View original document

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

Opinions expressed in the Economic Review do not necessarily reflect the views of
the management of the Federal Reserve Bank of San Francisco, or of the Board of
Governors of the Federal Reserve System.
The Federal Reserve Bank of San Francisco's Economic Review is published quarterly by the Bank's
Research and Public Information Department under the supervision of Joseph Bisignano, Senior Vice
President and Director of Research. The publication is edited by Gregory J. Tong, with the assistance of
Karen Rusk (editorial) and William Rosenthal (graphics).
For free copies of this and other Federal Reserve publications, write or phone the Public Information
Department, Federal Reserve Bank of San Francisco, P.O. Box 7702, San Francisco, California 94120.
Phone (415) 974-3234.



Bank Stock Performance Since the 1970s .................... 5
Jack Beebe


Current Fiscal Policy: Is it Stimulating Investment or
................................................. .
Adrian W. Thro op


Inflation, Supply Shocks and the Stable-Inflation Rate
of Capacity Utilization
...... ..45
Rose McElhattan


International Debt with Unenforceable Claims ......... ..64
Jiirg Niehans

Editorial Committee:
Frederick T. Furlong, Joseph Bisignano, Hang-Sheng Cheng,
Michael Keeley, Michael Hutchison and Bharat Trehan.



Jack Beebe*

Since 1979, a turbulent economy and an environment of deregulation have raised concern over bank risk. In the last two years,
the proliferation ofproblem loans has heightened this concern. The
following study presents empirical evidence on risk and returns of
the stocks of 82 U. S. bank holding companies with assets over $1
billion each. Judging from stock performance, the post-1979 economic and deregulatory environment, at least until 1982, was not
unfavorable to bank holding companies with assets over $1 billion
and was favorable overall to those with assets ranging from $1
billion to $10 billion. Since 1982, there has been a sharp downward
valuation on average in the equities of the very large bank holding
companies-those with assets over $10 billion. Statistical analysis
suggests that domestic energy losses and Latin American debt exposures may be largely responsible.

Since 1979, turbulence in financial markets
and changes in the regulatory environment
have raised concern over bank risk. The fiveyear span saw two recessions and wide variations in real growth, inflation, and interest
rates. Uncertainty also has extended wen beyond U. S. markets, as high interest rates, dramatic changes in exchange rates and relative
commodity prices, and worldwide economic
slumps have helped to bring on potential foreign debt crises.
At the same time that the external economic
environment has been volatile, the deregulation
of U. S. banking has proceeded at a rapid pace.
Beginning with the money market certificate in
1978 and proceeding through the Depository

Institution Deregulation and Monetary Control
Act of 1980 and the Depository Institutions
Deregulation Act of 1982, deposit rate ceilings
have been all but eliminated for banks and
thrifts while other nonbanking institutions have
taken on bank-like powers.
In an earlier paper (Beebe, 1983), the author
addressed the question of whether or not the
equity risk of large bank holding companies
(hereafter, "banks") had increased in the 197982 period as a result of the change in the economic and deregulatory environment. 1 That
study found that neither the risk nor returns of
the stocks of banks with assets over $1 billion
seemed to have been affected adversely by the
post-1979 monetary and deregulatory environment, at least through mid-1982. On the contrary, the study found some evidence of a decline in risk-sensitivity for the group consisting
of the largest banks, those with assets over $10
Since 1982, bank risk has received renewed
notice. Problem domestic loans have prolifer-

*Vice President and Associate Director of Research, Federal Reserve Bank of San Francisco. The author thanks Tom Iben and Maureen
O'Byrne for their excellent research assistance
and Fred Furlong for many helpful comments
and suggestions.

ated, particularly within sectors suffering significant declines in relative prices, such as energy, construction, real estate, agriculture, and
timber. Moreover, defaults and reschedulings
of loans to foreign corporations and governments have become realities in some cases and
sobering possibilities in others. Financial markets weathered the failures of Drysdale Government Securities in May 1982, Penn Square
Bank less than two months later, the purchase
of Seafirst Corporation by BankAmerica Corporation, and the "failure" of Continental Illinois Bank. During 1984 alone, an estimated
seventy-nine commercial banks failed, the largest number in any year since 1938.
In light of these recent developments, the
present study looks again at bank equities in

thepost"1979 environment with an emphasis on
thy period since 1982. In this study, the equities
of 82 major U.S. bank holding companies
("banks") with year-end 1981 assets of over $1
biHionareanalyzed to determine whether or
not theperiodthiough September 1984 depicts
abnormal risk or returns. For the largest 24
banks, stock returns since 1982 are related statistically to total debt exposures to the Latin
American countries of Argentina, Brazil, Mexico and Venezuela.
In the sections that follow, there is first a
brief description of the major events that could
have affected bank stocks since 1979, then a
description of the statistical procedures employed, and finally a presentation of the empirical results and conclusions.

I. Events Since 1979
Since 1979, a number of important developments have unfolded that could have had dramatic effects on the equity risk and returns of
large banks. Table 1 gives a chronological list
of a number of such events. However, the
changing environment of banking is better
understood in the context of broader developments. Several such developments have occurred since 1979: (1) From October 6, 1979
through approximately October 1982, the Federal Reserve's short-run operating procedures
targeted nonborrowed reserves rather than the
federal funds rate or borrowed reserves; (2) the
1979-82 period was characterized by considerable interest-rate volatility; (3) in March
1980, Congress passed the Depository Institutions Deregulation and Monetary Control Act
(DIDMCA) which, among other changes, extended NOW accounts to banks and thrifts on
a nationwide basis and called for the phase-out
of deposit rate ceilings; (4) since 1979, the economic environment has been one of volatile inflation (on the downside as well as the upside)
and relative prices (particularly in world commodity markets such as oil), recessions in the
U.S. and abroad, and high real interest rates
worldwide; (5) the financial environment has

seen an increased number of defaults and failures among thrifts, government securities dealers, and banks.
It would have been difficult to say a priori
how this complex combination of events would
affect the equities of large U.S. banks either in
absolute terms or in relation to the stock market as a whole. The outcome would have depended not only on how the events actually unfolded, but also on the extent to which banks
had anticipated or hedged against them,
through ex ante portfolio and operational policies, and the degree to which intervention and
protection by the regulators was perceived as
important by the market. It would have been a
plausible belief that turbulence within the economy would have increased the risk and depressed the prices of bank stocks held at the
time. However there was no reason to presuppose that bank stocks would have been affected
more adversely than the stock market in general. The impact of deregulation on the equities
of large banks might have been expected to be
positive, because of some combination of reduced risk and/or increased returns (see Beebe,
1983). Whether on balance regulatory protection was perceived to have increased cannot be

Events Affecting .U.S. Banks


determined for sure, but the 1980 rise in the
deposit insurance limit from $40,000 to
$100,000 certainly was one corroborating
factor. 2
Given the many collinear impacts of questionable direction and degree, the strategy employed below to gauge any change in bank cap-

ital risk is to examine the behavior of bank
equity risk and returns without offering a full
model that explains risk and returns with exogenous variables. As a partial explanation,
however, a model is estimated that relates stock
returns since 1982 to Latin American debt exposures for the largest 24 banks.

II. Tests for Equity Risk and Return

In the empirical section that follows, bank
stock prices are analyzed first in an absolute
sense to point out the degree of actual price
variation in critical periods. Then, bank stock
returns are compared with returns on the S&P
500, which serves as a proxy of the "stock market." The empirics employ the single-index
market model from the finance literature. This
model postulates that capital risk sensitivity can
be represented by the equity "beta" or the measured sensitivity of the firm's (or portfolio's) equity return with respect to the return on the
market bundle of risky assets (originally,
Sharpe, 1963).3 Precisely because it is measured
in relation to an index of risky assets, beta represents sensitivity to commonly experienced, or
nondiversifiable (often called "systematic")
risk. According to the capital asset pricing interpretation of the single-index market model,
assets with a high beta should have a high expected return because such assets have a high
degree of nondiversifiable risk (originally,
Sharpe, 1964).
In its simplest form, the single-index model

excess of the risk-free rate of interest.
Again, returns are monthly price returns using month-end closing prices
exclusive of dividends and the riskfree rate of interest. (Calculated as
described above for BK.)
"excess" or "risk-adjusted" return for
the sample period-i.e., in excess of
the return earned for taking on nondiversifiable risk, as measured
through beta.
the elasticity of the bank stock price
with respect to the S&P 500 (interpreted as the sensitivity to nondiversifiable or "systematic" risk).
error term (interpreted as non-market-related, or residual, risk).
For the "average" stock in the S&P 500, the
value of beta will be 1.0 by definition. Stocks
with true betas above 1.0 carry above average
nondiversifiable risk and, according to the capital asset pricing model, will have above average ex ante expected returns. Since the model
predicts that only nondiversifiable risk will
yield positive expected returns, the ex ante expected value of alpha in the model is zero. 5
However, the ex post measured value of alpha
may differ from zero because it will reflect the
impact of new information (surprises) on the
stock's price during the period of estimation.
Several questions can be addressed using the
above model. Questions relevant here are: (1)
whether the price of bank equity is more or less
sensitive to nondiversifiable (or "systematic")
risk than the average equity in the S&P 500
(that is, what is the extent to which bank betas
are greater than, equal to, or less than one?);
(2) whether there are significant shifts in beta;



percentage return for the individual
bank stock over the period t in excess of the risk-free rate of interest.
For this study, time periods are
monthly intervals and returns are
price returns calculated from monthend closing prices, exclusive of dividends and the risk-free rate of interest-that is,

(P t



Pt t-l




percentage return on the S&P 500 in

Rrisk-free .


and (3) whether, during turbulent periods,
bank stocks actually have significant ex post
positive or negative alpha (that is, whether new
information leaves positive or negative effects
on bank stock prices after adjustments are
made for the stock~s normal co-movements with
the stock market).
Beta gives a measure of risk or co-movement
with the overall stock market while alpha gives
a measure of return in excess of that associated
with beta. If the market perceives bank equities

to be hedged (or protected by government)
against systematic risk, bank betas will be low.
If the market interprets new information received during the estimation period to be adverse to banks, the estimate of alpha will be
negative. New information conceivably could
affect both the stock's beta (systematic risk sensitivity) and alpha (value beyond that risk sensitivity). The fonowing analysis focuses primarily on whether or not developments since 1979
have affected these parameters.

IU. Empirical Evidence
2. From their depressed levels in 1975, the two
groups of smaller banks generally performed
strongly over the 1976-84 period. This
strong performance may have been due to
the fact that regional banks benefitted from
anticipated and actual deregulation, particularly of consumer deposit rate ceilings.
3. Since early 1983, the group of $10+ billion
banks has had a widely different price performance from that of the other two groups.
Since May 1983, stocks in the group of $10+
billion banks have declined in price an average of 8 percent, while average stock
prices in the $5-10 billion and $1-5 billion
groups each have risen 16 percent. It is plausible that the poor performance of the largest banks since early 1983 resulted from increasing investor concern over foreign loans.
Although the indices in Figure 1 give an overall picture of the performance of large bank
equities over the 12-year period, it is possible
to use the market model to separate the risk
and return measures of bank-stock performance. Table 2 gives estimates of bank stock
betas for the fun twelve-year period and for
subperiods of approximately three years in
length. Despite uncertainty as to a representative estimate of beta in the 1979-81 period (see
the footnote to the table), it is apparent that
the $10 billion and $5-10 billion banks tend to
have average betas above one and the $1-5 billion banks, below one. Moreover, beta tended
on average to decline in the middle of the period for all three groups and then to rise again
for the two groups of largest banks.

The data consist of month-end closing common equity prices for 82 bank holding companies ("banks") with total assets over $1 billion
as of year-end 1981. Twenty-two of these banks
have assets over $10 billion, 17 have assets of
$5-10 billion and 43, assets of $1-5 billion. 6
Figure 1 depicts stock-price levels for the S&P
500 and for equally weighted indices of the
three bank groups. For the full period of over
12 years, the price returns of equities for the
three groups of banks generally have kept up
with those of the S&P 500, although there were
some subperiods that were marked exceptions. 7
Some of the most noteworthy of the trends
in bank stock prices in Figure 1 are worth highlighting and examining here:
1. All bank stocks and the S&P 500 experienced significant declines in value during
1974. The S&P 500 declined by 34 percent
between January and September of 1974,
while the bank stocks declined by even
more. Moreover, the $1-5 billion and $510 billion banks suffered severe and longlasting downward adjustments in value relative to the S&P 500 and the $10+ billion
banks. It is difficult to pinpoint the cause of
the long-lasting effect. It could have been
due to interest-rate exposures from mortgage holdings, although this conclusion is
contradicted by the relatively low betas for
these two bank groups (shown later). Disintermediation attributable to consumer deposit-rate ceilings and loan defaults in nondiversified lending portfolios are other
possible explanations.


BKt = a 0 D0 +
D x + a2 D7
+ p0 (SPt) + 3S (SP, x D.)
+ 3 s2 (SP, X D2) + et






a 0> l?a 2



As described in the introduction, bank stock
prices did not seem to reflect increased bank
risk until about 1982, when domestic and inter­
national lending risks became paramount. To
test for more than one shift in the post-1979
environment relative to the 1972-79 period, the
following variation of the market model [Equa­
tion (1)] allows both alpha and beta to shift at
1979:10 and again at 1982:01:



one for 1972:08 - 1979:09 and
zero thereafter.
one for 1979:10 - 1981:12 and
zero otherwise.
one for 1982:01 -1984:09 and
zero otherwise.
the estimates of alpha for the
1972:08-1979:09, 1979:101981:12, and 1982:01-1984:09
periods, respectively.
the estimate of beta for the
1972:08-1979:09 period.

Figure 1
Monthly Stock Levels 1972.07 To 1984.09
(month-end price levels, excluding dividends)


Betas for Bank Stocks, 1912:08-1984:09
Mean Estimates from Individual Bank Regressions
All Banks










$10+ Billion






$5-10 Billion






$1-5 Billion






lBeta for the boxed-in period is estimated for the 36 months, 1979:01-81:12, excluding the 12 months 1980:07-81:06.
During the excluded 12-month period, bank stocks moved contrary to the S&P 500 in a way that gives spurious and
misleadingly low estimates of the bank stock betas. For !he excluded 12-month period, mean bet;!s were .07, - .22, .01,
and - .02 for the four groups, respectively, and median R2 values were zero or negative. Median R2 values for all estimates
shown in the table ranged between .20 and .54.



Risk and Returns of Bank Stocks
Mean Coefficient Values and Median Test Statistics
From Individual Bank Regressions
1972:08-1984:09 with Shifts at 1979:10 and 1982:01

All Banks

$10+ bil
$5-10 bil

$1-5 bil






(.01 )


(- .29)

(.41 )







(- .45)







(.04 )

- .57
(8.04 )* (-2.13)*






(- .45)






- .02

(- .17)




Coefficient values reported are means of the estimated values for the individual bank regressions in the group. Other
statistics reported are median values. Figures in parentheses are median t-statistics from the individual bank regressions in
the group. Asterisks denote significance of the median t-statistic at the 90% confidence level (one-tailed test for 130 and
two-tailed test for alphas and for 13shifts)' 13shift values and their t-statistics are relative to the base period 130 value. Alpha
values and t-statistics are relative to zero. Alpha values and the standard error of the estimate are expressed as monthly
percentage rates of change at monthly rates.
lBeta shift for the boxed-in period is estimated for the 1979:10-1981:12 period, excluding 1980:07-1981:06. See the note
to Table 2.
2Median t-statistics for 130 using Ho: 130 = 1.0 are - .31, .73, - .09, and -.71 for the four groups, respectively. Therefore,
median 130 estimates do not differ significantly from 1.0, the (weighted) average beta for the S&P 500.




alpha is insignificant, a value that differs from
zero can have a large cumulative effect. In
ure2, the cumulative effect of alpha (and the
error term) is plotted over the 1982:01-1984:09
period. In the table, the actual stock price for
eachindex is plotted against a "market-related"
price, where the latter is that price that would
compensate stockholders for market-related
risk, as hypothesized by the capital asset pricing
model. The "market-related" returns are calculated as follows: 13

the estimate of beta for
1979:10-1981:12 relative to
130 (that is, beta shift for the
second period).
the estimate of beta for
1982:01=1984:09 relative to
130 (that is, beta shift for the
third •period relative to the
first period).

Equation (2) was estimated separately for
each of the 82 bank holding companies. Table
3 presents mean estimates of the coefficients in
Equation (2) and median estimates of the test
statistics for the groups of individual banks. 8
For the $10+ billion banks, the mean beta was
1.15 in the pre-October-1979 period, a figure
that is well above the weighted average beta of
1.0 in the S&P 500. 9 For the $10 + billion
banks, beta declined in the period between
1979:10 and 1981:12 (the decline being significant if the 12-month 1980:07-1981:06 period is
included), but then rose again in the post-1981
subperiod. For the other two groups, there is
no evidence of a significant shift in beta. Even
for the $10+ billion banks, it is hard to conclude that beta shifted significantly given the
uncertainty inherent in the 1979:10-1981:12 estimate of beta. 1o
New information received about a bank
could affect the estimates of both beta and alpha. However, if a bank were to announce that
some of its loans had just become subject to
certain default, it is possible that beta might be
largely unaffected while alpha would be affected negatively because the market value of
the bank's capital would decline by the present
value of the default. 11 As predicted by the efficient market hypothesis and the capital asset
pricing model, investors' ex ante expectations
of alpha are that it will be zero over any future
period. However, ex post observations could
exhibit positive or negative alpha depending on
new information received during the holding
(estimation) period.
Estimates of alpha are reported in Table 3.
As expected for fairly long periods, median
t-statistics for the alpha estimates indicate that
alpha is insignificant. 12 However, even though










estimated "market-related"
return at month t for the jth
individual bank (or bank
estimated beta for the jth
bank (or bank group) over
the 1982:01-1984:09 period
(130 plus 13S2 in Equation
the actual return for the S&P
500 at month t

The vertical spread at a point in time between the two series in each frame of Figure 2
is interpreted as the cumulative effect of new
information unrelated to beta from 1981:12 up
to that point. The group of $10 + billion banks
performed worse than would have been required to compensate for their beta and movements in the S&P 500, while the other two
groups performed better than their betas and
S&P 500 movements would have suggested.
Much of the differentials in performances occurred in early 1983.
The strong performances within the $1-5 billion .and $5-10 billion groups suggest that the
sharp decline in interest rates between July and
November 1982 and/or further deregulation of
consumer deposits (the money market deposit
account, MMDA, of December 1982) may have
been instrumental in raising the market's valuations of these banks. When the MMDA was
first implemented, regional banks that did not
have access to the prime national CD market
maintained that the new account would lower
their marginal costs of funds. Many of these


The fact that the $10 + billion banks had a
negative alpha over the 1982:01-1984:09 period
while the other two groups had positive average
alphas suggests that there is some factor distinguishing the group of largest banks from the
otllertwogroups.••Onepossibility is that foreign
loan exposures of the largest banks may have
affected their stock prices significantly since
1982. As a test of this hypothesis, the following
regression was run on a cross-section of the
largest 24 banks-the 22 banks in the $10+ billion group plus the two largest banks in the $510 billion group,14

banks normally had paid well above the national rates for jumbo CD's, holding company
paper, and other marginal funds, and the
MMDA would attract marginal funds at a substantially lower rate. (In contrast, many small
b;lnks and thrifts claimed tbat,althoughthe
MMDA might lower their marginal cost of
funds, it might also raise the average cost of
funds since they still had considerable amounts
of 5IJ2 percent passbook savings accounts on the
books. Thus, for small banks and thrifts not
induded in this study, the MMDA might have
resulted in a negative alpha.)

Figure 2
Bank Stock Price Levels
-Actual vs. Market-Related
A. All Banks

B. $10+ Billion Banks

1981 :12 = 100


1981 :12 = 100


Actual ...



75 ........._ ............_ _.........._ _.....


75 10-..._........_ _........_ - - 1 1



C. $5-10 Billion Banks
1981 :12 = 100



D. $1-5 Billion Banks
1981 :12 = 100





Actual .... ~~

75 ......







a + b

Latin loans
Capital + e

The value of the Latin American loan coefficient in the second line of Table 4 indicates
that a Latin loan/capital ratio of 1.0 instead of
zero would have lowered the average monthly
stock price return over the 1982:01-1984:09 pe-




the individual bank alpha for
1982:01-1984:09 as estimated by
Equation (2) (see Appendix Table Al for data across the top 22
individual banks).

L~in ~o~ns

month period, the overall compounded effect
would have been a 40-percent negative impact
on a bank's stock price. Equally important, the
dummy for the two banks with heavy energy
loan exposure is also large and highly significant- - 3.2 percent per month. This result, together with significant negative U2 values for
Seafirst and Continental Illinois of - 5.1 percent and - 5.8 percent per month (in Appendix
Table AI), suggests strongly that the problems
of banks with the largest negative performances
were related to energy loans rather than to
Latin American debt exposure. This conclusion
is not surprising, since the energy loan problem
resulted in a sizeable number of actual defaults
and chargeoffs while the foreign lending problem up to this point has resulted primarily in
reschedulings and fears of default. Moreover,
because the foreign loan problem affects almost
all large banks, the market might expect more
government protection in the event of a crisis
than with the energy loan problem.

total loan exposures to
Argentina, Brazil, Mexico, and Venezuela for
March 31, 1984, divided
by primary bank capital
for the same date (see
Appendix Table Al for
the data).

The results of the cross-section fit for Equation (4) appear in the first line of Table 4. The
ratio is insignificant and the iP is about zero.
However, two banks of the twenty four were
extreme outliers in the pattern of residuals,
both with heavy domestic energy loan exposures. When the aberrant residuals of these two
banks were "explained" by a single zero-one
dummy, the Latin American loan exposure ratio became significant and the R2 rose to .65
(the second line in Table 4).

latin loan Exposure Related To Alpha for 1982:01-1984:09

Estimated Across the Largest Twenty-Four Banks for March 31, 1984


latin loans


(- .72)



(- 3.53)

( -6.61)




The Latin loan exposure ratio consists of total non-local-currency loans to Argentina, Brazil, Mexico and Venezuela as
reported in the March 31, 1984, country exposure report (FFIEC-009(a») divided by primary bank capital as of March 31,
1984 (FDIC Call Report). (See Appendix Table Al for the data.) The dependent variable is el2 in Appendix Table AI.
The dummy is one for InterFirst, Dallas, and First City, Houston, and zero otherwise. It is used to capture major energyloan exposure.
In addition to the 22 banks in the $10+ billion size group (Appendix Table AI), the two largest banks in the $5-10 billion
group were included-North Carolina National Bank Corp. and Republic New York Corp. They had !X2 values of 1.3
percent and -0.6 percent and Latin loan/capital ratios of .19 and .55, respectively (lead-bank loan exposure divided by
lead-bank primary capital).


( __ 0.4% + 1.1 %), or exactly in line with the
estimates of <X2 of 0.7 and 0.8 percent per month
for the other two bank groups.
These crude estimates imply that once the
major energy lenders are omitted, the estimate
of the. effect of Latin •American. loan. expoSures
(obtained from a cross-section estimate within
the $10 + billion size group) explains the <X2
differential between this group and the other
two groups. (Banks within the other two groups
tend to have little or no Latin American

It is illuminating to estimate how the stocks

of the 22 banks in the $10+ billion size group
might have performed had they had no Latin
American loans. For the 22 largest banks, the
average Latin loan/capital ratio is 1.02. This
suggests that La.tin American loansl'ladan.. impact on the average <X2 for the $10 + billion size
group of - 1.1 percent per month (- 1.03 x
1.02). The average estimated <X2 for the $10 +
billion banks in Table 3 was -0.4 percent per
month. Without Latin American loans, therefore, <X2 might have been 0.7 percent per month

IV. Conclusions
In the early post-1979 period (1979:101981:12), considerable uncertainty wasfound in
the estimated values of beta-a measure of the
sensitivity of equity returns to systematic, or
nondiversifiable, bank risk. By the latter part
of the post-1979 period (1982:01-1984:09), average beta values were close to the values that
prevailed over the 1972:08-1979:09 period. We
can conclude that the post-1979 period of economic and monetary uncertainty and financial
deregulation has had no significant impact on
average on the betas of bank holding companies with assets over $1 billion. The largest
banks ($10 + billion) still have betas that average well over 1.0, while the smaller banks
($1-5 billion) have betas that average well below 1.0.
Judging from stock price performance, we
can conclude that, at least until 1982, the post1979 economic, monetary, and deregulatory environment was not unfavorable on average to
banks with assets over $1 billion, and may have
been favorable overall to bank holding companies with assets ranging from $1 billion to $10
billion. However, since 1982 there has been a
sharp downward valuation, on average, in the
equities of the very large bank holding companies, those with assets over $10 billion.
Crude statistical analysis suggests that the negative performance of the $10 + billion bank
holding companies is explained by domestic energy loan losses and Latin American debt
Once the banks with very heavy loan losses
are removed from the sample of largest banks,

a highly significant relationship across banks
appears between negative stock performance
(negative alphas) in the post-1981 period and
total debt exposures to the Latin American
countries of Argentina, Brazil, Mexico, and
Venezuela. On average, this negative relationship is enough to account for the poor stock
performance of the 22 bank holding companies
with assets over $10 billion, taken as a group,
as compared to the 60 other bank holding companies with assets of $1 billion to $10 billion,
which tend to have little Latin American debt
There are two caveats to keep in mind regarding the evidence in this study. First, it
might be very misleading to extrapolate the results to smaller bank holding companies (say,
with assets under $500 million) or to thrifts.
These institutions normally have very different
portfolios and markets than do the large bank
holding companies. Second, it is plausible that
an increase in implicit regulation or government protection has affected the stock prices of
large bank holding companies since 1979. Certainly, the increase in the deposit insurance
limit from $40,000 to $100,000 in March of 1980
had a favorable impact. But we cannot be sure
of the market's perception, on balance, of other
changes in regulatory protection, such as the
explicit policy changes of the FDIC, first toa
partial payout on large deposits and then to de
facto protection of all deposits and even nondeposit liabilities in the case of Continental
Illinois. 15


ApPENDIX Table A1
Risk and Returns of Bank Stocks, $10+ Billion Banks, Reported Individually
1972:08-1984:09 with Shifts at 1979:10 and 1982:01
Bank Holding Company

BankAmerica Corp.
Chase Manhattan
Manufacturers Hanover
J. P. Morgan
Chemical New York
First Interstate
Bankers Trust
First Chicago
Security Pacific
Wells Fargo
Crocker National
Marine Midland Banks
Mellon National
Irving Bank Corp.
InterFirst, Dallas
Northwest Bancorp., Minn.
Texas Commerce, Houston
Republic of Texas, Dallas
First City Bancorp, Houston
NBD Bancorp., Detroit
Bank of New York Co.





0.5% -1.7% 1.2%1.19*
-0.1 -0.2
0.4 1.26*#
0.3 -0.6
1.0 1.20*
0.2 -0.6
0.1 1.16*
0.1 1.18*
0.0 -0.4 1.38*#
0.5 1.11*
0.2 -0.1
0.2 1.43*#
0.6 -1.1
0.9 1.51*#
0.4 -0.5 ~1.3 1.39*#
1.2 -0.1
0.1 -0.3 1.15*
0.7 -0.4
0.6 -2.7* 1.26*#
0.4 -1.1 -0.3 1.32*#
1.5 -0.2
1.3 -1.2 1.62*#
1.6 -2.7* 1.18*
0.4 -2.0*







- .48
- .46

.06 .40
.56* .31
-.28 .44
.52 .34
.54 .27
.09 .40
.01 .36
-.13 .38
.23 .46
.08 .30
-.44 .43
.60* .43
.31 .45
-.14 .33



Latin Loans2


Addendum (excluded from $10+ billion bank index)

Seafirst Corp.
(through 6/83)




Continental Illinois
(through 9/26/84)



-5.1* 1.30*





.78* .38




Alpha values and the standard error of the estimate are expressed as monthly percentage rates of change at monthly rates.
Asterisks denote significance at the 90-percent confidence level (one-tailed test for [30 and two-tailed tests for alphas and
for [3shifts). 13shift values and t-statistics are relative to the base period 130 value. Alpha values and t-statistics are relative to
zero. #Denotes that [30 is statistically different from 1.0 at the 90-percent confidence level (two-tailed test).
lEstimatespf beta in the boxed-in area are for 1979:10-81:12, excluding 1980:07-81:06 (see the note to Table 2).
2The ratio for First Interstate was supplied directly to the author by the holding company and it represents holding company
Latin American loans divided by holding company capital. The other ratios in the column must be read with caution. Latin
American loan figures are from the country loan exposure report; all of the institutions reported on a lead-bank basis
except Citicorp, Bankers Trust, NBD Bancorp, and Continental Illinois which reported loans on a holding company basis.
The denominator for all ratios in the final column, except that of First Interstate, is primary capital of the lead bank. The
proper ratio in all cases would treat both loans and capital on a holding company basis. BankAmerica owns Seattle First
Bank. First Interstate, InterFirst, Northwest, Texas Commerce, Republic of Texas, First City and NBD are all multibank
holding companies.
Source: Equation (2) in the text. Latin American loan exposure is total non-local-currency loans to Argentina, Brazil,
Mexico, and Venezuela as of March 31, 1984 (country loan exposure report, FFIEC-009(a)). Capital is primary
bank capital from the March 31, 1984, FDIC Call Report.


1. From Octoper 6, 1979, through mid- tolate-1982, the
FederaIHeseI"Ve'l:!9peratiI1QproceduresWer!!fqirectedexplicitly to\NardcontrQllingnonborrowed re~e"veS;O¥!!lr the
short-run while allowing tl1efederal funds rate to. fluctuate
over a relatively wide range. In other periods, the Fed has
tended to hold the short-run variability of the federal funds
rate· to within a narrowband.

6. The.$10+billion group includes the population of bank

tiQnsofiseafirs{ Corp., •which was acquired by BankAmerica Corp. in July 1983, andContinental Illinois ....,hich
failed in the Summer of 1984 and was re-formed in September 1984. (BankAmerica Corp. equity excludes Seafirst
prior to July 1983 andiincludesit thereafter;) Separate resuits are reported in Table Al for· Seafirst (through JUly
1983) and for Continental Illinois through September 26,
The $5-10 billion and $1-5 billion groups are samples of
the populations in those size classes, where the choice of
bank holding companies in each group depended on data
availability. Data are from the Data Resources DRI-SEC
database. They were screened by the author to correct
errors and to exclude bank holding companies for which
trading appeared to be infrequent.

The der!!l9ulationofbanking,particularlYdepOSit-rate deC
regulation and the extension of checkable •. deposits to
thrifts, was greatly accelerated by passage of the Depository Institutions Deregulation and Monetary Control Act of
March 1980 (DIDMCA). Passage of such a bill became
anticipated by the market as early as late-1979. Over the
1980-83 period, deposit rate ceilings on consumer accounts at panks and thrifts. were effectively eliminated. The
Depository Institutions Deregulation Act of October 1982
(Garn-St Germain Act) furthered deposit rate der!!l9ulation
by requiring that the Depository Institutions Deregulation
Committee create an account at banks and thrifts that
would be competitive with money market mutual funds. The
Act also gave the regulators more leeway in arranging
takeovers of failing banks and thrifts, and put into legislation much of the deregulation of thrift holding company
powers that had been implemented by the Federal Home
loan Bank during 1982.

7. Because dividends of the bank stocks were not available, returns throughout (inclUding the S&P 500) exclude
dividends. To the extent that bank stock dividends differed
from those of the S&P 500, total return differentials would
differ from those implied by Figure 1. The author can only
infer the possible bias introduced by omitting dividends
from the fact that the estimated alphas for the three bank
groups over the full period are very close to zero. This fact
suggests that the omission of dividends does not affect
average return differentials significantly. See Footnote 5.
Estimated mean alpha values over the full 1972:081984:09 period for the three bank groups are 0.1 ($10+
billion banks), 0.2 ($5-10 billion banks), and 0.1 ($1-5
billion banks), with median t-statistics of .27, .30, and .30

2. For evidence on the effect of regulatory protection, see
Brickley and James, 1984.
3. In theory, the market bundle of risky assets should include bonds, real estate, and other forms of wealth. However, empirical tests of the market model almost always
use a broad stock market index such as the S&P 500 because reliable market-value indices of other risky assets
are not available.

8. Since the paper focuses on the average results of individual banks, the regressions are run on individual bank
data and the mean coefficients of the individual bank
regressions are reported for each group. Grouping the
banks into portfolios and then running one regression for
each group would seriously overstate the t-statistics because grouped data would reduce the standard errors by
diversifying away much of the variance in individual bank
data. Median test statistics (R2, Cl'e' and t-statistics) are
reported for the same groups of individual bank regressions. Medians are used for test statistics because mean
test statistics are not appropriate for confidence tests such
as the Hest.
9. As noted in the footnote to Table 3, the median beta
value is insignificantly different from 1.0.

4. Dividends paid out during a particular month should be
included with price returns to obtain total returns. Because
of data limitations, dividends are omitted from the bank
stock returns throughout the study. For consistency, they
also are omitted from S&P 500. The exclusion of dividends
affects average returns and estimates of alpha, but it has
little effect on the estimated betas because almost all of
the monthly variations in total stock returns are in the
The capital asset pricing model specifies that the relationship between returns in Equation (1) will be linear as long
as returns are specified in excess of the "risk free" rate.
Thus, all returns of the bank stocks and the S&P 500 in
the empirical analysis are net of the return from holding to
maturity Treasury bills which, at the beginning of the month,
have only one month left to maturity. The Treasury bill with
one month to maturity gives the purest risk-free one-month
rate of interest because it is free of default and interestrate risk. Data for the one-month risk free rate through 1982
are from the CRSP database, University of Chicago, with
1983-84 upqates constructed by the author.

10. For earlier hypotheses and tests of the beta for large
money center banks, see Beebe, 1977 and 1983.
11. Beta would rise if the default changed the systematic
risk-sensitivity of the remaining portfolio. It also would rise
somewhat because the market value of capital would decline and hence capital leverage would rise. For a given
systematic risk of bank assets and liabilities, beta of the
bank's equity is sensitive to equity leverage.

5. Wh!!lre dividends are ignored, the ex ante expected
value of alpha will be the dividend differential between the
bank stock and the S&P 500.

12. The longer the estimation period for alpha, the more
likely the estimate is to be zero. There are two reasons for


this result: (1) any mispricing of securities (that is, market
inefficiency) is likely to be very short-lived; and (2) the
longer the time period, the less likely it would be for new
information to be correlated. For isolated banks,
alpha estimates in Table 3 are significant in some cases.
See Appendix TableM.

Beebe,Jack, "A Perspective on Liability Management and Bank Risk," Federal Reserve
Bank of San Francisco Economic Review,
Winter 1977.

• leVel~fQftheindi()l:lsinfig­
urE;l2 are Calculated by setting the price •level. in 1981 :12
equal to 100 and then cumulating the. monthly market-related returns that are derived from Equation (3).
14. The 24 banks in the cross-section sample include the
22 banks in the $10+ billion size group (Appendix Table
A1) plus North Carolina Bank Corporation and Republic
New York Corporation, which are the two largest banks in
the $5~ 10 billion group. Seafirst and Continental Illinois
are excluded from the sample.
15. For an analysis of the market's perception of spillover
effects from Continental Illinois, see Furlong, 1984.

.•.•• .•·•· .• ·.··l3ankcapita'fliskinthfl. Ppst-l!:j79
M()netary .and •Regulatory Environment,"
FederalReserve Bank of San Francisco
I;conomicReview, Summer 1983, NO.3.
Brickley. James A. and Christopher M. James,
"Access to Deposit Insurance and the Stock
Returns. of Financial Institutions," unpubliShed draft (Universities of. Utah and Oregon,respectively), August 1984.
Furlong,Frederick, "Market Responses to Continental Illinois," Federal Reserve Bank of
$anFrancisco Weekly Letter, August 31 ,
Sharpe, William, "A Simplified Model for Portfolio
Analysis," Management Science, January
- - , "Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of Risk,"
Journal of Finance, September 1964.


Adrian W. Throop*

Although a major objective of current fiscal policy is to stimulate
capital formation and productivity growth, the policy is internally
inconsistent for this purpose. On the one hand, investment in plant
and equipment has been promoted by accelerated depreciation allowances and liberalized investment tax credits. On the other, the
cost of capital is being raised by the impact of large federal demands
for credit on interest rates. Econometric simulations of the effects
of alternative fiscal policies indicate that the net effect of current
fiscal policy is actually to stimulate consumption rather than
saving were forthcoming. The Economic Recovery and Tax Act of 1981 therefore also contained a reduction of personal income tax rates
by a cumulative 23 percent over three years,
partly on the theory that the resulting decline
in marginal tax rates would stimulate a large
increase in the private saving rate.
The actual outcome has been quite different,
however. "Bracket creep," caused by
nominal incomes, made the actual cut in marginal tax rates for households considerably less
than 23 percent. Also, a rise in the after-tax
return to saving due to tax cuts may theoretically either increase or decrease the saving rate;
and since 1981 the private saving rate (personal
plus business) has been relatively stable. 1 Finally, the large budget deficits resulting from
tax cuts and spending increases has meant a
fall, rather than a rise, in the national saving
rate (private plus government). As a result, real
interest rates have been bid up, and capital has
been attracted from abroad. Although foreign
capital inflows reduce the pressure on domestic
real interest rates, the fact remains that these
higher real rates have tended to offset the stimulatory effect of the tax incentives for business

In 1981, the Reagan Administration embarked upon a bold program for dealing with
the problems of high inflation and stagnant economic growth. Over the two previous decades,
the inflation rate in the U.S. economy had risen
from near zero to double digit levels, and the
rate of growth of labor productivity had fallen
from an average of around 2.5 percent per year
in the 1950s and early 1960s to about 0.5 percent in the late 1970s. The Administration's
program consisted of tax and spending reductions as wen as regulatory reform to stimulate
saving, investment and work effort, and a commitment to monetary stability to bring the rate
of inflation down.
A main feature of the Administration's fiscal
policy was the set of tax incentives for business
investment provided in the Economic Recovery
and Tax Act of 1981. However, these tax incentives alone could not stimulate more capital
formation in plant and equipment without reducing other kinds of investment unless greater
*Senior Economist, Federal Reserve Bank of
San Francisco. Research assistance was provided by Sharon Tamor and Roger Weatherford.

To provide a perspective on the net impact
that the Reagan Administration's fiscal policy
is actually having, this article measures the permanent effect of fiscal policy on investment and
consumption by simulating its effects with an
econometric model. The simulation results indicate that current fiscal policy is actually "proconsumption" rather than "pro-investment."
Because the private saving rate is estimated to
be only modestly affected by the tax cuts, the
net effect of the tax cuts has been to stimulate
consumption. The expansion in consumption is
being financed largely by borrowing from
abroad, with the counterpart of this being a decline in net exports. The effect on business investment in plant and equipment is about neutral, as increases in real interest rates almost
exactly offset the stimulus to investment. Current fiscal policy, however, is not neutral in its
effects on other types of investment as it is tending to reduce residential and inventory
The article is organized as follows. Section I
describes a procedure for estimating the permanent effect of fiscal policy on the consumption-investment mix. In Section II, the change
in fiscal policy occurring since 1981 is measured

in terms of changes in average tax rates, marginal tax rates, and expenditures of the federal
government. In Section III, we provide a
thumbnail sketch of the econometric model that
is used for silllulating the effect of this change
i~fis~al policy. This description stresses the responses of various sectors to real after-tax interest rates. A more detailed description of the
model can be found in the Appendix. Although
this model is relatively small, its key relationships are similar to those embodied in most
large-scale structural econometric models. In
Section IV, we perform two experiments in
counterfactual history, corresponding to two alternative fiscal policies that might have been
followed. The first of these shows the effect on
the consumption-investment mix of a continuation of fiscal policy as it existed at the beginning of 1981. In addition, we consider the consumption-investment mix resulting from the
same tax cuts for business, but with no change
in the aspects of fiscal policy that affect federal
spending on goods and services and personal
taxes and transfer payments. Lastly, in Section
V we present a summary of the results and
some policy conclusions.

I. Theoretical Analysis of the Effect of Fiscal Policy
When analyzing the effects of fiscal policy on
the consumption-investment mix of the economy, it is important to distinguish between possible temporary impacts and permanent ones.
The temporary effects of an increase in the fiscal deficit depend upon the policy response of
the Federal Reserve. Increased demands for
credit flowing from a larger fiscal deficit put
upward pressure on real interest rates. The
Federal Reserve can temporarily alleviate this
pressure by supplying more funds to the credit
markets through the creation of a larger stock
of money. At some point, however, the expansion of aggregate demand resulting from the
monetary accommodation of fiscal deficits generates a higher level of prices, which then reduces the real stock of money and pushes real
interest rates back up. Given the economy's capacity to produce, the level of real activity will

ultimately be no different, but real interest
rates will be permanently higher. Alternatively,
the inflationary effects of the fiscal deficit can
be avoided if the Federal Reserve immediately
reduces the stock of money by enough to move
real interest rates to their equilibrium leveL
Over a longer span of time, the economy's
capacity to produce will be altered by the effect
of fiscal policy on supplies of capital and labor.
Of particular importance is the supply of capital
per worker, and hence real income per capita.
To enhance growth in the stock of capital per
worker, fiscal policy must raise investment relative to consumption.
The problem can be analyzed more pre~cisely
within the context of an explicit macroeconomic model. For this purpose, we utilize the
well-known IS-LM framework. The IS curve
defines the equilibrium conditions that deter20


mine the demand for goods, and the LM curve
shows the conditions corresponding to equilibrium between the supply and demand for
money. By Walras' law, equilibrium in the markets for goods and money implies equilibrium
inflation is
aUov.'ed to vary in this analysis, for simplicity
and without loss of generality we abstract from
the effects of changing inflationary expectations and,
inflation pn~m:iunls
in interest rates. 3
Real rather than nominal interest rates drive
real aggregate spending. But since movements
in real and nominal interest rates are assumed
to be the same in this analysis, the equilibrium
level of real GNP depicted by the IS curve depends only upon the nominal interest rate.
Also, the IS curve includes the adverse effect
of higher interest rates on net exports, and
hence aggregate spending, that operates indirectly through an appreciation of the exchange
With respect to the LM curve, the demand
for real money balances depends upon the nominal interest rate and real GNP. Given the nominal stock of money determined
the monetary authority, the real stock of money varies
inversely with the price leveL The LM curve
assumes a given stock of real money balances
and shows combinations of the nominal interest
rate and real GNP that generate an equilibrium
between the supply and demand for real money
The permanent effect of fiscal
upon what happens to the consumption-investment mix at the fun employment level of real
GNP. In Figure 1, we focus on an economy initially operating at a fun employment level of
GNP, denoted by Yr. Suppose that cuts in personal and business taxes similar to those undertaken by the Reagan Administration result
in a larger budget deficit. The higher level of
consumption and investment
at any
interest rate shifts the IS curve to the right,
from lSI to IS2. If the nominal stock of money
is unchanged, the economy will move to a
higher interest rate and real GNP at iz and yz.
But that is
a temporary adjustment. Because the economy is operating beyond full em-

Figure 1
Interest Rate



Y1 Y2 Y3
Real GNP
ployment, an excess demand for labor drives
up wages and prices until the excess demand is
eliminated. As wages and prices rise, the real
stock of money declines, and the LM curve
shifts to the left, from LMI to LM2. Thus, the
permanent equilibrium generated by the new
fiscal policy is at the same level of real GNP
and a higher interest rate, or Yl and i3 •4 In this
equilibrium, the decrease in expenditures produced by higher interest rates exactly offsets
the original stimulus to expenditures from the
tax cuts. 5
The net effect on the composition of aggregate demand depends upon the interest-sensitivity of various types of expenditures in comparison to the size of the stimulus that they
received from the tax cuts. For example, if consumption benefits from the tax cuts but is not
at all sensitive to interest rates, various types
of investment would have to contract even if
they had been stimulated by the tax cuts. Or if
only one type of investment spending benefited, the major burden of rising interest rates
would have to faU on other types of investment.
The Federal Reserve can do nothing to
change the permanent effect of fiscal policy on
the real economy. If it tries to peg the interest
rate at i 1 to prevent a decline in interest-sensi22

tive expenditures, it only generates a greater
amount of inflation. At LM3, the nominal
stock of money would initially be higher than
it was at LM1. The process of inflation would
then proceed until the real stock of money faUs

sitive components of aggregate expenditure at
the full employment level of real GNP. This
permanent effect of a change in fiscal policy is
associated with the impact on expenditures of
the difference in real interest rates between the

to LM2. Although the increase in the price
level would be greater than before, the permanent level and composition of GNP would
be unaffected. Alternatively, inflation can be
avoided altogether if the Fed reduces the nominal stock of money to shift the LM curve to
LM2 immediately. Whatever the monetary action, the permanent effect of a change in fiscal
policy on the composition of GNP is found at
the level of interest rates that generates the
same level of real GNP as before.
This analysis points the way to a procedure
for measuring the permanent effect of fiscal
policy on the economy. The permanent effect
of a larger fiscal deficit is to raise the level of
real interest rates and impact upon interest-sen-

composition of expenditures at full employment
is approximately the same as that which would
occur at neighboring values of GNP.
To simulate the permanent effects of alternative fiscal policies that might have been followed in the 1981-84 period, we therefore allowed changes in fiscal policy to alter the
composition, but not the level, of real GNP at
each point in time. Real interest rates and the
exchange rate are allowed to adjust to generate
unchanged levels of total real spending and real
GNP. The resulting changes in the consumption-investment mix, at historical levels of real
GNP, then become an approximate measure of
the permanent effect of fiscal policy on the

U. Measuring the Permanent Effect of a Change in Fiscal Policy
a component of their real after-tax interest
rates and, therefore, affects expenditures on
consumer durables and housing. Similarly, corporate taxes influence the real cost of capital in
the business sector and, hence, expenditures on
inventories and nonresidential fixed investment. An unchanged fiscal policy would not alter the marginal tax rates that affect these expenditures, and would not shift the IS curve for
this reason.
Second, an unchanged fiscal policy requires
federal outlays and receipts not to change as a
fraction of GNP at a high level of employment.
With unchanged government receipts and expenditures, as well as unchanged effective marginal tax rates, there would be no shift in the
IS curve. Thus, the composition of aggregate
demand and output would not be affected by
fiscal policy.
It might appear that there could be inconsistencies in this dual criteria for an unchanged
fiscal policy. For example, if marginal tax rates
are higher than average rates, as in fact they

This article looks at fiscal policy in terms of
its effects, as opposed to the specific instruments of policy in the form of laws. Fiscal policy defined in terms of its effects may be altered
even when there are no legislated changes. The
increase in taxes as a proportion of GNP that
occurs as a result of normal economic growth,
and also from inflation in the absence of tax
indexing, are examples. Conversely, legislative
changes may be required just to keep the effects
of fiscal policy from changing as, for example,
when taxes have to be cut in order to keep revenues from
as a fraction of GNP. In this
context, an unchanged policy is one with an unchanged impact on the composition of economic activity at a high level of employment.
From a macroeconomic point of view, an unchanged fiscal policy has two dimensions.
First, there should be no change in effective
marginal tax rates that would alter economic
incentives. In the structural model of aggregate
demand that we use for the policy simulations,
the average marginal tax rate for households is

generally are, normal growth in the economy
with fixed marginal rates would tend to raise
tax receipts as a proportion of GNP. However,
an unchanged fiscal policy-one with a neutral
effect on the composition of GNP over time~
could· be maintained· by reducing .average tax
rates without changing marginal rates. In the
case of personal income taxes, this could be
done by increasing the standard deduction. The
extent of progressivity in the tax structure is
much less for corporations, but here too, the
average tax rate could be reduced without
changing the marginal tax rate on the cost of
new investment.

top marginal individual income tax rate was reduced from 70 percent to 50 percent. However,
the resulting change in the average marginal tax
rate for individuals was smaller because of the
bracket creep caused by rising nominal incomes
overthethree·year period; The tax cuts and
bracket creep combined to reduce the average
marginal tax rate for individuals from 30.4 percent to 27.1 percent by 1984. 6 In the experiments in counter-factual history described in
Section III, an unchanged fiscal policy is simulated by holding the average marginal tax rate
for households at the 30.4-percent level from
1981 through 1984.
The Tax Act of 1981 also contained substantial reductions in effective tax rates on the cost
of business fixed investment without, however,
changing the corporate tax rate on net income.
These tax cuts applied not only to business
plant and equipment, but also to rental housing. First, the Accelerated Cost Recovery System (ACRS) was introduced, which replaced
the previous system of basing tax lives on expected useful lives.? For most assets, the new
tax lives are considerably shorter than their

Marginal Tax Rates
The first dimension of fiscal policy that we
consider is changes in marginal tax rates. As
shown in Table 1, rising nominal incomes combined with a progressive tax system raised the
average marginal personal tax rate from 21.2
percent to 30.4 percent between 1965 and 1980.
The Economic Recovery Tax Act of 1981 reduced personal income tax rates by a cumulative 23 percent over three years. In addition,

First Dimension of Fiscal Policy

Average Marginal
Tax Rate for the
Individual Income Tax

Effective Corporate Tax Rate on Cost of
Equity Financed Investment1















IThis effective tax rate equals 1- u
1 , as discussed in Box 2.
Sources: Barro and Shahasakul (1983), Board of Governors of the Federal Reserve System, and Data Resouces, Inc.


economic lives. Second, the 1981 Tax Act increased tax credits on investment in equipment. 8 The Tax Equity and Fiscal Responsibility Act of 1982 took back part, but by no means
all, of these tax cuts for business as part of a
package to reduce the size of the federal budget
The extent of the changes in the effective tax
rate on the cost of business fixed investment
before and after the Reagan tax cuts are shown
in Table 1 for equipment, commercial and industrial structures, and rental housing. 9 In the
1950s, the effective tax rates on different types
of business fixed investment were fairly similar.
However, in the 1960s and 1970s, large disparities developed. The effective tax rate on investment in equipment dropped as a result of
legislated changes, while tax rates on investment in commercial and industrial structures
and rental housing went up due to reductions
in the present value of depreciation caused by

the higher nominal interest rates accompanying
higher inflation. The Tax Act of 1981 reduced
effective tax rates on the cost of equity-financed
capital investments by 6 to 20 percentage
points, though it did little to remove the large
disparities between rates on different classes. In
Section III, an unchanged fiscal policy for the
1981-84 period is simulated by keeping effective tax rates on the various classes of business
fixed investment the same as they were at the
end of 1980.
Government Spending and
Disposable Income

When simulating the effects of fiscal policy
changes, it is necessary to consider that observed movements in federal outlays and receipts are partly due to changes in the level of
economic activity and partly due to other factors. The federal budget measured on a high
employment basis removes the cyclical varia25

tions in outlays and receipts caused by deviations from a.specified high employment rate of
unemployment. 12 The changes in outlays and
receipts that are left are attributable to normal
growth in theeconomy,inflation, or legislated
lays and receipts that deviate from those required to maintain a constant proportion to
high •employment GNP constitute •the second
The federal government's high employment
deficit rose from 0.9 percent of high employment GNP in 1981 to 1.7 percent in 1982, and
to 2.5 and 3.0percentin 1983 and the first half
of 1984, respectively. However, for an unchanged fiscal policy, federal spending on goods
and services, transfer payments, and taxes

should be kept at unchanged proportions of
QNP measured on a high. employment basis.
Thus, in. simulating an unchanged fiscal policy
for 1981--84, federal spending on goods andservices was reduced by the difference between
actual spending and • whatspending wpqld have
been if its ratio to high employment GNP at the
beginning .. of • 1981 .had.been maintained. As
shown in Table 2 the required adjustment is
small-less than $5 billion, in 1972 dollars, in
all but one quarter. 13
A similar procedure was used for. adjusting
the level of personal taxes and transfer payments, and consequently disposable personal
income. Items in the federal budget that affect
the difference between personal disposable income and GNP are separated into two com-

Second Dimension of Fiscal Policy

Adjustments for
Unchanged Fiscal Policy
and Tax Cuts
for Business Only
Federal Government
Spending on
Goods and Services





Adjustment for
Tax Cuts
For Business Only



State and local
Spending on
Goods and Services


Percent of


Percent of



























Adjustment for
Fiscal Policy

Percent of



Percent of







ponents----{;orporate profits taxes and everything else. Corporate profits taxes must be
treated separately because of the interaction
between them and dividend payments, which
feed back to personal income. Lower corporate
profits taxes • increase ··disposable··perscmalilF
come dollar-for-dollar if the resulting increase
in after~tax corporate profits is all paid out in
dividends, but have no impact on disposable
personal income if it all goes into retained
In recent years, dividends have averaged
one-half of after-tax corporate profits. Ordinarily, an increase in profits must be sustained
for several years for this 50-percent payout ratio
to be fully realized. However, if firms believe
an increase in profits is permanent because it
has been caused by a change in the tax law, then
the speed of adjustment would be quicker. Statistical analysis reveals that the Reagan tax cuts
have, in fact, operated in this way. The ratio of
dividends to after-tax profits was not significantly depressed in the 1981-84 period even
though profits soared.
As one component of an unchanged fiscal
policy, the ratio of corporate profits taxes to
GNP on a high employment basis should remain the same. To simulate this part of an unchanged fiscal policy, we reduced the historical
series for disposable personal income by 50 percent of the difference between historical corporate profits taxes and what they would have
been if their ratio to GNP in 1981 had been
maintained on a high employment basis. This
procedure is consistent with the observation
that the payment of dividends actually adjusted
fairly quickly to maintain the desired long-run
payout ratio.
The remaining items in the federal budget
that contribute to the difference between disposable personal income and GNP tend to alter
disposable income dollar-for-dollar. These aggregate to the sum of personal income taxes,
contributions to social security, and federal indirect business taxes less federal government

transfer payments to persons (including interest
payments) and net subsidies to federal governmententerprises. To simulate an unchanged fiscal •policy, the· historical series on· disposable
personal income is, therefore, reduced further

by thedifferellcebetweell.··this aggregate and
what it would have been if its ratio to GNP at
the beginning of 1981 had been maintained on
a· high employment .basis.
As shown in Table 2, the adjustment to disposable personal income becomes very large by
1984. In fact, except for the earlier part ofthe
1981-84 period, the entire stimulus to aggregate
demand from the Reagan economic program
has come from its impact on taxes and transfer
payments,as· opposed to federal spending· on
goods and services. With an unchanged fiscal
policy, disposable personal income would have
been $39.1 billion lower, in 1972 dollars, by the
second quarter of 1984; this difference equals
2.3 percent of GNP.
An additional adjustment is required for federal grants-in-aid .to state and local governments. Whereas, for an unchanged fiscal policy
they would stay at the same ratio to GNP as at
the beginning of 1981, by the second quarter of
1984 actual grants-in-aid were $11.2 billion, in
1972 dollars, less than this measure. About 20
percent of this reduction took the form of a
decline in payments to persons, while the remainder was for spending on goods and services. It is assumyd that during the time period
studied, other spending and taxes at the state
and local level were not affected. Therefore, for
stimulation of an unchanged fiscal policy the
portion of the adjustment for grants-in-aid
going to persons is added back into personal
disposable income, and the remaining portion
is added to state and local spending on goods
and services. Table 2 shows this adjustment to
state and local government spending on goods
and services and also includes the effect of federal grants-in-aid in the adjustment to disposable personal income.


III. A Structural Model of Real Aggregate Demand
In this section, we provide a thumbnail
sketch of the econometric model that is used
for simulating the permanent effects of fiscal
policy,streSsingtesponseSof thevariolls Sectors
to real interest rates. A more detailed presentation is provided in the Appendix. 14 The theory underlying the model follows. the mainline
neo-Keynesian view embodied in most largescale structural econometric models, with particular attention being paid to the way that real
interest rates enter into the cost of capital for
specific types of investment. In the short run,
the slow speed of adjustment of wages and
prices allows monetary policy to influence real
interest rates, which, in turn, are a prime
mover of aggregate demand through their impact on various types of investment expenditures. Further effects on demand occur through
changes in consumption spending induced by
changes in income and accelerator effects on
investment expenditure. However, in the longrun real interest rates are determined by the
balance between saving and investment.
The model can be solved for an equilibrium
level of real aggregate demand and output,
given the level of real interest rates and other
exogenous variables. Used in this way, it can
forecast real GNP and its components on the
basis of a projected path for real interest rates.
Alternatively, it can be used with a separate
aggregative forecast of real GNP and prices to
make forecasts of sectoral activity and the level
of real and nominal interest rates. 16 For this
second purpose, one solves for the path of nominal and real interest rates that produces the
projected path of real GNP. Our exercises in
counterfactual fiscal history employ the latter
approach. We assume the path of real GNP to
be unaffected by alternative fiscal policies since
we are interested only in permanent effects.
A schematic overview of the structural model
of aggregate demand is provided in Figure 2.

The components of expenditure that are affected directly by real interest rates include inventory investment, consumer expenditures on
dllrablegoods, ··l"eSiderttial. <construction, .<and
nonresidential fixed investment. Net exports
areinfluenced indirectly through the impact of
the. differentictl between U.S. and foreign real
interest rates on the real exchange rate. The
components of expenditure that are not affected by real interest rates are government
spending on goods and services and consumer
expenditures on nondurables and services. The
latter depends only on permanent disposable
income, and the former is an exogenous policy
An increase in the fiscal deficit is associated
with increases in government spending, consumption (through personal tax cuts), or iuvestment (through business tax cuts), or some
combination of these. With a given level of real
GNP, an amount of interest-sensitive private
spending equal to the spending generated by
the increase in the fiscal deficit must be
"crowded out" by a rise in interest rates. Interest rates rise because of the government's extra
borrowing in the credit market. The sectors of
domestic investment that get "crowded out" the
most by a fiscal deficit are those that are most
sensitive to iuterest rates. And whether crowding out falls more heavily on domestic investment or foreign investment depends upon the
response of the international value of the dollar
to real interest differentials between the United
States and other countries.
In the Appendix, we discuss the estimated
responses of the various sectors of aggregate
demand to real interest rates and the real exchange· value. of the dollar. An understanding
of the model will help in following the results
of alternative fiscal policy simulations. However, those who are not interested in further
details at this point can skip the Appendix.



IV. Simulations of Alternative Fiscal Policies
variables are the 6-month commercial paper
rate, the average yield on newly issued AA corporate bonds, the S&P earnings to price ratio
oncommon stocks, and the real exchange value
of the V. S. dollar. The real variables of interest
include total personal consumption expenditures, residential fixed investment, nonresidentiaLfixed investment, and net exports. These
real variables are all measured in 1972 dollars.
The 6-month commercial paper rate directly affects spending on consumer durables and residential construction, and indirectly affects nonresidential fixed investment through the yield
on bonds and the return to equity. In addition,
the real short-term interest differential between
V.S. and foreign markets drives the real exchange value of the dollar, which, in turn, affects net exports.

Simulations of the permanent effects of alternative fiscal policies are summarized in
graphical form in this section. The historical
errors in each equation of the econometric
model of aggregate demand were first added
back in to allow a simulation of the model to
replicate history exactly. Then, simulations of
two alternative fiscal policies were performed,
allowing interest rates and the foreign exchange
value of the dollar to adjust in such a way that
real GNP would be unaffected in each period. 17
The first of these simulations is for an unchanged fiscal policy. It holds marginal tax rates
constant at their values at the end of 1980, corresponding to the data in Table 1, and also
makes the adjustments to government spending
on goods and services and disposable personal
income shown in Table 2. Spending of state and
local governments on goods and services increases about $9 billion, in 1972 dollars, by the
first half of 1984 because of increased federal
grants-in-aid. But there is virtually no change
in federal spending on goods and services. Disposable personal income is reduced by over $40
billion, in 1972 dollars.
The second of the simulations considers the
degree to which investment spending on plant
and equipment would have been stimulated if
the 1981 tax cuts had been limited to the business sector only. For this simulation, the marginal tax rate for households is held constant at
its value for the end of 1980, but marginal tax
rates on investment in equipment, structures,
and rental housing take on their actual values.
The adjustment to government spending on
goods and services is the same as in the first
simulation; and, as in that simulation, disposable personal income is adjusted downward by
the amount of the tax cut that households otherwise would have received. However, the
downward adjustment to disposable personal
income is less than in the first simulation because business tax cuts raise after-tax corporate
profits, and hence dividend payments.
We focus particularly on the effects of these
changes on four financial variables and four
real variables in the economy. The financial

An Unchanged Fiscal Policy
During 1981, there was little difference between the effects of the Reagan Administration's fiscal policy and those of an unchanged
fiscal policy. The 5-percent cut in personal taxes

Chart 1A
6-Month Commercial
Paper Rate


Fiscal Policy









in October was largely offset by bracket creep,
and the business tax cuts had not yet begun to
stimulate business investment. spending. Consequently, as shown in Chart lA,the commercial paper rate corresponding to an unchanged
fiscalpo!icy is little different froIl1 the adual
rate during the initial period. After the middle
of 1982, however, the effects of the Reagan
Administration's fiscal policy became more evident. Personal income tax rates were cut by 10
percent in July of 1982 and again in July of
1983. And the liberalization of depreciation
rules and the investment tax credit began to
affect business investment. With an unchanged
fiscal policy, the 6-month commercial paper
rate in the simulation fans 6V2 percentage
points below its historical value by the third
quarter of 1982. The effects of interest rates on
expenditure build over time, however, so that
the difference between the simulated commercial paper rate and its actual value then shrinks.
By the first half of 1984, the simulated commercial paper rate corresponding to an unchanged fiscal policy averages about 4 percentage points less than the actual-at around 6
percent instead of 10 percent (on a discount

Chart 1C
Earnings-Price Batio
on Common Stock





Fiscal Policy



Unchanged ~
Fiscal Policy





In this simulation, real after-tax interest rates
are reduced to about the same level as in earlier
business cycle expansions. For example, using
the real after-tax 6-month commercial paper
rate as a gauge, its nominal level on a discount
basis would be 6 percent rather than 10 percent
by the first-half of 1984. On an annual yield
basis, and after taking into account the deductibility of interest costs from personal· income
taxes, .its nominal level would be 4.5· percent
instead of 8.1 percent. Subtracting an expected
inflation of around 4.5 percent gives a •real
after-tax commercial paper rate close to· zero
instead of its actual value of 3.6 percent. A year
into prior postwar business cycle expansions,
the real after-tax 6-month commercial paper
rate was also near zero, averaging 0.1 percent.
Thus, current fiscal policy is having a marked
effect upon real after-tax interest rates . Furthermore, if fiscal policy had not been changed
in 1981, the permanent effect of this would
have been toreduce real interest rates to historicaHy Mrmallevels.· This fact sugggests that
the Federal Reserve has not anowed the fiscal
stimulus to generate much more of an increase
in real GNP than would otherwise have ·occurred, consistent with a policy of offsetting the

Chart 1B
Yield on
Corporate Bonds (AA)


Fiscal Policy


Chart 2A
Personal Consumption

Chart 10
Real Exchange Value
of U.S. Dollar
1980-1982 Average



Billions of 1972 Dollars




Fiscal Policy







inflationary effects of the fiscal stimulus. IS
In the model, arbitrage in financial markets
transmits the reduction in the commercial paper rate to the yield on corporate bonds and
the required return on equity. By the first half
of 1984, an unchanged fiscal policy would have
reduced the AA corporate bond rate by 2.7 percentage points (Chart 1B) and the earnings-toprice ratio on common stocks by 3 percentage
points (Chart Ie). Because prices are unchanged, the reduction in short- and long-term
interest rates corresponds to a decline in the
differential between U.S. and foreign real interest rates. This produces a substantial reduction in the real exchange value of the dollar,
even though the effect on the interest differential is diminished by the tendency, included
in the simulation, of foreign central banks to
match part of the movement in U.S. real interest rates. By the first half of 1984, with an unchanged fiscal policy, the real exchange value
of the dollar in the simulation would be nearly
15 percent lower than it otherwise would have
been, bringing it back to the levels of late 1981
(Chart 1D).
The effects of these changes in financial vari-




abIes on the real variables of interest are shown
in Chart 2A-D. According to the simulation, an
unchanged fiscal policy reduces personal consumption expenditures by about $24 billion by
the first half of 1984 (Chart 2A).19 This result
is due to the effect of lower disposable income
under an unchanged fiscal policy, which
strongly dominates the effect of lower interest
rates on spending for consumer durables.
By far, the largest offset to the reduction in
total consumption expenditures with unchanged fiscal policy is net exports; they are
about $14 billion higher by 1984 (Chart 2B).
The higher net exports are produced by the effect of lower interest rates in depreciating the
exchange value of the dollar. It is important to
recognize, however, that for reasons unrelated
to fiscal policy (for instance, expectations by
foreign investors of greater monetary stability
in the U.S.), the dollar appreciated by twice as
much between 1980 and 1981 as it did from 1981
and 1984. The lagged effects of this earlier appreciation plus other factors, such as the cyclical position of the U.S. and the effect of LDC
debt burdens in reducing the demand for U.S.


first—about a 12-percent increase. However,
tending to offset this interest rate effect is the
reduction in disposable income associated with
an unchanged fiscal policy. Over time, the re­
duction in disposable income grows, and the
impact on interest rates weakens. By the first
half of 1984, the stimulus to residential invest­
ment from an unchanged fiscal policy is only $2
Offsetting forces are also at work in the case
of nonresidential fixed investment (Chart 2D).
The real cost of capital investment in this sector
is equal to a weighted average of real debt and
equity costs plus the physical rate of deprecia­
tion, all multiplied by one plus the effective tax
rate. The average effective tax rate on the cost
of capital for nonresidential fixed investment
was reduced from 27 to 16 percent by the 1981
Tax Act. An unchanged fiscal policy would
have rescinded this tax cut. But offsetting the
effect of higher business taxes is a lower yield
on corporate bonds and a lower required return
to equity due to the generally lower levels of
interest rates. This offset is nearly complete
throughout the simulation period. By the first

exports, also depressed net exports. As a result,
even with an unchanged fiscal policy, net ex­
ports would have fallen by $40 billion—or
nearly 3 percent of GNP from the beginning of
1981 to the first half of 1984. Thus, although
fiscal policy has had a significant effect on net
exports, we estimate that it accounts for only
one-third of the decline in this sector since the
beginning of 1981.
We turn next to residential fixed investment
(Chart 2C). The increase in personal tax rates
associated with an unchanged fiscal policy di­
rectly reduces the real after-tax interest cost of
mortgage finance, but at the same time, also
reduces personal disposable incomes. Besides
these direct effects, an unchanged fiscal policy
lowers the cost of mortgage credit still further
through its indirect effect in lowering interest
rates in general. The initial effects of an un­
changed fiscal policy on interest rates are larger
than the ultimate effects because of the lagged
response of investment spending and net ex­
ports to interest rates. The relatively large ini­
tial reductions in interest rates stimulate resi­
dential investment to the extent of $5 billion at

Chart 2C
Fixed Investment

Chart 2B
Net Exports
Billions of 1972 Dollars

Billions of 1972 Dollars


Tax Cuts for Business Only
As the first simulation clearly shows, the ef­
fects of reductions in taxes and increases in
transfer payments for households have over­
whelmed the tax incentives provided to busi­
ness. Personal income tax cuts have increased
the federal government’s demands for credit by
more than they raised private saving, thus put­
ting upward pressure on interest rates. The re­
sult is that real debt and equity costs have risen
by enough to offset the reduction in the cost of
capital investment in plant and equipment that
would otherwise have occurred.
Since a major objective of current fiscal pol­
icy was to promote capital formation and pro­
ductivity growth, it is interesting to examine the
extent to which business investment would have
been stimulated if current fiscal policy had not
included any net tax benefits for households.
This is the purpose of the second simulation,
which mirrors an unchanged fiscal policy except
that effective tax rates on the cost of business
investment are reduced and corporate profits
are allowed to increase correspondingly.
The results of this policy of tax cuts for busi­
ness only on the commercial paper rate is con­
trasted with the results of an unchanged fiscal

Chart 2D
Fixed Investment
Billions of 1972 Dollars

half of 1984, nonresidential fixed investment is
actually very slightly higher, by $0.6 billion,
with an unchanged fiscal policy compared to
the current fiscal policy, a major objective of
which was to promote higher capital formation
and growth in productivity.
To summarize, the overall longer term effect
of current fiscal policy, compared to an un­
changed one, has been to stimulate consump­
tion rather than investment. At a given level of
real GNP, nearly two-thirds of the current stim­
ulus to consumption is being offset by a decline
in net exports, and, consequently, a corre­
sponding decline in net foreign investment. The
remaining offset to higher consumption takes
the form of declines in residential investment,
inventory investment, and government spend­
ing on goods and services. Most strikingly, de­
spite the tax advantages for plant and equip­
ment investment in the 1981 Tax Act,
nonresidential fixed investment actually has
tended to be somewhat lower, rather than
higher, as a result of higher real interest rates.
In all respects, then, current fiscal policy has
been pro-consumption rather than pro­

Chart 3A
6-Month Commercial
Paper Rate


policy in Chart 3A. Compared to an unchanged
fiscal policy, the 6-month commercial paper
rate would be at 8 percent instead of 6 percent
by the first half of 1984. Thus, compared with
an unchanged fiscal policy, tax cuts for business
raise the commercial paper rate, but not to its
actual value of 10.1 percent. The effects on the Billions of 1972 Dollars
corporate bond rate and the return on equity
are similar. The AA corporate bond rises to
12.0 percent from 10.9 percent by the first half
of 1984, but remains below its historical value
of 13.6 percent. The earnings-to-price ratio on
common stocks rises from 6.8 percent to 8 per­
cent, but not to its historical value of 9.8 per­
cent for that period. These higher interest rates
in comparison to an unchanged fiscal policy,
push up the real exchange value of the dollar
about 4 percent, or by about a third of the dif­
ference between its actual value and that cor­
responding to an unchanged fiscal policy.
Personal consumption is initially stimulated
by the impact of the business tax cuts on cor­
porate dividends. But, as the effect on interest
are displaced the most, followed by residential
rates builds, expenditures on consumer dura­
investment and inventory investment.
bles are depressed to such an extent that by the
This simulation illustrates the point that
first half of 1984 total personal consumption ex­
more saving is required in order to have more
penditures are almost exactly the same as in the
investment. Business tax cuts by themselves are
case of an unchanged fiscal policy. Therefore,
able to increase investment in plant and equip­
any stimulus to spending for plant and equip­
ment partly because foreign saving inflows in­
ment must come at the expense of other types
crease by more than government saving is re­
of investment.
duced, but also because the increase in plant
As shown in Chart 3B, although nonresidenand equipment investment comes partly at the
tial fixed investment rises, compared to an un­
expense of other domestic capital formation. If
changed fiscal policy in this simulation, its in­
we wanted to increase investment in plant and
crease is limited by the increase in interest
equipment without contracting other kinds of
rates. By the first half of 1984, nonresidential
capital formation, business tax incentives alone
fixed investment is $6 billion, or 3 percent
are not enough. They have to be combined with
higher. But this expansion occurs only because
reductions in government expenditures and/or
other types of investment are crowded out. Net
tax increases that boost national saving through
exports (and therefore net foreign investment)
an increase in government saving.

Chart 3B
Fixed investment


V. Summary and Conclusions
personal disposable incomes and the relative insensitivity of consumption to interest rates.
By far the largest offset to the increase in
personal consumption is a decline in net exports. This decline occurs because the real foreign exchange value of the dollar is quite sensitive to the difference between U. S. and
foreign real interest rates. As the fiscal deficit
drives up domestic real interest rates the U.S.
dollar appreciates, which in turn reduces the
volume of net exports. The simulation thus
clearly demonstrates that, in an open economy
with floating exchange rates, the crowding out
of investment by a fiscal deficit primarily takes
the form of a reduction in net foreign investment or, equivalently, in net exports. Without
this response, we estimate that short-term interest rates would have increased approximately 3 percentage points more, and the required return to equity, about 2 percentage
points more. As a result, investment in plant
and equipment actually would have been lower
than with an unchanged fiscal policy.
Since a major objective of current fiscal policy has been to promote capital formation and
productivity growth, we also examined the extent to which business investment would have
been stimulated if the current fiscal policy had
not included any net tax benefits for households. In this simulation, the rise in interest
rates is by more than half. Although these
higher interest rates reduce expenditures on
consumer durables, the increase in disposable
personal incomes due to higher dividend payments out of larger corporate profits boosts
consumption to the point of leaving total consumption expenditures unchanged. Nonresidential fixed investment does rise under this
policy, but mainly at the expense of net foreign
investment. If we desire an increase in plant
and equipment investment without a contraction in other kinds of capital formation, business tax incentives alone are not enough. They
must be combined with reductions in government expenditures and/or tax increases that
boost national saving through an increase in
government saving.

A major objective of the Economic Recovery
and Tax Act of 1981 was to stimulate greater
capital formation and productivity growth. The
Reagan Administration's original program
called for large tax reductions and even greater
spending reductions to achieve a balanced
budget by 1984. Federal spending on goods and
services actually didn't rise relative to high employment GNP because cuts in non-defense
spending offset the defense build-up. But the
growth of transfer payments to individuals and
the cuts in personal taxes greatly increased the
credit demands of the federal government.
The result has been a fiscal policy that has
been internally inconsistent for the purpose of
promoting capital formation. On the one hand,
investment in plant and equipment has been
promoted by accelerated depreciation allowances and liberalized investment tax credits that
have reduced effective tax rates on the cost of
capital for this type of investment. On the
other, the cost of capital is being raised by the
effect of large federal demands for credit on
interest rates.
This article has measured the permanent net
impact of these opposing forces by simulating
the effects of alternative fiscal policies on consumption and investment at unchanged levels
of real GNP. Higher capital formation would
eventually boost real GNP at any level of employment by raising labor productivity. But, for
this to happen, the ratio of investment to GNP
must first be increased. Simulations of this kind
tell us whether fiscal policy is working in the
desired direction.
Our simulation indicates that current fiscal
policy is actually promoting consumption rather
than investment. We estimate that the shift in
fiscal policy that has occurred since the beginning of 1981 is having no effect on nonresidential fixed investment. The reason is that the permanent effect of current fiscal policy on real
interest rates is just about equal to the size of
the stimulus from the tax cuts for business investment. In contrast, the permanent effect of
current fiscal policy on consumption has been
very substantial because of its large boost to

the dollar bear a strong resemblance to the simulated, longer term, effects of current fiscal policy. In fact, we estimate that real interest rates
would now be close to historically normal levels, at current levels of real GNP, if fiscal policy
had been unchanged since 1981. This suggests
that the Federal Reserve has not allowed the
fiscal stimulus to generate much more of an increase in real GNP than would otherwise have
occurred, consistent with a policy of offsetting
the inflationary effects of the fiscal stimulus.

Although our simulation estimates only the
permanent, or longer term, effect of current
fiscal policy on real interest rates and the composition of GNP, these effects can show up in
the short-run as well if the Federal Reserve is
successful in pursuing an anti-inflationary monetary policy. In such a case, the potential inflationary effects of the fiscal stimulus would be
quickly offset by monetary restraint; and real
interest rates would rise fairly immediately to
their equilibrium level. The current high level
of real interest rates and real exchange value of

This Appendix describes the structural model
of real aggregate demand in greater detail, and
explains the estimated reponses of the various
sectors of aggregate demand to real interest
rates and the real exchange value of the dollar.
Table A.1 brings these estimates together. Each
of the main sectors of aggregate demand is considered in turn.

by 2.8 percent after 11 quarters. Both the real
after-tax bond rate and the return to equity respond strongly to movements in the real shortterm interest rate. A one percentage point
change in the real short-term interest rate
moves the real after-tax corporate bond rate by
47 basis points after 11 quarters, and the return
to equity by 63 basis points.

Nonresidential Fixed Investment
The equations for nonresidential fixed investment follow the neoclassical theory of investment, as developed by Jorgenson. 20 In the
neoclassical theory, capital is viewed as being
substitutable for other factors of production, so
that firms respond to the relative price of capital in making their decisions to invest in capital
goods. The per period payment for the use of
a capital good is its "rental," or "user," cost,
which was discussed in Box 2. Firms invest in
fixed capital to bring their actual stock of capital into alignment with their desired stock,
which, in turn, depends upon final sales and
capital's rental cost.
Firms finance about one-third of fixed investment with debt and two-thirds with equity
capital. In the model, a permanent increase of
one percentage point in the weighted average
of the real after-tax bond rate and the return
to equity would currently depress real investment in equipment by $3.8 billion in 1972 dollars, or 2.8 percent, and reduce investment in
structures by $1.5 billion in 1972 dollars, or also

Inventory Investment
Inventory investment in the model follows a
stock adjustment process, modified by the effects of surprises in sales. 21 Such surprises result
in unintended investment or disinvestment.
The desired stock of inventories relative to
sales depends upon the real after-tax short-term
interest rate. However, the speed of adjustment
is much faster than for business fixed investment. One-half of the adjustment is estimated
to occur within one-quarter, and over 90 percent of it within a year. A one-percentage point
increase in the real after-tax short-term interest
rate is estimated to reduce inventory investment by $2.1 million in 1972 dollars at current
values of sales, or by 11 percent, within the
current quarter.
Personal Consumption Expenditures
Consumption functions in this model are
based upon Friedman's (1957) permanent income hypotheses. Permanent income is an anticipated long-run measure of income. The difference between permanent and current

income is caned transitory income. Permanent
disposable income is calculated as a 16-quarter
distributed lag on current disposable income,
with geometrically declining weights. According to the permanent income hypotheses, the
flow of consumption is simply a function of permanent disposable income. In the case of the

consumption of nondurables and services, expenditures are approximately the same as the
flow of consumption, and so depend only upon
permanent income. However, for durables,
consumption expenditures and the flow of consumption are quite different;

Estimated Response to a One Percentage Point Change in a
Real Interest Rate or the Real Exchange Rate


Type of Rate


Length of Lag
in Quarters

Nonresidential Fixed
a) Equipment

Real After-Tax
Bond Rate or Return
to Equity

-2.8 percent


b) Structures

Real After-Tax
Bond Rate or Return
to Equity

-2.8 percent


Inventory Investment

Real After-Tax
6-Month Commercial
Paper Rate

- 11.0 percent


Consumer Durables

Real After-Tax
6-Month Commercial
Paper Rate

1.4 percent


Residential Fixed Investment

Real After-Tax
6-Month Commercial
Paper Rate

- 5.4 percent



Real Exchange Value
of U.S. Dollar

- 1. 2 percent



Real Exchange Value
of U.S. Dollar

0.5 percent


Real After-Tax Bond Rate

Real After-Tax
6-Month Commercial
Paper Rate

47 basis points


Return to Equity

Commercial Paper Rate

63 basis points


Real Exchange Value
of U.S. Dollar

Real 6-Month Commercial
Paper Rate less Foreign
Real Short-Term Interest Rate

Foreign Real Short-Term
Interest Rate

Real 6-Month Commercial
Paper Rate


8 percent

40 basis points



Expenditures on consumer durables are
viewed as following a stock-adjustment process,
in which the desired stock of durables depends
upon permanent income and the relative price
of durables. The most volatile part of the rel-

$3.2 billion in 1972 dollars at current levels of
income, or by 5.4 percent, after 3 quarters.
Also, the model suggests that nearly all of the
transitory disposable income that is not spent
on consumer durables is spent on housing.

of interest. Also, any windfall of transitory income acts to speed up the stock-adjustment
process. 22 The normal elimination of a discrepancy between the desired and actual stock of
durables requires a reduction in financial saving, which may be limited by the fact that much
is contractual. In contrast, a
windfall of transitory income can be spent relatively rapidly on desired durables without decreasing the accumulation of financial assets.
According to estimates of the model, about
60 percent of transitory income ends up being
spent on consumer durables after 3 quarters.
Thus, movements in current income can significantly alter expenditures on consumer durables
through the allocation of transitory income,
even though permanent income is affected very
little. The proportional effect of the real aftertax short-term interest rate on expenditures for
consumer durables is weaker than in other sectors. A one-percentage point increase in this
rate, if sustained for 3 quarters, depresses
spending on consumer durables by $2.4 billion
in 1972 dollars at current levels of income, or
by 1.4 percent.

uses dummy variables to capture the impact of
periods of disintermediation at financial institutions that were caused by Regulation Q ceilings on· the interest paid on deposits. As deposits dropped off when market interest rates
rose above Regulation Q ceilings, the availability of credit to housing was restricted to a
greater extent than indicated by the market
level of real interest rates. 24 However, by the
end of 1983, deposits subject to Regulation Q
interest rate ceilings constituted only 21 percent
of all small time and savings deposits at banks
and thrifts. Consequently, at the present time,
the effect of tighter credit conditions on residential construction works almost exclusively
through movements in market levels of real interest rates.
Wojnilower (1980) and some other financial
market analysts have argued that monetary policy's ability to control spending is greatly weakened by the reduced effectiveness of Regulation
Q. This view holds that the demand for credit
is highly inelastic with respect to interest rates,
and that it is subject to volatile expectations.
Extraordinary and unacceptable increases in interest rates that are damaging to the health of
financial institutions are needed to slow credit
demands. However, our estimates indicate
that, even in the absence of effective Regulation Q ceilings, the response of residential construction as well as other types of activity to
changes in real interest rates is very substantial.

Residential Fixed Investment
Residential investment is also assumed to folIowa stock adjustment process in the context
of the permanent income hypothesis. The desired stock of housing depends upon the size of
permanent income and the rental cost of capital
for housing. Because the tax treatments of
owner-occupied housing and rental housing difa weighted average of the rental cost of
capital for the two types of housing units is employed. 23 The real after-tax, short-term interest
rate affects the terms of mortgage credit,
which, in turn, enter into the rental cost of capital. We estimate that a sustained I-percentage
point increase in the real, after-tax short-term
int,erest rate reduces residential construction by

Net Exports and the Real Exchange Rate
Since the shift in 1973 to the managed floating of exchange rates, real interest rates have
had an additional channel of influence on aggregate demand. An increase in real interest
rates generates capital inflows that cause the
real value of the dollar to appreciate. This, in
turn, reduces the contribution of net exports to
the level of aggregate demand. 25


Exports are modeled as a function of the rest
of the world's GNP and the real exchange value
of the dollar, on a trade-weighted basis. Imports are related in a similar fashion to U.S.
GNP and the real exchange value of the dollar.
The influences ofGNP on exports and imports
are mostly contemporaneous. However, the responses of exports and imports to changes in
the real exchange value of the dollar take much
longer, with significant effects lasting for 7
quarters. As earlier studies have found, the response of exports is greater than that for imports. 26 A sustained one-percentage point increase in the real value of the dollar at current
levels of income is estimated to reduce exports
by 1.7 billion in 1972 dollars, or by 1.2 percent,
and to increase imports by .7 billion in 1972
dollars, or by 0.5 percent.
The approach to modeling the real exchange
value of the dollar follows the asset of view of
the exchange market. In this view, asset prices
adjust quickly to clear the foreign exchange
market, and expectations playa central role in
the determination of the short-run equilibrium
exchange rate. Trade flows help mainly to tie
down long-run expectations and also may influence demands for assets in some degree. 27 The
basic equilibrium condition in this view is that
the expected percentage change in the exchange rate over any period equals the difference between nominal returns on securities at
home and those abroad with maturities of the
same period. If this condition does not hold,
investors will bid the value of the exchange rate
to the point where it does. It is easily shown
that a similar relationship would hold between
the expected change in the real exchange rate
and the difference between the real returns on

securities. Thus, an increase in real interest
rates in the United States raises the real value
of the dollar to the point where its expected
depreciation in the future is equal to the differential between U.S. and; foreign real interest
According to our estimated model, a I-percentage point change in the real short-term interest rate differential in favor of the United
States, sustained over a period of 11 quarters,
raises the real exchange value of the dollar by
a full 8 percentage points. Also, at current levels of income, an unanticipated increase in net
exports of 1 billion in 1972 dollars appreciates
the dollar by 0.4 percent by causing the expected value of the real exchange rate to be
revised upward to that degree.
Finally, the model takes into account the responses of foreign central banks to movements
in U. S. real interest rates. In the period of managed floating, it is estimated that foreign central
banks have tended to respond to each I-percentage point change in the U.S. real shortterm interest rate with a 40-basis point change
of their own. Moreover, we estimate about the
same response during the 1981-84 period. This
kind of partial response tends to minimize the
impact of changes in U.S. real interest rates on
foreign GNP. If, for example, a fiscal deficit in
the U.S. produces higher real interest rates and
therefore leads to an appreciation of the U.S.
dollar, net exports from foreign countries
would increase, creating inflationary pressures
abroad. However, if foreign real interest rates
rise to partially match the change in U.S. rates,
foreign investment expenditures would be reduced; and this would tend to stabilize real aggregate demand and GNP abroad.


1. Existing research on the relation between the saving
rate and tax incentives is well summarized in Chapter 3 of
Bosworth (1984). The stability of the private saving rate
since 1981 is discussed in Bisignano (1984).

to 3 years a 6 percent investment tax credit and all other
equipment a 10 percent credit. However, in 1982 allowable
depreciation on all equipment was reduced by 50 percent
of the investment tax credits taken.
S.The \Jl'ldeflyil1~l"allles()fll,i, Cll1dkWereestimCltedby
Data Resources, Inc. and the staff of the Board ofGovernors of the Federal Reserve System.

2. Initial budget cut proposals prepared by the Office of
Management and Budget, a report on the proposed tax
reductions issued by the Treasury, and the White House
paper discussing the four major elements of the program
are contained in a paper issued by the Executive Office of
the President (981).

10. For the derivation, see Jorgenson (1963) or Hall and
Jorgenson (1967).
11. The effective tax rate on equity financed investment is

3. Inflation premiums in interest rates have actually been
highly variable in recent years. For a way of incorpprating
variable inflation premiums into the standard IS-LM framework, see Keran (1984).

not the same as the effective rate of taxation on all investment because a portion is .financed by debt and nominal interest costs on debt are deductible from the corporation's taxable income. Thus, for example, while the
current effective tax rate on equity-financed nonresidential
fixed investment averages about 14 percent (derived from
a weighted average of Table 2), the effective rate of taxation on all such investments is near zero. See Auerbach
12. For the empirical analysis, we use the high employment bUdget estimates, based on a 6 percent unemployment rate, that are maintained by the Bureau of Economic
Analysis of the Department of Commerce. Recently, the
Bureau has developed a cyclically adjusted bUdget based
on the realized trend in real GNP, rather than an estimate
of potential output at a specified unemployment rate. This
has the advantage of automatically allowing for demographic and other changes that may affect the unemployment rate associated with full employment. However, use
of this approach for relatively recent years requires a forecast of future real GNP to generate the trend, which introduces a different kind of uncertainty. Since the full employment rate of unemployment probably has not changed
much over the 1981-84 period, the high employment
budget based on a 6 percent unemployment rate was chosen. For descriptions and estimates of these two versions
of the high employment budget, see de Leeuw, et al (1980)
and de Leeuw and Holloway (1982, 1983).

4. Estimates differ as to the time required for the full adjustment of prices and complete "crowding out" of interestsensitive expenditures by fiscal. policy in the absence of
any change in nominal money. According to the St. Louis
reduced-form model of Anderson and Carlson (1970), such
full crowding out takes place within 4 quarters. However, in
the different reduced-form model of McElhattan (1982) it
takes 5 or 6 years before real GNP returns to its original
level. Similar differences also exist in estimates from large
structural models. In the FMP model used by the staff of
the Board of Governors of the Federal Reserve, the length
of time required for full crowding out, given nominal money,
is 2 to 4 years; but for some other structural models it is
much longer. See Ando and Modigliani in Stein (1976) and
Fromm and Klein (1973).
5; Presumably the Administration's plan was to increase
the proportion of investment relative to consumption to secure more growth over time. Even if the fiscal deficit is
accompanied by lower marginal tax rates that stimulate
work effort and thereby increase the level of full employment GNP, the effect on the consumption and investment
mix of the economy would be similar although total economic activity would be increased. In terms of Figure 1, the
leftward shift of the LM curve resulting from either monetary policy or price adjustments would be less if full employment GNP were raised compared to the case where
full employment output is unchanged.
6. Estimates of the average marginal tax rate for individuals are from Barro and Shahasokul (1983). Updates for
1981-83 were obtained from the Economic Research
Group of Goldman Sachs. The figure for 1984 was estimated by the author.

13. The detailed breakdown of the high employment
budget at a 6 percent unemployment rate that is used to
derive this and other elements of Table 2 is provided in
Holloway (June 1984, September 1984).
14. For a complete description, see Throop (1984).
15. Examples of such large scale neo-Keynesian models
include Eckstein (1983), Evans (1969), and Federal Reserve Board (1983).

7. Under ACRS, any depreciable asset falls into one of
four classes and is given a tax life of 3, 5, 10, or 15 years.
These shorter tax lives were effective immediately, Clnd depreciation schedules were to become more accelerated
during a five yar phase-in period. But the latter change was
rescinded in 1982.

16. Reduced-form models capable of making such forecasts of real GNP and the price level include Anderson and
Carlson (1970) and McElhattan (1981).
17. Actual values of real GNP could not be reproduced
exactly in each period due to the dynamics of the model.
Real interest rates and the exchange rate affect aggregate
demand with distributed lags. Thus, only a fraction of the
total effect of these variables on aggregate demand occurs
in the first period. To offset a large change in fiscal policy
exactly, as for example in 1982.3 and 1983.3, a very large
change in real interest rates would be needed. However,
in subsequent periods, the lagged effects of the initially
large change in interest rates would have to be offset, re-

8. Before 1981, 10 percent of the value of investment in
equiPment could be deducted from corporate taxes for
equipment with a life of 7 years or more. A 3-to-6 percent
deduction could be taken on equipment with lives of between 3 and 7 years, and no investment tax credit was
allowed on equipment with a life of less than 3 years. The
1981 Tax Act gave equipment with a recovery period of up


quiring further large movements of interest rates in the opposite direction. To reproduce real GNP in each period
could require larger and larger changes in interest rates
over time.

This last hypothesis has as its counterpart the notion of
the inflation-adjusted high employment budget, which reduces the high employment deficit by the rate of inflation
multiplied by the privately held stock of federal debt. See,
for example, Cagan (1981) and Eisner and Pieper (1984).
The increase in the inflation-adjusted high employment
bUdget deficit between 1981 and 1984 was even larger
than the change in the unadjusted high emplqyment budget
because of the decline in inflation, implying a greater effect
on cqnsumption. It shifted from a surplus equal to 1.3 percent of GNP to a deficit of 1.6 percent, compared to a
change in the unadjusted deficit from 0.9 percent of GNP
to·3 percent. However, the stability of the private saving
rate contrasts with the behavior of saving that would be
consistent with this inflation adjustment to the budget. As
inflation and inflation premiums on government debt rose
in the 1970s, the private saving rate should have increased. More recently, as inflation has declined, the private saving rate should have fallen. In practice,however,
the private saving rate appears to have been relatively impervious to these influences. See Bisignano (1984).

This is an example of the general problem of instrument
instability. See, for example, Holbrook (1972). The resulting
cycles could imply negative values .for interest rates and,
in any case, would make it difficult to compare the longer
term impacts of different fiscal policies. A degree of
smoothing of interest rates was therefore required. Still, the
average deviation of simulated GNP from historical GNP
for the entire 1981-84 period is less than $3 billion, in 1972
dollars; by the first half of 1984, the average deviation is
less than $1 billion.
18. This point can be demonstrated by referring to Figure
1. Starting from a historically normal level of interest rates
at i 1 and Y1 and considering the shift in the IS function from
lSi to IS2 due to the tax cuts, suppose that real GNP is
allowed to increase, say along LM1 with a given (real)
money supply, to Y2' Then the interest rate corresponding
to an unchanged fiscal policy of lSi at Y2 would be below
the normal level at i1 • But if the Federal Reserve prevented
higher inflation in the short-run by a policy action to shift
the LM curve from LM1 to LM2, then the interest rate corresponding to an unchanged fiscal policy would be the
same as the normal level i1 , as actually observed.

In yet another alternative view of the consumption function,
personal tax cuts would have little or no effect on consumption and interest rates. This is the ultrarationality hypothesis recently argued by Barro (1974), in which a tax
cut causes households to raise their saving rate in order
to pay the higher taxes that will be required for servicing
the government debt in the future. In this case, all of a tax
cut would be saved so that there would be no effect whatsoever on consumption. However, empirical studies of
short-run consumer spending do not generally support this
view. See Buiter and Tobin (1979) and Feldstein (1982).

19. Since the simulated change in consumption, and consequent impact on interest rates, is the single most important effect of the change in fiscal policy, the results are
sensitive to the form of the consumption function. As explained in the Appendix, consumption in the model is a
function of permanent disposable personal income and real
after-tax short-term interest rates. Permanent disposable
income is measured by a 16-quarter distributed lag on actual disposable income. This is a fairly standard formulation, but other approaches are possible. Most of these alternatives would make the impact of the change in fiscal
policy on consumption and interest rates even greater than
indicated in the present simulation.

20. The basic theory and its application are described in
Jorgenson (1963) and Hall Jorgenson (1967, 1971).
21. The classic papers on this type of investory model are
Metzler (1941), Lowell (1961), and Darling (1959).
22. On the role of permanent and transitory income in the
stock adjustment process for consumer durables, see Juster and Wachtel (1972) and Darby (1975).

First, if households anticipate their tax changes to be permanent, their perceived permanent income may change
more qUickly than the adaptive construction based on a 16quarter distributed lag. Second, if the public includes corporate earnings in its notion of permanent income, whether
paid out in dividends or not, the effect of the tax cuts for
business and households on consumption and interest
rates would be somewhat greater than indicated here because of the additional effect of retained earnings on permanent income. Third, and extending household rationality
somewhat further, households may perceive that the inflation premiums in interest payments on the federal debt are
in reality the repayment of principal in real terms, and so
should be saved. Thus, changes in disposable income due
to changes in these inflation premiums would not affect

23. For a useful discussion of the application of the neoclassical theory of investment to owner-occupied and rental
housing, including the specific taxes applicable to these
sectors see Ott, Ott, and Yoo (1975).
24. For a theoretical demonstration of this point, see Lombra (1984).
25. The classic works emphasizing this link are Fleming
(1962) and Mundell (1963).
26. See, for example, Feldman (1982) and Warner and
Kreinin (1983).
27. The approach used is basically a simplification of
Hooper and Morton's (1982) extension of the sticky price
monetary model of exchange rate determination developed
by Dornbusch (1976) and Frankel (1979). For a general
survey of the asset view of exchange rates, see Shafer
and Loopesko (1983).


Anderson, Leonall C., and Carlson, Keith M. "A Monetarist
Model of Economic Stabilization," Monthly Review,
Federal Reserve Bank of St. Louis, April 1970.
Auerbach, Alan J. "Corporate Taxation in the United
States," Brookings Papers on Economic Activity,
Barro, Robert. "Are Government Bonds Net Wealth?"
Journal of Political Economy, November-December
Barro, Robert J. and Shakosakul, Chaipat, "Measuring the
Average Marginal Tax Rate from the Individual Income
Tax," Journal of Business, October 1983.
Bisignano, Joseph. "Impervious Saving Behavior," Weekly
Letter, Federal Reserve Bank of San Francisco, September 28, 1984.
Bosworth, Barry P. Tax Incentives and Economic
Growth. Washington, DC: The Brookings Institution,
Buiter, William and Tobin, James. "Debt Neutrality: A Brief
Review of Doctrine and Evidence," in George M. von
Furstenberg, ed., Social Security Versus Private
Saving. Ballinger, 1979.
Cagan, Phillip. "The Real Federal Deficit and Financial
Markets," The AEI Economist, American Enterprise
Institute for Public Policy Research, November 1981.
Darby, Michael. "Postwar U.S. Consumption, Consumer
Expenditures, and Saving," American Economic Review, May 1975.
Darling, Paul. "Manufacturing Inventory Investment,"
American Economic Review, December 1959.
deLeeuw, Frank, Thomas M. Hollowway, Darwin G. Johynson, David S. McClair, and Charles A. Waite. "The
High Employment Budget; New Estimates: 1955-80,"
Survey of Current Business, November 1980.
de Leeuw, Frank and Holloway, Thomas M. "The High Employment Budget: Revised Estimates and Automatic
Inflation Effects," Survey of Current Business, April
Dornbusch, Rudgier. "Expectations and Exchange Rate
Dynamics," Journal of Political Economy, December
Echstein, Otto. The DRI Model oHhe U.S. Economy. New
York: McGraw-Hili, 1983.
Eisner, Robert and Pieper, Paul J. "A New View of the
Federal Debt and Budget Deficits," American Economic Review, March 1984.
Evans, Michael K. Macroeconomic Activity: Theory,
Forecasting, and Control. New York: Harper and
Row, 1969.

Executive Office of the President. America's
ning: A Program for Economic Recovery, February
Federal Reserve Board. The Quarterly Econometric
Model. Mimeographed, 1983.
Feldman, Robert A. "Dollar Appreciation, Foreign Trade,
and the U.S. Economy," Quarterly Review, Federal
Reserve Bank of New York, <:>""1111"0:;',
Feldstein, Martin. "Government Deficits and Aggregate Demand," Journal of Monetary Economics, January
Fleming, Marcus. "Domestic Financial Policies Under
PaFixed and Flexible Exchange Rates," I.M.F.
pers, November 1962.
Frankel, Jeffrey. "On the Mark: A Theory of Floating Exchange Rates Based on Interest Rate Differentials,"
American Economic Review, September 1979.
Friedman, Milton. A Theory of the Consumption Function. Princeton: Princeton University Press, 1957.
Fromm, Gary and Klein, Lawrence R. "A Comparison of
Eleven Econometric Models of the United States."
American Economic Review, May 1973.
Hall, Robert C. and Jorgenson, Dale W. "Tax Policy and
Investment Behavior," American Economic Review,
June 1967.
Hall, Robert C. "Application of the Theory of Optimum Capital Accumulation," in Gary Fromm (ad.) Tax Im~en­
tives and Investment Spending. Washington, D.C.:
Brookings Institution, 1971.
Holbrook, Robert. "Optimal Economic Policy and the Problem of Instrument Instability," American Eccmomic
Review, March 1972.
Holloway, Thomas M. Cyclical Adjustment of the Federai
Budget and.Federal Debt: Detailed. Methodology
and Estimates. Bureau of Economic Analysis, U.S.
Department of Commerce, June 1984.
- - . "Cyclical Adjustment of the Federal Budget: Revised and Updated Estimates; 1981 :1-1984:2" (mimeographed). Bureau of Economic Analysis, U.S. Department of Commerce, September 1984.
Hooper, Peter and Morton, John. "Fluctuations in the Dollar: A Model of Nominal and Real Exchange Rate Determination," Journal of International Money and Finance, 1982, 1.
Jorgenson, Dale W. "Capital Theory and Investment Behavior," American Economic Review, Proceedings,
May 1963.


Juster, F. Thomas and Wachtel, Paul. "Anticipatory ana
Objective Models of Durable Goods Demandi' American Economic Review, September 1972.
Keran, Michael


Bank of San Francisco, Summer 1984.
1922~1983;'E~()nomicReview, Federal

Lombra, • RCiYI'Tl()lld.·.·.E: ."JhElGhl:l.l'lSiIl91'l()leofRe~li:lhd
Nominal Interest Rates;'EcollomicReview, Federal
Reserve Bank of KansasCity,February1984.
Lovell,Mi~haeIA "Manufacturer's Inventories, .• SalEls Ex..

pectations, • CJ.nd the Acceleration Principle;' E~ono­
metrica, JUly 1961 .
Mcelhattan, Rose. "The REl.sponse of Real Outpl,ltandlnflation to Monetary Poli~y," Economic Review, Federal Reserve Bank of San Francisco, Summer 1981.

- -•. "On Federal'ld their Economic lropact,"
Econol11ic Review, Federal Reserve Bank of San
Francisco, Summer 1982.
Metzler, Lloyd A "Thl3 Nature CJ.nd Stability of Inventory
Cycles;' Review of Economics and Statistics, August 1941.
Mundell, Robert. "Capital Mobility and Stabilization Policy
Under Fixed and Flexible Exchange Rates;' Canadian
Journal of Economics and Political Science, November1963.
Ott, David J.,Attiat F. Ott, and Jong H. Yoo. Macroeconomic Theory, New York: McGraw-Hili, 1975, Ch. 6.
Shafer, Jeffrey A. and Loopesk, Bonnie E. "Floating Exchange Rates After Ten Years;' Brookings Papers on
Economic Activity, No.1, 1983.
Stein, Jerome (ed.). Moneterism. New York: Elsevier
North-Holland, Ive., 1976.
Throop, Adrian
"A Structural MocJel of Real Aggregate
Demand;' Working Paper in Applied EcotlOmi.c Theory and Econometrics, No. 84-03 Fl3deral Reserve
Bank of San Francisco, July 1984.
Warner,. Dennis and Kreinen, Mordechai. "Determinants of
In~ernCJ.tionaITradeFlows," JheReview of Economics and Statistics, February 1983.
Wojnilower, Albert M. "The Central Role of Credit Crunches
in Recent Financial History;' Brookings Papers on
Economic ActiVity, No.2, 1980.



Rose McElhattan*

Conventional Phillips Curve models emphasize the relationship
between inflation and the unemployment rate. From these models,
analysts derive the natural rate or stable-inflation unemployment
rate. This is a rate which, if maintained, is associated with no
change in the inflation rate. In this paper we focus upon inflation
and the capacity utilization rate, and derive a stable-inflation capacity utilization rate which is about 82 percent (with its 95 percent
confidence interval between 78.5 and 83.5 percent.) Evidence is
presented that capacity utilization is a more informative inflationary
signal than the unemployment rate.

The rate of inflation declined substantially
between 1981 and 1983, from 9.6 percent to 3.8
percent. This reduction accompanied a recession that was the worst since the Second World
War in terms of unemployment and unused
manufacturing capacity. The jobless rate averaged 9.7 percent and the capacity utilization
rate 71.1 percent in 1982. We must look back
to 1941 to find a comparable unemployment
rate of 9.9 percent, and to 1975 for the previous
post-war low of 72.9 percent in the capacity utilization rate.
After paying such substantial real costs to
bring inflation down, a concern has arisen that
subsequent economic growth may start another
inflationary spiral. Just how far can growth proceed before inflationary pressures are likely to
rebuild? This is the major question addressed
in this paper.

Our starting point in answering this question
is a traditional pricing model in which prices
are determined as a mark-up on unit production costs. In an earlier version of this model
(McElhattan, 1978), I estimated the inflationary impact of excess demand pressures, as measured by capacity utilization in U.S. manufacturing industries, and found that, on average,
during the 1954-1977 period, stable inflation
was associated with a capacity utilization rate
of about 82 percent. Demand pressures tended
to raise inflation when utilization rates rose
much above 82 percent; inflation tended to fall
when utilization rates fell below that critical
Since that earlier paper, the U.S. has experienced sharp and repeated changes in energy
prices and substantial changes in the international value of the dollar. In addition, the capacity utilization series has been revised. The
objectives of this paper therefore are to update
and expand the earlier model by adding supplyside shocks, and to determine the degree to

*Senior Economist, Federal Reserve Bank of
San Francisco

Section III provides estimates of the Stable-Inflation Rate of Capacity Utilization from the
expanded inflation model and compares these
to earlier estimates. Section IV compares capacity utilization and unemployment rates as
signals of impending inflation. Section V discusses economic and policy implications of the
stable-inflation capacity utilization rate concept, and the final section provides a summary.

which the stable-inflation capacity utilization
rate may have changed.
Section I presents an overview of the basic
inflation model and the determination of the
stable-inflation rate of capacity utilization. Section II considers the estimation of inflation and
the inflationary impact of the following supplyside shocks: Nixon-era wage and price controls,
changes in the real price of crude oil, and
ch;anl~es in the international value of the dollar.

I. The Inflation Model
The inflation equation used in this paper is
derived from a traditional price mark-up model
found in most econometric macro forecasting
models and studies of inflation. 1 The model itself determines an aggregate inflation rate as
measured here by changes in the GNP Implicit
Price Deflator. Inflation behavior is described
in terms of a wage and a price equation. The
price equation relates prices to a mark-up on
standard production costs, with wages as the
major cost component. Wages, in turn, are determined by excess demand in labor markets
and by
The mark-up of final product prices over production costs is related to excess demand pressures in final product markets. These pressures
are most often measured by the GNP gap between actual and potential GNP or by capacity
utilization rates. As aggregate demand builds
and utilization rates increase, the mark-up increases as final product prices adjust to eliminate excess demand. The higher mark-up also
may reflect noncompetitive pricing behavior by
some firms that feel they can raise prices without a serious loss in sales during periods of inCH~as:mg demand. 2
In a typical price mark-up equation, the aggre:gal:e inflation rate (IR) is determined by
ch:an~~es in standard unit labor costs, measured
chalng(~S in nominal wage~ CW) and a trend
rate of labor productivity (T), and by excess
aggregate demand, expressed here as a function
utilization, f(CU):

In Equation 1, dots over the variables indicate rates of change in that variable, and the
subscript, t, refers to a period of time. Upward
pressure is placed upon inflation when capacity
utilization increases; that is, the change in IR t
with respect to f(CU t) is positive.
In Equation 2, the rate of change in nominal
wages of labor are determined by the expected
inflation rate (IR *), by the trend rate of growth
of labor productivity and by excess demand in
labor markets, expressed as a function of the
unemployment rate, h(u).



a21IR; + a23Tt - h(u)


Rising unemployment places downward pressure on wages, and a decrease in the unemployment rate represents an increase in wage
pressures, that is, hi < O. According to Equation 2, inflation-adjusted wage changes
CW - a21IR *) will rise in proportion to labor
productivity for given levels of unemployment.
Substituting Equation 2 into Equation 1, we
obtain Equation 3, in which the inflation rate
is determined by expected inflation, labor productivity and the two excess demand variables,
unemployment and capacity utilization rates:


a l2 a21 IR; + (a l2 a23 - a13)Tt
- al2 h(u)t + f(CU)t


Focus on Capacity Utilization
Because of the high correlation between the
unemployment and capacity utilization rates,
empirical estimation of Equation 3, or specifi-



cations similar to it, generally include only one
of these variables, the unemployment rate. 3
The resulting negative relationship between the
unemployment rate and inflation, popularly
known as the Phillips Curve, has received wide

viding the data. Such judgment, nevertheless,
represents an economic concept that bears on
pricing decisions, just as inflation expectations,
which also are difficult to measure, bear on
pricing decisions. Moreover, the capacity utilization series has had a stable and close correlation with changes in the inflation rate
throughout the post-war period. It therefore
merits serious consideration as an empirical signal of inflation.

In our model we focus upon the inflationary
consequences associated with capacity utilization and drop the unemployment rate because
the capacity utilization rate provides a more reliable signal of inflation than the unemployment rate, as shown in Section IV. Consequently, in our model, capacity utilization
proxies for aggregate demand pressures in general as these affect both the price mark-up and
the determination of wages.
Major changes that occurred in the post-war
period have altered the unemployment rate associated with a given degree of inflationary
pressure. For instance, many economists contend that demographic changes since the mid1960s (particularly the presence of more women
and young people in the labor force) have resulted in a higher average "natural" unemployment rate (a rate consistent with no change in
inflation). There remains, however, a good
deal of disagreement and uncertainty among
economists over how much the natural rate of
unemployment has changed.
This uncertainty about the natural rate of
unemployment has public policy as well as academic implications. Some economists argue
that it has led to some inflationary bias in past
policy decisions. There was a natural tendency,
they argued, to err on the side of underestimating the unemployment rate consistent with
stable inflation and, therefore, to advocate policies which in retrospect were too stimulative
and inflationary.4 If this assessment were corthe use of capacity utilization to gauge inflationaJ'y pressures may be helpful. It would at
least serve as an independent check on assessments of inflation based on unemployment
The capacity utilization data do not represent
capacity measurements in some absolute or engineering sense. 5 Instead, they depend to a degree on the judgment of the respondents pro-

In view of the above considerations, we may
rewrite Equation 3 with capacity utilization as
the sole excess demand variable. In addition,
we express the general form, f(CU), as approximated by the linear relationship, bo + b 1 CUt,
to obtain Equation 4.
a12 a21 IR; + (a12 a23
a13)T t
+ bo + b 1 CUt
We regard inflation expectations, IR *, as a
weighted average of past actual inflation. This
is a general specification of the formation of
inflation expectations and a common one in
price mark-up models; it is shown in Equation





(IR t -





L b = 1, b

i ;?:


0 for all i.

Substituting Equation 5 into Equation 4
yields the reduced form equation of the wageprice sector of a more complete model of the
U.S. economy. In the reduced-form specification, Equation 6, the t~end rate of change in
productivity of labor (T) and capacity utilization (CU) are regarded as exogenous variables.

IRt = a12 a21Lbi IRt - i + (a12 a23 - a13)Tt

+ bo + b1 CUt


Equation 6 provides the short-run relationship between inflation and its determinants. A
stable, long-run relationship exists only if the
value of (a12a21) is less than unity. Under that


the value of goods and services produced in the
U.S. Therefore, it directly excludes the value
of imports. For example, although the value of
imports is included in personal consumption
spending, it is subtracted from GNP in the import account. As a result, imports have no direct net effect on GNP or the Deflator. However, changes in the price of imported items
may be correlated with changes in the GNP Deflator to the extent that changes in foreign
prices, through competitive pressures, lead to
changes in the U.S. prices of traded products
produced in the U. S.
Equation 8 includes a function of the vector
Z to incorporate these supply-side shock variables as determinants of U.S. inflation.

condition, any gap between inflation and its expected rate, for given values of capacity utilization and productivity, will become smaller
over time. Eventually, actual and expected inflation will be the same and will be associated
with a specific capacity utilization rate. In the
context of unemployment and inflation, the
condition that (alZa21) is less than unity implies
a stable, long-run Phillips Curve. By the same
reasoning, a permanently lower capacity utilization rate would be needed to achieve a permanently lower inflation rate.
On balance, econometric evidence since the
mid-1970s suggests that the value of aZl is unity
and that the value of al2 = al3 = 1. This leads
to the result that (al2aZl) is unity. 6 In the case
of the coefficient, aZb the unity estimate suggests that inflation expectations are fully reflected in wages over time (see Equation 2).
According to Equation 1, the estimate al2 =
al3 = 1, suggests that the relevant long-run determinant of inflation is the rate of change in
standard unit labor costs (W - T) that is fully
incorporated in final prices.
These considerations enable us to rewrite the
reduced-form inflation model (Equation 6) in
terms of the difference between actual and expected inflation:

(al2a23 - a l3 ) Tt
+ bo + b l CUt + k(Z)
In the short-run, supply-side price shocks are
likely to be positively correlated with inflation.
Such shocks would have no transitory effect on
the aggregate price level if other prices were
perfectly flexible (and if the productive capacity
of the economy were unaffected). However, we
generally do not observe perfectly flexible
prices over short periods of time so some positive correlation between movements in supplyside price shocks and the aggregate price level
appears most likely.
What ultimately happens to U.S. prices following a price shock depends upon whether
there are any related changes in (a) real GNP
and/or (b) the money supply. We assume first
that there is no long-run effect on the level of
real GNP. Given this assumption, the ultimate
impact on the GNP Deflator will depend upon
the monetary response to the initial price
We may distinguish two types of monetary
responses to a one-time decrease in the relative
price of crude oil. (A price increase has the
same effects, but of opposite sign.) In the first
case, the decrease in the general price level due
to the oil price decline is met by no change in
the money supply. This case is illustrated in Figure 1 in a simple aggregate demand-aggregate
supply paradigm of economic behavior.

IRt - IR;

IR t - IR;
al3) Tt + bo + b l CUt
Equation 7 indicates that a stable relationship
exists between the difference of actual and expected inflation and capacity utilization and labor productivity.

= (al2aZ3

Adding Supply-Side Shocks
Since the early 1970s, a number of events
have significantly affected U.S. prices. These
include the Nixon-era wage and price controls,
changes in the international value of the dollar
and the OPEC changes in the price of oil. We
shall refer to these events as supply-side shocks.
In the context of our model, the latter two
events affect final prices through their impact
on the the mark-up of domestic prices over
costs. We recall that the inflation rate is measured by the GNP Implicit Deflator. The GNP
Deflator is a value-added concept-it measures


quantity demanded, Q1, may be sold at a lower
aggregate price, P* in Figure 1. That price level
cannot be sustained, however. Over time,
prices on other goods and services would rise.
These revisions produce upward shifts in AS'
baqk to its original position, AS. At that final
point, the price level for the quantity, Q1, is
determined by the unchanged money supply
and is equal to its original level, P.
In Figure 2, the shift in aggregate supply is
met by a decline in the money supply. As a
result, aggregate demand will shift downward,
say, to AD'.7 An equal shift in demand and
supply denotes a fully "accommodative" monetary policy in the sense that the contraction in
money constrains the level of prices for other
goods from rising and thus prevents the aggregate supply schedule from shifting from AS'
back to AS. Ultimately, the price level is lower
(P') following the negative shock, its decline
determined by the extent of the monetary
These cases illustrate the transitory effect of
price shocks on domestic prices. Aggregate
price changes occur during the transition period
following the supply-side shock. In the subsequent equilibrium (p* or P' in the prior ex-

Figure 1





In this paradigm, the aggregate demand
schedule (AD) slopes downward and to the
right, signifying that greater quantities of output are demanded at lower price levels. This
occurs because lower prices mean larger real
money balances in private wealth portfolios,
which stimulate demand in general. The quantity of output a nation can produce is ultimately
constrained by the quantity and quality of real
economic factors, including the productivity of
its labor and population growth. In the longrun, because of these real constraints, the quantity of real GNP produced is regarded as independent of the price level. These considerations
are expressed in the vertical, long-run supply
curve, LRS. For short periods of time, however, greater quantities of output may be supplied at higher prices, as expressed in the upward sloping aggregate supply curve, AS.
Changes in price expectations or supply-side
price shocks, will shift the AS schedule.
Beginning at full-employment equilibrium, a
decline in the OPEC price of crude oil means
a downward shift in AS, say, to AS' in Figure
1. In the absence of a monetary response, aggregate demand does not shift. The downward
shift in aggregate supply means that the original

Figure 2







the supply-side price shocks have no long-run
effects on the level of real GNP-an assumption some economists have challenged in the
case of oil-price shocks. s In the event that a
permanent change in a supply-side variable alters the level of potential output, the price level
will be changed in the long-run from what it
otherwise would be independent of whether
there is any monetary accommodation. Nevertheless, as in our example, even if the supplyside shock has a permanent effect on the price
level, it would not permanently change the rate
of inflation.

amples), there will be no further change in the
price level, or in the rate of inflation, as a result
of the supply-side shock. A partially accommodative monetary policy will lead to a final
price level somewhere between the P* and the
pi of our two examples, and would require
some upward shift in the aggregate supply
schedule from AS'. For how long, and by what
magnitude, prices and the rate of inflation in
the U. S. respond to different supply-side
shocks is an empirical matter discussed in the
following section.
The preceding discussion has assumed that

II. Estimation of the Inflation Model
equal to unity. This indicates that the relevant
measure in pricing decisions is standard unit labor costs. In the wage equation, a23 also is expected to equal unity, reflecting that, on average, the rate of change in real wages is equal
to the rate of change in labor productivity. Consequently, the productivity parameter is expected, from an economic point of view, to
have a zero value.
With the estimated coefficient of productivity
not significantly different from zero, and the
coefficient on the previous year's inflation
equal to unity, we may re-write the general inflation model (Equation 8) in terms of the
change in the inflation rate:

The general form of the inflation model is
expressed in Equation 8. All estimations in this
paper use annual data since 1954, unless otherwise specified. The expected rate of inflation
in Equation 8 was replaced with the value of
the inflation rate lagged one period. This was
done because in estimations of Equation 6 only
the previous year's inflation rate is statistically
significant, as was found in our earlier paper
(see Appendix 3). Adding more values of past
inflation did not significantly improve the determination of inflation according to the "F"
test. Moreover, we found that the estimated
coefficient associated with the previous year's
inflation rate is not significantly different from
unity. 9
In addition, in this section, the re~ults are
reported with the productivity term (T) omitted. The reason for doing so is that changes in
labor productivity, measured by the trend rate
of growth in real GNP per hour of employment,
has no significant effect on the rate of inflation.
This result accords with our earlier study and
is generally found in estimations of reducedform equations similar to the above. The
economic significance of this result may
best be understood by referring to the components of the productivity parameter in Equation 8. That parameter is derived from the parameters in the price mark-up and wage equation, (al2a23 -al3)' As noted above, the
parameters al2 and a13 from the price equation
generally are found in empirical research to be


IR t - IR t - 1
= bo + blCUt + k(Z)t
Regression results of Equation 9 are provided
in Table 1, at first without supply-side shocks
(Column I) and then with wage and price controls (Column II), changes in real oil prices
(Columns III and IV), and changes in the international value of the dollar (Columns V and
VI) progressively introduced. The dependent
variable is the change in the inflation rate, with
the latter measured by the year-over-year percentage change in the GNP Implicit Price Deflator. Capacity utilization, CU, is measured as
the annual rate in total U.S. manufacturing, a
series published by the Federal Reserve Board
of Governors.




Regression Results for Change in the Inflation Rate
(Measured by the GNP Implicit Deflator*)




C (bo)



( -4.0)

(- 5.4)

(- 3.7)

( -4.4)

( -3.3)

( -4.2)







( -1.0)

( -1.6)

( -1.4)

(- 2.0)

( -1.9)


















Capacity Utilization Rate
CU (b l )
Wage/Price Controls "On"
Wage/Price Controls "Off"
Change in Real Price of Oil

DIPE Lagged One Year

( -0.1)

DIPE Lagged Two Years


Acceleration in Relative Price of Oil
DDIPE Lagged One Year

Acceleration in Exchange Rate of U.S. Dollar
DDEX of U.S. Dollar
DDEX Lagged One Year
Stable Inflation Capacity Utilization Rate
Cue (= - bo/b l )
95% Confidence
Summary Statistics















* Estimation period is 1954-83 for equations (I) through (IV) and 1959-1983 for equations (V) and (VI). Figures in
parentheses are t-statistics.
** Nominal exchange rate
*** Real exchange rate
t The derivation of confidence limits for the ratio of two estimators is described in Scadding (1973).


Wage and Price Controls
In Equation II, Table 1, wage and price controls are included in the determination of the
change in the inflation rate. These controls
were applied in several stages from August
1971 through April 1974, when they were removed entirely. The "on" effect is represented
by the dummy variable WPON, which is unity
in 1972 and zero elsewhere, and the "off" effect, by the dummy variable WPOFF, which is
unity in 1974 and 1975. 10
Controls, according to the estimates in Column II, tended to lower the measured inflation
rate about 1.0 percentage points in 1972; their
removal tended to increase inflation about 2.7
percentage points in both 1974 and 1975. Other
studies also have found a greater price increase
when controls were removed than a price decrease-reduction in inflation-when they
were imposed. However, such estimates are
suspect since a number of other events influenced the economy in 1974 and 1975. If not
explicitly included in the estimation, their influence will tend to be captured by the dummy
variable. The dollar depreciation in 1973 represents one such influence that, due to adjustment lags, could have increased domestic inflation the following year. Another important
influence was the OPEC quadrupling of oil
prices beginning in December 1973, which led
to unprecedented increases in domestic energy
prices shortly thereafter. We will consider first
the introduction of energy shocks, and then
changes in the international value of the dollar.

Equation III of Table 1 adds the contemporaneous, first and second year lagged values of
the percentage change in the real price of crude
petroleum (DIPE). (Additional lags did not
add significantly to the estimation.) These
chang¢sin relative prices are statistically
nificant in the determination of inflation. Induding them reduced the standard error of the
estimation from 1.02 percentage points (Equation II) to .74 percentage points.
It is important to consider the sum of the
coefficients associated with changes in the real
price of oil. A sum significantly different from
zero implies that a one-time change in the real
price of crude would have a permanent impact
on the rate of inflation. Rather, we would expect a one-time change in oil prices to have a
transitory impact on the rate of inflation as far
as its direct effect on price indices is concerned.
In addition, there may be indirect effects that
are longer lasting. Depending on how monetary
policy responds, the shock may get embedded
in inflation expectations, for example. This indirect effect, however, would be caught by the
lagged inflation terms in our regressions. Similar reasoning applies to the indirect effects
caused by the price shocks' impacts on the effective capacity utilization rate.
We therefore tested whether the sum of the
coefficients in Equation III of Table 1 is significantly different from zero. We did so by specifying accelerations in the real price of crude oil
in Equation IV of Table 1: DDIPE = DIPEDIPE - l' The accelerations were entered contemporaneously and with a one-year lag. This
constrained estimation was then compared with
Equation III. The "F" test statistic indicated
that there is no significant difference between
the two estimations at the 95 percent level of
significance. In light of this, we concluded that
the statistical results indicate oil prices have
only temporary inflation effects. We subsequently specified the energy variable in later
estimations as accelerations in the real price of
oil. 11
Table 2 illustrates the estimated impact of a
one-time 10 percent increase in the real price
of crude oil upon the aggregate price level, the
rate of inflation and the change in the rate of

The Real Price of Crude Oil
To estimate the impact of oil price changes
upon the GNP deflator, we used a measure of
the real (or relative) price of crude oil that consists of the ratio of the producer price index for
crude petroleum to the aggregate producer
price index. The producer price index is not
based on a value added in production concept,
as is the Deflator, so it reflects price changes
in crude petroleum used in the U.S. whether
imported or domestically produced. The data
on the annual percent rates of change in the
relative price of crude oil used in this section
are presented in Appendix 1.

The International Value of the Dollar
Changes in the international value of the dollarhaverecently been included in price-markup models 14 because they are believed to affect
the mark-up of domestic prices over domestic
prodllction.·.costs.In • • addition; changes .·in. the
international value of the dollar may affect directly the determination ofwages. For instance,
competitive pressures may induce workers· in
industries competing in foreign trade· to change
their wage demands. Consequently, in our redllced-form model, which combines the price
mark-up and wage equations, changes in the
international value of the dollar may affect both
the mark-up on domestic costs and domestic
costs themselves.
The international value of the dollar may be
expressed in nominal or real terms. The nominal value is simply the trade weighted average
of the dollar's value in terms of foreign currencies. The "real" exchange rate is obtained by
adjusting the nominal rate for differences in domestic and foreign prices. It is the real exchange
rate that appears more relevant in the pricemark-up and wage equations because appreciation of the dollar does not necessarily lead to
cheaper foreign products if foreign prices have
risen proportionally to offset the appreciation.
However, since no convention appears in the
literature establishing the appropriate measure
of exchange rates in empirical estimation, we
present empirical results for both the· nominal
and real exchange rates.

inflation according to the estimates of Equation
IV of Table 1.
During the first year of the shock, the price
level and inflation rate increase .36 percentage
points. The cumulative effect on the price level
is.8percentagepoints; "Whichis reached in the
second year. The rate of inflation is .36 percentage points higher in the first year and .44
percentage points higher in the second year; it
shows no effect of the .energy .price change in
the third year. The volatility in the inflation
rate is illustrated by the change in the inflation
the first and second
rate. After increasing
year by a total of .44 percentage points, the
inflation rate declines .44 percentage points in
the third year. 12 This fairly abrupt change in the
inflation rate is associated with the full adjustment to the oil shock that has occurred in the
level of the price deflator.
Inclusion of these energy prices in the estimation substantially reduces the estimation errors, particularly after 1974. Also, as shown in
Table 1, after including changes in the real price
of crude oil, less of the increase in prices in
1974 and 1975 is attributable to the removal of
wage and price controls, "WPOFF." Before the
consideration of energy prices, the removal of
controls was associated with an increase in the
inflation rate of 2.67 percentage points in both
1974 and 1975 (Equation 2). By expressly incorporating real energy prices in Equation 4,
this is reduced to .93 percentage points. 13

Effects of One-Time Ten-Percent Rise in the Real Price of Crude Oil*

(percentage points)

Chauge in the
Inflation Rate


Change in
Price Level

First Year of the Shock




Second Year




- .44



Third Year

* Based on Equation IV of Table 1. Inflation measured by GNP Implicit Deflator


nificance according to the Fstatistic) to the determination of U.S. prices when both the
contemporaneous and one-year lagged accelerations are included.· Based on the results from
Equation VI, a 10 percent real appreciation of
the dollaris c()l'felatedWithchangesirith6 level
and rate of change in the GNP Implicit Deflator, as shown in Table
According to this estimated response in the
U.S. price level, the period required to adjust
to changes in the dollar's exchange rate is two
years. Within the fil'styearafteralO-percent
appreciation, the U.S. price levelis .6 percentage points below what it would. be otherwise.
By the second year, it is 1.7 percentage points
lower. The rate oUnflation is correspondingly
.6 percentage points lower within the first year
of appreciation, and it is 1. 1 percentage points
lower than it otherwise would be during the
second year. The inflation rate during the third
year shows no effect related to the one-time
change in the dollar's international value.
The change in the inflation rate illustrates the
variation in inflation in response to changes in
the international value of the dollar. The inflation rate is .6 percentage points lower within
the first year than it was the year before the
dollar's appreciation. It continues to fall in the
second year by .5 percentage points. By the
third year, after the price level has fully adjusted, the inflation rate increases by 1. 1 percentage points. In that third year, the inflation
rate is back to what it would have been if the
dollar had not appreciated.

We anticipate that a one-time change in the
dollar's exchange rate will lead to transitory
changes in the U.S. inflation rate but not to a
permanent change. This suggests .that the. estimated coefficients associated with changes in
the• dollar's • •. exchao.gerate.inour •in;flation
model should sum tozero,as was the case for
oil price shocks.• Wetherefore constrained the
sum to zero by specifying accelerations in the
exchange rate, as we did in/the previous estimation for the effects of oil. price shocks. We
let .DEX represent the yearly percentage
change in the exchange rate of the dollar. The
acceleration in the exchange rate, DDEX, thus
Was the difference in the yearly percentage
changes in DEX, DDEXt = DEXt DEXt - 1 .
Table 1 provides the empirical results from
the estimation of Equation 9 using two values
of the dollar's exchange rate-the nominal effective bilateral exchange rate, EXB, and the
real effective bilateral exchange rate, REXB.
Data for both EXB and REXB are provided in
Appendix 215 •
We anticipate that an appreciation (depreciation) of the dollar would be correlated with a
decline (increase) in U.S. prices, and therefore
that the sign associated with the estimated parameters would be positive. Equations V and
VI in Table 1 include the acceleration in the
two exchange rates in the contemporaneous
year (DDEX) and lagged one year (DDEX_ 1).
The estimated coefficients are all of the expected sign. In each case, the exchange rate
adds significantly (at the 5 percent level of sig-



Effects of a Ten-Percent Appreciation in the Real Bilateral Exchange Rate*
(percentage points)

Change in the
Inflation Rate

Inflation Rate

Percentage Change in
the Price Level

Year of the appreciation




Second Year




Third Year




• Based on Equation VI of Table 1. Inflation measured by GNP Implicit Deflator


IU. Estimates of the Stable-Inflation Capacity Utilization Rate
A purpose of this study is to estimate the
stable-inflation capacity utilization rate with
the additional data since 1977 and the expanded
specificiltion which inclucl~s both energy price
shocks and changes in the international value
of the dollar. The additional variables played a
significant role in determining inflation, particularly since mid-1975 as we have discussed
above, and, if not expressly included in the
econometric specification, could bias our estimation. Moreover, the additional data cover a
period in which inflation was extraordinarily
Following McElhattan (1978), the stable inflation capacity utilization rate, cue, is estimated as
cue = - bo fbI ,

lier study for the period 1954-1973. The model
for that period included the simple relationship
only betwe~n the change in the inflation rate
anq the level of capacity utilization. The saI11e
simple relationship, but with an expanded estimation period of from 1954-1977, also
yielded an 81.9 percent stable-inflation capacity
utiliZiltion rMe. However, as indicated by the
wider 95 percent confidence interval, the additional data introduced greater uncertainty
with regard to the population value of the stable-inflation capacity utilization rate. That uncertainty is related to the sharp variations in
inflation that occurred at the same time as the
removal of wage and price controls, the qua.drupling of the OPEC oil prices and the sizable
changes in the international value of the dollar.
In our current study, with annual data from
1959-1983 and with the inclusion of the three
types of price-shocks mentioned above, the stable-inflation capacity utilization rate estimate is
81.7 percent. In addition, the precision of this
estimate is improved over the 1954-1977 period as illustrated by the fairly narrow 95 percent confidence interval of 78.5 percent to 83.6
The stability of the estimate for cue over
time is particularly notable given the behavior
of inflation. After the mid-1970s, inflation was
much higher and more variable than in the earlier period. Inflation averaged 3.0 percent per

where bo and b 1 are estimated from Equation
9. Since capacity utilization is positive, this suggests that the constant term in Equation 9
should be negative and the coefficient on CU
should be positive, which they are in Table 1.
Referring to Equation 9 again, this estimate of
cue also assumes that supply shocks have no
inflationary impact in the long run, which is
consistent with the results reported in Table 1. 16
The regression estimates in Table 1 indicate
a· remarkable stability in cue with respect to
model specification. In addition, cue has been
stable over time. As shown in Table 4, the estimate of this rate was 81.9 percent in my ear-

Estimates of the Stable-Inflation Capacity Utilization Rate
(Annual Data)
Estimation Periods*



Stable-Inflation Capacity
Utilization Rate




95 % Confidence Limits




* Equations associated with these estimates:

1954-1973, see McElhattan (1978), Equation (1), Table 1
1954-1977, see McElhattan (1978), Equation (2), Table 1
1959-1983 Equation VI, Table 1


year between 1954 and 1973. Its standard deviation during that time was .1.5 percentage
points, or 50 percent of its .mean value. Over
the entire 1954__1983 period, the variation in
inflation rose to 61.4 percent of its mean value.

According to our estimates, neither the substantially. higher .averctge illflation rate nor the
increased uncertainty with regard to that rate
apparently altered the stable-inflation capacity
utilization rate.

Comparison of Capacity Utilization and •Unemployment Rates in. the
Inflation Model: 1959-1983

Equation I

Capacity Utilization Rate

Equation III

( -4.2)


Equation n

( -2.4)



Unemployment Rate



( -1.9)

( -1.3)

( -1.9)


Wage & Price Controls:

- .329
( -2.7)



Acceleration in the
Relative Price of Oil:
Current year







Current year




Last year




Last year



Acceleration in the Real
Bilateral Value of the Dollar:

Stable-Inflation Rate
95% Confidence Intervals





Summary Statistics
Standard Error







IV. Comparison of Capacity Utilization and Unemployment Rates
as Inflation Signals
cent. This appears too wide a range to provide
a very useful policy guide.
'fhe statistical•• F-testenablesus todetennine
whether adding capacity utilization to Equation
II to obtain Equation III would significantly improve the determination of changes in the inflation rate. The F~statistic of 7.4 compared to
the critical value of 4.5 suggests that capacity
utilization does add significantly to the determination of changes in inflation, above and beyond any information provided by the unemployment rate.
However, when unemployment is added to
Equation I to get Equation III, the comparable
F-statistic is only .24. This means that the civilian unemployment rate does not provide any
statistically significant information once we use
the capacity utilization rate in the same determination of changes in the inflation rate. We
interpret this as evidence that capacity utilization has been a more informative inflationary
signal than the unemployment rate.

We suggested earlier that as a signal of inflation, capacity utilization may be a more reliable
policy guide than the. unemploymentrat~
examine that proposition more formally, we
next. compare the.. performance of unemployment and capacity utilization rates in determininginflation. For convenience, Equation Iof
Table 5 repeats Equation VI from Table 1. In
Equation II of Table 5, the unemployment rate
replaces the capacity utilization rate as the
proxy for excess demand. In Equation III, both
the capacity utilization and unemployment
rates are included.
In. comparing Equations I and II of Table 5,
we find that the standard error of the regression
is less and the correlation is higher when capacity utilization is the proxy for excess demand
in the estimation. The larger standard error and
uncertainty associated with the unemployment
rate estimates also is reflected in the fairly wide
95 percent confidence interval for the natural
rate of unemployment, 5.1 percent to 9.1 per-

V. Inflation In the 1980s and Policy Implications
by the increase in the relative price of oil which,
in 1980 and 1981, was 70 percent. The fast rising oil price alone pushed inflation up by 2.8
percentage points in 1980 and 1981. Other factors (representing the estimation error) added
.2 percentage points. On balance, inflation increased .9 percentage points between 1979 and
1981, although the dollar appreciated and the
economy was producing below its potential.
Between 1981 and 1983, the inflation rate reversed course, declining from 9.6 percent to 3.8
percent. That reduction was associated with a
fall in capacity utilization to a post-war low in
1982 and a partial recovery during 1983. The
relatively low capacity utilization over those
years reduced the inflation rate by 2.5 percentage points. In addition, the real price of crude
oil dropped almost 20 percent. This decline,
coupled with the fact that the large energy price
increases in 1980 and 1981 had largely worked

This section details how the change in inflation in the 1980s is explained by the inflation
model as estimated in Equation VI Table 1. In
1980 and 1981, inflation increased continuously
and reached a record high of 9.6 percent in
1981, despite weak aggregate demand that was
working to reduce the inflation rate in those
years (illustrated in Table 6). Capacity utilization averaged 79.6 percent and 79.4 percent in
1980 and 1981, respectively, holding the inflation rate down .3 percentage points in each
year. In addition, the continued appreciation of
the dollar added downward pressure on inflation. The 13 percent (as measured by the real
effective bilateral exchange rate) increase in its
value is estimated to have reduced inflation by
a total of 1.5 percentage points over the 198081 period. However, the depressing effects of
both the domestic economy and the international value of the dollar were more than offset

their way through to a higher price level by
1983, resulted in a sharp deceleration in the
change in the price of crude oil. Alone, the
deceleration in energy prices decreased theaggregate inflation rate by a total of 4.3 percentage points in 1987an44983.
The international value of the dollar continued to appreciate in 1982 and 1983 a
substantially smaller amount than in 1981. This
meantthat.the depressing effects on theaggregateinflationrate in 1983 were significantly less
than in .1981 and .1982. On balance, the contribution of an appreciating dollar to holding
down inflation was 0.9 percentage points less
in 1983 than in previous years.
Over 1982 and 1983, our inflation model
overestimated the 5.8 percentage point decline
in inflation by only .3 percentage points. The
results clearly highlight the importance of economic slack and supply-side price movements
in both the decline and volatility of inflation
since 1981. The inflationary model of excess
demand and supply-side shocks explains the
sharp decline in inflation-even as the economy was experiencing its strongest cyclical recovery in the post-war period.
This discussion illustrates the applicability of
conventional inflation models to describing the
degree and volatility of inflation. It also illus-

tratesthe relevance of the important, but temporary, inflationary effects of supply-side
shocks, such as changes in the real price of
crude oil and in the international value of the
dollar, to macro policy decisions. In essence,
a~y .inflationary•• increase•• • duemasupply-si<ie
p(:iceshock will be temporary. Once the aggregate .price level adjusts to a higher relative
price, the inflation rate will drop levels
that would have existed without those shocks
and which reflect aggregate demand pressures.
Conversely, any decline in inflation associated
with an appreciation of the dollar or decline in
the real price of crude oil will be temporary.
Once the aggregate price level has adjusted for
lower relative prices, the aggregate inflation
rate will increase, reflecting the fact that the
benefits from the lower energy and/or traded
goods prices are over. Ultimately, the inflation
rate will reflect aggregate demand pressures in
the domestic economy. According to our estimates, adjustments to supply-side shocks take
about two years to work their way through the
price level.

Inflation Since 1979


Changes in
the Inflation Rate

Percentage Point changes in the inflation rate due to:


Real Price
Crude Oil







































In. this paper, we have derived aninflation
model that is the reduced-form of conventional
wage and price equations of a more complete

during the Nixon Administration, changes in
the international value of the dollar, and
changes in the real price of crude oil. We found

contains at
two excess demand measures: the unemployment rate,which
proxies for slack in the labor markets, and the
capacity utilization rate, .which proxies for excess demand in final product markets. Because
of the close correlation between the two, either
unemployment or capacity utilization may
serve as a general measure of excess demand
in the economy.
Conventional Phillips Curve models emphasize the relationship between inflation and the
unemployment rate. We focus upon inflation
and the capacity utilization rate. From the conventional Phillips Curve, analysts derive the
natural rate or stable-inflation unemployment
rate. This is a rate which, if maintained, is associated with no change in the inflation rate.
Similarly, we have found a stable-inflation capacity utilization rate, and estimated this rate
to be about 82 percent (with its 95 percent confidence interval between 78.5 percent and 83.5
In this paper, we also have introduced supply-side shocks into the conventional correlation between inflation and excess demand.
These shocks include wage and price controls

price of crude oil
termination of inflation during
Changes in the inflation rate over the
years are due almost as much to
rary price shocks as to the fUlld~lml~ntal
lation between excess demand and mtlilHon.
The negative relationship
ployment and inflation has rec:ehred
tion, but we believe the capacity
may be a more reliable indicator of inflation.
Our belief rests on the observation thatthe stable-inflation capacity utilization rate has remained steady over time, making it a reliable
standard. In contrast, a good deal of uncertainty surrounds the estimate of the natural rate
of unemployment. This uncertainty may have
led to some inflationary bias in past policy decisions to the extent that policymakers and others have tended to underestimate the natural
rate and recommend policy actions which in
retrospect were too stimulative. The use of capacity utilization rates to gauge inflationary
pressures therefore also may be helpful as an
additional check on the inflation assessments
based on unemployment measures.


ApPa:NDIX 1
Percentage Change in the ReaFPrice of Crude Oil

(Producer Price Index)




- .2
- 2.7
- 1.2





Percentage Change in Bilaterally Weighted Exchange Rate (EXB)
















- 1.6
- 5.1


- 7.7
- 4.7

Percentage Change in the Real Bilateral Exchange Rate (REXB)






Source: Board of Governors of the Federal Reserve System, MPS database.
REXB = PDIEXB*PF where PD represents U.S. prices
PF foreign prices.








(Dependent Variable: Rate of Inflation as Measured by the GNP Implicit Deflator
Estimation Period: 1954-1982*)






( -4.0)

Capacity Utilization Rate
Inflation Lagged One Year
IR_ 1
Wage/Price Controls "On"

Wage/Price Controls "Off"
Deviation of Productivity from Trend













( -1.3)

(- .7)

( -1.27)

Inflation Lagged Two Years
IR_ z
Inflation Lagged Three Years
IR_ 3


Summary Statistics













* t-statistics in parentheses.



1. See references in McElhattan, 1978, and in the international study by Tavlas.
2. For a view of the mark-up model that describes competitive market behavior, see Nordhaus. For a view that
pertains to noncompetitive behavior, see Modigliani (1958).

cent increase in the real price of crude oil in the first period
mElans that. in the second period the oil price decelerates
by 10 percent. Relating this to the coefficients for DDIPE
and DDIPE -.1 in Equation IV, the 10 percent acceleration
in the first year and the subsequent 10 percent deceleration
leads to a net second year impact on the change in inflation
of (.44
.36 =) .08.

3. See, for instance, Modigliani and Papademos, in which
unemployment rather than capacity utilization is used in an
inflation model.

13. Recently, it has been argued that due to differences in
adjustment costs, inflationary responses will vary according to whether the real price of energy is increasing or
falling (see Promboin). According to this argument, an energy price increase may render obsolete some portion of
the capital stock. Also, an energy price decline may render
obsolete certain capital items that might be energy-efficient
but too expensive to operate if energy costs become less
important. However, there is no reason to believe a priori
that these costs are the same and that, therefore, energy
price changes have symmetric inflationary effects. To test
this hypothesis, Equation III from Table 1 was modified to
include a variable that is equal to the change in energy
prices when the change has been positive, and zero elsewhere. This added variable was entered contemporaneously and with the first and second year lagged values.
The results indicate that there was no significant difference
in the impact of changes in the real price of crude oil related
to the sign of the change. The implication is that adjustment
costs to energy shocks are not statistically different during
the estimation period depending upon whether the energy
shock increased or lowered the real price of crude oil.

4. See
5. For a description of how the capacity utilization rate series are constructed see the Federal Reserve Bulletin of
February 1978 and July 1963.
6. See the discussion in Gordon and in Frye and Gordon.
7. For a discussion, See Dornbusch and Fischer.
8. See, for instance, John Tatom and Tavias' study of the
OECD countries.
9. In recent studies of inflation models, Frye and Gordon
have suggested that the productivity measure relevant in
the price equation may differ from that in the wage equation
because opposing sides in labor negotiations have different views of productivity. If that is the case, Frye and Gordon illustrate that the relevant measure of the deviation of
productivity from its trend has a small but statistically significant effect upon inflation in the contemporaneous
quarter. I have considered a similar measure in the estimation of Equation 1, in Appendix 3. DPR represents the
deviation of actual from trend productivity, but it is not statistically significant in my estimation. Adding additional lags
does not change that result. This finding may be due to
inflation estimates that use annual data rather than quarterly data as did Frye and Gordon's work. Annual data allows time for offsetting quarterly effects to occur. Not finding productivity statistically significant, I have dropped the
productivity term from further estimates and discussions of
the inflation model in this paper.
10. A possible alternative to the dummy variable technique
has been suggested by Blinder, as described in Frye and
Gordon. Blinder constructed a variable to represent the
impact of controls. It is equal to the fraction of the CPI
subject to price controis in each month, based on government records for the period between August 1971 and May
1974. However, Frye and Gordon have compared the
Blinder methodology with the simple dummy variable approach similar to that used in this paper and concluded that
the Blinder series provided neither a better fit nor an evaluation 01 the controls that differs from the simple dummy
variable approach. Therefore, I have applied only the
dummy variable technique in assessing these controls.
11. This specification constrains the sum of coefficients of
the rates of change in real oil prices, DIPE, to be zero. A
constrained estimation based on this priori knowledge provides more efficient estimates than unconstrained estimates. This is our reason for continuing estimations with
accelerations in real oil prices.

14. The MPS model incorporates the real biiateral exchange rate and Gordon's work uses the effective multilateral exchange rate.
15. The exchange rate data is taken from the Board of
Governors of the Federal Reserve System database for
the MPS model. The Bilateral Exchange rate is a 10-country weighted index.
For the period 1955-1976 the series is a geometrically
weighted average of exchange rates with the following
countries. The weights are average bilateral trade shares
(shares of trade with U.S.).

The exchange rate data were taken from IMF, International
Finance Statistics.

16. In McElhattan (1978), the change in inflation was expressed as a function of the difference between actual capacity utilization (CUt) and the stable inflation capacity utilization rate (cue).

12. The change in the inflation rate in the second year is
only .08 percentage points because the one-time 10 per-




b 1 (CUt - cue).

By incorporating the parameter
the equation can be written as
DIRt = bo





cue in the constant term,
b 1 CUt,

This leads to the expression

Promboin, Ronald L., "Energy Shocks and the U.S. Economy: How Much has OPEC Helped Recently," Business Economics, July 1984.

Federal Reserve Measures of Capacity and Capacity Utilization, Board of Governors of the Federal Reserve
System, February 1978.
Frye, Jan and Robert J. Gordon, "The Variance and Acceleration of Inflation in the 1970's: Alternative Explanating Models and Methods," NBER Working Paper
551, September 1980.

Scadding, John, L., "Inflation: A Perspective from the
1970s," The Economy in the 1980's, Michael Boskin,
editor, Institute for Contemporary Studies, San Francisco, California. 1980.

Gordon, Robert J., "Inflation, Flexible Exchange Rates,
and the Natural Rate of Unemployment," NBER Working Paper 708, July 1981.

"The Sampling Distribution of the Liviatan Estimator of the Geometric Distributed Lag Parameter,"
Econometrica, Vol. 41, May, 1973.

McElhattan, Rose, "Estimating a Stable-Inflation Capacity
Utilization Rate," Economic Review, Federal Reserve
Bank of San Francisco, Fall 1978.

Tatom, John A., "Energy Prices and Short-Run Economic
Performance," Federal Reserve Bank of St. Louis,
Economic Review, January 1981, Volume 63, NO.1.

Modiglianl, Franco, "New Developments on the Oligopoly
Front," Journal of Political Economy, Vol. 66, June

Tavlas, George S., "A Model of the Inflationary Process in
Six Major O.E.C.D. Economies: Empirical Results and
Policy Implications," Journal of Policy Modeling, (5),

Modigliani, Franco and Papademos, Lucas, "Monetary Policy for the Coming Quarters: The Conflicting Views,"
New England Economic Review, Federal Reserve
Bank of Boston, (March/April 1976).
Nordhaus, William D., "Recent Developments in Price Dynamics," In The Econometrics of Price Determination
Conference, Oct. 30-31, 1970, Washington, D.C.,
Board of Governors of the Federal Reserve System
and the Social Science Research Council.


by JOrg Niehans*

What is called the "international debt crisis", like every complex
problem, has many aspects. The present paper concentrates on one
of them-the unenforceability of foreign claims. Recent developments make the investigation of the implications of unenforceable
claims for the international credit market particularly worthwhile.
The resulting picture of international loans to developing countries
is inevitably incomplete, but it is nevertheless revealing and, in part,

Prevailing views on the international debt
crisis are based on the notion that the international loans in question are essentially similar
to private domestic loans. In particular, the
"conventional wisdom" assumes that debtors
can be expected to service their debt to the
limit of their ability. From this assumption, the
current crisis appears to be a consequence of a
diminished ability to pay, partly due to higher
interest rates and partly to deteriorating economic conditions. This view raises the hope
that a decline in interest rates and a return to
prosperity would cause the crisis to fade away.
It is argued in this paper that the conventional wisdom overlooks a fundamental difference between private domestic loans and the
international loans in question. Economic
thinking about borrowing and lending is commonly based on the paradigm of debt enforceable through the law. In a cash transaction,
there is an immediate quid pro quo-the con-

tract is self-enforcing. In a credit transaction,
the two parts of the exchange are separated in
time. As a consequence, there have to be procedures to enforce the fulfillment of the contract. Because ethics are usually not enough,
enforcement becomes a problem. Western civilization has solved this problem through an
elaborate system of contract and bankruptcy
laws. If a debtor defaults on his obligations, he
forfeits collateral, that is, his assets can be attached, impounded, or turned over to his creditors by a bankruptcy court. By such means, the
debtor is put under strong pressure to live up
to his obligations, no matter how onerous, to
the limit of his abilities. Without these legal
enforcement mechanisms, modern credit systems could not have developed.
The circumstances are very different for international bank loans to developing countries.
Most of these loans are made to governments
or guaranteed by governments. To the extent
they are loans to private firms, exchange control tends to subject them to a calculus similar
to that for government loans. There is usually
no collateral and, at least in practice, no access
to bankruptcy courts. While the seizure or attachment of assets is conceivable, it is rarely

*Professor, University of Bern and Visiting
Scholar, Federal Reserve Bank of San Francisco. Comments from staff economists and the
research assistance provided by Paula Crosswright were valuable and greatly appreciated.

feasible. This makes such loans legally unenforceable. I It is true that concern about "political" repercussions, about "gunboat diplomacy"
or about disruption of trade may, to a certain
extent, substitute for legal remedies, but they
seem relativelY weak. World opinion today
tends to side with the debtors rather than the
creditor banks. 2
The analysis in this paper is based on the
assumption that such penalties do not exist at
all. Its specific contribution is the development
of an "unorthodox" model in which concern
about the future availability of credit is the only
deterrent to default. No doubt this radical assumption does not do justice to the complexities of reality because at least some traces of
enforceability are often present, and even the
principles of ethics may be of some help. There
are, indeed, historical cases of governments
faithfully repaying their foreign debts over decades. Recent developments strongly suggest,
however, that it is worthwhile to investigate the
conditions under which the international credit
system might still be viable under this radical
In the last few years, the implications of
unenforceability have found growing attention. 3 The present paper is written in the belief
that these implications are not yet fully understood. Section I outlines the statistical contours
of the problem. Section II shows that unen-

forceable claims do not necessarily lead to crisis. Sections III and IV analyze, respectively,
the objectives of rational debtors and creditors.
Using a specific model of creditor strategy, Section V pays particular attention to the relationship between the rate of interest and the rate
of growth. Section VI extends this analysis to
the initial "overshooting" of the long-run debt
level. The exposition uses verbal, graphical and
mathematical arguments. A reader not worried
about mathematical precision can obtain the
main content of the paper by concentrating on
words and graphs.
Throughout the paper, it will be assumed that
international lending, while important for the
levels of output and consumption at any moment, has only a negligible influence on the
rate of economic growth over decades, the latter depending mainly on population, natural resources and technological progress. Although
this assumption may not be strictly valid, the
inaccuracies involved do not seem large enough
to invalidate the conclusions. Uncertainty
about future developments, although obviously
important in reality, also is disregarded in this
paper. The paper thus does not purport to provide a complete theory of lending with unenforceable claims, but concentrates on certain
aspects that seem to be important from the
point of view of current debt problems.

I. Statistical Contours
As a preface to the theoretical argument, this
section presents some statistical contours of the
current debt crisis. These contours are intended
to show that the interpretation developed in
this paper, although based on a somewhat radical assumption, is consistent with important
stylized facts. The data are taken from the
World Debt Tables (External Debt of Developing Countries), published by the World
Bank, 1983-84 and the first supplement. From
1978 to 1982, as Table 1 and Chart 1 show, disbursements of private loans through financial
markets to public debtors or debtors with public
guarantees were in the range of $39 to $48 bil-

lion. Repayments of principal moved roughly
between $14 and $18 billion. Neither of these
flows displayed a pronounced trend. Interest
payments, however, increased dramatically
year by year from $6 billion in 1978 to $24 billion in 1982. As a consequence, net transfers to
the governments of developing countries
through the international banking market, representing the net cash flow to the debtors, declined from $19 billion to less than $2 billion.
With little exaggeration, it can be said that net
transfers came to a standstill.
This was immediately followed by the outbreak of the international debt crisis. Based on

Chart 1
Developing Countries:
Public and Publicly Guarantf3ed Debt to Private
Creditors Through Financial Markets, 1973-1982
Billions of Dollars





10 ----


1974 1975 1976 1977 1978 1979

1980 1981


Developing Countries: Public and Publicly Guaranteed Debt
to Private Creditors Through Financial Markets, 1973-1982.

(Billion Dollars)
Interest Payments
Principal Payments
Total Debt Service
Net Transfers
















Source: World Bank, World Debt Tables: External Debt of Developing Countries, 1983-84 and First Supplement,
Washington 1984.


from 1973 to 1982. It is significant that their
amount, after reaching a high point of $11 billion in 1978, actually turned negative in 1982.
It is also important to note that this negative
balance was far from alarmingly large. The
overall burden of the debt cannot have been
very heavy.
The Latin American countries can be divided
into those with rescheduling between 1974 and
1982 and those without. The difference is revealing. The 15 rescheduling countries present,
although at a somewhat lower level, the same
picture as Latin America as a whole. The 10
"good" debtors, however, show roughly stable
and positive transfers since 1977; in the last
three years, their net transfers actually increased. This observation fits nicely into the
aggregate picture. 5
According to another hypothesis, it is precisely the prospect of negative transfers that
triggered the debt crisis. This hypothesis is consistent with the effective unenforceability of
most of these claims because it is an essential
aspect of unenforceable claims that the debtor
cannot be forced to accept negative transfers.
If, and when, they appear on the horizon, he
defaults. The present paper hypothesizes that
this hypothesis is largely correct.
It is true that most applications for debt renegotiation are accompanied by arguments intended to show that acute balance-of-payments
problems make it impossible for the borrower
to pay his debts. Such arguments,
should not be accepted at face value. Balance-

the International Monetary Fund's (IMF) chronology of bank-debt restructuring cases (IMF,
1983), there were 3 completed cases in each of
1978 and 1979, 6 cases completed in each of
1980, 1981 and 1982, and 20 cases completed in
1983 by early October with 8 cases still under
negotiation. 4
It is difficult to believe that the association
of the virtual vanishing of net transfers and the
wave of reschedulings was a mere coincidence.
The nature of their causal connection, however,
is not obvious. According to one possible hypothesis, borrowers asked for restructuring because rising interest rates, together with deteriorating terms of trade and worldwide
recession, had made the burden of their debt
unbearably heavy. Evaluated against the statistical contours, this argument is not convincing. It is true that total debt service, consisting
of payments on principal and interest, reached
a high amount. The relevant burden of international debt, however, is the net transfer obtained by deducting from debt service the disbursements on new loans. Since these
disbursements were also very high, there was,
until 1982, not a single year with a negative net
transfer. Far from carrying an intolerable debt
burden, debtors received net benefits throughout, albeit on a rapidly declining scale.
The preceding argument was based on aggregate data for all developing countries. A more
detailed picture can be obtained by focusing on
Latin America and the Caribbean. Table 2 and
the corresponding Chart 2 present net transfers

Latin America and the Caribbean: Net Transfers from Private Creditors
Public Debtors through Financial Markets, 1973-1982

(Billion Dollars)
All Countries
































Source: World Bank, World Debt Tables: External Debt of Developing Countries, 1983-84 and First Supplement,
Washington 1984.


of-payments crises result primarily from a
country's own policies. They can be produced
at will, virtually overnight, simply by overvaluating the exchange rate. 6 The fact that IMF
lending, despite the collapse of the Bretton
Woods system, still is largely conditional on a
balance-of-payments crisis creates an incentive
for a country to let itself slip into such a crisis

whenever IMF lending is desired. 7
The rough statistical contours presented in
this section, although far from proving anything, suggest that net transfers, defined as the
difference between new loans and debt service
on old loans, may be of crucial significance for
international solvency. The following sections
elaborate on the theory behind this idea.

Chart 2
Latin America and the Caribbean:
Net Transfers, 1973-1982
Billions of Dollars


Millions of Dollars




All Countries ...









Rescheduling ...




0 1 - - - - - - - - - - - - - - - - -.. . . . . .

-4 "'"----a._
1973 1974


_r.- _







II. Lending Without Budget Constraints
This section considers the long-run feasibility
of lending with unenforceable claims. Goods
exchanged in the market are, at market prices,
of equal value. Their exchange is nevertheless
of mutual advantage because the goods, despite
their equivalence, have different utility for the
buyer and the seller. Similarly, the future payments promised by a borrower, properly discounted, have a present value equal to the
amount he borrows; the present value of all his
cash flows is zero. This is the intertemporal
budget constraint for loans. Despite this constraint, the loan is regarded as advantageous by
both parties because a dollar may have different
utility depending on the time at which it is
An essential implication of an intertemporal
budget constraint is that once the loan is disbursed, its present value to the borrower turns
negative. For example, once a homeowner has
received a mortgage loan, he has to make payments to the bank for years. Economically,
therefore, there is a virtually irresistible incentive to default. To counteract this incentive,
loan contracts have to be enforced by collateral
and bankruptcy courts.
Such a budget constraint, by analogy with individualloans, is often postulated for an economy as a whole (McDonald, 1982; Sachs, 1983;
Sachs and Cohen, 1982). As a matter of fact,
the unenforceability of loans means that there
is no way of imposing this constraint. The paradigm of unenforceable debt thus implies that
borrowing is not subject to the familiar intertemporal budget constraint. The constraints
that take its place are the main topic of this
In the absence of a budget constraint, a councan forever borrow more each year than
what it needs to service its outstanding debt.
International lending begins to look like a
"Ponzi scheme." Economists, conditioned to
the paradigm of enforceable claims, tend to regard such a state of affairs as intrinsically unsustainable. As a matter of fact, it may conceivably be sustained, although only under
stringent conditions. Even firms, if they grow

rapidly enough, may end up never repaying
their aggregate debt. The same applies to growing economies. While the budget constraint is
faithfully observed for every single loan, the
present value of aggregate cash flows to the
debtor may well be positive and even infinite.
The possibility of such a case over an indefinite period depends largely on the relationship
between a country's rate of growth and the rate
of interest. 8 Consider a country in balanced
growth at rate g with net foreign debt A. Debt
expands at the rate a = (dA/dt)(l/A). There is
clearly no reason why a should not be positive
in such an economy for an indefinite period.
Specifically, in a world economy in balanced
growth, there is a continuing capital flow from
the creditors to the debtors, growing exponentially at the same rate as debt. Aggregate debt
is never repaid.
The net cash flow to the debtor country depends on the difference between new debt and
the interest payments on the existing debt. It
corresponds to the net transfers of Section 1. If
there is a world interest rate i, the cash flow is






(a - i) A.


Disregarding other service items, this is equal
to the trade deficit. Depending on a and i, c
may be positive or negative, which means that
debtor countries can have trade deficits or
Is it possible for a debt expansion rate different from the interest rate to be maintained
forever? The answer depends largely on the relationship between a and g. A rate of debt expansion significantly different from the rate of
economic growth would imply that debt either
increases beyond any limit relative to national
income or else virtually shrinks away. It is unlikely, therefore, to be maintained for very
long. Within the framework of balanced
growth, one is dearly forced to assume a = g.
The above question thus pertains in essence to
the relationship between the rate of growth and
the rate of interest.
In balanced growth with g = i, the net cash

flow is zero at all times and so is the trade balance; interest payments are continuously reinvested. With g < i, the debtor needs a trade
surplus to finance the excess of interest payments over new lending; cash flows are negative and their present value is equal to outstand~
ing debt. 9 In the case g > i, the debtor enjoys
a positive cash flow forever, which finances a
permanent trade deficit; the present value of

its international indebtedness by more than its
interest payments virtually forever, thereby finallcing a· permanent trade deficit. lO Are the
creditors cheated by the debtors in such a case?
This· analogy to fraudulent bankruptcy would
bellllSplaced;" In successive generations, each
individual creditor obtains full repayment. It is
just that .the number of such creditors grows
continuously. If the world suddenly came to an
erid, the last generation of creditors would indeed regret having made those loans, but its
regret would be shared by those who hold other

There seems to be no general reason to rule
out any of these possibilities. In particular,
there is nothing intrinsically unsustainable in a
situation in which a growing economy increases


The Borrower's Objective

The concept of balanced growth helps to clarify one's thoughts on debt problems. The
world, however, is not in balanced growth. This
raises the question, under what conditions can
a debt crisis generally be avoided, despite the
unenforceability of claims. In the present section, this question is considered from the borrower's point of view. The analysis assumes that
the rate of interest is given by the market, but
that the available amounts of loans may be limited. Although potentially important, risk premia depending on the borrowing strategy of the
particular country are disregarded.
If a country with unenforceable debts is able
to maintain a positive cash flow at all times,
then its aggregate borrowing is not subject to
the usual efficiency criteria according to which
the marginal return on investment must be no
lower than the rate of interest. If loans, in effect, turn into gifts, they cannot be excessive
from the borrower's point of view. The overriding objective of borrowing becomes the maximization of the present value of future cash
This Objective can be formalized by postulating that the finance minister in the borrowing country at time zero, in choosing the time
profile of debt, A(t), seeks to maximize

f[ ~~ -

iA(t)] e-itdt.

If this expression is infinite for several paths,

he .presumably will prefer the path that approaches infinity, in some sense, as fast as
In pursuing this objective, the borrower is
constrained not by the cost of future debt service but by the willingness of creditors to lend.
If the interest rate is assumed to be given by
the world market, this constraint expresses itself in a quantitative limitation of the loan supply. The nature of this constraint will be discussed in the following section.
The essential point is that the optimal strategy of the debtor may well entail default. 11
More specifically, a "crisis" will occur if, and
when, the present value of cash flows becomes
negative over future time spans of any length. 12
In more formal terms, default will occur at time
T if

L[ ~~ -

iA(t)}-i(t - T)dt < 0


for aUe from T to infinity. The unenforceability of contracts means precisely that the debtor
cannot be compelled to accept the prospect of
paying out more than he receives for an indefinitefuture. 13
The foregoing argument is illustrated in Figure 1. The blue curve describes a path of expected annual cash flows from now into the
indefinite future. These cash flows are concep-



tually similar to· the. net transfers in Tables 1
and 2 and Charts 1 and 2 except that (1) they
relate to future periods and (2) the dollar value
for each year is assumed to be discounted back
to the present time at the appropriate interest

the left, but these do not trigger default because
the debtor expects positive cash flows of even
larger size in the future. 14 In the other case,
represented by the dark lines, the upper curve
has a maximum that is never surpassed. At that

curve by cumulating the annual cash flows from
time zero to a given future year. It thus represents the present value of all future cash flows
over a given future time period. For the moment, one may imagine that the shape of these
curves is imposed on the borrower from the
Under what conditions will the debtor, confronted by these curves at time zero, detect a
future time T at which default is advantageous?
The two cases of solvency and default are illustrated by two variants of each curve. In the
solvency case, represented by light lines, the
upper curve has no (global) maximum. Whatever the present value of future cash flows up
to a given time, there is a later time promising
a still higher value. There may indeed be subperiods of negative cash flows, illustrated on

further servicing of debt
Inreality, of course, the shape of these curves
is not given to the debtor but depends in part
on the debtor's own policies. It is in the debtor's
interest to choose those policies that would
raise the upper curve as high as possible. In
particular, a path with infinite present value is
always better than any default path since the
latter's present value is finite. This means that
the debtor will often do his best to escape default. In contrast a path with solvency is not
necessarily better than default because, although it never declines, it may never reach the
level of the default path at the crisis point. This
possibility is illustrated by the dotted solvency
path in relationship to the black default path.
It means that there may be no incentive for the
debtor to remain solvent.

Figure 1
Net Cash Flows
$ (Discounted)



Cumulative Cash Flow



intendtW result of such negotiations is a further
increase in the cash fl6wto the borrower, pushingthe maximum further up. As Sachs (1982)
pointed out, outright repudiation is most likely
if the creditors cannot be brought to the barga.iningtable;This is consistent with the observation that repudiation was frequent when most
lending took the form of bonds whose numerousowners could not speak with a common
voice. Today, with most lending done through
banks, default usually appears in the form of
renegotiation. (but renegotiation does not necessarily involve default).

The shape of the curves <Uso depends on circumstances beyond the debtor's control. Thus,
a change in interest rates may produce a downturn where none was anticipated before. Even
more importantly, the shape of the cash flow
curve. is heavily influenced by the policies of
the lender. This aspect will be taken up in Section IV.
In general, default will not take the form of
an outright· repudiation of existing obligations.
The.rationaI debtor will, instead, use the threat
of repudiation to induce his creditors to negotiate a rescheduling of debt, a lowering of interest rates, and an extension of new loans. The

IV. Strategic Planning for Creditors
If, with unenforceable contracts, debtors
have, loosely speaking, an unlimited demand
for loans, what is the appropriate strategy for
creditors? To prevent default, creditors have to
plan aggregate lending in such a way that the
present value of future cash flows to the debtor
remains positive forever. This necessary condition for solvency can be formalized as

L"'T ~~ - iA(t) }-i<! -T)dt > 0

that the other creditors will continue to lend on
an ever-increasing scale. It is difficult to state
precisely what this condition requires. For the
present purpose, it is enough to draw attention
to tW0a.~pects.First, the confidence of creditors will>be difficult to maintain unless the path
of outstanding debt can be expected to merge
into a path of balanced growth. Clearly, an explosive debt profile with a debt ratio rising beyond any limit would not be acceptable, but a
debt ratio asymptotically approaching a finite
limit may inspire confidence. IS This idea can be
form<Uized by postulating that the rate of debt
expansion eventually should approach the rate
of growth.
Besides the expansion rate of debt, the creditors will have to consider the level of debt at
a given time. This level determines the size of
the net cash flows, positive or negative, relative
to the national incomes of debtor and creditor
countries along a given growth path. With enforceable debt, the main criterion for judging
the optimal level of debt is the condition that
the marginal social product of the loan shall be
at least as high as the rate of interest. With
unenforceable claims, this criterion is clearly
not relevant. In fact, with unenforceable claims
there are no established criteria for optimal
debt levels. The decisive consideration seems
to be the share of their portfolios creditors are
willing to hold in the form of unenforceable


for all T. It will be c<Uled the solvency constraint. 16 There must never be a moment at
which default would pay in this scenario. The
larger the present value of the cash flow at any
moment, the larger is the safety margin against
This constraint imposes no limit on lending
at a particular time. Rather, it relates to the
shape of the whole lending profile. Specifically,
it requires that the cumulative cash flow curve
in Figure 1 never pass an all-time maximum.
By and large, more rapid debt expansion in the
early periods makes default more likely at a
later time. The more slowly debt expands in the
early stages, the better is the chance of avoiding
a crisis. This intuitive argument will be elaborated upon below. 17
The solvency constraint, although necessary,
is not sufficient for avoiding a debt crisis. In
addition, each creditor must have confidence

loans. This will be calleq the "acceptable" level
of debt here. From the point of view of the
viability of the credit system, precise determination of the acceptable level is of little importance. The important thing is that it is not sub-

Developed by analogy to domestic indicators of
creditworthiness, they put the emphasis on the
debtor's ability to pay. This is appropriate for
enforceable claims because, if a debtor is able
to . pay, the .law can make him willing. With
unenforceabl<:; • • claims, ·.however, •the···. ability to
pay loses its crucial relevance for solvency because a rational debtor, no matter how able to
pay, will not do so unless he can borrow the
required funds. The decisive factor, therefore,
is the willingness of creditors to lend. 2o
For the same reason,even drasticimprovementsineconomic conditions, in the debtor's
terms of trade, or in his export performance
may fail to resolve the debt crisis. If default was
in the debtor's interest in adversity, it may still
be in his interest in prosperity, no matter how
much his ability to pay may have improved.
With unenforceable claims, solvency depends
on the collective action of the creditors and not
on the economic strength of the debtor.

A lending strategy along these lines requires
coordinated planning. 19 With unenforceable
claims, decentralized decision-making is likely
to lead to crisis. The reason is that continued
debt service Oll. each loan depends crucially on
continued net lending by all lenders. This creates an externality somewhat analogous to a
congestion problem. Even full information of
all participants will not, in itself, prevent market failure in such cases. In the absence of concerted action, international lending with unenforceable claims is a game of "devil take the
This is probably the basic reason that the familiar country risk indicators performed so
poorly in guiding the banks' lending activities.

V. The Growth/Interest Relationship
Considering some particular debt profiles
will make the preceding argument more concrete, albeit at the loss of some generality. This
section concentrates on the significance of the
relationship between interest rates and the rate
of economic growth.
To relate the present discussion to that of
Section II, it is convenient to begin with an analytically trivial case. Suppose at time t = 1, a
country with a growth rate g has no foreign
debt. With the world interest rate at i, it is now
perceived, both at home and abroad, that under
the given economic conditions the acceptable
level of foreign debt as described in Section IV
is A(l) = Aoe g • It is also perceived that, in the
long-run, the rate of debt expansion cannot differ from the rate of the country's growth. There
is, therefore, a growth path for acceptable debt:

= Aoegt .

the new net lending each year will be equal to
the annual increase in A,

= gA egt



while interest payments on the outstanding
debt are
iA = iAoegt.
The net cash flow, therefore, is
c(t) = (g-i)Aoegt.
If the growth rate exceeds the rate of interest,
there is a positive cash flow forever. As long as
parameters do not change, there will not be a
debt crisis. If the two rates are equal, the situation is just barely viable. With a growth rate
lower than the interest rate, the cash flow is
forever negative-the borrowing country
would immediately default.
Figure 2 depicts this last situation. The two
curves represent, respectively, the growth path
of new lending and the. growth path of interest
payments. With g < i, the first curve lies below


Suppose at t = 1, the country obtains loans
in the full amount of Aoe g , thus "jumping" instantaneously to the growth path. In this case,

the second throughout. The shaded area between the curves measures (negative) cash
This case, in all its simplicity, suggests the
consequences of an increase in interest rates.
Supposeinterestrates so far have lain wen below growth rates, but that they now rise while
growth rates remain constant. Borrowers and
lenders would encounter the case in which i >
g, with the expectation that this will remain so
for a long time. With unenforceable debt, the
result would be an immediate debt crisis with
widespread default. The same result would hold
if, at given interest rates, the growth prospects
In reality, the adjustment of debt to a higher
acceptable level usually takes time. This idea
can be formalized by postulating that debt follows the path given by

Figure 2
Debt Adjustment



where a = Aoe g is set equal to acceptable debt
at time t = 1, so that A(1) = O. The particular
specification of the adjustment term aft is arbitrary. The important point is that actual debt,
beginning at zero, approaches its acceptable
level monotonically as shown by the debt curve
in Box 1.
New lending is determined by taking the derivative of debt,

oe + 2'



They are initially zero and approach their
growth path from below.
The resulting cash flow is the difference between new lending and interest payments,

= (g-i)Aoegt + i a + ~ .




sider the case g = i as illustrated in Box 1. In
this case, the growth paths of lending and interest payments coincide. New lending, represented by the blue arrow, approaches this
growth path asymptotically from above,
whereas interest payments, represented by the
black arrow, approach the growth path from
below. Cash flows, measured by the vertical
distance between the arrows, never become
negative although they diminish over time.
Their present value, therefore, is positive
throughout. The debtor will never find it profitable to default.
This is even more true if g > i. In this case,
the growth path of lending lies above the
growth path of interest payments. As a consequence, the lending curve and the interest
curve, after first moving closer together, eventually diverge. Cash flows, after passing a, will increase beyond any finite limit. In
this case, even unenforceable claims are quite
In the opposite case, with g < i, unenforceable claims are not safe. As shown in Box 1,
the growth path of lending now runs below the
growth curve of interest payments. New lend-

Initially, at t = 1, new lending is far above its
growth path, gAoe gt , but approaches this path
in a V-shaped curve without ever crossing it.
Interest payments, on the other hand, are





-d - g



The profile of cash flows depends crucially
on the relationship between g and i. First con-


Monotonic Debt Adjustment

Cash Flows

»» New Lending Growth Path
™ interest Growth Path
« Positive Cash Flows
mm Negative Cash Flows


unenforceable claims have already been acquired and the system has reached a point such
as T in the diagram for g < i? If lenders, under
the. pressure of increasing default risk, succumb
to the temptation to raise interest rates and to
curtail lending, they will precipitate the crisis
they are trying to protect themselves against.
The only way to forestall a crisis seems to be a
combination of lowering the interest rate to a
level not in excess of long-run growth and continued lending at a level slightly in excess of
interest receipts. The cost of such a strategy to
the creditors would have to be weighed ag:linst
the costs of default.

ing, therefore, while initially exceeding interest
payments, will sooner or later fall below them.
In the initial stages, positive cash flows give
debtors a strong motive to service their debt.
Eventually, however, the cash flow is bound to
become negative forever. With unenforceable
claims, therefore, the debtor will find it advantageous to default. This leads to strong advice
for the lending banks: do not acquire unenforceable claims unless the borrower's rate of
economic growth exceeds the rate of interest.
(Clearly, the validity of this conclusion is limited by the assumptions underlying the present
What should the creditors' strategy be if

VI. Debt Overshooting
The preceding cases are characterized by a
debt path that starts below the acceptable level
but approaches that level monotonically from
below. Debt is never too high compared to its
growth path. In reality, debt may initially expand so rapidly that it overshoots its growth
path, and one is inclined to suspect that this is
what happened in the late 1970s under the pressure of lenders' competition. It will be shown
that this may lead to a debt crisis even though
the rate of interest does not exceed the rate of
Suppose creditors permit loans to expand according to

It can easily be ascertained that new lending is

initially, at t = 1, above its growth path, which
is given by gAoe gt . At t = 2b/(b - a), the lending curve declines below the growth
reaching its maximum shortfall compared to the
latter at t = 3b/(b - a). Thereafter, it gradually
approaches its growth path from below (see
Box 2). The important point is that, with overshooting, new lending will eventually run below
the growth path.
The path of interest payments,
iA = gA = gA e gt




has the same shape as the debt
with an
amounts multiplied by g = 1. It crosses the debt
growth path at t = b/(b - a) and reaches the
maximum deviation from the latter at t = 2bl
(b - a). With debt overshooting, therefore, the
path of interest payments is bound to lie above
the growth path of debt except in the
This means that the cash flow to the '-',",'J'''',
while positive in the early stages, becomes negative and remains negative after the change.
With unenforceable claims, overshooting
makes a debt crisis inevitable. The
overshooting, as measured by the parameter b,
the larger the negative cash flows become and
the more acute the debt crisis grows. Extending
this reasoning to the case g > i would be

= A egt + b-a - Q ( 1 3 )

+ g -t- - gt2


where b - a > O. In this case, the debt profile looks like the debt curve in Box 2. At
t = b/(b-a), the curve crosses the debt growth
path. It reaches the maximum amount of overshooting at t = 2b/(b-a) and gradually approaches the debt growth path thereafter.
There may actually be a phase of absolute decline in debt, but this is not necessary.
To describe the implications of debt overshooting for cash flows, the following discussion is restricted to the case g = i. New lending


Box 2
Debt Overshooting

straightforward. In that case, a rate of interest
safely below the rate of growth, despite over­
shooting, may avoid a crisis but can lead to one
as well. It is evident that with g < i, there is no
chance of avoiding a crisis.
The policy conclusions suggested by this
analysis of debt overshooting, provided the un­
derlying assumptions are regarded as realistic,
may be summarized in three rules.
(1) If, in a country with a growth rate safely
above the interest rate, a higher ratio of
foreign debt relative to national income is
perceived to be acceptable, this ratio
should never be permitted to overshoot its
acceptable level or a repayment crisis
would emerge. Once this rule has been vi­
olated, efforts to slow down the excessive
debt expansion are likely to precipitate the
(2) If debt, even though it is not yet overshoot­
ing its acceptable level, is seen to be on an
overshooting path (as at P in the crisis man­
agement diagram in Box 2), lending should
switch to a monotonic (non-overshooting)
adjustment path as described in Section V.
This would involve an immediate lowering
of the rate at which debt is allowed to ex­
pand. Gradually, debt would approach its
growth path from below. The same is true
for interest payments. New lending, in con­
trast, would be instantaneously reduced,
but its future decline also would be reduced
and possibly even eliminated. In any case,
new lending would remain continuously
above its growth path. As a consequence,
positive cash flows, and thus solvency,
could be maintained. This is illustrated in
Box 2, where the light curves represent the
original paths whereas the dark curves de­
scribe the revised paths.
(3) Once the crisis point has been reached (as
at Q in Box 2), a feasible emergency strat­
egy may consist of a combination of a low­
ering of interest rates to a level clearly be­
low the rate of growth and continuous
relending of interest payments. If compe­
tently executed over many years, such a
strategy, while achieving no overnight mir-



tional negotiations, although perhaps without clear insight into its underlying

acles, could gradually lead debt back to a
sustainable path. Such a strategy may actually be emerging from current interna-

VII. Concluding .Flernarks
Practical men of affairs, bankers, financial
writers and policymakers, can often be heard
to say, at least in less guarded moments, that,
"of course," a large part of the bank loans to
governmel.1ts of developing countries will
never, in the aggregate, be repaid. This paper
tries to develop the implications of this notion
for the viability of the international credit
To focus. attention on the essential aspects of
the problem, the analysis in this article was
based on the assumption that concern about the
future availability of credit is the only deterrent
to default. Any other penalties, both legal and
extra-legal, were disregarded, as was the ethical
maxim that contracts shall be honored. In reality, such penalties may exist and even ethics
may have some force. To the extent that they
do, the basic assumption of this paper may be
one-sided and the conclusions derived from it

Despite this limitation, the paradigm of
unenforceable claims seems to shed light on important aspects of the recent "debt crisis". Over
the last decade, banks have acquired vast claims
on foreign governments, claims whose enforceability is weak and in many cases virtually nonexistent. In doing so, the banks probably did
not fully understand the implications of unenforceability. They also did not recognize the significance of the relationship between a country's economic growth and interest rates. In
addition, the apparent profitability of the loans
during the early stages seems to have induced
the banks to "overshoot" the sustainable debt
level. It is hard to believe that the last was not
an important contributing factor to the recent
debt crisis. From this point of view, the analysis
of international debt under the radical assumption of unenforceable claims may perhaps make
some contribution toward the prevention of future debt crises.

1. The basic ideas of this paper are summarized in Niehans (1984).

6. This point is rightly stressed by Aliber (1980).
7. The moral hazard inherent in IMF lending policies is
analyzed in Vaubel (1983). Tying the schedule of debt service payments to exports, as is sometimes proposed,
would create still another moral hazard.

2. Despite the most severe pressures, including partial occupation, vanquished Germany after WWI could not be
compelled, on balance, to bear the burden of reparations.
Quite to the contrary, by defaulting on her foreign debt,
which amounted to roughly twice the cumulative reparation
payments, she extracted a vast transfer of resources in her
favor (Schmidt, 1934, p. 82 f.).

8. Inflation is not considered in this paper: Nominal rates
are thus equal to real rates throughout.
9. If debt in period t is e 91A", the corresponding cash flow,
discounted to the initial period, is (g
i) e(9 iliA". The
present value of cash flows, therefore, is

3. Among the most illuminating contributions are Aliber
(1980); Eaton and Gersovitz (1981 a,b); Sachs and Cohen
(1982); Sachs (1983); Cline (1983); and Swoboda (1984).
Important suggestions can be found in Wallich (1982).
4. Germany in the 1920s presented strikingly similar contours. Before 1929, new foreign lending, with the exception
of 19~6, far exceeded interest payments and reparations.
In 1929, .the net transfer became negative, followed by defaUlt in the Summer of 1931 (Schmidt, 1934, p. 111).


i) e(9




Ao .


10. It may be noted in this context that for a debtor country
in balanced growth, the "Golden Rule" for maximum consumption per capita requires a rate of growth in excess of
the rate of interest. The reverse is true for a creditor country. (For the world as a whole, unequal growth rates would,
of CO\.lrse, be inconsistent with balanced growth.)

5. It should be noted, however, that Mexico and Uruguay
differ from .the group of reschedUling countries in that they
enjoyed s\.lbstantial positive transfers in recent years.
ThrQu~n rl;)sched\.lling, these countries seem to have increased the net benefits from foreign lending long before
these benefits were near the vanishing point.

11. The optimality of default also has been discussed in
the finance literature (for example, in Van Horne, 1976),
but there, with enforceable claims, optimality is seen from
the point of view of the creditor, while here, with unenforceable claims, it relates to the debtor.


17. Eaton and Gersovilz (1981 b, p. 13) argue that unenforceable loans would be rationed to a sub-optimal level.
From the point of view of the present analysis, there is no
general reason for this to be the case. Even inefficiently
high lending might satisfy the solvency constraint.

12. Essentially the same criterion for default is used by
Eaton and Gersovitz (1981 b, pp. 11-12).
13. Sachs and Cohen (1982) postulate a default penalty
in the amount of a certain percentage of gross national
product. It is the essence of unenforceability, however, that
there are no effective penalties of this sort.

18. There is evidently an analogy between this problem
and that of "bubbles" in financial markets.

14. This was pointed out by Eaton and Gersovitz (1981 b).

19. This point is forcefully made in De Grauwe and Fratianni (1984).

15. Eaton and Gersovitz (1981 a,b) argue that the benefits
from future borrowing are higher, and the likelihood of default therefore smaller, the more future income is expected
to fluctuate.Their benefit/cost calculus thus relates to the
traditional borrowing criteria. From the point of view of the
present analysis, these criteria are immaterial. Once the
cash flow is reversed, no amount of net benefits in the
sense of Eaton and Gersovltz would dissuade a country
from defaulting on its unenforceable debts.

20. Strictly speaking, with fiat money and floating exchange rates a country is always able to pay its debts
because (1) it can print its own money and (2) this money
can be exchanged for foreign currency at some exchange
rate, although perhaps at very onerous terms (similarly,
Sachs and Cohen, 1982, p. 22). A historical example is
provided by German reparation payments in 1921,
promptly followed by suspension (Bresciani-Turroni, p.
93 f.; Graham, 1930, p. 30 f.; Stolper, et ai, 1967, ch. IV).

16. Solvency, In the present context, is meant in the sense
of maintaining debt service. In the sense of positive net
worth, the concept has no relevance for unenforceable
government debt.

Aliber, Robert Z. "A Conceptual Approach to the Analysis
of External Debt of the Developing Countries." World
Bank Staff Working Paper, No. 421 (October, 1980).

Niehans, JQrg. "Strategic Planning with Unenforceable
Claims." International Conference on Strategic Planning in International Banking. Rome, May, 1984.

Bresciani-Turroni, Costantino. The Economics of inflation. Translated by M.E. Sayers, London: Allen and
Unwin, 1937.

Sachs, Jeffrey, and Cohen, Daniel. "LDC Borrowing with
Default Risk." Working Paper. No. 925. National Bureau of Economic Research, 1982.

Cline, William R. "International Debt and the
4. Institute for International Economics. Washington,
D.C. (September, 1983).

Sachs, Jeffrey. "LDC Debt in the 1980's: Risk and Reforms." In Crisis in the Economic and Financial
Structure. Edited by Paul Wachtel. Lexington, Massachusetts: Lexington Books, 1982.

De Grauwe, Paul, and Fratianni, Michele. "The Political
Economy of International Lending." International Economics Research Papers, No. 42. Catholic University
of Louvain, 1984.

- - . "Theoretical Issues in International Borrowing."
Working Paper No. 1189, National Bureau of Economic Research, 1983.
Schmidt, Carl T. German Business Cycles 1924-1933.
New York, National Business of Economic Research,
Stolper, Gustav.; Hauser, Karl.; and Borchardt, Knut. The
German Economy 1870 to the Present. Translated
by Toni Stolper. New York: Harcourt, Brace, 1967,
Chapl. IV.

Eaton, Jonathan, and Gersovitz, Mark. "Debt with Potential
Repudiation: Theoretical and Empirical Analysis." Review of Economic Studies, Vol. 48 (2). 1981 a.
- - - . "Poor-Country Borrowing in Private Financial Markets and the Repudiation Issue," Princeton Studies in
International Finance. Vol. 47 (June 1981 b).

Swoboda, Alexander K. "Debt and the Efficiency and Stability of the International Financial System." UnpUblished, 1984.

Graham, Frank D. Exchange, Prices and Production in
Hyper-inflation: Germany 1920-1923. Princeton:
Princeton University Press, 1930.

Van Horne, James C. "Optimal Initiation of Bankruptcy Proceedings by Debt Holders," Journal of Finance. Vol.
31, No.3, 1976.

International Monetary Fund. Recent Multilateral Debt
Restructurings with Official and Bank Creditors.
Occasional Paper, No. 25 (December, 1983).

Vaubel, Roland. "The Moral Hazard of IMF Funding," The
World Economy. (September, 1983).
Wallich, Henry C. "A U.S. Regulator's View of the Debt
Rescheduling Problem." World of Banking. May!
June 1982.
World Bank, World Debt Tables-External Debt of Developing Countries, 1983-84, and First Supplement. Washington, D.C. 1984.

McDonald, Donogh, C. "Debt Capacity and Developing
Country BorroWing: A Survey of the Literature." International Monetary Fund Staff Papers. Vol. 29, No.4
(December, 1982).