16 results in

25See Kocherlakota (1996) for a recent review. Although habit
persistence helps to account for the observed average of the
premium in equity over risk-free debt, it does not account well
for the volatility of these variables. For a further discussion,
see Boldrin, Christiano, and Fisher (1997) and Heaton (1995).

a^- = v/0(Lc+L,Y.
The argument in that note that L and L. cannot move in the
same direction does not work with this utility function.
20Baxter’s model is a convenient vehicle for illustrating an issue
that has to be confronted in macroeconomic models generally.
The text provides an illustration of Baxter’s assumption that
durable goods and market goods are substitutes. However, it is
just as easy to think of examples in which they are complements.

Federal Reserve Bank of Chicago

26See Constantinides (1990) and Sundaresan (1989) for more
evidence on the plausibility of habit persistence preferences.

27In the Hornstein and Praschnik (1997) modification, the
output of the consumption sector is C + m, where m is interme
diate goods sent to the investment good sector. Suppose the
marginal utility of market consumption is 1/C. Then, the value
of the output of the consumption sector is (C + m)/C = 1 + m/C.
Note that this jumps with a rise in C as long as m rises by a
greater percentage than C. With m/C sufficiently procyclical,
it is possible for employment in the investment and consump
tion good sectors to move up and down together over the cycle.

28 We are very grateful for instructions and advice from Mike
Kouparitsas on how to analyze the input-output data.

29We do not have an index of hours worked for this sector. In
stead, we used LHAG, which is Citibase’s mnemonic for number
of persons employed in the agricultural industry. We obtained a
measure of comovement for this variable in the same way as
for the other variables.
30The least squares regression line through the data in figure 6 is
p, = 0.48 + 1.35/ . Thus, if a sector was not connected to the
investment sector at all (that is, I = 0), employment in that
sector would still exhibit substantial procyclicality (that is,
Pv = °48)

31 Such an exercise could be pursued by building on the models
in Long and Plosser (1983) and Horvath (1998a, b). To our
knowledge, comovement in the sense studied in this article has
not been investigated in these models.
32A slightly different mechanism, whereby a firm’s expectation
that other firms will be inactive leads all firms to be inactive
was analyzed by Shleifer (1986) and Dupor (1998) and summa
rized in note 13.
33For example, Benhabib and Farmer (1994, 1996) incorporate
strategic complementarities by way of an externality in the
production function. Because their production function is of
the Cobb-Douglas form, the argument in note 16 applies to
these models too. In particular, in these models, employment
in the production of consumption and investment goods must
move in opposite directions over the business cycle.

34The literature on the potential for expectations to be selffulfilling is large. Influential early papers include Azariadis
(1981), Bryant (1983), Cass and Shell (1983), Cooper and
John (1988), Diamond (1982), Farmer and Woodford (1984),
Shleifer (1986), and Woodford (1986, 1987, 1988, 1991).
More recent contributions include Benhabib and Farmer
(1994, 1996), Christiano and Harrison (1998), Cooper and
Haltiwanger (1990, 1996), Farmer and Guo (1994), Gall
(1994), and Schmitt-Grohe (1997).
35An example of a negative externality is the pollution that is
generated as a byproduct of a manufacturing process.
36For an analysis of the case where there are information exter
nalities and timing is under the control of managers, see Chamley and Gale (1994). They find, as one might expect, that there
is a tendency to delay decisions under these circumstances.

37We are grateful to Henry Siu for pointing this out to us.

38The example is similarly sensitive to the assumption that
people view the signals they receive as equally reliable to the
signals received by others. It is possible that, in practice, the
type of individual making investment decisions has greater
confidence in her ability to interpret signals than her counter
parts at other firms. This is the implication of empirical evi
dence that suggests that these types of people are overly
confident in their own abilities. See Daniel, Hirshleifer, and
Subrahmanyam (1998), and the references therein for further
discussion. According to them, (p. 5-6): “Evidence of overconfidence has been found in several contexts. Examples in
clude psychologists, physicians and nurses, engineers, attorneys,
negotiators, entrepreneurs, managers, investment bankers,
and market professionals such as security analysts and eco
nomic forecasters. Further, some evidence suggests that experts
tend to be more overconfident than relatively inexperienced
individuals.”

39A small subset of the literature on information externalities
includes Banerjee (1992), Bikhchandani, Hirshleifer, and
Welch (1994), Caplin and Leahy (1994), and Chamley and
Gale (1994).
40See Romer (1996) for a review.

41Let e(w) be the amount of effort a worker expends per hour,
given the hourly wage rate, w. The efficiency wage is the value
of w that maximizes e(w)/w. One type of e function that guar
antees that this has a maximum for 0 < w < °o is one in which e,
when expressed as a function of w, has an S shape: convex for
w near zero and turning concave for larger values of w (see
Romer, 1996). The optimal e(w)/w is the slope of the ray
drawn from the origin, tangent to the concave part of the e
function. At the optimum, the elasticity of effort with respect
to the wage is unity, that is, e'(w)w/e(w) = 1. Optimality requires
that, when evaluated at the efficiency wage, the second deriva
tive of e with respect to w, is negative.

42The algebra underlying this analysis is simple. Let the produc
tion function be/(e(w)L, K, z), where eL is the total amount of
effort expended in L hours of work, z is a shock to technology,
and K is the stock of capital. We assume that the derivative of
/in its first argument is positive and strictly decreasing in eL
and increasing in z. Revenues net of labor costs are f(e(w}L, K, z)
- wL. The firm maximizes this with respect to w and L. It is
convenient, however, to adopt a change of variables, X = wJ,
and let the firm choose X and w instead. Then, the revenue
function is

Evidently, maximizing this with respect to w is equivalent to
maximizing effort per dollar cost, e(w)/w with respect to w. For
a further discussion of this maximization problem, see the pre
vious note. Maximization with respect to A implies:
ffeL,K,z)e —w,

that is, the marginal product of labor must equal the wage rate.

43The marginal product of labor curve in figure 7 graphs
/(e(w*)Z,, K, z)e(vv*) as a function of L, holding K fixed. Here,
w* is the efficiency wage rate. The curve marked marginal
product of labor' graphs / (e(w*)L,K,z)e(w*) for z>z.
44These observations motivate why efficiency wage theory is
sometimes viewed as a way to fix another set of counterfactual
implications of the standard real business cycle model: that
wages tend to fluctuate too much and employment too little
over the business cycle.
45This argument implicitly assumes that the stock of capital
used by a firm, once put in place, cannot be shifted to another
firm. The assumption guarantees that a positive technology
shock which drives up the marginal productivity of labor curve,
mustbs accompanied by a rise in employment if marginal pro
ductivity is to remain unchanged. If capital were mobile between
sectors, this could even be accomplished with a fall in labor, as
long as capital in that sector fell by an even greater percentage.
The standard real business cycle model assumes that capital is
freely mobile between sectors. Thus, the intuition in this arti
cle is based on two modifications to the real business cycle
model: incorporation of efficiency wages and sectoral immobili
ty of capital. The second of these is not sufficient to produce
business cycle comovement. This is because the argument in
note 16 holds even if capital is immobile between sectors.

Economic Perspectives

46In addition to verifying the logical coherence of efficiency
wage theory as an explanation of comovement, there are two
empirical issues to be investigated. How hard is it to monitor
worker effort? If it can be monitored easily, then efficiency
wage theory is irrelevant. Also, if the penalty for being fired
for shirking is enormous, workers will behave as if they are
being monitored continuously, and once again the theory
becomes irrelevant. For a further discussion of these issues,
see Alexopoulos (1998).

47We stress that the intuition developed here relies on two assump
tions—efficiency wages and sectorally immobile capital.
48To be precise, suppose e(w, D) is the effort supplied by work
ers when the wage rate is w and unemployment duration is D.
At the efficiency wage, e (w, D) < 0. Also, we assume e12(w, D)
= 0. Totally differentiating the first order condition for the ef
ficiency wage, wefiv, D~)/w = 1, with respect to w and D, and im
posing the restrictions on e12 and en yields the result, dw/dD < 0.
49In the literature, what we have called the worker’s effort
function, e, is referred to as the “incentive compatibility
constraint.”

50Alexopoulos and Gomme have reported to us privately that
their models are only partially consistent with business cycle
comovement. In both cases, employment in the consumption
and investment sectors is positively correlated, but investment
in these two sectors is negatively correlated. However, both
models assume that capital can be transferred instantaneously
across sectors in response to a shock. The analysis here suggests

that sectoral capital immobility may be important for obtain
ing comovement.
51For an introduction to the literature on sticky prices and wages,
see Romer (1996). To see why countercyclical markups might
help, recall the key equation in note 16, used to show why hours
worked making consumption goods and hours worked making
investment goods in a standard real business cycle model must
move in opposite directions. Aversion of that model with mar
ket power, for example, the model of Rotemberg and Woodford
(1992), implies that it is the ratio of the marginal product of
labor to the markup that must equal the marginal rate of substi
tution between consumption and leisure. That is, that equation
must be modified as follows:

a C _
yC
ffTc~ (\-Lc-Lf
where p is the markup of price over marginal cost. Cancelling
consumption on the two sides and rearranging, we get
—(1-4-i,)5 = Li/,.

V
Suppose a boom occurs, driving up L. If p falls, as in the
Rotemberg and Woodford model, then it is possible for L to
rise too. (For another model with countercyclical markups see
Gali [1994]). See Murphy, Shleifer and Vishny (1989) for a
conjecture about how limited intersectoral labor mobility,
together with credit market restrictions, may help account
for comovement.

REFERENCES

Alexopoulos, Michelle, 1998, “Efficiency wages,

unemployment, and the business cycle,” Northwest
ern University, manuscript.

_________ , 1997, “Shirking in a monetary business
cycle model,” Northwestern University, manuscript.
Azariadis, Costas, 1981, “Self-fulfilling prophesies,”
Journal ofEconomic Theory, Vol. 25, pp. 380-396.
Banerjee, A., 1992, “A simple model ofherd behav

ior,” Quarterly Journal ofEconomics, Vol. 107, pp.
797-817.
Baxter, Marianne, 1996, “Are consumer durables im

portant for business cycles?,” Review of Economics
and Statistics, Vol. 78, No. 1, February.
Benhabib, Jess, Richard Rogerson, and Randall
Wright, 1991, “Homework in macroeconomics:

Household production and aggregate fluctuations,”
Journal ofPolitical Economy, Vol. 99, pp. 1166-1187.
Benhabib, Jess, and Roger E. A. Farmer, 1996, “Indeter

minacy and sector-specific externalities,” Journal of
Monetary Economics, Vol. 37, No. 3, June, pp. 421—443.

Federal Reserve Bank of Chicago

_________ , 1994, “Indeterminacy and growth,” Jour
nal ofEconomic Theory, Vol. 63, pp. 19-41.
Bikhchandani, Sushil, David Hirshleifer, and Ivo
Welch, 1992, “A theory of fads, fashion, custom, and

cultural change as information cascades,” Journal
ofPolitical Economy, Vol. 100, No. 5, October,
pp. 992-1026.
Boldrin, Michele, Lawrence J. Christiano, and Jonas
D. M. Fisher, 1997, “Habit persistence and asset re

turns in an exchange economy,” Macroeconomic
Dynamics, Vol. 1, No. 2.
_________ , 1995, “Asset pricing lessons for model
ing business cycles,” National Bureau of Economic
Research, working paper, No. 5262.
Burns, Arthur, and W. C. Mitchell, 1946, Measuring
Business Cycles, Studies in Business Cycles, No. 2,
New York: National Bureau of Economic Research.

Bryant, John, 1983, “A simple rational expectations
Keynes-type model,” Ouarterlv Journal of Econom
ics, Vol. 98, No. 3.

Caplin, Andrew, and John Leahy, 1994, “Business as
usual, market crashes, and wisdom after the fact/M/nerican Economic Review, Vol. 84, No. 3, pp. 548-565.

Cass, David, and Karl Shell, 1983, “Do sunspots
matter?Journal ofPolitical Economy, Vol. 91,
pp. 193-227.
Chamley, Christophe, and Douglas Gale, 1994, “In

formation revelation and strategic delay in a model of
investment,” Econometrica, Vol. 62, No. 5, September,
pp. 1065-1085.

Daniel, Kent, David Hirshleifer, and Avanidhar Subrahmanyam, 1998, “Investor psychology and securi

ty market under- and overreactions,” Northwestern
University, Kellogg Graduate School, manuscript,
October 12.
Danthine, Jean-Pierre, and John B. Donaldson, 1995,

“Non-Walrasian economies,” '^Frontiers of Business
Cycle Research, Thomas F. Cooley (ed.), Princeton,
NJ: Princeton University Press, chapter 8.
Diamond, Peter, 1982, “Aggregate demand manage

Chari, V. V., and Patrick J. Kehoe, 1997 “Hot money,”

ment in search equilibrium,” Journal ofPolitical
Economy, Vol. 90, No. 5, pp. 881-894.

F ederal Reserve Bank of Minneapolis, staff report,
No. 228.

Dupor, William, 1998, “Aggregate fluctuations and

Christiano, Lawrence J., and Jonas D. M. Fisher,

production complementarities,” University ofPennsylvania, Wharton School, manuscript, July.

1998, “Stock market and investment good prices:
Implications for macroeconomics,” Federal Reserve
Bank of Chicago, working paper, July, available on
the Internet at www.econ.nwu.edu/faculty.html.

_________ , 1995, “Tobin’s q and asset returns: Impli
cations for business cycle analysis,” National Bureau
of Economic Research, working paper, No. 5292.
Christiano, Lawrence J., and Terry Fitzgerald, 1998,
“The band pass filter: Optimal approximations,”
Northwestern University, manuscript in process,
available on the Internet at www.econ.nwu.edu/
faculty.html.
Christiano, Lawrence J., and Sharon Harrison, 1998,
“Chaos, sunspots, and automatic stabilizers,” Jour
nal ofMonetary Economics, forthcoming.

Einarsson, Tor, and Milton M. Marquis, 1997, “Home

production with endogenous growth,” Journal of
Monetary Economics, Vol. 39, pp. 551-569.
Farmer, Roger E. A., and Michael Woodford, 1997,

“Self-fulfilling prophecies and the business cycle,”
Macroeconomic Dynamics, Vol. 4, pp. 740-769.
Farmer, Roger E. A., and J.-T. Guo, 1995, “The

econometrics of indeterminacy: An applied study,”
Journal of Monetary Economics.

_________ , 1994, “Real business cycles and the ani
mal spirits hypothesis,” Journal ofEconomic Theory,
Vol. 63, pp. 42-73.
Gali, Jordi, 1994, “Monopolistic competition, business

Constantinides, George, 1990, “Habit formation: A

cycles, and the composition of aggregate demand,”
Journal ofEconomic Theory, Vol. 63, pp. 73-96.

resolution of the equity premium puzzle,” Journal of
Political Economy, Vol. 98, pp. 519-542.

Gomme, Paul, 1998, “Shirking, unemployment, and

Cooper, Russell, and Andrew John, 1988, “Coordinat

aggregate fluctuations,” International Economic
Review, forthcoming.

ing coordination failures in Keynesian models,”
Quarterly Journal ofEconomics, Vol. 103, August,
pp. 441^463.

Hansen, Gary, 1985, “Indivisible labor and the business

cycle,” Journal ofMonetary Economics, Vol. 16, pp.
309-327.

Cooper, Russell, and John Haltiwanger, 1996, “Evi

dence on macroeconomic complementarities,” Review
ofEconomics and Statistics, Vol. 78, No. 1, February,
pp. 78-93.

Heaton, John, 1995, “An empirical investigation of

_________ , 1990, “Inventories and the propagation
of sectoral shocks,” American Economic Review, Vol.
80, No. 1, March, pp. 170-190.

Hornstein, Andreas, and J. Praschnik, 1997, “Inter

asset pricing with temporally dependent preference
specifications,” Econometrica, Vol. 63, pp. 681-717.

mediate inputs and sectoral comovement in the busi
ness cycle,” Journal of Monetary Economics,
pp. 573-595.

Economic Perspectives

Horvath, Michael, 1998a, “Sectoral shocks and aggre
gate fluctuations,” Stanford University, Department
of Economics, manuscript, February.

_________ , 1986, “Theory ahead of business cycle
measurement,” Carnegie-Rochester Conference Series
on Public Policy, Vol. 25, No. 0, Autumn, pp. 11-44.

________ , 1998b, “Cyclicality and sectoral linkages:
Aggregate fluctuations from independent sectoral
shocks,” Stanford University, Department of Eco
nomics, manuscript, April.

Romer, David, \996,AdvancedMacroeconomics,
New York: McGraw-Hill.

Huffman, Gregory, and Mark Wynne, 1998, “The

role of intratemporal adjustment costs in a multisec
tor economy,” .Journal ofMonetary Economics,
forthcoming.
King, Robert, Charles Plosser, and Sergio Rebelo,

1988, “Production, growth, and business cycles: I.
The basic neoclassical model,” Journal ofMonetary
Economics, Vol. 21, No. 213, pp. 195 232
King, Robert G, and Mark W. Watson, 1996, “Mon
ey, prices, interest rates, and the business cycle,”
The Review of Economics and Statistics, February,
pp. 35-53.

Rotemberg, Julio, and Michael Woodford, 1992,

“Oligopolistic pricing and the effects of aggregate
demand on economic activity,” Journal of Political
Economy, Vol. 100, December, pp. 1153-1207.
Rupert, Peter, Richard Rogerson and Randall
Wright, 1995, “Estimating substitution elasticities

in household production costs,” Economic Theory,
Vol. 6, No. 1, June, pp. 179-193.
Sargent, Thomas, 1979, Macroeconomic theory, New
York: Academic Press.

Schmitt-Grohe, Stephanie, 1997, “Comparing four
models of aggregate fluctuations due to self-fulfilling
expectations,” Journal ofEconomic-Theory, Vol. 72,
No. 1, January, pp. 96-147.

Kocherlakota, Naravana, 1996, “The equity premium:

It’s still a puzzle,” Journal of Economic Literature,
Vol. 34, No. 1, March, pp. 42-71.

Shapiro, Carl, and Joseph E. Stigliz, 1984, “Equilibrium
unemployment as a worker discipline device,” American
Economic Review, Vol. 74, June, pp. 443 444.

Koopmans, Leonid, 1974, The SpectralAnalysis of

Time Series, New York: Academic Press.

Shleifer, A., 1986, “Implementation cycles,” Journal
ofPolitical Economy, Vol. 94, pp. 1163-1190.

Kydland, Finn E., and Edward C. Prescott, 1982,

Long, John, and Charles Plosser, 1983, “Real busi

Sundaresan, Suresh M., 1989, “Intertemporally de
pendent preferences and the volatility of consump
tion and wealth,” Review of Financial Studies, Vol. 2,
pp. 73-89.

ness cycles,” Journal of Political Economy, Vol. 91,
pp. 39-69.

Woodford, Michael, 1991, “Self-fulfilling expectations

“Time to build and aggregate fluctuations,” Econometrica, Vol. 15,No.6,pp. 1345-1370.

Lucas, Robert E., Jr., 1981, “Understanding business

cycles,” in Studies in Business-Cycle Theory, R. E.
Lucas, Jr. (ed.), Boston: MIT Press, reprinted from
Karl Brunner and Allan H. Meltzer (eds.), 1977, Stabi
lization of the Domestic and International Economy,
Vol. 5, Carnegie-Rochester Series on Public Policy,
Amsterdam: North-Holland Publishing Company.
Murphy, Kevin M., Andrei Shleifer, and Robert W.
Vishny, 1989, “Building blocks of market clearing

business cycle models,” NBER Macroeconomics
Annual 1989, Olivier J. Blanchard and Stanley Fischer
(eds.), Boston: MIT Press.
Prescott, Edward, 1991, “Real business cycle theory:

What have we learned?,” Revista de Analisis Economico, Vol. 6, No. 2, November, pp. 3-19.

Federal Reserve Bank of Chicago

and fluctuations in aggregate demand,” in New Key
nesian Economics, Vol. 2, Coordination Failures
and Real Rigidities, N. Gregory Mankiw and David
Romer (eds.), Boston: MIT Press.
_________ , 1988, “Expectations, finance, and aggre
gate instability,” in Finance Constraints, Expecta
tions, and Macroeconomics, M. Kohn and S-C.
Tsiang (eds.).
__________, 1987, “Three questions about sunspot
equilibria as an explanation of economic fluctua
tions,” American Economic Review Papers and
Proceedings, Vol. 77, No. 2, May, pp. 93-98.
_________ , 1986, “Stationary sunspot equilibria in a
finance constrained economy,” Journal of Economic
Theory, Vol. 40, No. 1, October.

Index for 1998
Title & author

Issue

Pages

BANKING, CREDIT, AND FINANCE
Lessons from the history of money

Frangois R. Velde

First Quarter

2-16

Second Quarter

2-20

Second Quarter

21-32

Second Quarter

33-52

Fourth Quarter

2-11

Deposit insurance reform in the FDIC Improvement Act:
The experience to date

George J. Benston and George G. Kaufman
Assessing the impact of regulation on bank cost efficiency

Douglas D. Evanoff
Access to FHLBank advances and the performance of thrift institutions

Lisa K. Ashley, Elijah Brewer III, and Nancy E. Vincent
Credit derivatives: Just-in-time provisioning for loan losses

James T. Moser
ECONOMIC CONDITIONS
Effects of personal and school characteristics on estimates
of the return to education

Joseph G. Altonji

Fhirst Quarter

65-79

First Quarter

17-43

Second Quarter

53-72

The decline of job security in the 1990s:
Displacement, anxiety, and their effect on wage growth

Daniel Aaronson and Daniel G. Sullivan
Trends in homeownership: Race, demographics, and income

Lewis M. Segal and Daniel G. Sullivan
The increasing importance of retailers’ inventories

Paula R. Worthington

Third Quarter

2-12

Fourth Quarter

56-83

First Quarter

46-64

Third Quarter

13-28

Fourth Quarter

12-34

Fourth Quarter

35-55

Third Quarter

29—43

Third Quarter

44-55

The business cycle: It’s still a puzzle

Lawrence J. Christiano and Terry Fitzgerald
INTERNATIONAL ISSUES
Are international business cycles different under fixed
and flexible exchange rate regimes?

Michael A. Kouparitsas
Understanding the Asian crisis: Systemic risk as coordination failure

David Marshall
Assessing the condition of Japanese banks: How informative
are accounting earnings?

Hesna Genay
Foreign growth, the dollar, and regional economies, 1970-97

Jack L. Hervey and William A. Strauss
MONEY AND MONETARY POLICY
How does an increase in government purchases affect the economy?

Martin Eichenbaum and Jonas D. M. Fisher
Real-time Taylor rules and the federal funds futures market

Charles L. Evans
To order copies of any of these issues, or to receive a list of other publications,
please telephone (312)322-5111 or write to:
Federal Reserve Bank of Chicago,
Public Information Center, P.O. Box 834,
Chicago, IL 60690-0834.

Economic Perspectives and other Bank publications are available on the World Wide Web at www.frbchi.org.
84

Economic Perspectives

Full text of Economic Perspectives (Federal Reserve Bank of Chicago) : Fourth Quarter 1998, Volume XXII, Issue 4

FRASER