PrivateInvestment PDF

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/227470073
Private Investment in Developing Countries: An Empirical Analysis
Article in IMF Staff Papers · March 1991

DOI: 10.2307/3867034 · Source: RePEc
CITATIONS READS
479 675
2 authors:
Joshua Greene Delano Villanueva

Singapore Management University Bangko Sentral ng Pilipinas (BSP)
17 PUBLICATIONS 763 CITATIONS 67 PUBLICATIONS 1,836 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Bangko Sentral ng Pilipinas: The Case of Successful Central Banking in a Development Context View project
Economic Adjustment and Growth: Theory and Practice View project
All content following this page was uploaded by Delano Villanueva on 03 May 2020.
The user has requested enhancement of the downloaded file.

INTERNATIONAL MONETARY FUND
S T A F F
P A P E R S
Vol. 38 No. 1 MARCH 1991
©International Monetary Fund. Not for Redistribution

EDITOR'S NOTE
The Editor invites from contributors outside the Fund
brief comments (not more than 1,000 words) on pub
lished articles in Staff Papers. These comments should
be addressed to the Editor, who will forward them to the
author of the original article for reply. Both the com-
ments and the reply will be considered for publication.
The term "country," as used in this publication, may not refer to a

territorial entity that is a state as understood by international law and
practice; the term may also cover some territorial entities that are not
states but for which statistical data are maintained and provided
internationally on a separate and independent basis.
® 1991 by the International Monetary Fund

International Standard Serial Number: ISSN 0020-8027
The U.S. Library of Congress has cataloged this serial publication as follows:
International Monetary Fund

Staff papers — International Monetary Fund. v. 1- Feb. 1950-
[Washington] International Monetary Fund.
v. tables, diagrs. 23 cm.
Three no. a year, 1950-1977; four no. a year, 1978-

Indexes:
Vols. 1-27, 1950-80, 1 v.
ISSN 0020-8027 = Staff papers — International Monetary Fund.
1. Foreign exchange—Periodicals. 2. Commerce—Periodicals.

3. Currency question—Periodicals.
HG3810.15 332.082 53-35483

CONTENTS
Vol. 38 No. 1 MARCH 1991
Long-Run Money Demand in Large Industrial Countries

JAMES M. BOUGHTON • 1
Private Investment in Developing Countries:
An Empirical Analysis
JOSHUA GREENE and DELANO VILLANUEVA • 33
Wages, Profitability, and Growth

in a Small Open Economy
BANKIM CHADHA • 59
The Transmission Mechanism

for Monetary Policy in Developing Countries
PETER J. MONTIEL • 83
The Budgetary Control and Fiscal Impact

of Government Contingent Liabilities
CHRISTOPHER M. TOWE • 109
An Inflation-Proof Tax System?

Some Lessons from Israel
EFRAIM SADKA • 135
Exchange Rate Expectations:

A Survey of Survey Studies
SHINJITAKAGI • 156
Currency Bands, Target Zones, and Price Flexibility

MARCUS MILLER and PAUL WELLER • 184

IV CONTENTS
Shorter Papers and Comments

The 1990 Reform of U.K. Local Authority Finance
TAMIM BAYOUMI • 216
Issues in Interest Rate Management
and Liberalization:
Comment on Leite and Sundararajan
STEPHEN H. AXILROD • 229
IMF Working Papers • 232

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
© 1991 International Monetary Fund
Long-Run Money Demand

in Large Industrial Countries
JAMES M. BOUGHTON*
The long-run properties of money demand functions in the large industrial

countries are examined under the hypothesis that the long-run functions
have been stable but that the dynamic adjustment processes are more
complex than those represented in most earlier models. The results
broadly support this hypothesis, but for certain aggregates they also call
into question some basic hypotheses about the nature of the demand
function, including, notably, that of homogeneity with respect to the price
level. [JEL 311, 211, 431]
HE REPUTATION of the aggregate demand function for money bal-

T ances has plummeted since the mid-1970s. Once viewed as a pillar of
macroeconomic models, it is now widely regarded as one of the weakest
stones in the foundation. The origins of this fall from grace are not hard
to find: the past two decades have witnessed a large number of financial
innovations and deregulatory measures in many countries, which have
dismembered traditional payments patterns and have rendered the iden-
tification of the line between money and other liquid assets all but
impossible. What is remarkable is not that the estimation of functions
relating money (conventionally defined) to other macroeconomic vari-
ables has become much more difficult, but rather that it remains possible
at all.
*James M. Boughton, an Advisor in the Research Department, holds ad-
vanced degrees from the University of Michigan and Duke University and was
formerly Professor of Economics at Indiana University. He has published two
books on monetary economics and numerous articles in economic journals.
The author would like to thank Charles Adams, Matt Canzoneri, Neil Erics-
son, Bob Flood, Paul Masson, Ben McCallum, George Tavlas, Mark Taylor, and
participants at seminars at Brown and Rutgers Universities for helpful comments
and suggestions.
1

2 JAMES M. BOUGHTON
In spite of these difficulties, a renaissance seems to be underway, or at

least a renewed effort among economists to uncover stable relationships.
Part of the basis for this revival has been the development of new econo-
metric approaches, especially those related to cointegration and error-
correction models. A rapidly growing literature—exemplified by the
work of David Hendry and his co-authors—has raised the possibility that
models that combine a conventional steady-state function with a com-
plex set of dynamics may be reasonably stable even over periods of
substantial institutional change. In a sense allied with this econometric
army, though scarcely on speaking terms with it, Milton Friedman and
Anna Schwartz (among others) have emphasized that one need not
throw out the long-run demand function simply because it cannot always
predict shorter-run developments. With appropriate allowances, they
argue, the qualitative characteristics of the underlying relationships have
changed but little over very long stretches of time.
The object of this paper is to examine the nature of the long-run
demand for money in the large industrial countries. The maintained
hypothesis is that money demand in these countries has remained stable,
but that the dynamic adjustment processes are more complex than those
represented in most earlier models. By "nature of the long run" is meant
the parameterization of the steady state associated with time-series esti-
mates of equations relating the stock of money to aggregate income,
prices, interest rates, and perhaps other variables.
Section I offers a selective review of the recent empirical literature, in
order to assess what seems to be known or not known on the subject.
Section II then examines various methods for estimating the long-run
component of money demand equations, and Section III applies the
preferred methodology to data for the five largest industrial countries.
Conclusions are summarized in Section IV. The Appendix presents the
complete dynamic regressions summarized in Section III.
I. State of the Art
What is the starting point: what do we think we know about the long-
run demand for money? There is, of course, a vast literature on this
subject, and no attempt will be made here to review it systematically.1
What is of primary interest is the literature that has focused on the
1
See Laidler (1985) and Goldfeld and Sichel (1990) for general reviews.

LONG-RUN MONEY DEMAND 3
effects of financial innovation and deregulation on the nature and stabil-

ity of money demand. In order to narrow the field a little further, this
discussion will focus on two aggregates that have received close attention
in the literature: Ml in the United States and in the United Kingdom.2
United States
The conventional point of departure for looking at Ml in the United
States is the classic study by Goldfeld (1976). In that era, the conven-
tional way to estimate the demand for money was to specify a dynamic
adjustment comprising a Koyck transformation of a stock-adjustment
process and a first-order serially correlated error term. The steady-state
solution of Goldfeld's "basic" equation of this type, estimated over the
1952-73 period, was
where m, y, and p are the logarithms of money, real gross national

product (GNP), and the GNP deflator; and rtd and rs are the interest
rates on time deposits and short-term securities (Treasury bills in this
case).3 Price homogeneity was imposed on theoretical grounds. The
long-run, real-income elasticity (0.629) is consistent with the Baumol-
Tobin proposition of economies of scale in holdings of cash. Both in-
terest rates are intended to represent substitute prices, so the negative
coefficients are as expected.
In estimating this equation, Goldfeld was interested in analyzing its
inability to predict the sharp rise in velocity that began in 1974. Although
it was then too early to judge whether the 1974-76 period constituted an
unusually large blip or a more permanent shift, subsequent research
generally has confirmed the notion that the demand for money has
become more difficult to predict, and that the problems have become
much more serious since the early 1980s. What has never been satisfacto-
rily determined is whether the problem relates primarily to the short-run
2
For a detailed analysis of the properties of the broader aggregates in these
two countries, see Friedman and Schwartz (1982). For cross-country compari-
sons covering all or most of the major industrial countries, see, among others,
Atkinson and others (1984), Boughton (1981), Boughton and Tavlas (1990),
Domowitz and Hakkio (1990), Fair (1987), and Leventakis (1990).
3
This equation, like all of those presented below, also includes a constant
term, which is omitted for simplicity. Except as noted, all equations were esti-
mated with quarterly data.

4 JAMES M. BOUGHTON
dynamics or whether one must re-examine the shape and stability of the
long-run demand function as well.4
Interestingly, re-estimation of Goldfeld's equation with an updated
sample (1963-88) produces a rather similar steady state:5
The time deposit rate used by Goldfeld is excluded here, owing to the
multiplicity of deposit rates in the latter part of the period. Nonetheless,
both the income elasticity and the semi-elasticity of the interest rate are
close to earlier values. To some extent, the problems with the equation
seem to reside in the dynamics; the adjustment rate—1 minus the coeffi-
cient on the lagged dependent variable—falls from 18 percent per quar-
ter in Goldfeld's sample to 8 percent in the update. More fundamentally,
however, the literature on this issue has called attention to possible
specification errors in the conventional approach, including both the
Koyck lag and the "correction" for first-order serial correlation.
Jumping ahead to the "modern" era, Michael Darby's buffer stock
approach (Darby (1972), Carr and Darby (1981), Carr, Darby, and
Thornton (1985)) provides a useful point of departure.6 This approach
was developed to explain short-run phenomena, but it is of interest to
determine whether the nature of the steady state is affected when the
adjustment process is modeled more extensively. The U.S. equation that
Carr and Darby estimate for the 1957-76 period (their Table 3) has a
very slow adjustment rate (1 percent per quarter, insignificantly differ-
ent from zero), and a rather high long-run, real income elasticity (1.87).
4
Major contributions on this issue have been made recently by Bordo and
Jonung (1987, 1990) and Lucas (1988); this literature suggests that the long-run
properties of money demand have been stable, but the authors get to that
conclusion by different paths. Bordo and Jonung argue that conventional formu-
lations of money demand have undergone secular shifts that are explained in
large part by institutional developments. Lucas examined Meltzer's (1963) esti-
mates that were derived from annual data on U.S. Ml over the 1900-57 period,
and argued that Meltzer's elasticities do not change when data are added through
1985. That conclusion was based, however, on constrained estimation; Lucas's
unconstrained ordinary-least-squares (OLS) estimates for the 1958-85 period
produced radically different elasticities. Whether the constraints are an appro-
priate way to deal with the different behavior of the data in the two periods
remains controversial.
5
All of the estimates presented in this paper have been made using PC-GIVE;
see Hendry (1989).
6
Another important vein of research (and source of controversy; see Hendry
and Ericsson (1989)) has been the work of Friedman and Schwartz (1982 and
forthcoming), using annual data from 1870 to 1975 for the United States and the
United Kingdom. Their tests, however, used M2 rather than Ml and so are not
included in this review.

The simple buffer-stock model thus calls into question the existence of a
long-run solution.7
A more general examination of the short-run dynamics was made by
Judd and Scadding (1982). Comparing conventional approaches with
several buffer-stock models, they found that the best results—not only
over the 1959-74 estimation period but also over the 1974-80 post-
sample simulation period—were obtained for Coats's (1982) model,
which postulates that prices adjust to the excess of monetary growth over
the previously expected rate of inflation. They estimated the steady state
of that model to have a real-income elasticity very close to one half, with
the usual imposition of unitary long-run price elasticity and with esti-
mated negative responses to short-term interest rates.
The very long adjustment rates in all of these partial-adjustment mod-
els raises the issue of whether the relationship between the level of the
stock of money and the levels of other macroeconomic variables is char-
acterized by unit roots. Consequently, a number of researchers have
examined the cointegration properties of money demand relationships.
In this vein, Baba, Hendry, and Starr (1985) estimated a general dy-
namic equation for U.S. Ml that has a steady state of the following form
(slightly simplified):
where rl is a long-term interest rate.8 As with most earlier models, this

equation was estimated subject to the prior constraint—in this case
based on initial estimation of a simplified equation—that the long-run
price elasticity was unity. The real income coefficient was initially esti-
mated to be close to 0.5, and that value was then imposed as being
consistent with the simple version of the Baumol-Tobin model.
The main departure that equation (2) makes from the conventional
approach is the inclusion of both short- and long-term interest rates.
7
More recent studies of the Carr-Darby approach have confirmed the long-run
problems with the model. Estimates by Carr, Darby, and Thornton (1985) using
the same 1957-76 data period but a slightly modified equation yielded a negative
adjustment coefficient, implying that a steady state did not exist. Boughton and
Tavlas (1990) estimated the original Carr-Darby equation over the 1973-85
period, using a modified estimate of unanticipated changes in money, and found
an adjustment rate of 4 percent per quarter (still insignificantly different from
zero), with a long-run income elasticity around 2.8.
8
This equation was estimated by a two-stage procedure; the imposed error-
correction term was m—p—Q.5y, and the steady-state interest rate effects
emerged in the second stage. Baba, Hendry, and Starr (1985) measured interest
rates in decimal rather than percentage form; the interest rate coefficients re-
ported in their paper have been divided by 100 for consistency with the other
equations reported here.

6 JAMES M. BOUGHTON
When the demand function is estimated with all arguments entering

contemporaneously, it is very difficult to sort out the effects of different
interest rates, owing to collinearity. In a more general specification, in
which short and long rates might, for example, enter with different lags,
disentangling their effects could be enhanced.9 This appears to be the
case with equation (2), where the short rate acts as an own rate with a
positive coefficient, and the long rate captures substitution effects. That
finding is puzzling, however, since the equation estimated by Baba,
Hendry, and Starr also included the yield on NOW accounts as a proxy
for the own rate.
Porter, Spindt, and Lindsey (1989) experimented with an error-
correction model that incorporated a trend, intended to capture various
exogenous innovations that help to economize on money holdings. They
found the trend to be significant, with a value close to -1 percent per
year over the 1961-86 period (their Table 3.3); allowing for this trend,
the real income elasticity was estimated to be approximately unity.10
United Kingdom
A convenient place to begin reviewing the modern study of the de-
mand for money in the United Kingdom is the study by Hacche (1974)
for the Bank of England, which determined that the narrow aggregate
(Ml) appeared to be more stable than the broad aggregate (M3).11
Hacche used a conventional function similar to that of Goldfeld; the
steady state of his estimated Ml equation (1963-72) was approximately
This equation is completely standard except for the inclusion of both

long- and short-term interest rates in collinear form.12
Artis and Lewis (1976) presented a number of estimates over almost
the same data period (1963-73), experimenting with different functional
forms, which suggested that the long-run income elasticity had risen
9
For the United States over the period 1963-88, the correlation coefficient
between the quarterly first differences of short- and long-term interest rates (the
six-month commercial paper rate and the yield on long-term government bonds)
is 0.73.
10
When one rate is lagged by one quarter, the correlation drops to 0.15.
These estimates were derived from disaggregated equations for currency
and demand deposits.
11
For reviews of earlier work, see Artis and Lewis (1976) and Coghlan (1978).
For a criticism of Hacche's methodology and further tests relating to M3, see
Hendry
12
and Mizon (1978).
Hacche used log(l + r) instead of the level of the interest rate; the two are
quite close (x 100) for low levels of r.

quite sharply in the early 1970s and probably exceeded unity. For exam-
ple, when they specified the equation in real per capita terms and in-
cluded a measure of interest rate volatility as an additional argument,
they calculated the long-run income elasticity to be 1.35. The same
equation truncated at 1971:4 yielded an elasticity of 0.75, which is close
to Hacche's estimate.
Artis and Lewis also reported equations in nominal form, which al-
lows the price elasticity to differ from unity but constrains it to equal the
real income elasticity. Those equations also produced elasticities that
shifted from less than unity to greater than unity with the lengthening of
the sample. But when Coghlan (1978) allowed both elasticities to range
freely and used unrestricted lags, he estimated the price elasticity to be
around 0.7 and the real income elasticity to be quite close to unity.
Perhaps the extreme low estimate for the income elasticity was obtained
by Cuthbertson and Taylor (1987). They estimated a Carr-Darby buffer-
stock model for U.K. Ml (1964-81), and found an elasticity of 0.32. In
contrast, Hall, Henry, and Wilcox (1989) estimated an error-correction
model (1963-87), allowed both price and income elasticities to vary, and
found both to be close to unity.
Muscatelli (1989) tested an error-correction model against forward-
looking buffer-stock models, and found that the error-correction model
consistently outperformed the alternatives. His preferred equation (esti-
mated over the 1963-82 period) had a steady state of the form
which is very close to Hacche's much earlier estimate (equation (3),

above), with the short-term interest rate here capturing the combined
effects of short and long rates in Hacche's equation.
Implications
This sketchy review of recent developments suggests a few conclusions

and some open issues to be explored. First, there are important interac-
tions between the specification of the short-run dynamics and the long-
run properties of the demand for money. The conventional partial-
adjustment model (including the Carr-Darby buffer-stock variant) is an
inadequate representation of the former and probably distorts the esti-
mation of the latter. Second, there is near-unanimous agreement that
the function is homogeneous in prices in the long run; most researchers
have simply imposed that condition, but most of those who have tested
for it have failed to reject the hypothesis. Third, there is no consensus on

8 JAMES M. BOUGHTON
the nature of the long-run income elasticity; estimates range from less
than one half to well above unity, both in the United States and the
United Kingdom. Fourth, while there is a clear negative relationship
between money demand and the level of market interest rates, the rela-
tive roles of short and long rates are less evident. Fifth, there is at
least preliminary evidence suggesting that the demand for money in the
United States has been characterized by a negative trend over the past
three decades.
II. Cointegration Analysis
One of the main sources of confusion regarding the parameterization

of the long-run demand function has been the variety of methods em-
ployed for obtaining approximately white-noise residuals in time-series
estimates. The most common method employed for many years was
simply to "correct" for first-order serial correlation through a Cochrane-
Orcutt or similar transformation. It is now known, however, that this
procedure covered up serious specification errors and led to the accep-
tance of excessive restrictions (see Hendry and Mizon (1978)). During
the past ten years, there has emerged a growing acceptance of more
complex dynamic structures, combined with prior analysis of static rela-
tionships through analysis of the cointegrating properties linking the
various arguments in the model; that is, tests of whether the levels of
variables are linked by a stable long-run relationship.
This section examines the implications for long-run money demand of
several approaches to cointegration analysis. In each case, the initial step
is to examine the stationarity properties of the data; only if the data are
(or can be transformed so as to be) stationary in differences is it appro-
priate to apply these procedures for estimating cointegrating relation-
ships. All of the data discussed here meet the usual tests for difference
stationarity.13
Two-Stage Estimation
Engle and Granger (1987) propose a two-step procedure under which

a cointegrating vector is estimated first, with the dynamics estimated
subject to that steady state. They demonstrate the general consistency of
13
One problem is that these tests have low power for rejecting stationarity
and may be misleading if the data are not pure autoregressive processes. For
a discussion and application, see Schwert (1987). By some tests, price indices
are stationary only in second differences; see, for example, Boughton, Branson,
and Muttardy (1989). For this study, however, they are treated as difference-
stationary.

the procedure, but they also emphasize the nonuniqueness of the solu-
tion; that is, there is a variety of methods for obtaining the first-stage
equation—and, in the multivariate case, there may be as many as n — 1
valid cointegrating vectors—and there is no clear basis for choosing
among them. As alternatives to simple static estimation (as originally
proposed by Engle and Granger), one could directly estimate the steady
state of a distributed-lag function, or one could arbitrarily select one of
the vectors estimated by Johansen's vector autoregression (VAR)
methodology discussed in the following subsection, or one could impose
prior values on the coefficients. Choosing among the results raises issues
of both efficiency and identification. In any case, the lagged residuals
from the cointegrating vector would be introduced as an argument (the
error-correction term) in the dynamic equation.
Regardless of how the first stage is obtained, the second stage is to
estimate
where X = (p, y, rs, r/), L is a general lag operator, and ECt is the time
series of residuals from the cointegrating vector. Equation (5) can then
be reduced to a parsimonious equation through the elimination of in-
significant terms and the imposition of constraints that hold to a reason-
able approximation.14 If and only if the levels of all variables vanish in
this reduction process will the initial estimate of the cointegrating vector
be accepted as the steady state (that is, the long-run demand function).
Hence, the specification of the dynamics cannot be treated as recursive
to the specification of the steady state.15
14
The algorithm used here for this specification process is as follows. First, for
each variable, drop the lag (or the current value) with the lowest f-ratio, as long
as the ratio is less than unity. Repeat this operation as necessary. Second,
eliminate further lags if t-ratios are below the 5 percent level, taking due account
of interactions and of the effects on the final specification. If eliminating a
variable has a noticeable effect on other coefficients or would qualitatively alter
the steady-state solution, further testing (including F-tests) may be needed to
determine if the variable should be retained. Third, test to determine whether
two or more lags can be combined to form simple or compound differences. For
example, if *, has a positive coefficient and xt-i has a negative coefficient of
similar magnitude, test to see if the two can be replaced by A/JC, without signifi-
cantly raising the standard error of the estimate. At each step, account is taken of
the shape of the equation residuals; problems such as instability, autocorrela-
tion, heteroskedasticity, or skewness may reflect a misspecification.
15
If the constraints in the cointegrating vector are valid, then ECt-\ should
capture all of the relationships among the levels of the variables; see Engle and
Granger (1987). But if the researcher imposes restrictions that seem to hold
approximately (such as price homogeneity), or omits key variables (such as
interest rates), or if there are different lags linking the levels, then EC may not be
a sufficient characterization of the steady state. These complications are dis-
cussed further, below.

10 JAMES M. BOUGHTON
The difficulty of applying the Engle-Granger procedure may be illus-

trated by reference to the Ml equation for the United States. When the
cointegrating vector is estimated through a least-squares regression re-
lating the levels of the variables in the model (with m as the dependent
variable), one obtains (1963:1-1988:4)
This static equation has coefficients that, for the most part, approximate
their a priori values: the price elasticity is close to unity, the real income
elasticity is less than unity, and the coefficient on the long-term interest
rate is quite small but negative. The only oddity is that the coefficient on
the short rate is not only positive (which is acceptable, given that a
portion of Ml pays interest), but larger than the coefficient on the long
rate.
Because of the high degree of autocorrelation in the errors of equation
(6), the standard errors of this regression are not meaningful. What is
clear, however, is that the coefficient estimates are highly sensitive to the
postulated structure of lags linking the variables. Suppose, alternatively,
that one starts by estimating an autoregressive distributed-lag (ADL)
regression with four lags on each variable:
where
Solving for the static long-run solution gives
Now the interest rate coefficients are in the "right" range, but the price
and income elasticities are not. It is noteworthy, however, that the
standard errors in equation (8)—which are meaningful because the rele-
vant dynamics have been included in the estimating regression—suggest
that one or both of the interest rates should be excluded from the
cointegrating equation.
The "best fit" for this type of ADL occurs when the short rate is
excluded:
Here, the coefficients move somewhat closer to their expected values,

but it still seems doubtful that equation (9) is a valid characterization of a

long-run money demand relationship, since the interest rate is barely
significant and the long-run price elasticity is well under unity.
In addition to this ambiguity, two-step estimation raises the difficulty
that the final dynamic process may be restricted relative to what would
emerge from a more general regression strategy.16 To illustrate this prob-
lem for a simple bivariate model, let the cointegrating vector be
and the dynamic error-correction process be
where D denotes a general differencing operator. In principle, these

differences can be made general enough to incorporate any adjustment
process. Nonetheless, because the specification of the dynamics is neces-
sarily data-based rather than a priori, the specification of equation (11)
may well be affected by the ordering of the estimation. In contrast, a
single-step approach could be written as
where L is a general lag operator on the levels of y and x. Depending on

the exact specification of the lag distributions in (11) and (11'), the two
estimates may well be nonnested, but in any event the final specifications
can be compared through encompassing tests, as discussed below.
Multiple Cointegrating Vectors: The Johansen Procedure
The initial development of cointegration analysis was aimed at bivari-

ate models. In that context, the question is whether a single relationship
exists and, if so, how it is specified. In the case of money demand
functions, the model is multivariate and there may exist multiple cointe-
grating vectors linking some or all of the included variables. Johansen
(1988) has devised a general procedure for the multivariate case, and the
test statistics have been generalized in Johansen and Juselius (1990). For
this test, consider a VAR on the detrended logarithms of money, the
price level, and real income, plus short- and long-term interest rates
(equation (7), above). The null hypothesis is that there are 5 unit roots in
this system (no cointegrating vectors). If that hypothesis is rejected, one
tests sequentially for additional cointegrating vectors. For the present
16
For evidence on the difficulties that can arise with application of the two-step
methodology, see Banerjee and others (1986).

Table 1. Tests for the Existence of Cointegrating Vectors
Elasticities3
Number of Price Real Short Long
Aggregate Significant Vectors13 Preferred Vector0 Level income rate rate
United States
Ml 3 of 5 3* 0.760 1.084 0.110 -0.027
Ml 2 of 4 2* 0.885 0.895 -0.045
M2 2 of 5 1*** 0.791 2.013 0.140 -0.153
M2 2 of 4 2* 1.091 0.917 -0.016
Japan
Ml 4 of 5 2*** 1.198 0.962 0.027 -0.028
M2 Iof5 1*** 0.625 1.423 0.038 -0.079
Germany
Ml 5 of 5 1*** 0.976 1.061 0.041 -0.036
M3 3 of 5 1*** 0.533 2.123 0.033 -0.027
M3 2 of 3 1*** 0.876 1.630
France
Ml 3 of 5 2*** 0.862 0.816 0.030 -0.050
M2 3 of 5 1*** 1.272 0.821 0.024 -0.060
United Kingdom
Ml 2 of 5 2* 0.753 1.709 0.125 -0.154
M3 2 of 5 1*** 0.534 3.280 0.090 -0.079
a
b
Semi-elasticities for interest rates.
The first number is the number of vectors /1 for which the trace of the eigenmatrix exceeds the 90 percent significance level, as
__.... j in
reported .... ™_,_,_ _ Johansen
^ ;in
Table A2 T_I andj Juselius
T__—i-___ (1990).'
^^\ The second number is the maximum number (when the matrix is full rank);
that
G
is, the number of variables included in the VAR.
The first vector is the one with the highest eigenvalue (and significance level), and so forth; (*), (**), and (***) indicate
significance levels of 90 percent, 95 percent, and 99 percent, respectively.

problem, it is also interesting to examine the coefficients of any signifi-

cant vectors to determine if they have the signs and order of magnitude
that are expected for a long-run money demand relationship.
A complication that arises is that the steady-state demand function, if
it exists, may exclude one or both interest rates. In that case, the
5-variable VAR may still be characterized by one or more cointegrating
vectors, none of which might have the desired characteristics. Further-
more, there is a strong likelihood of multiple cointegrating vectors,
because—in addition to the presumed long-run money demand rela-
tionship—the two interest rates are normally linked with each other and
possibly with other included variables through related demand functions.
The procedure followed here is therefore also to examine subsystems
that exclude, first, one interest rate (generally the short rate, which is
less likely to enter the long-run function),17 and then both interest rates.
The results of this exercise, with four lags included in each VAR, are
summarized in Table 1. The interpretation of this table may be illus-
trated by taking the first two rows (Ml for the United States) as an
example. For the full 5-variable VAR, there are 3 significant cointe-
grating vectors (at the 90 percent level or higher), of which the third (that
is, the least highly significant of those 3) comes closest to matching the
prior values on the coefficients. This vector is not very satisfying, how-
ever, because the price elasticity is low. Consequently, a second test has
been conducted without the short-term interest rate. In the second row,
2 of the 4 cointegrating vectors are significant, of which the second looks
reasonably as expected. Overall, the results shown in Table 1 do support
the hypothesis that these data sets are characterized by error-correction
representations, with steady states that could be interpreted as conven-
tional money demand relationships. There are, however, a few excep-
tions. First, for both M2 in Japan and M3 in the United Kingdom the
estimated long-run price elasticities are well below unity (0.6 and 0.5,
respectively). Second, for Ml in Germany the matrix may be full rank;18
in this case, although there is a steady state, there may not be a stable
dynamic adjustment process.
One way to employ these results would be to take the vectors in Table
1 as point estimates of the steady state, and incorporate the lagged
17
In most cases, bivariate tests indicate that short- and long-term interest rates
are cointegrated; the apparent exception is the United Kingdom, where the
Dickey-Fuller test just fails at the 90 percent level to reject the null hypothesis of
no cointegration. The long-term rates, however, have stationarity properties that
come closer to those of the other variables in the system; specifically, the levels
of short-term
18
rates have somewhat higher Durbin-Watson statistics.
If significance were restricted to the 95 percent rather than the 90 percent
level, then only 4 of the 5 vectors would pass the test.

residuals from those equations as arguments in a dynamic error-

correction equation linking changes in money to changes in the other
variables. There are, however, a number of difficulties with that proce-
dure. First, as already noted, the key parameters are not always consis-
tent with conventional priors regarding the shape of the long-run de-
mand function. Second, compounding this first problem, it is not always
obvious which of perhaps several candidates should be selected as the
most relevant cointegrating vector. Third, the estimated steady state
may change in the context of a more fully specified model, especially
when constraints have been imposed at the initial stage.
To illustrate the problems selecting among Johansen vectors and inte-
grating them as steady states, consider the 4-variable vector listed as the
second row of Table 1:
While these are consistent estimates subject to the constraints under

which they were estimated, there is no particular reason to expect them
to be consistent estimates of a more general model. For example, appli-
cation of equation (12) as the first stage of an error-correction model
leads to the following dynamic equation:
where dumSO is a dummy variable representing the temporary imposi-

tion of credit controls in the United States in 1980, R2 is the coefficient of
determination, a is the standard error, and DW is the Durbin-Watson
statistic.19 The error-correction term (EC,= |x, from equation (12)) is
significant, but the levels of y, p, rs, and rl are also all significant. Solving
for the steady state of equation (13) gives
This more general estimate implies price and income elasticities that are
rather different from those in equation (12), and it restores the short-
term interest rate. This solution is similar to the 5-variable vector at the
19
dum80 = 1 in 1980:1, -1 in 1980:3, and 0 otherwise.

top of Table 1, which suggests that the additional constraint in the

second row of the table is invalid. Thus, the estimation of the error-
correction model may help to choose among multiple cointegrating
vectors.
Single-Stage Estimation
The final alternative is to estimate the steady state implicitly, through

a general-to-specific regression strategy along the lines advanced by
David Hendry (for example, 1987). That is, starting from an ADL, such
as equation (7), one can eliminate the most insignificant lagged elements
and reduce the system to a more parsimonious equation in levels and
differences, and then solve directly for the steady state of that equation.
There are two main differences between this approach and the two-step
procedure. First, the error-correction term need not be limited to con-
temporaneous observations. That is, a cointegrating relation like equa-
tion (10) could in effect be estimated with noncontemporaneous (lagged
or led) data on x and y.20 Second, because the extraneous elements in the
original regression are eliminated and restrictions may be imposed on
the coefficients, there may be a substantial gain in degrees of freedom.
While neither of these differences is likely to be consequential asymptot-
ically, both may be important in small samples. The extension to non-
contemporaneous observations turns out to be especially important for
the problem at hand.
Applying the general-to-specific methodology (see footnote 14, page
20
Any equation in levels and differences can be transformed into one in which
the levels all appear with the same lag, simply by adding the appropriate differ-
ence terms subject to constraints on the parameters. To take a simple example,
the equation
may be transformed as
or as
where
Imposition of a single-lag structure is equivalent to estimating equation (B) or
(B') without the constraint. That procedure is inefficient and may lead the
researcher to drop terms that would be (implicitly) significant in equation (A). In
this example, one would be likely to find the coefficient on Ay_i to be
insignificant; dropping it, however, would produce an equation that would be
encompassed by (A).

9) to U.S. Ml and ignoring all changes yields the following relation

among levels:
With this lag structure, the short-term interest rate reappears with a
positive coefficient. If the model reduction is valid, equation (14) is
preferred to (8) or (9) because of the rise in degrees of freedom. The
inference is that the short rate was inappropriately excluded from equa-
tion (9) because of extraneous regressors, and from the Johansen esti-
mates because of the imposition of invalid constraints.21
One additional contrast between the Johansen procedure and this
general ADL methodology is that the latter may give different results
depending on the choice of normalization. Equation (14) was estimated
with the nominal money stock as the dependent variable. Normalizing
on real rather than nominal balances would make little difference as long
as the price elasticity was not constrained a priori. Normalizing on prices
rather than money, however, could make a more substantial difference;
the dynamic adjustment process would be affected, which could in
turn lead to a different estimate of the steady state. For U.S. Ml, the
renormalized-levels portion of the equation for the parsimonious ADL
with prices as the dependent variable is as follows:
There are some minor differences between (14) and (14'), notably in that
the real income elasticity is higher and the price elasticity lower. Qualita-
tively, however, a similar picture emerges.22
Yet another variant of the two-stage procedure is to impose some or
all of the coefficients of the error-correction term a priori. In particular,
since all theoretical models of money demand hypothesize long-run
homogeneity with respect to prices, it is natural to consider imposing
unitary price elasticity. In addition, a number of theoretical models
imply constraints on the range of acceptable values for the real income
elasticity. Commonly considered possibilities would include a simple
velocity model, used by, among others, Hendry and Ericsson (1989) for
M2 in the United Kingdom and by Hall, Henry, and Wilcox (1989) for
Ml in the United Kingdom:
21
It is tempting to try to draw causal inferences from the lag structure in
equation (14); it should be kept in mind, however, that the dynamic relationships
in the
22
full equation are much more complex.
The equation was also estimated with the short-term interest rate as the
dependent variable; that formulation did not produce results that could be
interpreted as a money demand function.

and the Baumol-Tobin model of economies of scale, used by Baba,

Hendry, and Starr (1985) for Ml in the United States:
It has already been seen that the restrictions in the Baumol-Tobin

model are rejected in a general ADL for U.S. Ml, notably in that the
income elasticity appears to be at least unity. The use of (15a) is there-
fore certain to dominate (15b) in any comparison on this data set. When
the residuals from equation (15a) are used as the error-correction term
in an ADL, such as equation (7), the price level is a significant additional
determinant with a negative sign, indicating again that the long-run price
elasticity is less than unity. For U.S. Ml, the steady state of this model is
or
In addition to these various models and approaches, there are many

other functional forms and estimation methods that could be explored,
including the addition of other arguments. Without attempting a com-
prehensive survey, it may be worth trying to incorporate the effects of
the financial innovations that have occurred during the past two or three
decades through the addition of a trend. As noted in Section I, there is
some evidence that money demand in the United States may have been
subjected to a negative trend over the postwar period, possibly as a
result of innovation or deregulation. When a linear trend (t) is added to
equation (7), the trend is significantly negative, and the steady state of
the final parsimonious specification is as follows:
The difficulty here is that the inclusion of the trend sharply raises the
estimated elasticities on income and prices to implausible levels, and it
eliminates the significance of the levels of the interest rates, which enter
only in the dynamic process.
Evaluation of Approaches
In view of the variety of methods available for estimating cointegrating
vectors, it is necessary to find a means of selecting among them. The
various steady-state relationships discussed above for Ml in the United
States are collected in Table 2, where it is immediately apparent that the

Table 2. Summary of Steady-State Estimates for U.S. Ml
Equation Model Estimate

(1) GoldfekT m =p + 0.629? - 0.197r'd - 0.56r5
(!') Equation (1) updated13 m=p+ 0.758? -0.053r*
(2) Baba, Hendry, and Starrc m =p + 0.5? + 0.046r* - 0.072r'
(6) Simple static equation m = 0.960/7 + 0.629? + 0.017^ - 0.006r7
(8) Static solution to 4-lag ADL m = 0.615p + 1.424? + 0.022^ - O.OSOr'
(9) Steady state of two-stage ADL estimation (model AD2) m = 0.702/7 + 1.225? -0.014^
(12) 4-variable Johansen vector m = 0.885p + 0.895? - 0.045r/
(13') Steady state of model JV2 (2-stage estimation with equation (12) m = 0.669p + 1.293? + 0.014^ - 0.029^
as error-correction term)
(14) Steady state of model EC1 (single-stage estimation) m = 0.652/7 + 1.338? + 0.015r* - 0.030r'
(14') Model EC1 estimated with p as dependent variable m = 0.522/? + 1.544? + 0.021r* - O.CMOr7
(14") Model EC1 with m-pas dependent variable (model EClc) m = 0.648/? + 1.364? + O.OlSr5 - 0.030r7
(16) Steady state of model PV2 (2-stage estimation with m-p-y m = 0.825/7 + ? + O.OlSr* - O.OSTr7
as error-correction term)
(17) Model EC1 with trend included (model EClb) m = 1.761/7 + 5.004? - 0.041*
Note: Sample period is 1964-8 $, except as noted. Equation numbers refer to those in the text, except for equation (14"), which
was
a
not shown explicitly.
b
Sample period 1952-73.
c

choice of estimation strategy makes a substantial difference for the quali-

tative interpretation of the steady-state relationship. Most estimates that
allow for nonhomogeneity with respect to the price level produce an
elasticity that is significantly less than unity, although that result holds
neither in the simple static equation (6) nor in equation (17), in which a
trend is included. Similarly, most but not all estimates generate a real
income elasticity that exceeds unity and a positive coefficient on the
short-term interest rate. Even the finding of a negative steady-state
relationship with the long-term interest rate, which is supported by
almost all models, is overturned when the trend is included.
These various models—at least in the final parsimonious form—are
almost all nonnested, so an appropriate test for dominance is the proce-
dure developed by Davidson and MacKinnon (1981), in which the pre-
dictions from one model are added to another: if the predictions from
model 1 are significant in model 2, but those of model 2 are insignificant
in model 1, then model 1 is preferred over model 2.23
Tests comparing several models for Ml in the United States and the
United Kingdom, and the two basic approaches for the other aggregates,
are summarized in Table 3. It would have been a pleasant surprise to find
that these tests conclusively favored one approach over the others; more
realistically, they do broadly support unconstrained single-stage estima-
tion over most alternatives. For Ml in the United States, the conclusion
is unambiguous: model EC1 is preferred over all other estimates. For Ml
in the United Kingdom, the tests imply that a more general model may
be required, since the predictions of model EC1 are significant when
added to the other three estimates, but the estimates from the other
models are also significant when added to model EC1. Comparisons of
single-stage and two-stage estimation for the other eight aggregates indi-
cate that the single-stage estimates are preferred in four cases; the two-
stage estimates in two cases (both of the French aggregates); and a more
general model in the remaining two. Thus, with the exception of France,
relatively little seems to be gained—and a possibly heavy cost could be
incurred—by two-stage estimation. The submodels included for the
United States in Table 3 are also of interest. Constraining the real
income elasticity to unity (model ECla) has little effect one way or
another,24 but estimating the model with real rather than nominal money
23
See Hendry (1989) for a summary of several other encompassing tests.
Those tests are programmed into PC-GIVE, but only for smaller data sets than
those used here.
24
Unconstrained estimation (equation (14)) yields a long-run real income
elasticity of 1.338, which is close to that in most other estimates (see Table 2).
However, an F-test fails to reject (at 95 percent) the hypothesis that the true
coefficient is unity.

Table 3. Encompassing Tests for Nonnested Models, 1964-88
(F-statistics)a
Additional Model5
Aggregate AD2 PV2 JV2 EC1 ECla EClb EClc
United States (Ml)
Basic Model
AD2 0.55 2.27 2.89* 2.08 2.85* 0.25
PV2 2.87* — 3.49* 4.41** 1.47 4.82** 0.00
JV2 5.26** 4.86** — 4.20** 4.18** 5.34** 0.55
EC1 2.09 1.63 0.01 — c
2.15 0.03
ECla 4.44** 1.48 1.30 3.00* — 5.04** 0.02
EClb 4.65** 4.03** 2.15 4.27** 3.99** — 0.00
EClc 5.45** 4.34** 1.26 3.59* 3.62* 4.79** —
United Kingdom (Ml)
Basic Model
AD2 — 0.90 0.51 4.91**
PV2 3.21 — 2.34 5.75**
JV2 2.58 3.45* — 6.27**
EC1 6.02** 4.72** 4.86** —

Other Aggregates AD2 vs. EC1 EC1 vs. AD.
United States
M2 20.36*** 5.45**
Japan
Mld 10.43*** 2.33
M2 62.73*** 0.01
Germany
Ml 19.76*** 6.46**
M3e 12.63*** 1.80
France
Ml 0.84 6.26**
M2 0.24 7.64***
United Kingdom
M3 17.01*** 2.59
a
Test of the significance of adding the predictions from the "additional" model to the regression of the basic model; see
Davidson and MacKinnon (1981). The null hypothesis is that the information from the additional model is already in the basic
model. The symbols (*), (**), and (***) indicate rejection of the null hypothesis at the 10 percent, 5 percent, and 1 percent levels,
respectively.
b
AD2 = two-stage estimation, with first stage estimated as a four-lag ADL; PV2 = same procedure as AD2 except first stage
imposed, with unitary elasticities on prices and real income and both interest rates omitted; JV2 = same procedure as PV2, except
that long-term interest rate included, and coefficients taken from Table 1; EC 1 = single-stage, general-to-specific modeling,
starting from the four-lag ADL; ECla = real income elasticity constrained to unity; EClb = trend included; and EClc = real
balances as dependent variable.
°For
d
this aggregate, ECla is nested in EC1.
e
Estimated in first differences.
Estimated with real balances as the dependent variable.

Table 4. Long-Run Elasticities: Final Estimates
Elasticity Semi-Elasticity
Price Real Short-term Long-term
Aggregate Model3 level income interest rate interest rate
United States
Ml EC1 0.652 1.338 1.487 -3.013
M2 EC1 1.0 1.0 -0.779
Japan
Ml ECld 0.831 0.831 -2.628
M2 EC1 0.387 1.604 1.777 -5.572
Germany
Ml AD2 1.0 1.199 -2.447 -1.482
M3 EClc 0.418b 2.240 2.180 -3.183
France
Ml AD2 1.0 0.434 -1.226 0.960
M2 AD2 0.685 1.965 -1.280
United Kingdom
Ml EC1 1.0 3.332 -10.343
M3 EC1 0.699 3.281 -3.441
a
AD2 = two-stage estimation, with first stage estimated as a four-lag ADL; EC1 = single-stage, general-to-specific modeling,
starting
b
from the four-lag ADL; EClc = real balances as dependent variable; and ECld = estimated in first differences.
Insignificantly different from zero.

balances as the dependent variable (model EClc) makes matters decid-

edly worse. Inclusion of a trend (model EClb) is either neutral or
superior to all other models except for the unconstrained single-stage
model.
The clearest conclusion to be drawn from this exercise is that there
does exist at least one cointegrating vector linking the variables that are
expected to enter the long-run demand function. This conclusion holds
for all ten data sets (two aggregates in each of five countries) being
examined. There is thus a prima facie case for estimating error-
correction models of the demand for these aggregates. It also is clear
that the imposition of prior constraints leads, in most cases, to inferior
performance. In spite of the very strong theoretical prior for price homo-
geneity, the imposition of that constraint can be quite misleading.
Whether a single- or two-stage estimation strategy is to be preferred is,
however, much less clear.
III. Final Error-Correction Estimates
The procedure adopted here for estimating error-correction models

for each of these aggregates is to pursue at least two approaches—the
general single-stage and two-stage strategies discussed above—and then
to select between them on the basis of both encompassing tests and
plausibility of the parameters. If both models have serious problems,
then related models are also estimated and compared.
The results of this approach are summarized in Table 4; the complete
dynamic regressions are listed in the Appendix. With two exceptions,
one or the other of the two basic models has been successfully estimated
for each aggregate. For M3 in Germany, single-stage estimation failed to
confirm the existence of an error-correction model; the final equation
contained only differences in the variables, except for the level of the
long-term interest rate. Two-stage estimation generated an acceptable
steady state, but the error-correction term was nonstationary (that is, the
hypothesis that the data are not cointegrated could not be rejected), and
the dynamic equation was encompassed by other models. The preferred
estimate was taken to be that of model EClc: single-stage estimation
starting with real balances as the dependent variable. For Ml in Japan,
none of the models produced an acceptable error-correction equation,
so the basic models were rerun in first-difference form.
The most striking finding from this exercise is that six of the ten
aggregates have long-run price elasticities that are significantly less than
unity. The reasons for this anomaly are not obvious, but there are two

explanations that probably account for at least part of it. The first is the
possibility of aggregation bias. For four of the five countries, the esti-
mated price elasticity is much closer to unity for narrow money than for
the broad aggregate; for the United States, the opposite holds. It may
be, therefore, that excessive (or, in the U.S. case, insufficient) aggrega-
tion is introducing errors in the parameter estimates. If, for example, the
output elasticities differ across components of the aggregates, constrain-
ing them to be equal could bias the estimates.
The second likely explanation for the low price elasticities is the short-
ness of the data sample (25 years). If the true underlying price elasticity
is unity but adjustment of the stock of money to repeated inflationary
shocks has been incomplete over the data sample, then the estimated
elasticity would be less than unity. To the extent that the central bank
has responded to increases in the price level by reducing the money stock
or that the price level has responded slowly in response to monetary
policies, lags would be increased and the downward bias would be aggra-
vated. In this case, the discrepancy would be greatest for the aggregates
that were the primary focus of monetary policy; this story would thus
suggest that U.S. monetary policy had been concerned more with Ml
during this period, while other countries had focused more on M2.
Sorting out the importance of these various factors would require further
research.25
There is virtually no evidence here in support of economies of scale
in cash holdings. That is, the microeconomic theories pioneered by
Baumol and Tobin do not seem to apply to the aggregate data: for eight
of the ten aggregates, the real income elasticities are unity or higher.
In all but one case (Ml in France), the total effect of interest rates (the
effect of a combined change in short and long rates) is negative, as
expected. In several cases, however, there is a significant term structure
relationship. In three cases, short-term interest rates have a positive
effect on money holdings in the long run, offset by a somewhat larger
negative effect from long-term rates.
Table 5 presents tests of the stability of the regression estimates over
the period 1972-88, which indicate that most of the equations have been
stable since at least the mid-1970s. These are TV-step Chow tests, which
25
Other factors may also help to explain the lack of homogeneity in the
regression estimates. First, the aggregate price indices may not measure the
true prices on which asset-holders' decisions are based. Second, there may be
inflationary-expectations effects that have not been modeled, such that a rise in
the price level would cause the desired stock of real balances to fall. These
factors, however, would not explain why the irregularity is much more present
for one aggregate than for the other in each country.

Table 5. Stability Tests, 1972-88

Date of Most
Aggregate Likely Instability3 F-Testb
United States
Mlc 1986:4 *2.55 (1.91)
M2 1972:3 *2.24(1.79)
Japan
Ml 1971:2 *2.34 (1.87)
M2 (1985:4) 0.63 (1.85)
Germany
Ml (1988:1) 1.88 (2.47)
M3 (1974:2) 0.69 (1.67)
France
Ml 1973:1 *2.15(1.71)
M2 1985:1 *2.05 (1.77)
United Kingdom
Ml (1983:2) 1.47 (1.68)
M3 1972:1 *2.28 (1.83)
Note: Af-step Chow tests for stability of recursive regression estimates. Each
regression is first estimated using data only through 1971:4, and an F-test is
computed against the null hypothesis that the same equation fits the remaining
observations (through 1988:4). The test is then repeated sequentially through
the end of the sample, with one observation added each time. The test statistic
has (N-T,T-k) degrees of freedom, where N is the total number of observa-
tions (100), Tis the number of observations in the truncated sample, and k is the
number
a
of regressors.
Date (T+ 1) at which the value of the F-statistic is maximized. Parentheses
indicate
b
that the shift is not significant at the 5 percent level.
The 5 percent significance level is given in parentheses; asterisks flag cases
where the null hypothesis of stability is rejected at that level.
°For this aggregate, the presence of the dummy variable (dumSO) necessitates
the use of restricted least squares for the recursive regressions. The dependent
variable was redefined as Am - 0dum80, where 0 is the full-sample coefficient
estimate.
test for sustained or permanent shocks to the relationships. In six of the

ten cases, there is a significant shift at some point. In four of those cases,
however, the shifts come in the early 1970s, around the time of the first
oil shock and the switch from fixed to more flexible exchange rates.
Some of the shifts may also be associated with definitional breaks in the
data.
In only two cases—Ml in the United States and M2 in France—is
there any evidence of sustained parameter instability after 1973. The
U.S. Ml equation has a very large residual in 1986:4, a period when Ml

was growing quite rapidly against all expectations, apparently reflecting

shifts in yields on included accounts relative to those on accounts ex-
cluded from Ml. That residual is large enough to destabilize the parame-
ter estimates around that time. The instability in the French M2 equation
reflects a sharp increase in the volatility of the data after the end of 1984,
which in turn may be attributable to the increased openness and flexibil-
ity of the French financial system around that time.26 When the equation
is estimated through the end of 1984 and is then used to forecast through
1988, the forecasts stay on track but miss some unusually large swings.
IV. Conclusions
This paper has examined the long-run properties of money demand

equations for five large industrial countries and has compared the per-
formance of equations specified under various approaches. Examination
of earlier estimates supported the hypothesis that the long-run elastici-
ties should be reasonably robust with respect to changes in the estima-
tion period or in the specification of the estimating equation, while the
short-run properties tend to be much less stable. The estimates pre-
sented here, however, raise questions about the robustness of the long-
run properties as well, and they suggest that some of the most commonly
accepted restrictions employed in the money demand literature may be
inconsistent with the data. These questionable properties include homo-
geneity with respect to the price level, unitary or less-than-unitary elas-
ticities with respect to real income, and restriction of the set of included
interest rates to either short- or long-term rates, to the exclusion of the
other.
The results discussed here are robust with respect to a variety of
estimation strategies. Notably, whether the dependent variable is nomi-
nal money, real balances, or prices is of little consequence. What does
matter is whether one imposes prior constraints on the dynamic process
or allows it to be driven by the data. Two-stage error-correction model-
ing, in which the errors from a cointegrating equation are used as an
argument in a dynamic adjustment equation, is generally outperformed
by a less restricted general-to-specific specification process. Here again,
however, the choice has little effect on the long-run elasticities.
26
From 1964 through 1984, the mean percentage quarterly change in M2 was
3.01 percent, with a standard deviation of 1.36 percent. For the period 1985-88,
the mean increase was 2.08 percent, with a standard deviation of 2.06 percent.
For a review of financial innovation and deregulation in France during the 1980s,
see Patat (1987).

APPENDIX
Dynamic Regressions
The following are the regression estimates that underly the steady-state elas-
ticities listed in Table 4. The sample period is 1964:1-1988:4. Each equation
includes a constant term (not shown), and the variables are defined as in the text;
kiX-j^Xt-j — X t - i - j . Coefficients on interest rates are multiplied by 100 for
clarity. Heteroskedasticity-corrected standard errors are given in parentheses.
The test statistics, in addition to R2 and the Durbin-Watson statistic, are as
follows:
a = standard error of the estimate (x 100), followed by the standard
deviation of the dependent variable in parentheses
AR4 = F-test for fourth-order serial correlation in the residuals
ARi-s = F-test for serial correlation, jointly for orders 1 through 8
X2 = Chi-square tests (with 2 degrees of freedom) for nonnormality in
the residuals.
Tests that reject the null hypothesis at the 5 percent level are marked with an
asterisk. Five of the ten equations show significant departures from the normal
distribution of the residuals (either leptokurtosis or skewness), and one of those
five also shows significant higher-order residual autocorrelation. All of these
cases also show significant parameter shifts (see Table 5), and the residual
problems probably reflect those shifts.
United States
Ml27
M2
27
dum80 = 1 in 1980:1, -1 in 1980:3, and 0 otherwise.

Japan
Ml
M2
Germany
Ml
M3
France
Ml

M2
United Kingdom
Ml
M3
REFERENCES
Artis, Michael J., and M.K. Lewis, "The Demand for Money in the United
Kingdom: 1963-73," The Manchester School of Economic and Social
Studies, Vol. 44 (June 1976), pp. 147-81.
Atkinson, Paul, Adrian Blundell-Wignall, Manuela Rondoni, and Helmut
Ziegelschmidt, "The Efficacy of Monetary Targeting: The Stability of De-
mand for Money in Major OECD Countries," OECD Economic Studies,
No. 3 (August 1984), pp. 145-76.
Baba, Yoshihia, David F. Hendry, and Ross M. Starr, "U.S. Money Demand,
1960-1984," prepared for presentation at the NBER Universities Research
Conference on Money and Financial Markets (Cambridge, Massachusetts,
1985).
Banerjee, Anindya, Juan F. Dolado, David F. Hendry, and Gregor W. Smith,
"Exploring Equilibrium Relationships in Econometrics through Static
Models: Some Monte Carlo Evidence," Oxford Bulletin of Economics and
Statistics, Vol. 48 (August 1986), pp. 253-77.

Bordo, Michael D., and Lars Jonung, The Long-Run Behavior of the Velocity of
Money: The International Evidence (New York: Cambridge University
Press, 1987).
, "The Long-Run Behavior of Velocity: The Institutional Approach Re-
visited, " Journal of Policy Modeling, Vol. 12 (Summer 1990), pp. 165-97.
Boughton, James M., "Recent Instability of the Demand for Money: An Inter-
national Perspective," Southern Economic Journal, Vol. 47 (January 1981),
pp. 579-97.
, William H. Branson, and Alphecca Muttardy, "Commodity Prices and
Inflation: Evidence from Seven Large Industrial Countries," NBER Work-
ing Paper 3158 (Cambridge, Massachusetts: National Bureau of Economic
Research, November 1989).
Boughton, James M., and George S. Tavlas, "Modeling Money Demand in
Large Industrial Countries," Journal of Policy Modeling, Vol. 12 (Summer
1990), pp. 433-61.
Carr, Jack, and Michael R. Darby, "The Role of Money Supply Shocks in the
Short-run Demand for Money," Journal of Monetary Economics, Vol. 8
(September 1981), pp. 183-99.
, and Daniel Thornton, "Monetary Anticipations and the Demand for
Money: Reply," Journal of Monetary Economics, Vol. 16 (September
1985), pp. 251-58.
Coats, Warren L., "Modeling the Short-Run Demand for Money with Exog-
enous Supply," Economic Inquiry, Vol. 20 (April 1982), pp. 222-39.
Coghlan, Richard T., "A Transactions Demand for Money," Bank of England,
Quarterly Bulletin, Vol. 18 (March 1978), pp. 48-60.
Cuthbertson, Keith, and Mark P. Taylor, "Monetary Anticipations and the
Demand for Money: Some Evidence for the U.K.," Weltwirtschaftliches
Archiv, Vol. 123 (1987), pp. 509-20.
Darby, Michael R., "The Allocation of Transitory Income Among Consumers'
Assets," American Economic Review, Vol. 62 (December 1972), pp. 928-
41.
Davidson, Russell, and James G. MacKinnon, "Several Tests for Model Specifi-
cation in the Presence of Alternative Hypotheses," Econometrica, Vol. 49
(May 1981), pp. 781-93.
Domowitz, Ian, and Craig S. Hakkio, "Interpreting an Error Correction Model:
Partial Adjustment, Forward-Looking Behavior, and Dynamic Interna-
tional Money Demand," Journal of Applied Econometrics, Vol. 5 (Janu-
ary-March 1990), pp. 29-46.
Engle, Robert F., and C.W.J. Granger, "Cointegration and Error Correction:
Representation, Estimation, and Testing," Econometrica, Vol. 55 (March
1987), pp. 251-76.
Fair, Ray C., "International Evidence on the Demand for Money," Review of
Economics and Statistics, Vol. 69 (August 1987), pp. 473-80.
Friedman, Milton, and Anna J. Schwartz, Monetary Trends in the United States
and the United Kingdom (Chicago: The University of Chicago Press, 1982).
, "Alternative Approaches to Analyzing Economic Data," * American
Economic Review (forthcoming 1991).

Goldfeld, Stephen, "The Case of the Missing Money," Brookings Papers on

Economic Activity: 1 (Washington: The Brookings Institution, 1976), pp.
683-739.
, and Daniel Sichel, "The Demand for Money," in Handbook of Mone-
tary Economics, Vol. 1, ed. by Benjamin M. Friedman and Frank H. Hahn
(Amsterdam; New York: North-Holland (1990).
Hacche, Graham, "The Demand for Money in the United Kingdom: Experience
Since 1971," Bank of England Quarterly Bulletin, Vol. 14 (September 1974),
pp. 284-305.
Hall, S.G., S.G.B. Henry, and J.B. Wilcox, "The Long-Run Determination of
the U.K. Monetary Aggregates," Bank of England, Discussion Paper
No. 41 (London: Bank of England, August 1989).
Hendry, David F., "Econometric Methodology: A Personal Perspective," in
Advances in Econometrics, Vol. 2, ed. by Truman F. Bewley (Cambridge:
Cambridge University Press, 1987).
, "PC-GIVE: An Interactive Econometric Modelling System," Version
6.0/6.01, University of Oxford, January 1989.
-, and Grayham E. Mizon, "Serial Correlation as a Convenient Simpli-
fication, Not a Nuisance: A Comment on a Study of the Demand for Money
by the Bank of England," The Economic Journal, Vol. 88 (September
1978), pp. 549-63.
Hendry, David F., and Neil R.Ericsson, "Econometric Analysis of UK Money
Demand," in Monetary Trends in the United States and the United Kingdom,
ed. by Milton Friedman and Anna J. Schwartz (Washington: Board of
Governors of the Federal Reserve System, July 1989).
Johansen, Soren, "Statistical Analysis of Cointegration Vectors," Journal of
Economic Dynamics and Control, Vol. 12 (June-September 1988), pp. 231-
54.
, and Katarina Juselius, "Maximum Likelihood Estimation and Inference
on Cointegration, with Applications to the Demand for Money," Oxford
Bulletin of Economics and Statistics, Vol. 52 (May 1990), pp. 169-210.
Judd, John P., and John L. Scadding, "Dynamic Adjustment in the Demand
for Money: Tests of Alternative Hypotheses," Federal Reserve Bank of San
Francisco Economic Review (Fall 1982), pp. 19-30.
Laidler, David, The Demand for Money (New York: Harper and Row, 3rd ed.,
1985).
Leventakis, John A., "Modeling Money Demand in Open Economies over the
Modern Floating Rate Period" (unpublished; Federal Reserve Bank of
St. Louis, 1990).
Lucas, Robert E., Jr., "Money Demand in the United States: A Quantitative
Review," Carnegie-Rochester Conference Series on Public Policy, Vol. 29
(Autumn 1988), pp. 137-68.
Meltzer, A.H., "The Demand for Money: Evidence from the Time Series,"
Journal of Political Economy, Vol. 71 (1963), pp. 219-46.
Muscatelli, V.A., "A Comparison of the 'Rational Expectations' and 'General-
to-Specific' Approaches to Modelling the Demand for Ml," Oxford Bulletin
of Economics and Statistics, Vol. 51 (November 1989), pp. 353-75.

Patat, Jean-Pierre, "Les nouveaux agr6gats monetaires en France: signification,

contenu et comportement," Cahiers feconomiques et Monetaires, Banque
de France (1987), pp. 59-99.
Porter, Richard D., Paul A. Spindt, and David E. Lindsey, "Econometric Mod-
eling of the Demands for U.S. Monetary Aggregates: Conventional and
Experimental Approaches," in Structural Change and Economic Modeling:
Papers and Proceedings of the 7th Pacific Basin Central Bank Conference on
Economic Modeling (Sydney: Ambassador Press, 1989), pp. 218-30.
Schwert, G. William, "Effects of Model Specification on Tests for Unit Roots
in Macroeconomic Data," Journal of Monetary Economics, Vol. 20 (July
1987), pp. 73-103.

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
Private Investment
in Developing Countries
An Empirical Analysis
JOSHUA GREENE and DELANO VILLANUEVA*
The effects of several policy and other macroeconomic variables on the

ratio of private investment to gross domestic product in developing coun-
tries during 1975-87 is analyzed. Econometric evidence indicates that the
rate of private investment is positively related to real GDP growth, level of
per capita GDP, and the rate of public sector investment, and negatively
related to real interest rates, domestic inflation, the debt-service ratio, and
the ratio of debt to GDP. The impact of most variables was greatest before
the 1982 debt crisis, but the ratio of debt to GDP has since become more
important. [JEL 023, 121, 221]
S INCE THE beginning of the 1980s developing countries have experi-

enced a pronounced slowdown in economic growth. The growth rate
of real gross domestic product (GDP), which for all developing countries
averaged 5.5 percent a year during the 1971-80 period, averaged only
3.3 percent a year during 1981-89 (International Monetary Fund (1989,
pp. 78-79)). On a per capita basis, the average growth in real GDP fell
from 3 percent a year during 1971-80 to less than 1 percent a year during
* Joshua Greene, a Senior Economist in the Developing Country Studies

Division of the Research Department, holds graduate degrees in both law and
economics from the University of Michigan. His research includes studies of
special issues affecting Africa and Eastern Europe.
Delano Villanueva is a Senior Economist in the Developing Country Studies
Division of the Research Department. He is a graduate of the University of the
Philippines and the University of Wisconsin and has published articles on finan-
cial reforms, saving, and growth.
The authors are grateful to Mohsin Khan, Peter Montiel, Jacques J. Polak,
Steven Symansky, and Philip Young for comments and suggestions, and to
Ravina Malkani for research assistance.
33

34 JOSHUA GREENE and DELANO VILLANUEVA
1981-89. High among the reasons for this slowdown has been a decline
in investment rates, which have been shown to be positively and signifi-
cantly related to real growth rates in a large sample of developing coun-
tries (see International Monetary Fund (1988)). Gross capital formation
in developing countries declined from an average of 26.5 percent of GDP
during 1981 to less than 23.5 percent during 1985-88 (International
Monetary Fund (1989, p. 80)).1
The decline in gross investment rates reflects many factors that have
affected most developing countries during the 1980s. These include,
among others, falling prices for primary commodity exports, a decline in
private external financing, the presence of a large stock of foreign debt,
and the implementation of adjustment programs designed to restore
balance of payments viability. While there may have been an overall
decline in investment, the ratio of gross investment to GDP has differed
substantially across countries and regions, remaining close to its 1981
level for developing countries in Asia and Europe while falling signifi-
cantly in other regions. Over time, there have also been important
differences among countries. For example, during the 1980s developing
countries with recent debt-servicing difficulties experienced lower rates
of gross capital formation than developing countries without such prob-
lems. Likewise, gross capital formation was, on average, greater for
developing countries specializing in manufactured exports than for coun-
tries exporting primary commodities—mostly minerals or agricultural
products.
These differences in gross capital formation across countries have
reflected variations in both public and private sector investment rates.
The importance of public sector investment has been underscored dur-
ing the 1980s, as the adoption of adjustment programs led many deve-
loping countries to reduce public sector investment activity as a way to
cut fiscal deficits. Nevertheless, because public sector investment in most
developing countries is effectively a policy variable, economists have
focused on private sector investment as being more susceptible to exten-
sive economic analysis. Also contributing to the interest in private in-
vestment activity is recent research suggesting that private sector invest-
ment has been more directly related to economic growth in developing
countries than has public sector investment (Khan and Reinhart (1990)).
Despite the recognition that private investment plays a critical role in
generating economic growth, there has been surprisingly little research
1
According to the same source, median levels of gross capital formation have
fallen even more sharply during this period, from 25.3 percent of GDP during
1981 to 20.3 percent of GDP or less during 1987-89.

PRIVATE INVESTMENT 35
on its determinants in developing countries. Stern (1989, p. 672), for

example, in his recent survey of development economics, notes that
"what determines investment" remains very much an open question in
research on economic growth. Among the few recent studies on the
subject is Blejer and Khan (1984), which examined the impact of govern-
ment economic policy on private investment in some 24 developing
countries. This study found that the level of private investment activity
was related positively to the change in expected real GDP, negatively to
excess productive capacity (the shortfall of actual GDP from its trend
value), and positively to the availability of funds for private investment
(as measured by the change in bank credit for the private sector and in
the level of private capital inflows). The study also found that the level of
private sector investment was a positive function of the trend level of
government investment, which was taken as representing investment in
infrastructure, but not of deviations from that trend. This finding sug-
gests that there is long-run complementarity of private to public sector
investment, but short-run substitutability, in the sense that short-run
increases in public sector investment appear to crowd out private sector
investment.
The present study is an attempt to learn more about the empirical
determinants of private investment activity in developing countries dur-
ing the post-1974 period. Following the approach taken in a recent study
of national saving behavior (Aghevli and others (1990)), this paper pro-
vides a preliminary look at how various macroeconomic factors have
affected private investment activity during this period in a number of
developing countries. Among the factors examined are the following:
(1) economic growth and per capita income level; (2) macroeconomic
stability (as represented by low inflation rates); (3) the level of real
interest rates; (4) the size of debt-service burdens (as measured by
debt-service ratios and the magnitude of external debt relative to GDP);
and (5) the rate of public sector investment. Because of the difficulty in
identifying the theoretically correct specification and obtaining the nec-
essary data, this paper does not attempt to build and estimate a full-scale
structural model of private investment in developing countries.2 Rather,
it is more of an exploratory data analysis. Nevertheless, the results of this
study may be useful in identifying the more fundamental relationships
between private sector investment and macroeconomic variables in these
countries, which can then be used to develop an appropriate model of
investment behavior in developing economies.
2
Examples of possible models are contained in Blejer and Khan (1984) and
Sundararajan and Thakur (1980).

Table 1. Private Investment in Selected Developing Countries, 1975-87
(As a percent of GDP)
f vlean Public Debt-
Sector Service
Mean Mean Mean Investment Difficulties
Country 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1975-87 1975-81 1982-87 1975-87* in 1986-88
Argentina 17.1 13.9 14.4 12.5 12.5 13.0 9.4 8.9 7.9 6.9 5.5 6.0 6.8 10.4 13.3 7.0 9.4 Yes
Bolivia 109 84 69 6.8 6.9 73 3.8 5.0 2.0 1.6 3.6 4.9 4.9 5.6 73 37 74 Yes
Brazil 20.9 16.1 15.7 13.6 7.7 14.5 12.0 11.7 9.7 9.5 10.5 12.3 12.8 12.8 14.4 11.1 8.6 Yes
Chile 3.2 4.8 7.6 11.4 12.6 15.6 17.5 6.5 4.9 7.3 6.8 7.0 10.0 8.9 10.4 7.1 6.6 Yes
Colombia 11 8 12 1 96 11 5 12.3 11.4 12.0 11.1 11.0 10.0 9.4 9.3 11.1 11.0 11 5 103 79 No
Costa Rica 15 1 14 2 15.6 17.3 14.7 15.2 13.1 11.6 13.7 12.3 12.8 15.1 14.2 15.4 13 1 71 Yes
Ecuador 14.5 13.0 13.4 16.8 14.5 14.1 11.7 13.0 8.6 9.0 9.5 9.8 14.6 12.5 14.0 10.8 8.8 Yes
Guatemala 12 1 14 0 13 0 144 124 98 8.4 8.8 59 5.8 8.3 8.1 9.5 10.0 120 77 50 Yes
India 96 103 103 10.6 10.7 10.5 10.7 10.6 9.7 9.4 9.4 10.4 11.0 10.2 10.4 10 1 9.1 No
Kenya 11.7 11.6 12.1 15.6 12.8 12.9 13.4 10.6 11.6 10.7 10.9 11.8 12.6 12.2 12.9 11.4 8.8 No
Korea, Rep.
of 193 18 8 20 7 24 5 25 8 23 4 19.8 21.7 23.5 23.3 21.8 22.4 230 22.2 21 8 226 69 No
Mexico 124 12.8 11 9 12.8 13.5 13.9 14.7 12.7 10.0 10.8 12.2 13.3 13.5 12.7 13.1 12 1 84 Yes
Pakistan 4.6 5.9 6.2 5.8 5.8 6.4 6.1 5.6 6.0 6.1 6.2 6.1 6.3 5.9 5.8 6.1 10.3 No
Peru 9.2 8.7 8.3 8.4 8.5 9.7 11.4 11.5 7.9 6.9 6.8 9.3 10.6 9.0 9.2 8.8 7.6 Yes
Philippines 18 6 177 166 16 8 18 0 184 18.5 18.1 18.8 14.5 11.4 9.7 10.6 16.0 178 13 9 60 Yes
Singapore 265 23 6 21 6 22 4 25 0 26.7 30.4 31.9 30.7 31.3 26.7 21.9 23.4 26.3 25 2 277 124 No
Sri Lanka 82 7.9 7.2 7.9 13.0 12.9 12.7 13.3 14.3 11.9 10.7 10.6 10.9 10.9 10.0 12.0 12.3 No
Thailand 17.7 16.1 18.6 17.6 18.0 16.3 15.8 15.3 15.9 16.1 14.6 14.1 17.3 16.4 17.2 15.6 12.1 No
Tunisia 12.3 11.6 10.9 11.7 12.0 13.3 14.8 15.4 14.2 13.6 12.3 10.5 9.0 12.4 12.4 12.5 16.3 No
Turkey 11 7 13 2 13 2 12.4 10.7 95 8.1 8.1 8.3 8.3 8.4 9.8 11.3 10.2 11.3 90 11 5 No
Uruguay 87 90 82 8.0 9.7 11.4 10.7 7.9 6.9 5.2 4.5 4.4 5.4 7.7 9.4 5.7 52 Yes
Venezuela 15.8 18.6 23.6 24.5 18.1 13.0 9.8 7.7 4.5 6.6 6.5 6.8 6.6 12.5 17.6 6.5 12.9 Yes
Zimbabwe 13.7 11.2 10.1 8.4 9.2 10.6 13.4 10.1 8.4 10.6 7.3 7.8 7.2 9.8 10.9 8.6 8.1 No
Meanb 13.2 12.8 13.4 13.5 13.3 13.4 13.1 12.1 11.0 10.8 10.2 10.4 11.2 12.2 13.2 11.0 9.1
Sources: Pfefferman and Madarassy (1989), International Monetary Fund (1989), and estimates by the authors.
a
Ratio of public sector investment to GDP.
b
Unweighted.

The paper is organized as follows: Section I reviews recent trends in

private investment activity across a group of 23 developing countries.3
Various patterns in private investment rates are identified, and these are
compared with public sector investment rates in the countries. Section II
reviews a number of hypotheses that have been advanced for explaining
differences in private investment rates. This section also contains a
casual examination of these hypotheses, comparing average levels of the
various indicators in countries with above- and below-average private
investment rates. The hypotheses are then tested econometrically in
Section III. Pooled time-series, cross-section equations are estimated,
relating private investment rates to a number of economic variables,
both for the entire sample period and for the predebt crisis (1975-81)
and more recent (1982-87) time periods. Section IV draws some
implications from these findings and offers suggestions for further re-
search. A statistical appendix describes the data sources for the paper.
I. Recent Trends in Private Investment in Developing Countries
Data on private investment rates for 23 developing countries over the

period 1975-87 were recently assembled in the World Bank Group (see
Pfefferman and Madarassy (1989)). These data, summarized in Table 1,
reveal several interesting patterns. First, there is a wide discrepancy in
private investment rates across countries. A few countries, in particular
the newly industrializing and rapidly growing Asian countries, exhibit
very high rates of private investment, often exceeding 20 percent of
GDP. At the other extreme, less affluent and more slowly growing
countries, such as Bolivia and Peru, have experienced much lower rates
of private investment, sometimes falling to less than 10 percent of GDP.
For most countries in the sample, private investment averaged between
10 and 15 percent of GDP during most of the 1975-87 period.
The data in Table 1 also indicate a significant decline in private invest-
ment activity between the first part of the observation period, 1975-81,
and the second part, 1982-87, which has been characterized by recurring
debt crises in a number of developing countries. For the 23 countries in
the sample, the average level of private investment activity decreased
from 13.2 percent of GDP during 1975-81 to 11.0 percent in 1982-87.
This trend is illustrated in Figure 1, which shows a significant decline in
average investment rates after 1981. Although private investment rates
3
The 23 countries are Argentina, Bolivia, Brazil, Chile, Colombia, Costa
Rica, Ecuador, Guatemala, India, Kenya, the Republic of Korea, Mexico, Pak-
istan, Peru, the Philippines, Singapore, Sri Lanka, Thailand, Tunisia, Turkey,
Uruguay, Venezuela, and Zimbabwe.

Figure 1. Average Levels of Private Investment

in 23 Developing Countries, 1975-87
(As a percent of GDP)
in a few developing countries, notably the Republic of Korea and Singa-

pore, increased during the 1980s, for most countries in the sample pri-
vate investment rates decreased. In several cases the decline was precipi-
tous. In Argentina, for example, private investment as a share of GDP
declined from an average of 13.3 percent in 1975-81 to 7.0 percent in
1982-87. In Venezuela the figures were 17.6 percent in 1975-81 and 6.5
percent in 1982-87. On balance, the difference in investment rates be-
tween the two periods was smaller for countries that did not undergo
rescheduling or incur debt-service arrears during the 1980s, such as
Colombia and India, than for countries that did have difficulties in
meeting scheduled debt-service obligations during this period, such as
Argentina and Bolivia. At the same time, investment rates also declined
noticeably for several countries that experienced higher debt-service
obligations during the 1980s, but were able to meet them without
rescheduling, such as Kenya and Zimbabwe.
The range of public sector investment rates for the countries in the
sample is smaller than that for private investment. As shown in Table 1,
public sector investment averaged between 5 percent and 13 percent of
GDP during 1975-87 in all of the 23 countries except Tunisia, with the
average for all countries being 9 percent. There is no obvious correlation

between high rates of private and public sector investment. For example,
public sector investment rates in Singapore and Thailand, two of the
countries with the highest private investment rates, were double those in
Korea and the Philippines, two other countries with relatively high levels
of private investment activity. Indeed, public investment rates in Korea
and the Philippines were lower than in many countries with smaller rates
of private sector investment. This disparity no doubt reflects the differ-
ent emphasis accorded public and private sector activity in different
countries. Nevertheless, public sector investment commanded a smaller
percentage of GDP than did private sector investment in all but five
countries in the sample (Bolivia, Pakistan, Sri Lanka, Tunisia, and
Turkey).
II. Factors Affecting Private Investment Rates
A number of hypotheses have been advanced to explain the variations

in private investment activity observed in developing countries.4 This
variety to some extent reflects uncertainty about the form of the private
investment function for these countries. The neoclassical flexible-accel-
erator model has been the most widely accepted general theory of invest-
ment behavior, and empirical tests of the model using data from several
industrial countries have been quite successful (see, for example,
Bischoff (1969, 1971), Mines and Catephoros (1970), Jorgenson (1967,
1971), and Clark (1979)). However, it has generally been hard to test this
model in developing countries, because key assumptions (such as perfect
capital markets and little or no government investment) are inapplica-
ble, and data for certain variables (capital stock, real wages, and real
financing rates for debt and equity) are normally either unavailable or
inadequate. Accordingly, research has proceeded in several directions,
in the process identifying a number of economic variables that might be
expected to affect private investment in developing countries. These
efforts, however, have not yet produced a full-fledged model of invest-
ment behavior in developing countries.
Theoretical Analysis
Because of the problems inherent in applying the standard neoclassi-
cal model to developing countries, one line of research, pursued notably
by McKinnon (1973) and Shaw (1973), has abandoned this model, ad-
4
For a more comprehensive analytical overview of private investment theory
and the impact of macroeconomic policies on private investment in developing
countries, see Serven and Solimano (1989).

vancing instead the hypothesis that private investment in developing

countries is positively related to the accumulation of domestic real
money balances. Underlying this hypothesis is the assumption that pri-
vate investors in these countries must accumulate money balances before
undertaking investment projects because of their limited access to credit
and equity markets. Because real money balances are directly influenced
by real deposit interest rates, there should be a positive relationship
between private investment and real interest rates in these countries.
This approach accordingly disregards the negative effect of higher real
rates on investment via increases in the user cost of capital that normally
follows from the neoclassical investment model.
Another line of research has attempted to retain the neoclassical
model, but to address the analytical and data problems involved in its
application to developing countries, in particular the lack of data and the
resource constraints facing private investors in developing countries
(see, for example, Sundararajan and Thakur (1980), Tun Wai and Wong
(1982), and Blejer and Khan (1984)). Applying the neoclassical model
leads to the conclusion that the private investment rate should be nega-
tively related to the real interest rate as a measure of the user cost of
capital.5 These studies also suggest that the rate of growth of real output
(real GDP) per capita should be positively related to the private invest-
ment rate, as is common in industrial countries.6
In addition to the real interest rate and real per capita growth rate, the
application of the neoclassical model to developing countries has led to
the public investment rate (the ratio of public investment to GDP) being
identified as a factor affecting the rate of private investment in these
countries (Blejer and Khan (1984)). However, at the theoretical level the
effect of public sector investment is ambiguous. On the one hand, public
investment activity may be complementary to and thus support private
investment, particularly where public investment involves useful infra-
structure—transportation systems, schools, water and sewage systems,
and the like. Projects in these areas tend to raise the expected rate of
return on private investment. On the other hand, public sector invest-
ment may detract from private investment activity to the extent that it
substitutes for or crowds out private investment. This may occur when
the investment involves parastatal enterprises producing goods that com-
pete with the private sector, or when heavy spending for public capital
5
The real interest rate is closer to the spirit of the neoclassical model than are
measures of the availability of financing, which some studies have used in the
absence of interest rate data.
6
This can be readily derived from a flexible-accelerator model with a fixed
relationship between the desired capital stock and the level of real output.

projects leads to high interest rates, severe credit rationing, or a heavier

current or future tax burden (Aschauer (1989)).
Besides the factors derived from the neoclassical investment model,
the domestic inflation rate has also been proposed as affecting private
investment rates in developing countries, where inflation is less often
correlated with a rise in economic output than in industrial countries
(Dornbusch and Reynoso (1989)). High rates of inflation adversely af-
fect private investment by increasing the riskiness of longer-term invest-
ment projects, reducing the average maturity of commercial lending,
and distorting the information content of relative prices. In addition,
high inflation rates are often considered an indicator of macroeconomic
instability and a country's inability to control macroeconomic policy,
both of which contribute to an adverse investment climate. Thus, the
domestic inflation rate should be negatively related to the rate of private
investment.
Private investment activity has been hypothesized as a positive func-
tion of income per capita because of the greater ability of higher income
countries to devote resources to saving. This ability is particularly im-
portant given the imperfection of capital markets, since it appears that
most investment projects must be financed, at least in substantial part,
through domestic savings.
Finally, the presence of large external debt burdens has also been
suggested as a factor reducing investment activity in three ways. First,
the higher debt-service payments associated with a large external debt
reduce the funds available for investment. Second, the existence of a
large debt overhang, in the form of a high ratio of external debt to GDP,
can reduce the incentives for investment, because much of the forthcom-
ing returns from investment must be used to repay existing debt, there-
fore acting as a tax on domestic investment (Borensztein (1989) and
Froot and Krugman (1990)). Third, if substantial external debt leads to
difficulties in meeting debt-service obligations, relations, with external
creditors may deteriorate, thus reducing the amount of trade financing a
country can obtain. This, in turn, may make it harder or more costly to
finance private investment, because imports play a major role in most
investment projects in developing countries, and most developing coun-
try imports are investment related (Mirakhor and Montiel (1987)).
Preliminary Evidence
As a first look at the evidence regarding the various factors discussed

above, it is interesting to compare the average values of these variables
in countries with above- and below-average private investment rates,

Table 2. Average Levels of Major Economic Indicators in Selected Developing Countries, 1975-87
Real
Private Real Per Capita Public Sector Per Capita Debt- External
Investment Deposit GDP Growth Investment Inflation GDP Service Debt-to-
Country Rate Rate Rate Rate3 Rate (in U.S. dollars) Ratio GDP Ratio
Singapore 26.3 3.6 5.57 12.4 2.8 5129.17 1.6 16.5
Korea, Rep. of 22.2 2.3 6.84 6.9 11.9 1615.43 19.3 43.1
Thailand 16.4 4.0 4.56 12.1 6.4 662.36 18.6 27.9
Philippines 16.0 -0.3 0.71 6.0 13.9 590.02 29.9 56.1
Costa Rica 14.2 -2.4 0.48 7.1 21.1 2502.51 31.9 43.6
Brazil 12.8 -1.5 1.60 8.6 112.4 2027.17 53.5 27.8
Mexico 12.7 -5.6 0.81 8.4 50.2 1920.99 56.6 42.9
Ecuador 12.5 -7.4 0.67 8.8 20.4 982.82 41.8 58.1
Venezuela 12.5 -2.2 -0.96 12.9 12.5 3105.61 26.0 55.6
Tunisia 12.4 -3.1 2.07 16.3 8.2 1138.78 17.9 46.5
Kenya 12.2 -2.7 0.54 8.8 12.3 328.18 19.6 37.0
Colombia 11.0 2.2b 1.97 7.9 23.3 1009.72 27.3 32.1
Sri Lanka 10.9 2.0 3.14 12.2 11.4 297.42 25.3 42.8
Argentina 10.4 -21.0 -1.53 9.4 259.3 2876.73 50.8 34.4
India 10.2 0.8 2.86 9.1 7.1 232.65 19.5 14.7

Turkey 10.2 -4.7 2.79 11.5 41.3 1186.33 32.7 31.8
Guatemala 10.0 -2.0 -0.37 5.0 12.4 1058.50 17.3 17.3
Zimbabwe 9.8 -3.7 -1.29 8.1 12.2 934.18 16.5 15.8
Peru 9.0 -21.2 -0.28 7.6 74.5 1064.28 54.3 56.0
Chile 8.9 4.9 1.08 6.6 71.6 1634.70 45.9 67.6
Uruguay 7.7 -0.3 1.49 5.2 56.5 1709.04 27.8 43.8
Pakistan 5.9 1.2 2.87 0.3 8.4 270.04 29.3 38.9
Bolivia 5.6 -26.7 -2.08 7.4 155.7C 728.46 34.1 62.9
High private
investment
countries'1 15.4 -0.9 2.1 9.8 24.7 1818.50 28.8 41.4
Low private
investment
countries6 9.1 -5.7 0.9 8.3 61.1 1083.50 31.7 38.2
Sources: Pfefferman and Madarassy (1989) and International Monetary Fund (1989).
a
As a percentage of GDP.
b
Based on the effective yield on a 90-day certificate of deposit.
c
Geometric average.
d
First 11 countries in the table.
e
Last 12 countries in the table.

and vice versa. This type of analysis is hardly definitive, because other
variables are not held constant in the real world and an apparent rela-
tionship between one factor and private investment rates may in fact
result from movements in a common underlying variable. Still, the infor-
mation does offer a preliminary look at whether countries with high
(equal to or greater than 12 percent of GDP) and low (less than 12
percent of GDP) private investment rates also differ in other economic
respects.
Table 2, which reports the average levels for a number of economic
indicators in the sample of 23 countries during the 1975-87 period,
provides support for many of the hypotheses outlined earlier. The data
suggest, for example, that interest rates may indeed affect private invest-
ment activity through their role in mobilizing domestic savings. As is
shown in Table 2, the average real deposit rate on midterm (6- to
24-month) deposits in countries with higher levels of private investment
was —0.9 percent, implying an average nominal deposit rate just less
than the mean rate of inflation. For the countries with lower private
investment rates, however, the average real deposit rate was —5.7 per-
cent. The data also suggest that the rate of real GDP growth per capita
was greater in the countries with higher private investment rates, averag-
ing 2.1 percent a year, more than twice the average for countries with
lower private investment rates. In addition, public investment in these
countries may on balance be complementary to private sector invest-
ment, as the average rate of public sector investment in the countries
with higher private investment activity (9.8 percent of GDP) was larger
than that for the countries with smaller private investment rates (8.3
percent of GDP).
The data in Table 2 support the view that high inflation rates may be
inimical to strong private investment activity, as the average inflation
rate in countries with higher private investment rates, about 25 percent a
year, was far below the average of 61 percent for countries with lower
private investment rates. The data also indicate that average per capita
GDP for the 11 countries with mean private investment rates above the
sample median of 12 percent of GDP was, at $1,818, nearly 70 percent
larger than the average for the 12 countries with smaller private invest-
ment rates. As for external debt, although countries with higher private
investment rates experienced somewhat higher external debt-to-GDP
ratios, these countries also had somewhat smaller average debt-service
ratios (29 percent versus 32 percent of exports of goods and services),
perhaps suggesting an ability to use borrowed funds more efficiently.
A second way of looking at the different hypotheses is to compare
private investment rates for groups of countries with higher or lower

Table 3. Average Private Investment Rates for Countries with Selected

Characteristics, 1975-87
(In percent of GDP)
Characteristic Private Investment Rate
Nonnegative real interest rate 13.0
Negative real interest rate 11.7
Higher growth3 12.8
Lower growth 11.9
Higher public investment rateb 12.0
Lower public investment rate 14.7
Higher inflation0 11.8
Lower inflation 13.6
Higher incomed 13.2
Lower income 11.1
Higher debt-service ratio6 11.9
Lower debt-service ratio 13.5
Higher debt-to-GDP ratiof 14.0
Lower debt-to-GDP ratio 11.7
Sources: Pfefferman and Madarassy (1989) and International Monetary Fund
(1989).
Note: Averages for countries in designated groups, with individual country
observations
a
weighted by three-year average of country GDPs.
b
Average growth rate of real GDP above 1.4 percent a year.
c
Average ratio of public sector investment to GDP greater than 8.4 percent.
d
Average annual rate of increase in consumer price index above 20 percent.
e
Average per capita GDP (1975-87) above $1,100.
Average debt-service ratio greater than 29 percent of exports of goods and
services.
f
Average ratio of external debt to GDP greater than 40 percent.
levels of the relevant economic variables. This is done in Table 3, using

average levels of private investment for each country weighted by the
country's GDP.7 The figures in Table 3 indicate that countries with
nonnegative real interest rates, higher average real GDP growth rates,
and lower average inflation and public investment rates also have higher
average rates of private investment. The same is true for countries with
higher average real levels of GDP per capita, and lower debt-service (but
not lower debt-to-GDP) ratios. These results are generally consistent
with theoretical expectations. Except for the findings regarding coun-
tries with higher and lower public investment rates, they are also consis-
tent with the data in Table 2.
7
As in International Monetary Fund (1989), weights were calculated on the
basis of each country's average GDP in U.S. dollars during the previous three
years.

III. Econometric Results
To examine more rigorously the various hypotheses outlined above,

equations for the private investment rate were estimated for the 23
countries in the sample, using a pooled time-series, cross-section ap-
proach. A detailed list of the variables and data sources appears in the
statistical appendix to this paper. Because the current values of the real
per capita growth rate, the per capita GDP level, and the debt-service
ratio may be affected by the private investment rate, lagged values of
these variables were used to reduce the possibility of simultaneous-equa-
tions bias in the coefficient estimates. In addition, the lagged value of the
ratio of external debt to GDP was employed, because the information is
usually available only for the end of the year and is therefore generally
known retrospectively. To capture the effects of country-specific factors,
a dummy variable for each country was included in the specifications.
Thus, the equations took the following form:
IP/Y=f[RI, GRt-tJPUB/GDP, CPI, INCt.l9 (DS/XGS),.^
(DEEBT/GDP)t-1,Z],
where
IP/Y = the ratio of private sector investment to GDP
RI = the real deposit interest rate, as measured by
the ratio (1 + NINT)I(1 + ECPI\ where NINT
is the nominal interest rate and ECPI is the
expected inflation rate
GRt-\ = the lagged percentage change in real GDP per
capita
IPUB/GDP = the ratio of public sector investment to GDP
CP/ = the percentage change in the country's con-
sumer price index
INCt-i = the lagged level of per capita GDP in current
U.S. dollars
(DS/XGS)t-i = the lagged ratio of external debt-service pay-
ments to exports of goods and services
(DEBT/GDP)t-1 1= the lagged ratio of the country's stock of exter-
nal debt to its nominal GDP
Z = a vector of country dummy variables, one for
each country in the sample, with the value of
each variable set equal to 1 whenever obser-
vations for that country were entered, and 0
otherwise.

For the real interest rate, three different variants were tried: one using
the current period value of the percentage change in the consumer price
index as the expected inflation rate; one using the previous year's value;
and one using the value of the year ahead, which is conceptually the
correct specification.8 The best results came from using the value of the
consumer price change one period ahead (that is, CPIt+i) to generate
the real interest rate, in line with the theoretically correct specification.
Because CPI and CP/,+i may both be affected by the rate of private
investment, instrumental variables were used for the real interest rate
and the current period's inflation rate, CPI.9
This equation was estimated over the entire 1975-87 time period. In
addition, separate equations were estimated for the two subperiods
1975-81 and 1982-87, to test for the effect of the post-1981 debt
crisis on the results. The results of all three equations are summarized in
Table 4. The table omits the results for the country dummy variables, all
of which were statistically significant at the 1 percent level. As indicated,
the overall fit of all three equations was fairly good, with R2 statistics
adjusted for degrees of freedom in the 0.7-0.9 range.
The results for equation (1) in Table 4, which reflects the entire
sample period, supported most of the hypotheses outlined earlier. The
estimated coefficient for the ratio of public sector investment to GDP
(IPUB/GDP) was positive and significant, suggesting that in this sample
public sector investment was on balance complementary to private sec-
tor investment activity. In addition, the lagged per capita real GDP
growth rate (GRt-i) was positive and highly significant, while the coeffi-
cients for the lagged debt-service ratio, (DS/XGS)t-i, and the lagged
debt stock, (DEBT/GDP)t-1? were both negative and statistically signifi-
cant at the 1 percent level.10 In addition, the estimated coefficient for the
inflation rate (CPI) was negative and highly significant, implying that a
higher inflation rate, other things equal, had a negative impact on the
private investment rate for countries in the sample. However, the esti-
8
The real rate of interest is thus defined as
with both NINT and CPI divided by 100; that is, a 10 percent nominal interest
rate is written as NINT = 0.10.
9
The instrumental variables included all the country dummy variables, the
current value of the public investment rate, and one-year lags of all the other
variables.
10
The significance levels reported here are for one-tailed tests, except for the
public investment rate and the real interest rate, for which two-tailed tests are
used.

Table 4. Regressions for Private Investment Rates
IPUBI (DEBT/ Number of

RI Gfl,-! GDP CPI INCt - ! (DS/XGS)t - 1 GDP),_! Observations R2 SEE
Entire sample period (1975-87)
(1) -0.083** 0.249*** 0.080***
(-3.06) (6.50) (2.03)
-0.004***
(-4.02)
0.084
(0.32)
-0.031*** -0.033***
(-2.55) (-2.99)
294 0.81 2.38
Predebt crisis period (1975-81)

(2) -0.078 0.210*** 0.359*** -0.024** 0.870* -0.043** -0.041* 156 0.79 2.34
(-1.62) (3.01) (3.17) (-2.62) (1.34) (-1.73) (-1.40)
Debt crisis period (1982-87)
(3) -0.019 0.209*** 0.061** -0.001*** -1.560*** -0.003 -0.052*** 138 0.92 1.56
(-1.03) (5.67) (2.05) (-3.02) (-2.87) (-0.21) (2.82)
Note: The dependent variable is the ratio of private investment to GDP, in percent. Figures in parentheses are estimated
f-statistics; R2 is the adjusted /?2-statistic; and SEE is the standard error of the estimate. One asterisk (*) indicates statistical
significance at the 10 percent level, two asterisks (**) indicate statistical significance at the 5 percent level, and three asterisks (***)
indicate significance at the 1 percent level. The coefficients of the country dummy variables, which are all statistically significant at
the 1 percent level, have been omitted from the table.

mated coefficient for the lagged value of GDP per capita (INCt-i) was
positive but insignificant.
Interestingly, the estimated coefficient for the real interest rate (RI)
was negative and statistically significant. The coefficient indicates that a
1 percentage point rise in the real interest rate would reduce the private
investment rate (which averaged about 12.5 percentage points) by less
than 0.1 percentage point. This overall result held even when observa-
tions in the sample were divided into cases with real interest rates above
and below a range of levels varying from minus 2 percent to plus
4 percent. This finding is more consistent with the neoclassical invest-
ment model than with the McKinnon-Shaw hypothesis, as it would sug-
gest that high real interest rates serve more to deter investment by
raising the user cost of capital than to promote investment by increasing
the volume of financial saving. Supporting this view is the recent finding
in Haque, Lahiri, and Montiel (1990) that the interest rate appears to be
negatively and highly significantly related to domestic investment rates
in a multi-equation macroeconomic model estimated over 31 developing
countries for much of the post-World War II period.11
The results of equations (2) and (3) in Table 4, which are estimated
over the 1975-81 and 1982-87 periods, respectively, suggest that the
findings reported in equation (1) mask rather different effects of certain
macroeconomic variables during the two subperiods. This is confirmed
by a simple F-test comparing the results in the two subperiods, which
yields an F-value of 4.4, compared to the critical value of F(30,234) =
1.86 to establish a significant difference in the estimated coefficients
between sample periods. In equation (2), for example, the estimated
coefficients for the public sector investment rate, inflation rate, and the
lagged debt-service ratio had the same sign and were larger in absolute
value than in equation (1). Indeed, the coefficient for the domestic
inflation rate in equation (2) was more than five times its absolute value
in the previous equation. This result would suggest that higher inflation
rates had decidedly more negative effects on private investment rates
during the 1975-81 period. In addition, the estimated coefficient on the
lagged value of GDP per capita (INCt-i) was considerably larger than in
equation (1) and became statistically significant at the 10 percent level.
By comparison, the estimated coefficient for the lagged ratio of external
debt to GDP, though still negative, was significant at only the 10 percent
level. This may reflect the generally low and stable ratios of debt to GDP
among developing countries during the pre-1982 years, as compared
11
A cross-country regression of private investment rates averaged over the
1975-87 period on a constant and each country's average real interest rate
yielded a positive but insignificant coefficient on the real interest rate.

with the ratios observed after 1981. It may also indicate high collinearity
between this variable and the lagged debt-service ratio.
In equation (3), which covers the 1982-87 period, most of the vari-
ables that were significant in equations (1) and (2) either had smaller
estimated coefficients or were no longer statistically significant. For
example, the coefficient for the real interest rate was barely one fifth of
its absolute value in equation (2) and was statistically insignificant. The
estimated coefficient for the domestic inflation rate was considerably
smaller in absolute value than in equation (1), although still significant at
the 1 percent level. The estimated coefficient for the public sector invest-
ment rate, though still positive and statistically significant, was smaller
than in the previous two equations, although the coefficient for the
lagged GDP growth rate was virtually the same as in equation (2). In
addition, the estimated coefficient for the lagged debt-service ratio was
considerably smaller in absolute value than in the other equations and
was no longer statistically significant. However, the estimated coefficient
for the lagged ratio of external debt to GDP increased in absolute value
and was significant at the 1 percent level. This suggests that the role of a
country's debt overhang became more important during the 1982-87
period, which may reflect the increasing gap during this period between
actual debt-service payments, as reflected in the debt-service ratio, and
contractual obligations. The estimated coefficient for lagged per capita
GDP, which was positive and marginally significant for the 1975-81
period, was negative and significant at the 1 percent level, contrary to
expectations. This may reflect the sharp decline in private investment
rates in higher-income Latin American countries after 1981.
In view of the significantly different results from equations (2) and (3),
it seemed useful to estimate a new equation for the entire sample period
that would allow the values of the coefficients for the various macroeco-
nomic variables to differ across time periods. This was done by creating
from each of the variables in the specification two variables, one contain-
ing only those observations for the 1975-81 period and the other holding
only those observations for the years 1982-87. The equation was then
estimated using this expanded set of explanatory variables, with the
country dummies left unchanged over the two subperiods.12 The results
of this equation are summarized in Table 5.
The figures reported in Table 5 confirm most of the results in equa-
tions (2) and (3) of Table 4. The estimated coefficients for the real
interest rate, lagged per capita growth rate, and domestic inflation rate
12
This essentially means assuming that the country-specific factors remained
invariant over time.

Table 5. Final Regressions for Private Investment Rates

Estimated Estimated Beta
Variable Coefficient ^-statistic Coefficient
RI
1975-81 -0.177*** -4.36 -0.249
1982-87 -0.061** -2.27 -0.112
GRt-1
1975-81 0.219*** 4.25 0.123
1982-87 0.218*** 4.57 0.131
IPUB/GDP
1975-81 0.091 1.35 0.086
1982-87 0.066* 1.74 0.073
CPI
1975-81 -0.008*** -4.78 -0.199
1982-87 -0.002*** -2.48 -0.120
INCt-1
1975-81 1.787*** 3.74 0.275
1982-87 0.625** 2.17 0.148
(DS/XGS)t-1 1
1975-81 -0.059*** -3.59 -0.212
1982-87 -0.026** -1.98 -0.101
(DEBT/GDP)t-l 1
1975-81 -0.034** -1.72 -0.111
1982-87 -0.037*** -2.74 -0.193
Note: The dependent variable is the ratio of private investment to GDP, in
percent. The number of observations is 294; the adjusted /?2-statistic is 0.83, and
the standard error of the estimate (SEE) is 2.24. One asterisk (*) indicates
statistical significance at the 10 percent level, two asterisks (**) indicate signifi-
cance at the 5 percent level, and three asterisks (***) indicate significance at the
1 percent level.
all had the same sign as before and were statistically significant at the 1
percent level or better. However, the coefficient for the real interest rate
during the 1975-81 period was now twice the size of the comparable
coefficient in equation (2), while that for the domestic inflation rate was
smaller than its equation (2) counterpart. The coefficient for the public
sector investment rate during the 1975-81 subperiod was now much
smaller than before and no longer significant, while the coefficient of this
variable for the 1982-87 subperiod was significant at only the 10 percent
level. The estimated coefficient for the lagged debt-service ratio was
again larger during the 1975-81 period, although significant during both
periods. By comparison, the coefficients for the lagged debt-to-GDP

ratio were uniformly smaller than in equations (2) and (3), although both
were significant at the 5 percent level or better. Interestingly, the esti-
mated coefficient for the lagged real GDP level was now positive and
quite significant for both the 1975-81 and 1982-87 periods, although it
was much larger during the earlier period. These results are more consis-
tent with the hypothesized relationship between income level and invest-
ment rate, while taking into account the falloff in private investment
rates in many higher-income Latin American countries after 1981. An-
other interesting finding reported in Table 5 is that for several variables
the differences in estimated coefficients between the two subperiods
were smaller than suggested from equations (2) and (3). For example,
the estimated coefficients for the lagged debt-to-GDP ratio in the two
subperiods became virtually the same, while the differences between the
coefficients for the lagged debt-service ratio and the domestic inflation
rate narrowed considerably. On the whole, these results suggest that
most of the macroeconomic variables affected private investment rates
in both the 1975-81 and 1982-87 subperiods. Except for the debt-to-
GDP ratio, however, the variables had more impact during the first
subperiod, before the debt crisis severely reduced investment rates in
most developing countries.
As a further indication of the relative importance of different variables
on private investment rates, Table 5 also reports the beta coefficients for
the macroeconomic variables in the final estimating equation. These
coefficients are unit-free and measure the relative impact of different
explanatory variables on the private investment rate. The beta coeffi-
cients indicate that changes in the real interest rate, lagged per capita
income level, and lagged debt-service ratio during the 1975-81 sub-
period had by far the largest relative impact on private investment rates,
about twice that of most other variables. Close behind these were the
effects of the domestic inflation rate during 1975-81 and the ratio of
debt to GDP during 1982-87. The next largest effect came from the
lagged per capita income level during 1982-87. The beta coefficients for
most other variables fell in a fairly narrow range, although those for the
public investment rate were distinctly smaller than the rest, implying less
impact on private investment rates than the other variables. Overall,
these results suggest that the impact on private investment rates of most
variables was larger during the 1975-81 subperiod than afterwards. The
notable exception was the beta coefficient for the lagged debt-to-GDP
ratio, which was substantially higher during the second subperiod. This
suggests that the debt overhang has become more important since the
onset of the debt crisis in 1982.

IV. Conclusions
The results of this study provide some support for the hypothesis that
private investment rates in developing countries are affected by impor-
tant macroeconomic variables. The econometric tests undertaken sup-
port the view that real interest and economic growth rates, the domestic
inflation rate, external debt burdens (either in the form of high debt-
service ratios or, following 1981, a high ratio of debt to GDP), and, to a
lesser extent, the public investment rate have all been significant deter-
minants of private investment rates in these countries during the post-
1974 period. Of these variables the domestic inflation rate and the exter-
nal debt burden appear to have had a negative impact on private
investment rates, while economic growth rates, the public investment
rate, and, particularly for 1975-81, the GDP per capita level have had a
positive effect. These results suggest that public sector investment has
been complementary to private investment in these countries. There is
also evidence that, in accordance with standard neoclassical theory, real
interest rates are negatively related to private investment rates.
There is evidence that several variables, in particular the real interest
rate, domestic inflation rate, per capita GDP level, and, to a lesser ex-
tent, the public investment rate, had a greater impact on private invest-
ment rates during 1975-81, before the onset of the debt crisis, than
afterwards. In addition, it appears that the way in which external debt
burdens reduced private investment changed between the 1975-81 and
1982-87 subperiods. During the former period, when most countries
remained current on their external debt-service payments, the debt-
service payments ratio was a more significant determinant of private
investment rates. During the second subperiod, when rescheduling and
external arrears became more common, the ratio of the stock of external
debt to GDP became equally if not more significant. On balance, these
results provide some support for the view that countries with higher
growth rates and income levels, more macroeconomic stability (in the
form of lower inflation rates), smaller debt burdens, and higher rates of
public investment have higher levels of private investment relative to
GDP. For the reasons mentioned earlier, however, these findings should
be considered suggestive, rather than providing strong evidence for the
various hypotheses discussed in the paper.
Because of the close links among saving, private investment, and
economic growth, it would seem useful to go beyond the partial equi-
librium framework of the present study and examine the interactions
among investment, saving, and growth in a general equilibrium model.

This could be done by applying a savings model (such as the one de-
scribed in Aghevli and others (1990)) and appropriate growth models
(see, for example, Otani and Villanueva (1990)) to develop a general
equilibrium framework in which separate equations for savings, private
investment, and growth are estimated simultaneously. Such a project
would greatly strengthen the current understanding of causal relation-
ships among these phenomena in developing countries. It might also
make it possible to develop more effective policy measures to strengthen
private saving and investment activity, and thereby raise the long-term
rate of economic growth.
APPENDIX
Data Sources for Variables

The data used in this study come primarily from three sources: Pfefferman and
Madarassy, "Trends in Private Investment in 30 Developing Countries" (1989);
International Monetary Fund, International Financial Statistics (IPS) data file;
and International Monetary Fund, World Economic Outlook (WEO). The first
of these sources provided the data on private and public investment rates for the
23 countries in the sample. Except for data on interest rates, all remaining data
are from IPS and WEO data files.
Data on interest rates were compiled from national sources. To focus on the
effect of real interest rates on private savings, interest rates on time deposits of 6-
to 24-month maturities at commercial banks were selected wherever possible,
with the specific maturity depending on the country. The data selected appear in
Table 6. Wherever possible, the maturities chosen were those in Hanson and
Neal (1986). Table 7 gives precise definitions of interest rates used.

Table 6. Nominal Interest Rates on Selected Time Deposits, 1975-87
Country 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987
Argentina 40.0 55.0 149.8 132.2 117.3 79.2 154.4 125.0 273.9 381.8 294.1 59.9 144.2
Bolivia 11.8 11.8 11.8 11.8 12.1 17.2 28.0 30.3 39.2 106.4 68.8 62.4 29.1
Brazil 26.5 38.0 44.4 47.8 48.8 60.3 101.4 112.5 157.5 242.3 249.8 75.5 401.0
Chile 277.3 202.1 93.9 57.4 45.1 37.4 40.8 47.8 27.9 26.1 31.6 19.0 25.2
Colombia 25.6 25.6 25.6 22.0 23.0 35.8 37.4 38.0 33.7 34.8 35.3 31.2 31.1
Costa Rica 21.5 21.5 21.5 21.5 21.5 21.5 21.5 21.8 23.1 22.0 20.0 18.0 17.5
Ecuador 0.0 0.0 0.0 0.0 0.0 9.0 9.0 13.0 16.0 22.0 22.0 24.8 32.0
Guatemala 9.0 9.0 9.0 9.0 9.0 9.0 10.0 12.3 9.0 9.0 9.0 10.3 11.0
India 8.0 8.0 6.0 6.0 7.0 7.5 8.0 8.0 8.0 8.0 8.5 8.5 9.0
Kenya 5.6 5.6 5.6 5.6 5.6 6.0 10.4 13.0 14.1 12.0 12.0 12.0 10.0
Korea, Rep. of 15.0 16.2 14.4 18.6 18.6 23.3 19.0 10.3 8.0 10.0 10.0 10.0 10.0
Mexico 10.0 9.6 12.4 13.0 13.8 21.2 30.9 56.5 53.4 44.4 70.7 95.7 96.0
Pakistan 8.1 8.5 9.1 9.5 9.5 9.5 10.9 10.6 9.5 10.9 9.8 8.8 8.9
Peru 7.0 9.0 14.0 28.8 31.5 31.5 52.0 55.0 57.5 60.0 32.8 41.9 29.5
Philippines 9.5 10.0 10.0 10.0 12.0 14.0 13.0 13.9 14.2 17.4 19.8 11.5 10.0
Singapore 6.2 5.5 5.3 5.7 6.9 9.2 10.7 7.9 6.7 7.2 5.5 4.1 3.5
Sri Lanka 7.5 7.5 10.0 15.0 15.0 20.0 20.0 15.0 15.0 14.0 12.0 8.5 8.5
Thailand 8.0 8.0 8.0 8.0 9.0 12.0 12.5 13.0 13.0 13.0 13.0 9.5 9.5
Tunisia 3.0 3.0 4.0 4.0 4.0 4.0 4.8 5.0 5.0 5.0 5.7 6.8 8.2
Turkey 9.0 9.0 9.0 11.3 17.4 26.6 49.2 50.0 42.5 45.0 50.2 52.0 45.7
Uruguay 21.0 44.0 48.9 53.1 43.0 51.5 46.9 50.6 66.1 71.1 84.1 61.2 62.8
Venezuela 6.0 6.0 6.0 6.0 6.0 11.3 14.9 14.8 13.9 12.5 10.5 8.9 8.9
Zimbabwe 5.0 4.8 4.6 4.3 4.3 4.4 12.0 10.5 14.2 10.5 10.5 10.3 10.2
Note: For definitions and sources of interest rate series, see text and Table 7 in the Appendix.

Table 7. Definition of Interest Rates

Country Interest Rate
Argentina 1975-76: maximum rates on annual savings deposits;
1977-87: interest rates on 30-day certificates of
deposit
Bolivia Minimum rate on peso-denominated, 1-year time
deposits
Brazil 1975-82: interest paid on bills of exchange at finance
companies; 1983-86: annualized interest rates on sav-
ings deposits; 1987: annual rates on time deposits
Chile 1975-87: annualized interest rates on 30-day time
deposits
Colombia Annualized rates on 90-day certificates of deposit
Costa Rica 1978-87 (data missing for 1975-77): "basic rate" on sav-
ings deposits at commercial banks
Ecuador 1975-85: interest rate on time deposits at commercial
banks and nonbank financial institutions; 1986-87:
rate on 90-day savings accounts3
Guatemala Maximum rate on savings deposits at commercial banks
India Lowest ceiling rate on 1- to 2-year deposits at commer-
cial banks
Kenya Interest rate on 9- to 12-month deposits at commercial
banks
Korea, Republic of Rate on commercial bank deposits of more than 1 year
Mexico Interest rate on 6-month deposits at commercial banks
Pakistan 1975-80: weighted average of interest rates on 6- to 12-
month deposits; 1981-87: interest rates paid on profit/
loss-sharing accounts
Peru 1975-86: interest rates on 6- to 12-month time deposits
at commercial banks with head offices in Lima; 1987:
rates on 91- to 180-day certificates of deposit at com-
mercial banks
Philippines Rates on 6- to 12-month deposits at commercial banks
Singapore Interest rate on 1-year deposits at commercial banks
Sri Lanka Actual or minimum interest rates on 1-year deposits at
commercial banks
Thailand Rates on 1- to 2-year deposits (1985-87: ceiling rate)
Tunisia 1975-81: maximum rate on 6- to 12-month deposits;
1982-87: rate on 3-6 month deposits
Turkey Interest rates on 12- to 24-month deposits
Uruguay Average interest rates on deposits of 6 months or more

Table 7 (concluded}.
Country Interest Rate
Venezuela Rates on time deposits of 6 to 12 months (1979: rate on
6-month deposits)
Zimbabwe Interest rate on deposits of 12 months or moreb
a
b
Ratesfor 1975-79 taken as the same as in 1980.
Data from Reserve Bank of Zimbabwe, Quarterly Economic and Statistical
Review (December 1988).
REFERENCES
Aghevli, Bijan B., James M. Boughton, Peter J. Montiel, Delano Villanueva,
and Geoffrey Woglom, The Role of National Saving in the World Economy:
Recent Trends and Prospects, Occasional Paper 67 (Washington: Interna-
tional Monetary Fund, March 1990).
Aschauer, David A., "Does Public Capital Crowd Out Private Capital?" Journal
of Monetary Economics, Vol. 24 (September 1989), pp. 171-88.
Bischoff, Charles W., "Hypothesis Testing and the Demand for Capital Goods,"
Review of Economics and Statistics, Vol. 51 (August 1969), pp. 354-68.
, "Business Investment in the 1970s: A Comparison of Models," Brook-
ings Papers on Economic Activity: 1 (Washington: The Brookings Institu-
tion, 1971), pp. 13-58.
Blejer, Mario I., and Mohsin S. Khan, "Government Policy and Private Invest-
ment in Developing Countries," Staff Papers, International Monetary
Fund, Vol. 31 (June 1984), pp. 379-403.
Borensztein, Eduardo, "Debt Overhang, Credit Rationing, and Investment,"
IMF Working Paper WP/89/74 (Washington: International Monetary Fund,
September 1989).
Clark, Peter K., "Investment in the 1970s: Theory, Performance, and Predic-
tion," Brookings Papers on Economic Activity: 1 (Washington: The Brook-
ings Institution, 1979), pp. 73-113.
Dornbusch, Rudiger, and Alejandro Reynoso, "Financial Factors in Economic
Development," American Economic Review, Papers and Proceedings, Vol.
79 (May 1989), pp. 204-209.
Froot, Kenneth, and Paul Krugman, "Market-Based Debt Reduction for Devel-
oping Countries: Principles and Prospects," (unpublished; Cambridge,
Massachusetts: National Bureau of Economic Research, January 1990).
Hanson, James A., and Craig R. Neal, "The Demand for Liquid Financial
Assets: Evidence from 36 Developing Countries," (unpublished; Washing-
ton: World Bank, August 1986).
Haque, Nadeem U., Kajal Lahiri, and Peter J. Montiel, "A Macroeconometric

Model for Developing Countries," Staff Papers, International Monetary

Fund, Vol. 37 (September 1990), pp. 537-59.
Hines, A.G., and G. Catephoros, "Investment in U.K. Manufacturing Industry
1956-67," in The Econometric Study of the United Kingdom: Proceedings,
ed. by Kenneth Hilton and David E. Heathfield (London: Macmillan,
1970), pp. 203-24.
International Monetary Fund, World Economic Outlook (April 1988).
, World Economic Outlook (October 1989).
Jorgenson, Dale W., "The Theory of Investment Behavior," in Determinants of
Investment Behavior, ed. by Robert Ferber (New York: National Bureau of
Economic Research, 1967).
, "Econometric Studies of Investment Behavior: A Survey," Journal of
Economic Literature, Vol. 9 (December 1971), pp. 1111-47.
Khan, Mohsin S., and Carmen M. Reinhart, "Private Investment and Economic
Growth in Developing Countries," World Development, Vol. 18 (January
1990), pp. 19-27.
McKinnon, Ronald I., Money and Capital in Economic Development (Washing-
ton: The Brookings Institution, 1973).
Mirakhor, Abbas, and Peter Montiel, "Import Intensity of Output Growth in
Developing Countries, 1970-85," Staff Studies for the World Economic
Outlook, World Economic and Financial Surveys (Washington: Interna-
tional Monetary Fund, August 1987).
Otani, Ichiro, and Delano Villanueva, "Long-Term Growth in Developing
Countries and its Determinants: An Empirical Analysis," World Develop-
ment, Vol. 18 (June 1990), pp. 769-83.
Pfefferman, Guy, and Andrea Madarassy, "Trends in Private Investment in 30
Developing Countries," IFC Economics Department Discussion Paper,
World Bank Group, July 1989.
Serven, Luis and Andres Solimano, "Private Investment and Macroeconomic
Adjustment: An Overview," PPR Working Paper WP 339 (Washington:
World Bank, December 1989).
Shaw, Edward S., Financial Deepening in Economic Development (New York:
Oxford University Press, 1973).
Stern, Nicholas, "The Economics of Development: A Survey," Economic Jour-
nal, Vol. 99 (September 1989), pp. 597-685.
Sundararajan, V., and Subhash Thakur, "Public Investment, Crowding Out, and
Growth: A Dynamic Model Applied to India and Korea," Staff Papers,
International Monetary Fund, Vol. 27 (December 1980), pp. 814-55.
Tun Wai, U, and Chorng-Huey Wong, "Determinants of Private Investment in
Developing Countries," Journal of Development Studies, Vol. 19 (October
1982), pp. 19-36.

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
Wages, Profitability, and Growth

in a Small Open Economy
BANKIM CHADHA*
Issues raised by the evolution of a rapidly growing small economy from a

labor-intensive, low-technology production base to a capital-intensive,
high-technology, knowledge-and-skill-intensive emphasis as it approaches
the limits of its resource constraints in the labor market are examined. A
model of endogenous growth for a small open economy that is driven by
increases in labor productivity from learning-by-doing and that allows for
the dynamic acquisition of comparative advantage is developed. In this
framework the effects of policies and exogenous shocks on the direction
and pace of restructuring are investigated. [JEL 110, 121, 410]
NAN ATTEMPT to obtain more satisfactory explanations for sustained

N differences in growth experiences, both over time and across coun-
tries, much attention has been focused recently on endogenous sources
or engines of growth. It has been increasingly recognized that models of
endogenous growth hold the potential for explaining aspects of growth
that the standard neoclassical aggregate growth model of Solow (1956),
and its many variants are either directly at odds or ill-equipped to deal
with.1
An important prediction of the neoclassical model is that initial condi-
tions or disturbances have no long-run implications for growth. Regard-
less of an economy's initial per capita endowment of capital, it will
* Bankim Chadha is an Economist in the External Adjustment Division of the

Research Department. He holds a doctorate from Columbia University. The
author thanks Fabrizio Coricelli, Robert Corker, Michael Dooley, Mohsin
Khan, Paul Masson, Susan Schadler, Ichiro Otani, and Ranjit Teja for useful
comments.
1
For a discussion of the failure of the neoclassical growth model to explain
various stylized facts of growth, and its limitations in general, see, among others,
Romer (1986,1987a, 1987b), Lucas (1988), Helpman (1988), and Shleifer (1989).
59

60 BANKIM CHADHA
converge to the same steady state per capita capital stock, after which
growth in per capita output is determined purely by exogenous techno-
logical progress. There is, therefore, no room for an analysis of alterna-
tive phases or stages of growth induced by particular initial conditions.
Moreover, the role for government intervention in the neoclassical
model is limited. The model predicts, for example, that policies that
succeed in increasing savings rates will increase the long-run level of per
capita output and consumption, but will have no sustained impact on the
growth rate of per capita output.
This paper analyzes a particular episode in the growth experience of
Singapore—that of economic restructuring, which encompassed changes
in both the technique of production and in the composition of output.2
Singapore is a small, highly open economy whose development strategy
has passed through several stages. A remarkably successful low-wage,
export-led period of industrialization from 1966-79 transformed it from
a labor-surplus economy to one where domestic labor constituted an
important constraint on growth.3 By the late 1970s, a development strat-
egy based on labor-intensive and relatively low-wage, export-driven
growth was proving to be unsustainable for several reasons. First, full
employment of the labor force was generating upward pressure on
wages. With limited prospects for growth in the domestic supply of
labor, a continuation of the strategy would have required a growing
reliance on imported foreign labor. This was ruled out as a socially and
politically viable option. Second, the Government realized that wage
restraint had resulted in excessive investment in, and retention of, labor-
intensive activities. The maintenance of low wages, it was felt, had
hindered the natural process of economic upgrading and restructuring
with technological progress, because it encouraged investment in rela-
tively labor-intensive, low-technology goods. Third, the small wage in-
creases in Singapore had been accompanied by lower growth of
labor productivity than in other newly industrializing economies
(NIEs), where wage increases had been more rapid. It was felt that
preventing wages from rising to higher levels had stunted productivity
2
For a review of Singapore's growth experience, see Lim (1984,1988).
3
With rapid economic growth, which averaged 9.4 percent a year from
1970-79, a large pool of unemployed labor was gradually absorbed into the work
force; between 1970 and 1979 the unemployment rate fell from 6.0 percent to 3.4
percent—close to the natural rate of unemployment, which is estimated at 3
percent. During this period real wage increases typically lagged behind growth in
labor productivity. It is estimated, for example, that while real product wages in
manufacturing increased at an average rate of 2.3 percent a year from 1975-79,
labor productivity grew by 4.4 percent a year. See Singapore, Yearbook of
Statistics (various issues).

WAGES, PROFITABILITY, AND GROWTH 61
growth by reducing incentives for labor-saving investment and organi-

zational rationalization.
In 1979 the Government adopted an economic restructuring strategy,
designed to shift the structure of production from low-technology, labor-
intensive activities to technology-and-skill-intensive, higher value-added
activities. This shift was intended to place the economy on a sustainable
growth path by allowing it to economize on the use of labor and to enter
markets for high-technology and knowledge-and-skill-intensive goods
and services. It was perceived that if this restructuring were not brought
about, the labor constraint and consequent wage pressure would eventu-
ally lead to a deterioration in competitiveness and a declining share in
traditional export markets. Moreover, the move toward high-technology
goods would increase labor productivity and thus output.
Three complementary policy initiatives were involved in the restruc-
turing strategy. First, significant wage increases were encouraged by the
National Wage Council (NWC) to compensate for the previous wage
restraint, which may have held wages at an artificially low level.4 This
policy of allowing wages and, more generally, labor costs to rise in excess
of productivity growth has been referred to as a "high-wage" or "wage-
correction" policy.5 It was envisaged that, with an increase in the relative
price of labor, capital would be substituted for labor, and consequently,
low-wage, labor-intensive activities would be phased out. Concomi-
tantly, to reinforce this shift, labor supply measures were adopted to
limit the inflow of foreign workers into lower-paid unskilled occupations,
and eventually to phase foreign workers out. The measures recom-
mended by the NWC raised labor costs for all employers, but the form of
the recommendations, which comprised a uniform lump-sum increase
and a percentage increase, also affected wage relativities by increasing
the lowest wages proportionately more than higher wages.6 Thus, firms
4
The NWC, a tripartite body comprising representatives from the Govern-
ment, employers, and trade unions, was formed in 1972 with the mandate of
recommending specific quantitative wage guidelines each year. The tripartite
nature of the council ensured wide acceptance and implementation of its recom-
mendations.
5
In June 1979 the NWC recommended a general wage increase amounting to
approximately 14 percent of average monthly wages. In accepting the NWC's
recommendations, the Government made it known that similar wage increases
were planned for the following two years. The measures to increase wages
coincided with systematic increases in employer contribution rates to the Central
Provident Fund from 16.5 percent of the wage bill in 1978 to 25 percent by 1984.
For an analysis of the short-term effects of the wage-correction policy, see Otani
and Sassanpour (1988).
6
The lump-sum dollar increase in the NWC recommendations for 1979/80 and
1980/81, for example, constituted approximately half of the total 14 percent
recommended increase in wages.

62 BANKIM CHADHA
engaged in low-skill, labor-intensive activities experienced the largest

increase in labor costs.
Second, restructuring was perceived to be constrained by an inade-
quately skilled labor force, and the Government therefore embarked on
an ambitious and successful program for increasing labor skills. In addi-
tion to various adult education and worker training programs, the Skills
Development Fund (SDF) was formed in 1979. Financed by a tax on
employers, the SDF was designed to provide incentives to employers to
upgrade the skills of their employees and to increase on-the-job train-
ing.7 Third, incentives were offered by the Economic Development
Board, to encourage investment in technology-and-skill-intensive activi-
ties, and to promote automation of existing production facilities.
While some restructuring of output toward knowledge-and-skill-inten-
sive and high-technology activities had already taken place in the 1970s,
the pace of restructuring accelerated considerably after the adoption of
the wage-correction policy in 1979 (Figure 1). The share of financial and
business services, representing relatively skill-intensive activities, which
was virtually unchanged during 1971-79, rose from 18 percent in 1979 to
almost 29 percent by 1987. The share of electronics, a relatively high-
technology activity, also increased sharply. Accompanying the restruc-
turing of output was a shift in the composition of the labor force toward
more highly skilled and higher-paid activities.8
In assessing the economic restructuring policies, one needs to address
several key questions. First, if the policies had not been undertaken,
would restructuring have come about naturally? Second, while restruc-
turing was expected to raise the level of productivity and, hence, the level
of output, is there any reason to believe that it might have also resulted
in a permanent increase in the potential rate of growth of the economy?
Third, what effects did changes in wage differentials and in the rate of
return to capital have on the direction and pace of restructuring?
For a labor-constrained and otherwise naturally resource-poor econ-
omy such as Singapore at the end of the 1970s, productivity increases
must eventually become the primary source of sustained growth in the
supply of output. Productivity of the raw labor force can be increased
7
At the time of the introduction of the SDF in 1979, the employer's contribu-
tion rate was set at 2 percent of the wage bill; it was subsequently raised to 4
percent
8
in 1980.
From 1979-87 the share of professional, managerial, and administrative
workers in the labor force, representing the highest skill and pay occupational
category, rose significantly from 11 percent to 17 percent; the share of produc-
tion, transport, and other manual workers, representing the lowest skill and pay
occupational category, fell from 39 percent to 35 percent. See Singapore, Year-
book of Statistics (various issues).

Figure 1. Structure of Output
Source: Singapore, Yearbook of Statistics, 1987 and 1988.

Note: Share of GDP at current market prices.

64 BANKIM CHADHA
directly by investment in human capital, as emphasized recently by

Otani and Villanueva (1989); or it can result from a process of "learning-
by-doing." The potential importance of learning-by-doing in endogen-
ously determining productivity growth was first advanced by Arrow
(1982). Subsequently, Bardhan (1970) and Krugman (1987) have empha-
sized its role in dynamically determining comparative advantage, and it
is identified by Lucas (1988) as a potential, endogenously determined,
source of growth. In the following, the questions raised above on Singa-
pore's economic restructuring program will be addressed in a simple
model incorporating learning-by-doing.
Section I extends Lucas's (1988) model of endogenous growth to
examine the process of restructuring that took place in Singapore in the
1980s. The model shows that in the absence of exogenous shocks or
government intervention, an economy will, over time, tend to one of two
long-run equilibrium growth paths—a "high-growth" path and a "low-
growth" path—depending on the initial endowment of resources in each
sector and, hence, historical comparative advantage. Thus, any govern-
ment intervention that results in a diversion of resources from one sector
to another can affect the pattern of growth and trade over time, indeed
even to the extent of putting the economy on a high-growth or low-
growth trajectory. The analysis presents an example of a change in the
level of aggregate output, resulting from a shift in its composition, induc-
ing a sustained change on the potential growth rate of the economy.
Effects of stylized versions of actual policies followed in Singapore and
exogenous shocks are analyzed. Section II offers concluding remarks.
I. A Model of Endogenous Growth and Restructuring
The model developed here builds on Lucas (1988). Consider a small

open economy producing two (baskets of) traded goods, the outputs of
which are denoted by Ql and Q2, and the prices of which are determined
in the rest of the world. Initially, for simplicity, abstract from the pres-
ence of physical capital.9 The population, or labor supply, is assumed to
be constant. The two goods are produced by a technology of the Cobb-
Douglas type, with diminishing returns to labor
9
The implications of the presence of physical capital and the role of invest-
ment are discussed below in an extended version of the model.

Here, hi is the skill level or human capital specialized in the production

of good i, where i = 1, 2; N denotes the size of the labor force, and /; is
the share of the labor force employed in the production of good i.
The effect of the skill level or human capital, h\, is assumed to be
entirely external to any of the large numbers of perfectly competitive
firms in each industry. This positive externality cannot be captured by
any single firm; that is to say, the production of each firm depends on the
average skill level in that industry. It is assumed that skills are acquired,
according to
and a dot over a variable denotes its derivative with respect to time. The
growth of the skill level should be interpreted as occurring due to learn-
ing and can be interpreted as learning-by-doing. The rate of growth of
skills in equation (2) is a positive function of both the speed of learning,
represented by 8/, and the effort or resources devoted to producing good
i, which is assumed to be related to the proportion of the labor force
employed in the production of good i. Good 1 will be referred to as the
high-technology good and good 2 as the low-technology good. It is
posited further that the speed of learning is greater in the high-technol-
ogy sector than in the low-technology sector, so that 81 > 82.
Equation (2) implies that the economy's production possibility fron-
tier shifts out over time, with experience gained by the labor force
resulting in an increase in the skill level and, hence, productivity of the
labor force. Note that the form of these learning equations implies
constant returns to experience. This seems counterintuitive, in that one
would expect learning-by-doing or the acquiring of skill in any particular
activity to occur rapidly at first, then more slowly, and then not at all.
The constant returns to learning in (2) should be interpreted as repre-
senting an environment in which innovations are constantly occurring
and being adopted, so that learning is interpreted not only as permitting
things to be done better, but also results in better things being done.
Viewing learning as encompassing the adoption of innovations provides
a justification for the assumption of a higher speed of learning in the
high-technology sector, since innovations are likely to occur at a rela-
tively more rapid rate in the sector.
It is assumed for simplicity that labor is perfectly homogeneous and
mobile across sectors, and that wages are perfectly flexible. Full employ-
ment therefore implies

Figure 2. Dynamic Path of Economy

For our purposes it is convenient to allow for a subsidy on wages in sector

1, granted at the rate TI, and a tax on wages in sector 2, levied at the rate
T2. Then combining firms' first-order conditions for profit maximization,
and substituting in the full-employment condition (3), yields
which can then be used to solve for the share of labor in sector 1 as a
function of the relative skill level in sector 1 and the exogenous variables
of the system, so that
In equation (4b), Ht denotes the ratio of skill levels or human capital;

that is, Ht = h}lhJ\ P equals P2IP\ and denotes the internationally given
relative price of good 2 in terms of the numeraire good 1. Signs under-
neath arguments in the 1} function indicate signs of the partial deriva-
tives, and are straightforward to derive from (4a). The learning or skill
accumulation equations can be combined to obtain a relative learning
equation and, with substitution of the full-employment condition, yield
Figure 2 graphs the possible dynamic paths of the economy repre-

sented by equations (4b) and (5). The growth of skills in each activity is
determined by both the speed of learning and the effort or resources—
that is, proportion of the labor force—devoted to each activity. Since the
speeds of learning are posited to be different in the two sectors, there
will exist distributions of the labor force between the two sectors such
that the ratio of skill levels remains exactly constant over time: where,
for example, the effect on the growth of the relative skill level of a
smaller share of labor devoted to producing the high-technology good is
offset exactly by the higher speed of learning in that activity. In Figure 2
the Ht = 0 line denotes such a locus of points where relative skill levels
remain constant. The arrows indicate the direction of movement of the
relative skill level, //„ when the economy is off the Ht = 0 line. The curve
OL represents the general equilibrium employment share in sector 1, 1],
as given by equation (4b). Under the assumptions here the economy is
always on this curve, and the double arrows on the curve thus indicate
the actual path of the economy at any relative skill level Ht. The two
curves intersect at a critical relative skill level, Hct, in the production of
the high-technology good. Given an initial ratio of human capital in the

68 BANKIM CHADHA
two sectors, H0, and in the absence of exogenous shocks to the system,
the future path of production is completely determined. Unless the
economy is initially endowed with a relative human capital ratio of
exactly Hct the economy will, over time, naturally traverse either up the
OL ray, eventually specializing in the production of the high-technology
good 1, or it will move in the opposite direction, specializing eventually
in good 2. If the economy specializes in good 1, the growth rate of
aggregate output is, from equations (1) and (2), 81, which is, by assump-
tion, greater than 82, the growth rate if the economy specializes in good
2. Thus, movements in Figure 2 toward L can be identified as converging
to a high-growth path, and movements toward O as converging to a
low-growth path. It can be shown that at any point in time along a
transition path—that is, at any point on the OL curve—the rate of
growth of output increases monotonically from 82 at O to 8X at L.
The production possibility frontier (PPF) shifts out, over time, pro-
portionately in favor of the good that the country has an initial compara-
tive advantage in producing—that is, the good it produces relatively
more of—since its skill level will grow relatively more in the activity to
which it devotes larger resources. Thus, at unchanged relative prices,
except by pure chance, in the case where resources are initially appor-
tioned in the production of each good such that relative skill levels
remain constant, the economy will end up specializing in the production
of one of the two goods over time. Initial comparative advantage is thus
magnified over time. Since the speed of learning is posited to be greater
in the high-technology good, eventual specialization in that good implies
a higher steady-state growth rate.
The analysis presents a manifestation of popular notions of bottle-
necks to growth.10 There are no forces in the system just discussed that
would necessarily place the economy on a path converging to the high-
growth path. Whether the economy ends up on the high-growth path or
the low-growth path will depend upon its initial relative skill levels and
internationally given relative prices. Since the learning effects are as-
sumed to be external, agents do not take them into account, and the
high-technology good is thus underproduced, and production and trade
are determined by temporal or historical comparative advantage. The
market, left to itself, will not necessarily pick the high-growth path,
except by historical accident. There is thus a clear role for government
intervention. In particular, allocating labor toward the high-technology
good, relative to the free market solution, would result over time in the
10
Formally, there is a bottleneck to "high" growth, rather than growth.

acquisition of comparative advantage in the production of the high-tech-

nology good and a higher growth rate.
An alternative reason for the existence of differential potentials for
productivity increases that are external to any single firm is different
rates of technology transfer from the rest of the world to the two sec-
tors.11 Imagine, for simplicity, that "available" technological progress
occurs globally at a faster rate in the high-technology sector. Now sup-
pose that the (relative) speed of adoption of these innovations—that is,
the actual transfer of the technology—is a function of the (relative)
resources devoted to the production of each good. The analysis of differ-
ent rates of technology transfer from abroad is then equivalent to that
described above for endogenous productivity increases.
The role of skills upgrading is transparent in the framework of Fig-
ure 2. An exogenous increase in the relative skill level in the production
of the high-technology good would move the economy to the right along
the curve OL. Thus, a relatively skill-scarce economy can, by increasing
its skill level, cross the threshold value of Hct, and put itself on a self-
sustaining path of restructuring toward the high-technology good, lead-
ing eventually to complete specialization in the good, and a higher
steady-state growth rate.
Role of Wage Differentials in Economic Restructuring
A simple and convenient way to analyze the effect of an exogenous

decline in the relative wages paid in the high-technology sector is to
examine the effect of a tax on wages in the labor-intensive, low-technol-
ogy sector, or the effect of a subsidy on wages in the high-technology
sector.12 Consider the effect of an increase in the tax rate on wages in
sector 2, T2. Recalling equation (4b), an increase in T2 results in an upward
shift of the OL curve, as depicted in Figure 3, to OL', so that, as would
be expected, an increasing proportion of the labor force is employed in
the high-technology sector at any skill level. It now intersects the Ht = 0
line at a lower relative skill level, Hct. This implies that if the economy
initially had a relative skill level between Hct and Hct, it would have been
on a path converging to the low-growth path; after the shift in the OL
11
This is particularly relevant in the case of Singapore.
12
Although such a differential tax or subsidy was not actually implemented, it
is analytically equivalent to changing wages in the low-wage sector relative to
the high-wage sector from employers' point of view, and considerably more
tractable.

Figure 3. Effect of an Increase in the Tax Rate on Wages in the Low-Technology Sector 2

curve, skill levels between Hf and Hct become points converging to the
high-growth path. The mechanism by which this occurs is straightfor-
ward. A decline in the relative wage paid by producers of good 1 shifts
employment in favor of good 1, increasing the proportion of the labor
force devoted to producing the high-technology good. Increasing the
proportion of resources devoted to producing the high-technology good
increases the scope for learning or acquiring comparative advantage in
that activity. The locus of initial conditions converging to eventual
specialization in the high-technology good thus increases.
Effect of a Change in the Terms of Trade

Consider the effect of an exogenous increase in the internationally
given relative price of the high-technology good. This corresponds to a
decline in P and, given the sign of the partial derivative in equation (4b),
results in an upward shift of the OL curve in Figure 3 to OL', exactly as
in the previous exercise. An increase in the relative price of good 1 shifts
production in favor of good 1, increasing the proportion of the labor
force employed in sector 1, and hence expands the set of initial condi-
tions converging to eventual specialization in the high-technology good.
It follows directly that any commercial policy—for example, a tariff—
that shifts domestic producer prices in favor of the high-technology good
would have the same effect.
Role of Foreign Labor

Consider the effect of an increase in the labor force, N. Note that,
from equation (4b), 1} is a negative function of (N*~°). The actual direc-
tion of movement of the OL curve therefore depends on whether P < a.
If, as we have assumed here, the low-technology good 2 is relatively labor
intensive, then p is greater than a, and hence an increase in the supply of
labor will cause the OL curve in Figure 3 to shift down.13'14 This effect of
an increase in the labor force is a standard prediction of the familiar
Rybczynski theorem: an increase in the supply of labor increases by
13
This shift is not shown in Figure 3, since the effects are exactly opposite to
those depicted for the previous exercise.
14
To talk meaningfully about relative labor intensities, physical capital needs
to be included explicitly in the production functions. It is implicitly assumed that
the capital stock in each sector is fixed in this subsection and is therefore sup-
pressed from the notation. The next subsection endogenizes physical capital
accumulation.

72 BANKIM CHADHA
relatively more the output of the relatively labor-intensive good. Output

and employment accordingly shift in favor of the low-technology labor-
intensive good, thus contracting the set of initial conditions that converge
to eventual specialization in the capital-intensive, high-technology good.
Role of Investment
The rate of investment in physical capital plays an important role in
the restructuring of output.15 An increase in the stock of physical capital
would increase by relatively more the output of the capital-intensive
sector. Thus, in the above setting, where the capital-intensive good is the
high-technology good, an increase in the stock of physical capital will
shift the composition of output and employment in favor of the high-
technology good. Hence, an increase in the rate of investment can be
viewed as an alternative engine for the restructuring of output, while
additionally providing an independent source of economic growth.
In introducing physical capital accumulation into the model above,
some simplifying assumptions are made to keep the analysis tractable.
The case of interest is one where the high-technology sector is relatively
capital intensive, and productivity increases at a faster rate because the
speed of learning and, hence, the rate of skill accumulation is greater. A
simple and tractable way to maintain these assumptions is to take the
extreme case: capital is employed only in the high-technology sector;
and whereas, as before, in the high-technology sector skills are accumu-
lated by learning, the skill level in the low-technology good is constant
and normalized to equal unity. Production functions can then be written
as
where Kt represents the stock of physical capital. Note the slight change
in notation: the skill level in the high-technology sector is now denoted
by Ht. This is to highlight the fact that while formally Ht now represents
only the absolute skill level in sector 1, it can still be interpreted as the
relative skill level in that sector. As before, it is assumed that the rate of
15
The effect of changes in the composition of investment is, of course, trans-
parent. An increase in the share of total investment going into the high- technol-
ogy sector will, by altering relative capital stocks, shift production and employ-
ment in favor of the high-technology good.

growth of the skill level in sector 1, //„ increases with the proportion of
the labor force, 1], devoted to producing good 1. Specifically
It is assumed along with Kouri (1979) that for a small open economy,
net investment at home is an increasing function of the discrepancy
between the actual rate of return to capital, r,, adjusted for any taxes
levied at the rate <|>, and the exogenously given rate of return, r*, in the
international capital market:16
Assuming a competitive market for capital, the rate of return (rental)

to capital at home is given at any point in time by the marginal product of
capital in sector 1:
As in the previous subsection, first-order conditions for profit maximiza-

tion can be combined with the full-employment condition to yield
so that
Substituting (lOb) into (7), combining (8) and (9), and again substituting
in (lOb), yields a pair of dynamic equations in the skills level and the
stock of physical capital:
16
Kouri (1979) argues that such an investment function can be derived if
expectations are static and there are adjustment costs to investment. The tradi-
tional neoclassical model of investment, as developed by Jorgenson (1963),
posits the flow of investment to be a function of the difference between an
optimal capital stock, determined by maximizing the present value of the firm,
and the current level of the capital stock. The formulation here is adopted to
emphasize that in a highly open economy, investment at home is a function of
relative rates of return.

Figure 4. Dynamic Path of Economy with Endogenous Capital Accumulation

It follows that
and it can be established that
To establish (13b), differentiate (12) with respect to the capital stock,

and rearrange, so that
and, therefore,
Sign Sign
Now, differentiating (lOa) and rearranging
so that dkldK < 0. Given (13a) and (13b)
so that a phase diagram describing the possible dynamic paths of the

economy can be drawn, as in Figure 4.
Note that both the skill level, Ht, and the stock of capital, Kt, are
predetermined variables given by history at any point in time. As the
arrows indicate, there exists a locus of initial H, K combinations, labeled
CC in Figure 4, which places the economy on a path converging to a
steady-state combination of H,K.17 The curve CC corresponds in this
setting to the point Hct, in the previous subsection. Again, the initial
combination of skill level and capital stock determines the entire future
17
Formally, the Jacobian matrix, with elements defined in equations (13a) and
(13b), has a negative determinant, so that the system has one stable and one
unstable root and is therefore characterized by saddle-path stability; CC denotes
the saddle path in Figure 4.

76 BANKIM CHADHA
path of production, and there are three possibilities: if the initial //, K
endowment places the economy above CC, then the economy will, over
time, tend toward specialization in the capital-intensive, high-technology
good; if below CC, the economy will tend toward specialization in the
labor-intensive, low-technology good; if exactly on CC, the economy will
remain diversified, converging over time to H, K. Note that CC is down-
ward-sloping in Figure 4. This implies that a high stock of capital can
offset a low-skill level as a bottleneck, and vice versa, in attaining a
self-sustaining path to restructuring toward the capital-intensive, high-
technology good.
This expanded framework allows consideration of the two main com-
ponents of the wage-correction policy: a change in relative wages and an
increase in the overall level of wages. Consider first the relative wage
effect. As before, consider the effect of an increase in the rate of tax, T2,
on employment in the labor-intensive, low-technology sector 2. Given
the signs of the partial derivatives in equation (lOb), it follows that both
the Ht = 0 curve and the Kt = 0 curve shift down in the H, K plane, as
shown in Figure 5 where the original curves are depicted by dashed lines.
This implies that the CC curve, which separates initial conditions con-
verging to eventual specialization in the high-technology good from
those converging to specialization in the low-technology good, also shifts
down from CC1 to CC2. The shaded area in Figure 5 denotes the set of
points that originally would have resulted in the economy continuously
expanding its share of labor-intensive, low-technology goods, but now
become points leading to eventual specialization in the high-technology
good.
Consider now the effect of an increase in average real wages in excess
of productivity growth. Since such an increase would reduce the rate of
return to capital, a convenient way to model it is to examine the effect of
an increase in the tax, <(>, on the return to capital.18 In this case there is no
effect on the Ht = 0 curve, since c|> does not enter equation (11). From
equation (12), however, the £, = 0 curve shifts up or to the left in the
H, K plane as shown in Figure 6. Thus at each skill level, the rate of
investment is zero at a smaller level of the capital stock. Each point on
the original Kt = 0 curve represents a pair of skill and physical capital
levels, such that the tax-adjusted rate of return to capital was just equal
to that in the rest of the world. An increase in the tax rate thus requires,
18
Sufficient conditions for an exogenous increase in wage or labor costs to
reduce the rate of return to capital are that the labor supply curve be positively
sloped and that the capital stock be predetermined at a point in time. In the
tradition of two-sector trade models, it has been assumed, for simplicity, that
labor is inelastically supplied.

Figure 5. Effect of an Increase in the Tax Rate on Wages in the Low-Technology Sector
with Endogenous Capital Accumulation

Figure 6. Effects of an Increase in the Tax Rate on the Return to Capital

for any skill level, a lower level of capital, so that the marginal product of
capital rises sufficiently to equate the domestic tax-adjusted rate of
return to that given exogenously in the rest of the world. The upward
shift of the Kt = 0 curve implies a consequent upward shift of the CC
curve in Figure 6 from CC1 to CC2. In this case the shaded area denotes a
set of points that originally would have resulted in production converging
to specialization in the high-technology good but, as a result of the
decline in the rate of return to capital, become points converging to
specialization in the labor-intensive, low-technology good.
The questions raised at the end of the last section on restructuring can
now be answered. First, in the framework developed above, government
intervention can play a pivotal role in providing an initial impetus for
restructuring, which market forces may not naturally generate due to the
presence of external learning effects. A second question was whether
restructuring would simply raise the level of output or could influence
long-run growth potential. The framework developed here clearly sug-
gests scope for the latter if the high-technology and knowledge-and-skill-
intensive sectors have the inherent potential for higher productivity
growth. Finally, on the third issue—the effect of the wage-correction
policy on economic restructuring—the analysis suggests that there were
probably two opposing influences. The decline of relative wages in the
higher-paid occupations, on the one hand, tended to shift employment in
favor of these occupations, thus moving the economy onto, or further
along, a path to self-sustaining restructuring toward capital-intensive,
high-technology and knowledge-and-skill-intensive activities. The rapid
increase in real wages across the board, on the other hand, lowered the
rate of return to capital and thus lowered investment, which tended to
move the economy away from such a path.19
II. Conclusion
This paper has examined a particular phase in the growth experience

of Singapore—that of economic restructuring. To examine the process of
restructuring in general, a model of endogenous growth and restructur-
ing for a small open economy incorporating learning-by-doing was devel-
oped. A broad conclusion from this framework is that impediments to
19
If, however, investment at home is unresponsive to differences between
domestic and international rates of return—for political risk reasons, for exam-
ple—but is responsive to differences in rates of return across sectors, then an
increasing proportion of the investment that does take place is likely to go into
the labor-saving, capital-intensive sector, spurring restructuring.

80 BANKIM CHADHA
restructuring may exist because of the potential importance of external

learning effects, the benefits of which redound not to individual firms
but to a sector as a whole. Such externalities present bottlenecks to
market-driven restructuring and create a role for government interven-
tion. In particular, it was shown that a diversion of resources, even
temporary, can induce an acquisition of comparative advantage and
hence permanently affect the pattern of trade and growth. The frame-
work thus suggests that if high-technology and knowledge-and-skill-
intensive sectors inherently have the potential for higher productivity
growth, economic restructuring would not simply raise the level of output
but could permanently raise the rate of long-run growth.
The framework was employed to analyze the effects of, among other
things, the wage-correction policy on the direction and pace of economic
restructuring in Singapore. The results indicate that there were probably
two opposing influences. The decline of relative wages in the higher-paid
occupations, on the one hand, tended to shift employment in favor of
these occupations, thus moving the economy onto, or further along, a
path of self-sustaining restructuring toward higher-technology and
knowledge-and-skill-intensive activities. The absolute increase in aver-
age real labor costs across the board in excess of productivity growth, on
the other hand, lowered the rate of return to capital and thus lowered
investment, tending to move the economy away from such a path. How-
ever, note that in Singapore restructuring did not hinge on the net effect
of these two opposing forces alone, since it was boosted in addition by
the upgrading of skills and the promotion of investment in relatively
labor-saving and knowledge-and-skill-intensive activities. Indications
are that initial bottlenecks to restructuring in Singapore have been over-
come—a basic level of skills and trained manpower now exists, and a
critical mass of high-technology and knowledge-and-skill-intensive activ-
ities has been established—and comparative advantage in these activities
is likely to grow naturally.
It is worth emphasizing that the framework developed here for exam-
ining alternative growth strategies was for a small and highly open labor-
constrained economy. The policy conclusions reached, therefore, may
not be applicable for all countries, and it should be noted that a shift
toward production of high-technology goods is not sustainable for the
world as a whole. Two crucial assumptions were made. First, the analysis
was carried out on the assumption of a fully employed and constant labor
force. This seems accurate for Singapore at the end of the 1970s. For a
labor-surplus economy, however, a shift toward labor-saving means of
production may clearly not be appropriate in the presence of a large pool
of unemployed labor. Second, for a small open economy, world demand

was assumed to be infinite. A global shift in production toward any one

good would, in the absence of any change in tastes, undoubtedly shift the
terms of trade against the good, making its production less profitable
and limit the scope for growth.20
REFERENCES
Arrow, Kenneth J., "The Economic Implications of Learning by Doing," Re-
view of Economic Studies, Vol. 29 (June 1982), pp. 155-73.
Bardhan, Pranab K., "Optimum Trade Policy in a Model of Learning by Do-
ing/' Chapter 7 in Economic Growth, Development, and Foreign Trade
(New York: Wiley-Interscience, 1970).
Helpman, Elhanan, "Growth, Technological Progress, and Trade," Empirica-
Austrian Economic Papers, Vol. 15 (1988), pp. 5-26.
Jorgenson, Dale W., "Capital Theory and Investment Behavior," American
Economic Review, Vol. 53 (1963), pp. 247-57.
Kouri, Pentti, "Profitability and Growth in a Small Open Economy," in Inflation
and Employment in Open Economies, ed. by Assar Lindbeck (Amsterdam;
New York: North-Holland, 1979).
Krugman, Paul, "The Narrow Moving Band, the Dutch Disease, and the Com-
petitive Consequences of Mrs. Thatcher," Journal of Development Eco-
nomics, Vol. 27 (October 1987), pp. 41-55.
Lim, Chong-Yah, Economic Restructuring in Singapore (Singapore: Federal
Publications, 1984).
, and others, Policy Options for the Singapore Economy (Singapore; New
York: McGraw-Hill, 1988).
Lucas, Robert E., Jr., "On the Mechanics of Economic Development," Journal
of Monetary Economics, Vol. 22 (July 1988), pp. 3-42.
Otani, Ichiro, and Cyrus Sassanpour, "Financial, Exchange Rate, and Wage
Policies in Singapore, 1979-86," Staff Papers, International Monetary
Fund, Vol. 35 (September 1988), pp. 474-95.
Otani, Ichiro, and Delano Villanueva, "Theoretical Aspects of Growth in Devel-
oping Countries: External Debt Dynamics and the Role of Human Capi-
tal," Staff Papers, International Monetary Fund, Vol. 36 (June 1989),
pp. 307-42.
Romer, Paul M., "Increasing Returns and Long-Run Growth," Journal of Politi-
cal Economy, Vol. 94 (October 1986), pp. 1002-37.
(1987a), "Growth Based on Increasing Returns Due to Specialization,"
American Economic Review, Papers and Proceedings, Vol. 77 (May 1987),
pp. 56-62.
20
In this context, see Lucas (1988), who considers a world where, because of an
assumed Ricardian technology, each country specializes in the production of one
of two goods. His model predicts some interesting cycles in the pattern of
production and trade.

82 BANKIM CHADHA
(1987b), "Crazy Explanations for the Productivity Slowdown," NBER

Macroeconomics Annual, ed. by Stanley Fischer (Cambridge, Massachu-
setts: MIT Press, 1987).
Shleifer, Andrei, "Externalities as an Engine of Growth" (unpublished; Gradu-
ate School of Business, University of Chicago, 1989).
Singapore, Department of Statistics, Yearbook of Statistics (various issues).
Singapore, Ministry of Trade and Industry, Economic Survey of Singapore (vari-
ous issues).
Singapore, Report of the Economic Committee, The Singapore Economy: New
Directions (Singapore: Ministry of Trade and Industry, 1986).
Solow, Robert M., "A Contribution to the Theory of Economic Growth,"
Quarterly Journal of Economics, Vol. 32 (February 1956), pp. 65-94.

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
The Transmission Mechanism

for Monetary Policy
in Developing Countries
PETER J. MONTIEL*
In many developing countries the financial system is characterized by the

absence of organized markets for securities and equities, by capital con-
trols, and by legal ceilings on bank borrowing and lending rates—a situa-
tion that gives rise to parallel markets for foreign exchange and informal
loan markets. This paper analyzes how changes in monetary policy instru-
ments are transmitted to domestic aggregate demand in a financially re-
pressed economy. Such an analysis is necessary to understand how the
move to a more market-oriented system would affect the economy in the
short run. [JEL 121, 311]
MONETARY AND fiscal policies are both commonly accorded promi-

nent roles in the pursuit of macroeconomic stabilization in deve-
loping countries. It is presumed that the authorities in such countries
have access to monetary policy instruments and can manipulate them to
achieve desirable macroeconomic objectives. Curiously, however, in
spite of the prominence given to monetary policy in the developing
country setting, the transmission mechanism for monetary policy in a
typical developing country has not been studied extensively, and conse-
quently is not well understood.
By contrast, in industrial countries with highly developed financial
markets, a broad consensus on the nature of the transmission mechanism
* Peter J. Montiel is Assistant Division Chief of the Developing Country

Studies Division of the Research Department. He is a graduate of Yale Univer-
sity and the Massachusetts Institute of Technology.
The author would like to thank Jagdeep Bhandari, Dale Henderson, Mohsin
Khan, and Geoffrey Woglom for useful comments.
83

84 PETER J. MONTIEL
has existed for some time.1 In brief, an open market purchase by the
central bank leaves the private sector with too much money in its port-
folio relative to other assets. In re-establishing portfolio equilibrium,
agents bid up the price of durable assets, thereby lowering their respec-
tive rates of return. Since the market prices of durable goods such as
physical capital thus exceed their replacement costs, private agents at-
tempt to increase their stocks of such assets by increasing their demand
for newly produced units of the latter. In this way, an open market
purchase results in an increase in aggregate demand.
In developing countries, the structure of financial markets makes the
operation of such a mechanism much more problematic. In the first
place, the menu of assets available to private agents is limited. Orga-
nized securities markets in which the central bank can conduct open
market operations simply do not exist in many developing countries. By
and large, individuals can hold currency as well as demand and time
deposits issued by the banking system, and they can borrow from com-
mercial banks. Durable goods such as land and physical capital can be
held directly, but organized equity markets are small or nonexistent.
Capital controls and prohibitions on the holding of foreign exchange
limit the extent to which foreign assets may be held by domestic resi-
dents, although parallel markets for foreign currency often emerge in
response to such regulations, thereby allowing private agents to circum-
vent official controls, at least in part. Finally, even in the case of those
assets and liabilities available to individuals such as demand or time
deposits and bank credit, official restrictions typically determine the
interest rates paid and charged by financial institutions. As in the case of
foreign exchange, however, informal markets often emerge to evade
interest rate ceilings, resulting in financial disintermediation through
"curb" markets for loans.
A typical developing country financial structure, therefore, might be
one in which individuals can hold domestic currency, deposits with the
banking system, foreign currency, loans extended through a curb mar-
ket, land, and physical capital. These assets may be financed by the
individuals' personal net worth or by borrowing, both from the banking
system and through the curb market. Interest rates on the assets and
liabilities of the banking system are fixed by applicable legal norms,
while the price of foreign currency on the parallel market, as well as the
interest rate on curb market loans, is determined by prevailing market
conditions.
1
Reviews of this issue in an industrial country setting are provided by Tobin
(1978), Laidler (1978), and Blanchard (1980).

TRANSMISSION MECHANISM FOR MONETARY POLICY 85
The connection between the actions of the monetary authorities and

the state of aggregate demand in such an environment is not obvious. In
the absence of organized securities markets, open market operations are
out of the question. Nonetheless, the monetary authorities retain four
instruments of policy—the level of administered bank interest rates, the
required reserve ratio, the amount of credit extended by the central
bank to the commercial banking system, and intervention in the parallel
exchange market. The purpose of this paper is to explore, within the
specific developing country context described above, the channels
through which these instruments affect domestic aggregate demand, as
well as to identify the parameters that govern their effectiveness.
It should be stressed that an understanding of the nature of the finan-
cially repressed economy, and the way monetary policy works in this
setting, is essential for predicting the effects of financial liberalization
policies with which many developing countries are experimenting. There
is at present no clear consensus on the effects that allowing interest rates
to be market determined (or raising them to positive real levels), or
reducing reserve requirements, or eliminating the spread between the
exchange rates in the official and parallel markets—policies that are
central in financial reform packages—would have on the economy.2
When it comes to discussing the consequences of such policies, often the
framework adopted is one more akin to the industrial country model
discussed earlier.
In the past, explicit analysis of the transmission mechanism for mone-
tary policy under financial repression in developing countries has tended
to fall into two camps—the McKinnon-Shaw tradition3 and a "neostruc-
turalist" approach.4 Adherents of the McKinnon-Shaw view maintain
that raising controlled bank interest rates need not be contractionary,
because in a rationed regime the induced increase in saving will result in
an increased supply of credit that facilitates the financing of private
investment or working capital or both.5 By contrast, analysts of a
neostructuralist persuasion cite the importance of informal loan markets
when bank interest rates are subject to legal ceilings. They emphasize
the possibility that increases in bank interest rates will draw funds away
from such markets, thereby increasing the marginal cost of funds and
exerting contractionary effects on the economy from both the demand
2
See Villanueva and Mirakhor (1990) for a recent description of financial
reform policies.
3
See McKinnon (1973) and Shaw (1973).
4
Associated with Taylor (1981).
5
See, for example, Kapur (1976) and Mathieson (1980). Fry (1982) provides a
survey of such models.

86 PETER J. MONTIEL
and supply sides (in the latter case, due to the need to finance working
capital).6 Notice that for the first group the primary channel of transmis-
sion is through the direct effect of the controlled bank deposit interest
rate on private saving, whereas the second group focuses on effects that
are transmitted through the loan interest rate to interest-sensitive com-
ponents of demand (and supply).
By contrast, the analysis presented here identifies a number of addi-
tional channels that also play important roles in the developing country
context. These include the wealth effects induced by changes in the
degree of financial repression, as well as effects that operate through the
premium in the free exchange market, through changes in expectations
of future inflation, and through changes in the economy's net foreign
assets. Since some of these mechanisms are closely affected by expecta-
tions of future events, the analysis highlights the important role of such
expectations in this context.
This paper is organized as follows. The model employed to analyze
these issues is presented in the next section. Section II examines the
transmission mechanism under partial equilibrium conditions. The third
section examines the macroeconomic general equilibrium effects of
changes in monetary policy instruments. The final section summarizes
the results and suggests some possible extensions that might usefully be
undertaken.
I. A Model of Developing Country Financial Markets
The model developed below involves a portfolio balance framework

expanded to include a commodity market. The economy in question is
small and open with four types of agents: households, the government,
the central bank, and the rest of the banking system. The authorities
maintain an official exchange rate for current international transactions,
but prohibit private capital movements. To avoid complicating the analy-
sis, the commodity structure is deliberately kept simple—that is, a stan-
dard Mundell-Fleming structure is assumed, with a single domestic and a
foreign good. Furthermore, since the focus is on exploring the shifts in
the economy's aggregate demand curve induced by changes in monetary
policy instruments, the model abstracts from supply-side complications
by assuming that domestic prices are instantaneously flexible, so that full
employment holds continuously. In this setting, shifts in aggregate de-
mand will be reflected only in the domestic price level. As indicated
6
See van Wijnbergen (1983) and Buffie (1984).

above, expectations play an important role in the model, and it will be

assumed that agents possess perfect foresight, an assumption that en-
riches the transmission mechanism by allowing changes in future ex-
change rates and prices that are induced by current policies to affect
current macroeconomic outcomes.
Model Specification
In principle, private households have access to five assets: domestic

currency, bank deposits, curb market loans, foreign exchange, and bank
credit (which is, of course, a liability for households). There is no market
for either private or government securities, and the stock of physical
capital is assumed to be constant.7 To simplify the analysis, currency is
ignored, and bank credit and curb market loans are taken to be perfect
substitutes in household portfolios. The latter permits all rationing-
induced "spillover" effects (see Barro and Grossman (1976)) of changes
in the stock of bank credit to be concentrated in a single market—that is,
the curb loan market—by permitting bank credit and curb loans to be
treated as a single asset.
Households
With these assumptions, households' financial portfolios are taken to
consist of bank deposits (D), curb market loans, bank credit (C), and
foreign exchange (//>).8 The value of households' financial portfolio (A)
is given by A = D + sfP - C, where s is the domestic currency price of
foreign exchange traded in the free (parallel) market. Portfolio balance
requires
7
Allowing for investment would make the model much more complicated and
preclude an analytical solution. This is an important limitation, because the level
of investment is itself an important target for policy. However, as long as the
speed of adjustment of the capital stock to its desired value via investment/disin-
vestment is slow relative to the rate at which markets for financial assets adjust,
the analysis of the (short-run) behavior of aggregate demand presented below
will remain valid.
8
Alternatively, //> could be taken to be the stock of land. In essence, fP is an
asset with a flexible, market-determined domestic currency price, which is
traded in organized markets by well-informed agents. It is intended to represent
"inflation hedges," which figure so prominently in developing country policy
discussions and might best be considered a composite of highly substitutable
assets, such as land and foreign exchange.

88 PETER J. MONTIEL
where iL and iM are, respectively, the interest rate on curb market loans
and the (controlled) interest rate on bank deposits; s is the expected (and
actual) rate of depreciation of the parallel market exchange rate; and P
is the domestic price level. The signs of the first three partial derivatives
of each function reflect the assumption that all assets are gross substi-
tutes. Money (in the form of bank deposits) is assumed to be held strictly
for transactions purposes, so the level of real financial wealth enters as a
scale variable to satisfy adding-up constraints in the demand functions
for curb market loans and foreign exchange, but not for deposits.9 The
partial derivatives in equations (l)-(3) must satisfy the standard con-
straints:
In addition to choosing the composition of their portfolios, house-

holds must also determine their level of consumption (c). The latter is
taken to depend on the real loan interest rate and on the level of house-
hold resources:
Household resources consist of real financial wealth (W/P) and real

factor income. As indicated above, however, real output is constant
under present assumptions, so the (unchanging) level of real factor in-
come is omitted from the function c( ).10
In the context of financial repression, the calculation of real house-
hold financial wealth must take into account the implicit taxes and subsi-
dies that interest rate ceilings impose on households as creditors and
9
Real output, which tends to shift asset composition among money and all
other assets in response to transactions motives, is excluded from the asset
demand functions because the assumptions of slow capital stock adjustment and
full employment render the level of real output constant over the time frame of
the10 analysis.
Since the real interest rate already appears with a negative sign in equation
(4), including the present value of factor incomes among the arguments of the
function c ( ) would not qualitatively affect the analysis. Notice that this
specification rules out direct McKinnon-Shaw effects of the controlled interest
rate IM on private consumption via intertemporal substitution. The effects of this
omission are discussed below.

debtors with the banking system. Letting ic denote the controlled interest
rate on bank credit, individuals with access to such credit receive a
subsidy of (iL — ic)C —that is, the interest rate differential between the
curb loan and bank credit markets times the amount of bank credit
extended to individuals with such privileged access. The present value of
this subsidy is given by (iL — ic)C/iL, and this represents a net addition to
household financial wealth.11 It is convenient to define an index of finan-
cial repression, denoted r, by
That is, r is the present value of the subsidy, per unit of bank credit,
which is implied by the prevailing interest rate ceilings. Notice that since
binding interest rate ceilings imply ic î^r is bounded between zero and
unity. When interest rate ceilings are not binding, ic = iL and r = 0. As
the curb loan interest rate rises relative to the administered interest rate
i'c, the constraint becomes more and more binding, and r approaches
unity.
While households are subsidized as debtors under financial repres-
sion, they are taxed as banking system creditors. The present value of
this tax is given by [(im — im)/im]M, where im is the deposit interest rate
that corresponds to a loan interest rate of iL under the banks' zero-profit
condition (equation (9) below). This condition can be used to show that
(im ~ im)/im = r, so the degree of financial repression can be written
equivalently as a function of banks' lending or borrowing rates, and the
present value of the tax on depositors can be expressed compactly as rM .
Taking these taxes and subsidies into account, households' real finan-
cial wealth can be expressed as
Thus, the wealth effects of financial repression depend on whether

households are net creditors (D - C> 0) or debtors (D - C < 0) of the
banking system. When D - C> 0, for example, an increase in the de-
gree of financial repression (r) reduces household wealth, since the im-
plicit tax imposed on households by interest rate ceilings on deposits
exceeds the subsidy received by favored borrowers.
11
This is actually an approximation to the true present value of the subsidy,
which depends on the entire stream of future interest rates and credit. The
approximation is adopted for tractability.

90 PETER J. MONTIEL
The Banking System

Bank assets consist of reserves held at the central bank (R) and credit
extended to households (C). Their liabilities are the deposits held by the
public (D), and credit received from the central bank (B). The balance
sheet of the banking system is therefore given by
Banks hold no excess reserves. Given a required reserve ratio of a,

reserve holdings are thus given by
Reserves at the central bank pay no interest, but credit extended to the
banking system by the central bank carries an interest charge, which, for
convenience, is set equal to the interest rate that banks charge their
customers, ic. Under these conditions, the zero-profit condition for the
banking system is given by
i
The Central Bank
The central bank pegs the official exchange rate at a value s. All
international commercial transactions are settled at this rate. Denoting
the central bank's stock of foreign exchange reserves (measured in for-
eign currency) as /c, and the trade balance (measured in units of the
domestic good and taken to be an increasing function of the real ex-
change rate, (SAP)) as #(s/P), the stock of foreign exchange reserves
evolves according to12
The central bank's assets include both foreign exchange reserves and
credit to the banking system, while its liabilities consist of reserves held
by the banking system. Thus the central bank's balance sheet is
The Government
Since the central bank's loans to the banking system earn interest, this
income must be allocated in some way. It is assumed to be transferred to
the government, which then uses it to purchase domestic goods. Since
the analysis is concerned with monetary rather than fiscal policy, the
government has no other role in the model than to dispose of these
12
This assumes, of course, that foreign exchange reserves do not pay interest.

funds. Letting g denote real government spending on domestic goods,

the government budget constraint implies
(12)
Commodity Market Equilibrium
The model is closed by the condition that the market for domestic
goods must clear. Letting y denote domestic real output, and ft, the
excess demand for domestic goods, this condition is
The Model in Compact Form

Before solving the model presented above, it is useful to introduce
some additional notation that can be used to rewrite the simultaneous
portion of the model in a more convenient form. Specifically, let
e(= s/P) denote the official real exchange rate;/ = /c +fp the economy's
total net foreign assets; and d = sfs the ratio of the free market exchange
rate to the official rate, which will be referred to as the premium. The
rate of change in d is denoted by d. Using this notation, the equilibrium
condition in the curb loan market (equation (2)) can be rewritten as
In addition to using e and d, several substitutions have been performed

in deriving (14). Using equation (7) in the definition of A and then
substituting from (11) enables the term AIP in equation (2) to be re-
placed by e(fc + dfp) in (14). Equations (7) and (9) together imply that
C = B 4- (1 - a)Z) —that is, the amount of lending that the banking sys-
tem can undertake consists of deposits net of reserves plus loans from the
central bank. Substituting from equation (1) and using the resulting
expression to replace C in equation (2) yields (14), where b = B/s. A
similar procedure permits equation (3) to be written as
Next, by using equations (7) and (11) in (6), real household financial
wealth can be written as
Thus, the wealth effects of financial repression depend on the excess

of central bank lending to the banking system (B) over reserves held by
the banking system. Intuitively, this is because credit to households can
exceed deposits only if the resources made available to banks by the

92 PETER J. MONTIEL
central bank exceed the resources extracted by the central bank. Using
equations (8) and (1) and letting w = W/s and b = Bis, the preceding
equation can be rewritten as
Substituting this in the commodity market equilibrium equation (13) and

using the notation introduced above yields
where use is also made of the government budget constraint (12).

Equations (14)-(17), together with the balance of payments equation
(10) and the definition of the financial repression index r, given in (5),
constitute the simultaneous portion of the model. The monetary policy
variables are the administered interest rate on deposits iM, the required
reserve ratio a, central bank lending to the banking system b, and cen-
tral bank intervention in the free market. The latter is captured by the
stock of foreign exchange available to the private sector fp, which can be
altered by the central bank subject to the condition dfc = —dfP, since
the economy's net international indebtedness / is the state variable in
the system, with equation of motion given by (10) (since //> = 0). When
the central bank sells foreign exchange in the parallel market, it may do
so either at the parallel rate or the official rate. In the latter case,
households will reap a windfall.13 The endogenous variables in the sys-
tem are the curb loan interest rate /L, the degree of financial repression r,
the balance of payments /c, as well as both the levels of the real exchange
rate e and premium d and their expected and actual rates of change e and
d. It is important to note that the effects of the monetary policy instru-
ments on aggregate demand are captured by their effects on the official
real exchange rate e, since e = s/P, and changes in P reflect shifts in
aggregate demand, given the economy's vertical aggregate supply curve.
II. The Transmission Mechanism in Partial Equilibrium

To solve the model, one can begin by solving the portfolio equilibrium
conditions (14) and (15) for the curb interest rate iL and the rate of
change of the premium d as functions of the remaining endogenous and
exogenous variables. The results are
13
Sales at the official rate may be inadvertent. For example, the case of
"leakages" from the official to the parallel market arising from export underin-
voicing or import overinvoicing can be treated as a sale of foreign exchange to
the parallel market at the official rate. A dual-market model incorporating such
leakages is presented in Bhandari and Vegh (1990).

TRANSMISSION MECHANISM FOR MONETARY POLIC\ 93
with
or
and
with
or
where
and the alternative versions of i7 and d7 depend on whether central bank

sales of foreign exchange in the free market are at the official rate (given
as the first alternative) or the free rate (second alternative).
These partial derivatives can readily be given intuitive interpretations.

94 PETER J. MONTIEL
Consider, first, the effects of the three dynamic variables d, e, and/. An

increase in the premium reduces the curb interest rate, but increases the
rate of depreciation of the free exchange rate. The reason is that, at the
initial values of iL and d, the increase in the premium creates an excess
supply of foreign exchange and an excess asset demand for curb market
loans. To induce households to hold more foreign exchange in their
portfolios relative to curb loans, the expected rate of depreciation in the
free exchange market has to rise and the loan interest rate to fall.
Turning to the real exchange rate e, a real depreciation (increase in e)
corresponds to a reduction in the domestic price level. On impact, this
creates an excess asset demand for curb loans, both because the real
value of the portfolio of lending households increases and because the
real value of the resources provided to the banking system by the central
bank is higher. The interest rate on loans therefore falls. The impact ef-
fect on the free exchange market is ambiguous, however, as both the real
supply of foreign currency and the real demand for this asset increase.
Finally, with regard to net international indebtedness /, since an in-
crease in the economy's net international creditor position must be asso-
ciated with an increased real value of household financial portfolios,
households seek to hold both more foreign exchange and more curb
loans as assets. Thus, both the curb interest rate iL and the rate of
depreciation of the free exchange rate d fall when/increases.
Turning to the monetary policy variables, an increase in bank credit
increases the asset demand for loans. The loan interest rate iL falls, and
to prevent the emergence of an excess demand for foreign exchange as
households switch out of the loan market, d must fall as well. This
situation is reversed in the case of an increase in the required reserve
ratio a. As banks' supply of loans is reduced, the loan interest rate rises.
This, in turn, requires an increase in d to induce households to hold their
existing stock of foreign exchange. The effect of an increase in this stock
(through central bank intervention in the parallel market), however, is to
create an incipient excess supply of foreign exchange as households seek
to restore their desired portfolio composition. Consequently, an increase
in d is required to restore equilibrium. The increase in d tends to reduce
the asset demand for loans. If the foreign exchange is sold at the official
rate, however, this will be partly offset by a positive wealth effect on the
asset demand for loans due to the windfall reaped by households, render-
ing the effect on iL ambiguous. If the currency is sold at the parallel rate,
the latter effect is absent (the last term inside the parentheses in the first
version of both i7 and d-j disappears), and the loan interest rate rises.
Matters are also somewhat complicated with regard to changes in the
administered interest rate iM. On impact, an increase in iM will result in
an excess supply of foreign exchange, as funds are attracted into the

domestic financial system and away from the holding of foreigjj^ssets.

However, the net asset demand for loans may rise or fall. Though4bank
lending will rise as deposits increase, household lending will fall as
households shift funds away from the loan market and into deposits.
Since banks hold reserves whereas private lending agents do not, each
unit moved by households from the loan market into the domestic finan-
cial system reduces the net supply of loans.14 However, to the extent that
funds attracted to banks come out of foreign currency holdings instead,
the supply of loans rises. Thus, the key to the impact effect of changes in
iM is whether households primarily move funds out of the loan market or
out of foreign currency—that is, whether loans, on the one hand, or
foreign currency, on the other, are better substitutes for deposits. This
determines the sign of (F2L3 — F3F2) in is. In the "currency substitution"
case (in which foreign currency and domestic money are close substi-
tutes), the expansion of deposits would primarily be at the expense of
foreign currency, rather than loans, and in this case an increase in iM is
likely to result in an incipient excess asset demand for loans, causing iL to
fall (i5 < 0, because F2 and L3 are large in absolute value). Whether IL
increases or decreases, however, the effect on d will be positive.
To explore the determination of real private financial wealth in the
model, substitute for r in (16) from (5) and use equations (18) and (19).
This yields
with
or
14
This analysis is familiar from the neostructuralist literature. See van Wiin-
bergen (1983) and Buffie (1984).

96 PETER J. MONTIEL
The wealth effects of the variables considered above can be decom-

posed into three parts: (1) a direct wealth effect; (2) an effect that
operates through changes in the degree of financial repression; and (3)
an effect that operates through changes in the base on which the implicit
financial repression tax is levied.
Equation (20) indicates that wealth effects are one mechanism
through which monetary policy instruments may affect aggregate de-
mand. All of the monetary policy instruments under examination here
exert such effects. Both an increase in credit to the banking system (b)
and a sale of foreign exchange in the free market (fp) lower the portfolio
equilibrium loan interest rate. By reducing the degree of financial re-
pression, this increases real private financial wealth. An increase in
controlled interest rates also reduces the degree of financial repression.
In the currency substitution case (that is, is < 0), this is because market
interest rates fall while controlled interest rates rise. Even when market
interest rates rise (that is, i5 > 0), under certain conditions their rise will
be proportionately smaller than that of controlled interest rates, so that
JS/JL ~ I/*M > 0, and the degree of financial repression will be reduced—
generating positive wealth effects in this case as well.15 When the re-
quired reserve ratio rises, however, the proportionate increase in the
loan interest rate (IÎL) exceeds that of the interest rate on bank credit
(given by 1/(1 — a)) for small initial values of or, so the degree of financial
repression increases and the wealth effect is negative. This effect is
reversed if a is large, but only the former case is considered below.
Notice, finally, that in the case of foreign exchange sales at the official
rate, the direct wealth effect is positive. In addition, a positive wealth
effect arises from lower household deposit holdings, which reduce the
base of the financial repression tax. Since the sign of i7 is ambiguous in
this case, the rate of financial repression may increase or decrease, so
wealth effects arising from this source are ambiguous. Finally, if foreign
exchange is sold at the free rate, there are no direct wealth effects (the
first term in w7 disappears). Since iL rises, the degree of financial repres-
sion increases and the wealth effect from this source is negative. In this
case, the net wealth effect depends on the initial degree of financial
repression, and is more likely to be negative the lower the initial value
ofr.
Turning to the remaining variables in equation (20), changes in the
premium, the real exchange rate, and the economy's net foreign assets
all exert direct wealth effects in a straightforward manner, as well as
15
Essentially, what is required is that the semi-elasticity of the demand for
deposits with regard to the own interest rate im does not greatly exceed the cross-
semi-elasticity with regard to iL.

indirect effects similar to those of the monetary policy instruments

through their influence on the degree of financial repression. Both the
real exchange rate and the stock of net foreign assets also affect the base
on which the financial repression tax is levied. In particular, increases in
e and / are associated with increases in L, which increase the private
sector's net creditor position vis-a-vis the public sector. While this effect
tends to offset the otherwise positive wealth effects of increases in e and
/, inspection of w2 and w3 reveals that this offset is only partial, since it is
dominated by the direct wealth effect in both cases. Thus, w2 and w3 are
both positive.
The next step is to examine the properties of the commodity market
clearing equation (17). Using equations (18), (20), and (9) in (17), this
equation can be written as
with
Equation (21) is of central importance for understanding the transmis-

sion mechanism for monetary policy in this model. Notice that equation
(21) must hold at every instant of time, and that the effect of changes in
the monetary policy instruments on aggregate demand is the inverse of
their effect on the real exchange rate e (since e — s/P and changes in
aggregate demand are reflected one-for-one in changes in P). Consider,
for example, an increase in central bank credit to the banking system (b).
Thus, from (21)
The term in parentheses in the second line of equation (22) describes

98 PETER J. MONTIEL
the effect of changes in b on the demand for domestically produced

goods at a given P, while the coefficient e^1 captures the change in P (and
thus in e) required to restore equilibrium in the commodity market.
Equation (22) tells us that the effect of changes in b on aggregate de-
mand can be decomposed into four parts. The term e4 captures the
"autonomous" contribution of b to aggregate demand—that is, the ef-
fect of, say, an increase in b on aggregate demand after allowing for
portfolio reallocations but before allowing for any induced adjustments
in the premium d, in the expected future time path of the real exchange
rate (that is, e\ or in the economy's net foreign assets/. Changes in b
will, however, also induce adjustments in these other endogenous vari-
ables, which themselves affect demand and thus represent part of the
transmission mechanism through which monetary policy operates. In
particular, the term e^ddldb captures the effects of b on demand that
operate through induced changes in d, the term —deldb captures effects
which appear through induced changes in e (the inverse of the domestic
inflation rate), and e3 df/db measures effects transmitted through changes
in the stock of foreign assets.
Leaving aside these induced effects for the present, let us consider
more closely the autonomous component. This component is given by
the partial derivatives e4 through e-j under equation (20) for each of the
four monetary policy instruments, respectively. As these partial deriva-
tives indicate, the autonomous effect operates through three distinct
mechanisms—a real interest rate effect, a real wealth effect, and a fiscal
effect.
Consider first an expansion of credit to the banking system (an in-
crease in b). From equation (18), given d, e, and /, an increase in b
requires a reduction in IL to restore portfolio equilibrium (i'4 < 0). This
stimulates private demand through the real interest rate effect (c\ i'4 > 0).
Since iL falls, the degree of financial repression decreases, and this
increases real private wealth (equation (20)). Thus, the wealth contribu-
tion c2H>2 reinforces the interest rate effect. Finally, an increase in b
increases revenues for the public sector due to the interest charges on the
larger stock of credit extended by the central bank, and since this addi-
tional revenue, amounting to ice, is spent by the government on home
goods, demand increases for this reason as well.
An increase in the administered interest rate on deposits in principle
has an ambiguous effect on demand. Consider, however, the currency
substitution case mentioned earlier (is < 0; see equation (18)). The inter-
est rate effect (ci i5), which is the source of the ambiguity in the more
general case, is therefore positive under these circumstances. Since the
degree of financial repression falls (i5HL - \HL < 0) whether i5 is positive

(under the conditions mentioned previously) or negative, wealth rises,

and the contribution of the wealth effect is therefore always positive.
The fiscal effect is also positive, since central bank (and thus govern-
ment) income is increased when administered interest rates are raised
(eb/(l — cr) >0). Thus, either the currency substitution case or a suffi-
ciently weak positive effect of increases in iM on the loan interest rate iL is
sufficient to make the partial equilibrium effect of an increase in iM on
aggregate demand positive.16
Increases in the required reserve ratio increase the portfolio equilib-
rium loan interest rate, exerting a negative effect on demand through
this channel (ci i6 < 0). Moreover, the degree of financial repression is
increased as well, resulting in a negative wealth effect on demand. Due
to the increase in the administered credit interest rate ic, however, fiscal
revenue increases, thus providing a positive effect that functions as at
least a partial offset to the negative demand effects described above.
Though e6 could thus have either sign, only the case e6 < 0 will be consid-
ered in the remainder of the paper.17
The monetary authorities can bring about an increase in fp by selling
some of their foreign exchange reserves in the free exchange market.
The autonomous effect on demand is ambiguous in sign. If the foreign
exchange is sold at the free rate and the initial degree of financial
repression is sufficiently low (so w7 < 0), an increase in fp raises the
portfolio equilibrium loan interest rate, making the interest rate effect
c\ i-j negative. The consequent increase in the degree of financial repres-
sion exerts an additional negative wealth effect, which is not offset in this
case by the reduction in the base of the financial repression tax.
In addition to these autonomous effects, monetary policy instruments
induce changes in aggregate demand through their effects on the parallel
market premium, on expectations of inflation, and on the economy's net
foreign assets. Increases in both the premium and in the economy's stock
of foreign assets decrease the portfolio equilibrium loan interest rate,
which exerts positive interest rate as well as wealth effects (by reducing
the degree of financial repression) on demand. They also directly in-
crease real household financial wealth, which supplements the indirect
16
Notice that, if the direct McKinnon-Shaw effect of iM on consumption were
present, this would contribute to a negative interest rate effect on demand,
making the ambiguous case more likely. However, as the next section will show,
the presence of iM in the consumption function will not alter the qualitative
nature of the general equilibrium effects of changes in 1M.
17
The case e6 < 0 is of greater interest both because it is likely to be empirically
dominant and because the magnitude of the fiscal effects are tied to essentially
arbitrary assumptions about the initial size of b and the composition of induced
government spending.

100 PETER J. MONTIEL
wealth effect through the reduction in financial repression. An increase

in the expected rate of inflation, given the loan interest rate, lowers the
expected real interest rate and thereby stimulates demand for home
goods.
Because of these induced effects operating through d, e, and/, the
total effect of monetary policy instruments on domestic aggregate de-
mand differs from the autonomous effect described above. In order to
determine the net effect of policy instruments on aggregate demand, it is
necessary to take these indirect transmission mechanisms into account—
that is, to calculate the total effect of policy instruments, including
effects transmitted through changes in d, e, and/. These general equi-
librium effects can be derived from a solution to the model.
III. The General Equilibrium Transmission Mechanism
The model can be represented as a system of three differential equa-

tions in d, e, and/ The system consists of equations (19), (21), and (10),
respectively. There is a single predetermined variable in this system (the
stock of net foreign assets/), and two "jump" variables (d and e).
To solve the model, the system is first linearized around the steady
state defined by d = e = / = 0, which is denoted d*, e*,f*:
where
and, for symmetry, the function /has been defined as/(e) = x(e)/e (so
f2 = x'/e> 0). The properties of this equilibrium can be determined from
trace
det
Since the trace condition implies that the sum of the roots must be
positive, on the one hand, the system has at least one positive root. Since
the negative determinant means that the product of the roots is negative,
on the other hand, the number of negative roots is odd. This establishes
that the system contains one negative and two positive roots—that is, the

equilibrium (d*, e * , f * ) exhibits saddlepoint stability. Let X denote the

negative root. The solution of the model can be written as
Taking time derivatives of these expressions yields
Equation (25a) indicates that when the economy's net foreign assets are
below their steady-state level (/<>-/* <0), /will be rising (recall that
X<0); that is, a trade surplus will emerge. Similarly, from (25b), this
situation will be characterized by a real appreciation (that is, expected
and actual inflation in excess of the world rate, which has been taken to
be zero). With regard to the premium, according to equation (25c), its
behavior under these circumstances depends on the sign of (d3 + d2 X//2)/
(X — di). The denominator of this expression is negative, but since the
sign of d2 is ambiguous, so is the sign of the numerator. Some benchmark
cases are considered below.
To investigate the general equilibrium effects of monetary policy,
begin by deriving the steady-state effects of changes in monetary policy
instruments. Setting d = e = f = 0in (23) and solving for dd, de, and df
yields

Notice that, since under the assumptions of the model there is a unique
real exchange rate compatible with trade balance equilibrium, monetary
policy instruments cannot affect the steady-state real exchange rate (that
is, the second row of the coefficient matrix in equation (26) consists of
zeroes). The premium and the economy's stock of foreign assets, how-
ever, will be affected in the steady state, and all three variables will
deviate from their steady-state values during the transition. The dynamic
effects of each of the monetary policy instruments are now investigated
in turn.
Central Bank Credit to the Banking System
The steady-state effects of an increase in b are given by
Thus, an increase in central bank credit to the banking system reduces

the steady-state stock of foreign assets (a result familiar from the mone-
tary approach to the balance of payments), but has ambiguous steady-
state effects on the premium.
Equation (27a) permits the time path of domestic demand in response
to the increase in credit to be described. Since the initial stock of foreign
assets is predetermined, the initial level of/exceeds its steady-state value
/* (that is, /o-/* >0). From equation (24b), this means that e0<e*.
Since e* is unchanged, e must appreciate on impact—that is, the domes-
tic price level must rise when b is increased. The general equilibrium
effect of an increase in b on domestic demand is thus expansionary on
impact—that is, the economy's aggregate demand curve shifts to the
right.
How is this effect transmitted? From equation (22), we know that part
of the effect arises from the autonomous component described earlier.
Since/does not change on impact, none of the initial effect on demand
arises from a change in the economy's net foreign assets (the fourth term
in (22)). From equation (27b), we know that initially de/db>0. This
depreciation of the real exchange rate over time implies that the domes-
tic rate of inflation falls below its initial value (zero in this case). In itself,
this effect is contractionary (since it contributes to raising the real inter-
est rate), so it does not represent a vehicle for the transmission of
positive effects on demand. Unfortunately, it is not possible to establish

without further restrictions whether the free exchange rate d depreciates

or appreciates on impact. Thus, some portion of the expansionary im-
pulse may be transmitted through an increase in d.
Changes in Controlled Interest Rates

Using equation (26), the steady-state effects of an increase in the
controlled interest rate iM are18
Thus, an increase in the controlled interest rate increases the steady-

state stock of foreign assets and reduces the steady-state premium in the
free exchange market. These results have an important implication,
moreover. From equation (24b), the increase in /* implied by (28a)
means that, on impact, e > e*\ that is, the real exchange rate depreciates.
Since this can only be brought about by a reduction in the domestic price
level, this means that, even in the currency substitution case in which an
increase in the controlled interest rate iM reduces the loan interest rate iL,
the effect of this measure on aggregate demand is contractionary on
impact when general equilibrium interactions are taken into account.
To see how this can happen, notice that in the case of changes in iM,
equation (22) becomes
In the currency substitution case, e$ is positive (see equation (21)). From

equation (25b), the effect of a change in iM on e on impact will be
negative—that is, domestic inflation will exceed the world rate. As equa-
tion (29) indicates, this adds a further expansionary mechanism to supple-
ment the autonomous component e5. Finally, since/does not change on
impact, the last term in (29) is zero. Thus, the negative impact on
domestic demand must arise from the term eidd/diM. An increase in
administered interest rates exerts a contractionary effect on demand on
impact because it reduces the premium in the free exchange market. It
does so by increasing the relative attractiveness of holding assets in the
18
Equation (28a) can be signed unambiguously after substituting from equa-
tions (19) and (21).

form of deposits with the domestic banking system, thereby reducing

demand for foreign currency.
This example highlights the importance of a comprehensive treatment
of the transition mechanism in developing countries. In this case, taking
into account effects that operate through the unconventional mechanism
of the free exchange market reverses the sign of the effect of a key
monetary policy instrument on aggregate demand.
Changes in Reserve Requirements

From equation (26), the steady-state effects of an increase in the
reserve requirement ratio cr are given by
Equation (30a) indicates that an increase in the required reserve ratio will
increase the steady-state stock of foreign assets. As in the case of admin-
istered interest rates, however, the effect on the steady-state premium is
ambiguous in sign. Since (/0 — /*) < 0, equation (24b) indicates that the
real exchange rate depreciates on impact—that is, the increase in a has
contractionary short-run effects on domestic demand. Since the domestic
price level begins to rise immediately (equation (25b)) and since the
country's net foreign assets (/) do not change on impact, this demand
contraction represents the net effect of the autonomous component of the
transmission mechanism and the component operating through changes
in the premium. The autonomous component, which in this case is given
by the partial derivative e6 under equation (21), has already been shown
to be negative in sign. Though signing the separate contribution of the
change in the premium would require additional restrictions, it is clear
that general equilibrium repercussions through this channel can at most
dampen the effects of the autonomous component.
Intervention in the Free Exchange Market

The steady-state effects of central bank sales of foreign exchange in the
free market are given by19
19
Equation (31b) can be signed after substituting from equations (19) and (21).

Although the magnitudes of these expressions depend on whether the

foreign exchange is sold at the official or the market price, their signs do
not. A sale of foreign exchange in the free market (increase in Pf )
increases the steady-state stock of foreign assets (df/dfP> 0) and reduces
the steady-state premium (ddldfP< 0). This measure will thus be contrac-
tionary on impact once general equilibrium interactions are taken into
account, even though the sign of the partial equilibrium autonomous
component could not be determined unambiguously (see the partial
derivative e1 under equation (21)). As in previous cases, the short-run
contractionary impulse must operate through some combination of au-
tonomous effects tending to raise the real loan interest rate and of the
induced change in the premium.
IV. Summary and Conclusions
A typical financial environment for a developing country is character-

ized by (1) the absence of markets for domestic securities; (2) the pres-
ence of capital controls; and (3) legally determined interest rates on bank
assets and liabilities. Legal restrictions on foreign exchange and loan
transactions give rise to parallel markets for foreign exchange as well as
to informal markets for loans. The tools of monetary policy in this
environment consist of central bank credit to the banking system, the
setting of administered interest rates and required reserve ratios, and
intervention in the free exchange market. Both the management of
monetary policy in such a setting and an assessment of the macroeco-
nomic effects of financial reform measures such as lowering reserve
requirements or raising controlled interest rates require an understand-
ing of the transmission mechanism through which such instruments affect
aggregate demand.
The analysis above suggests that they do so in a number of ways other
than the conventional effects on interest-sensitive components of de-
mand operating through changes in market-determined interest rates
that have been emphasized in the neostructuralist literature. In this
paper, these have been divided into autonomous (partial equilibrium)
effects and effects that emerge through general equilibrium interactions.
The former include, in addition to interest rate effects, those that oper-
ate through changes in household wealth and through the government
budget. Wealth effects arise from the recognition that financial repres-
sion entails an implicit system of taxes and subsidies on households as
creditors and debtors of the banking system. Changes in monetary policy

instruments affect the effective degree of financial repression, and thus

the present value of these taxes and subsidies, both by changing the rate
at which financial repression taxes household portfolios of a given com-
position and by altering the composition of portfolios in ways that affect
the base to which the financial repression tax applies. Moreover, since
monetary policy changes affect the profits of the central bank, the fiscal
consequences of these changes represent a separate channel through
which the effects of policy on demand are transmitted.
General equilibrium effects emerge through the consequences of pol-
icy changes for the parallel market premium, the expected rate of infla-
tion, and the economy's stock of foreign assets. Since all these variables
affect the state of domestic demand, induced changes in their values
represent important additional channels through which monetary policy
may exert aggregate demand effects.
In the model specified here, the direction of change in domestic aggre-
gate demand on impact in response to changes in monetary policy instru-
ments is determined by a combination of autonomous effects and effects
that operate through the premium in the free exchange market. Regard-
ing the other two mechanisms of transmission, the effects of induced
changes in the stock of foreign assets are felt only over time (since this
stock evolves as a function of cumulative current account surpluses),
while expectations of future inflation tend in each case to dampen the
demand effects of the policy change. The reason is that expectations of
future inflation are expansionary. Since in this model the price level
tends to return to its steady-state level, an expansionary shock (one that
increases today's price level) will reduce expected future inflation, which
in itself has a contractionary influence. Conversely, a contractionary
shock will reduce today's price level, but only temporarily, and the ex-
pected recovery of prices is itself expansionary.
For two of the monetary policy instruments examined here—central
bank credit to the banking system and the required reserve ratio—the
short-run general equilibrium impact on demand is qualitatively the
same as the partial equilibrium impact operating through loan interest
rate, wealth, and fiscal effects. Whether the general equilibrium reper-
cussions through the free market premium and expected inflation aug-
ment or weaken these effects depends on the particular values given to
the parameters of the model. Thus, in empirical assessments of the likely
macroeconomic impacts of prospective monetary policy measures, these
general equilibrium mechanisms must be taken into account.
In the case of changes in administered interest rates and interven-
tion in the free exchange market, the full general equilibrium effects of
the measures differ from what would be predicted on partial equilibrium

considerations. Increases in administered interest rates prove to be

contractionary (that is, aggregate demand for home goods decreases),
even in the currency substitution case in which partial equilibrium
demand effects are positive. The reason is that the adverse wealth effects
of a lower premium in the free exchange market overwhelm the positive
contribution to demand emanating from lower loan interest rates.
Similarly, though the partial equilibrium consequences of a sale of
foreign exchange in the free market by the central bank are ambiguous,
this measure can also be shown to be contractionary when induced
general equilibrium mechanisms of transmission are incorporated into
the analysis.
For the purpose of identifying the nature of the macroeconomic inter-
actions that govern the impacts of monetary policy instruments on aggre-
gate demand in developing countries, the model presented here can
most usefully be extended in two directions. First, some stickiness in
price adjustment is likely to be empirically important in many cases, and
this would affect not only the price-output consequences of changes in
demand, but also the role of expectations of inflation in transmitting the
effects of monetary policy. Second, as indicated at the outset, it is
desirable to integrate capital accumulation explicitly into the model. It is
important to note, however, that while this would increase the complex-
ity of the general equilibrium analysis, the mechanisms of transmission
that have been identified here would remain relevant in the extended
model.
REFERENCES
Barro, Robert, and Herschel Grossman, Money, Employment and Inflation
(Cambridge: Cambridge University Press, 1976).
Bhandari, Jagdeep, and Carlos A. Vegh, "Dual Exchange Markets Under In-
complete Separation: An Optimizing Model," Staff Papers, International
Monetary Fund, Vol. 37 (March 1990), pp. 146-67.
Blanchard, Olivier, "The Monetary Mechanism in the Light of Rational Expec-
tations," in Rational Expectations and Economic Policy, ed. by Stanley
Fischer (Chicago: University of Chicago Press, 1980).
Buffie, Edward, "Financial Repression, the New Structuralists, and Stabiliza-
tion Policy in Semi-Industrialized Economies," Journal of Development
Economics, Vol. 14 (April 1984), pp. 305-22.
Fry, Maxwell J., "Models of Financially Repressed Developing Economies,"
World Development, Vol. 10 (September 1982), pp. 731-50.
Kapur, Basant, "Alternative Stabilization Policies for Less-Developed
Economies," Journal of Political Economy, Vol. 84 (August 1976), pp.
777-95.

Laidler, David, "Money and Money Income: An Essay on the Transmission

Mechanism," Journal of Monetary Economics, Vol. 4 (April 1978), pp.
151-91.
Mathieson, Donald J., "Financial Reform and Stabilization Policy in a Develop-
ing Economy," Journal of Development Economics, Vol. 7 (September
1980), pp. 359-95.
McKinnon, Ronald I., Money and Capital in Economic Development (Washing-
ton: The Brookings Institution, 1973).
Shaw, Edward S., Financial Deepening in Economic Development (New York:
Oxford University Press, 1973).
Taylor, Lance, "IS-LM in the Tropics: Diagrammatics of the New Structuralist
Macro Critique," in Economic Stabilization in Developing Countries, ed. by
William Cline and Sidney Weintraub (Washington: The Brookings Institu-
tion, 1981).
Tobin, James, "Monetary Policy and the Economy: The Transmission Mecha-
nism," Southern Economic Journal, Vol. 44 (January 1978), pp. 421-31.
van Wijnbergen, Sweder, "Interest Rate Management in LDCs," Journal of
Monetary Economics, Vol. 12 (September 1983), pp. 433-52.
Villanueva, Delano, and Abbas Mirakhor, "Interest Rate Policies, Stabilization,
and Bank Supervision ip Developing Countries: Strategies for Financial
Reforms," IMF Working Paper 90/8 (Washington: International Monetary
Fund, February 1990).

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
The Budgetary Control

and Fiscal Impact of Government
Contingent Liabilities
CHRISTOPHER M. TOWE*
Conventional fiscal accounting methodologies do not appropriately ac-

count for governments' noncash policies, such as their contingent liabili-
ties. When these liabilities are called, budget costs can be large, as evi-
denced by the U.S. savings and loan crisis. In general, deficit measures
may underestimate the macroeconomic impact of government policies,
promoting the substitution of noncash for cash expenditure and increasing
future financing requirements. This paper describes extended deficit mea-
sures to address the problem, but notes their limited practical value.
Nonetheless, some alternative methods of valuing contingent liabilities are
proposed to gauge fiscal impact and facilitate budgetary control. [JEL
320]
M UCH OF of the debate regarding the efficacy of conventional cash

flow measures of the deficit as indicators of fiscal impact has
ignored the fact that an increasingly significant instrument of govern-
ment policy—the adoption of contingent liabilities such as deposit in-
surance, social security and health insurance, and loan guarantees—does
not involve a current cash flow, but an obligation regarding possible
future cash flows.1 Conventional budget methodologies account for con-
tingent liabilities not when the obligation is incurred, but when the
actual expenditure is made.2 However, insofar as contingent claims on
* Christopher M. Towe, an Assistant in the Office of Executive Directors, was

an Economist in the Fiscal Affairs Department when this paper was prepared.
Comments by Mario I. Blejer, Adrienne Cheasty, and George Mackenzie are
gratefully
1
acknowledged.
For a useful summary of extended measures of the deficit, see Tanzi, Blejer,
and2 Teijeiro (1988) and Mackenzie (1989).
Note that in accrual-based methodologies expenditure is included when the
commitment is made with certainty. See, for example, Diamond and Schiller
(1988, p. 40).
109

110 CHRISTOPHER M. TOWE
the government are valued by the private sector in excess of any fees
charged in exchange, they may affect economic behavior when issued in
a fashion similar to a cash tax or subsidy. Therefore, the conventionally
defined budget deficit, by accounting for contingencies only when a cash
outlay is made, may misrepresent the government's current fiscal impact
and limit its analytic usefulness. Moreover, since the issuance of such
contingencies may have severe future cash flow implications, by relying
on conventional accounting methods, budgetary authorities may not be
provided with the means to adequately monitor and control the govern-
ment's overall fiscal position.3
This paper reviews the types of contingent liabilities governments
issue, and examines their accounting in conventional budget methodol-
ogies (in Sections I-III). The impact of government contingencies on
private sector behavior is then discussed, and means by which measures
of the fiscal deficit could be amended to account for contingencies are
reviewed (in Sections IV-VII). Although, in theory, deficit measures
could be defined that would include the fiscal impact of governments'
issue of contingent claims, it is argued that they would be based on the
choice of relatively extreme views regarding the macroeconomy. A sim-
ple alternative would be to require calculation of the change in the
degree to which a contingency program is funded during each budgetary
period. This would provide both an ancillary measure of the govern-
ment's impact on the economy and a device with which to gauge and
enforce budget discipline.
I. Government Contingent Liabilities

The distinction between governments' contingent and noncontingent
liabilities (for example, interest-bearing debt) is that the nominal obliga-
tion and the settlement date of the latter are fixed at the date of issue,
whereas with contingent liabilities, the contractual obligation of the
government is dependent, in its timing and amount, on the occurrence of
a particular event. A major proportion of governments' contingent li-
ability relates to social security programs—sometimes referred to as
annuity programs—such as state pension schemes and medical insurance
programs. These imply an obligation by the government to provide
3
For example, in the United States deficit targets have been legislated that
include the current, but "temporary," surpluses of the Social Security Adminis-
tration. For a description, see Ebrill (1990).

GOVERNMENT CONTINGENT LIABILITIES 111
financial assistance to the private sector that is contingent on various

criteria including need, disability, retirement, unemployment, or death.4
While it has been estimated that governments of 143 countries now
provide some type of social security program, the scope and coverage of
such programs differ widely.5 Coverage may be employment related, in
which benefits are contingent on length of previous employment and
earnings and benefits are usually partially funded through compulsory
contributions either by employers or employees or both. Nonetheless,
the central government usually contributes a major share of total rev-
enue. Less prevalent are systems in which coverage is universal and
benefits are untied to recipients' employment history.
Governments provide a myriad of other, nonsocial security-related,
programs that are primarily (but not exclusively) designed to stimulate
particular economic activities by reducing risk, rather than to provide
income support; these include loan guarantees, deposit insurance, mort-
gage guarantees, and trade and exchange rate guarantees. Unlike the
annuity programs described above, which are usually based on a princi-
pal of universal coverage, loan guarantees and other similar insurance
schemes are usually associated with the consumption of a particular
service that is deemed worthy of subsidy (for example, deposit insurance
and guarantees of student and housing loans).6
These programs encompass a broad range of characteristics.7 The
guarantee may cover 100 percent of the credit arrangement, or up to
some fraction or fixed amount. Most credit and deposit guarantee
4
The rationale for such programs is most often normative—that provisions
should be made to redistribute income to the needy or aged, or that households
must be coerced into saving for old age. Alternatively, insurance market imper-
fections that restrict the development of private insurance markets, such as
informational asymmetries between the insurer and the insured, are also viewed
as an important reason for providing social security. See Atkinson (1987) for a
description
5
of social security programs and their economic implications.
For a complete cross-country description of the typical characteristics of
social security programs, see U.S. Department of Health and Human Services
(1987).
6
Nonetheless, these schemes may also be seen as a means of achieving welfare
or redistributive goals. For example, increasing access to capital markets, which
would have been denied owing to informational asymmetries, is argued to be a
welfare-improving policy (see, for example, Mayshar (1984)). However, others
have argued that such guarantees promote an inappropriate adoption of risky
investments and are, therefore, inefficient (Chancy and Thakor (1985),
Bosworth,
7
Carron, and Rhyne (1987, p. 41), and Towe (1989)).
See Brau and Puckahtikom (1985) for a description and survey of exchange
rate guarantee systems. For a useful survey of the characteristics of credit guar-
antee schemes in a number of industrial and developing countries, see Levitsky
and Prasad (1985).

schemes require a fee or premium, which may be a one-time or annual

payment and is usually based on a percentage of the amount guaranteed.
Often these programs are "funded," in the sense that some attempt is
made to ensure the maintenance of a reserve that matches the expected
liability, and in some cases the program's liability is limited to the
amount of the reserve. Guarantee or insurance schemes may be volun-
tary—for example, loan guarantee schemes—or mandatory, as in the
case of deposit insurance.
In addition, a substantial share of governments' contingent liabilities
may be associated with the implicit guarantee of transactions of para-
statal agencies or sectors of the economy. These implicit guarantees are
distinct from those discussed above, in that no specified contractual basis
exists defining the government's liability. For example, government-
sponsored enterprises (GSEs), although wholly privately owned, may be
mandated to perform public policy.8 Because GSEs are constrained to
fulfill public policy objectives and so may be prevented from profit
maximizing, it has been suggested that governments face a "moral"
(rather than a legal) obligation to guarantee their debt. Similarly, al-
though a government may not be contractually bound to rescue indus-
tries or regions that suffer financial reverses, there may be a similar
implicit obligation. Salient examples in the United States include the
extension of substantial credit guarantees to Penn Central Railway (in
1970), Lockheed (in 1971), New York City (in 1975), and Chrysler Cor-
poration (in 1979).9
II. Funding Contingent Liabilities
A firm's contingent liabilities are considered funded if they are

matched by a reserve or charge against profits equal to the actuarial
value of the liability—that is, when the reserve equals the present dis-
counted value of expected payouts. The relevance of this distinction, as
often applied to pension and insurance funds, is whether or not the
balance sheet is sufficiently strong to ensure that the liability can be
repaid if the plan were terminated. Actuarial examination of a private
sector pension plan, for example, will require estimation of benefits
accrued to date—the actuarial liability—which represent the firm's cur-
rent obligation to current plan participants. The firm's contingent liabil-
ity is considered funded if this value is matched by reserve assets; the
8
For a discussion of the special relationship that GSEs enjoy with the U.S.
Government,
9
see U.S. Office of Management and Budget (1989, p. F-24).
The budget implications of these "bailouts" are described in Ippolito (1984,
Chap. 4).

difference between the actuarial liability and any reserve assets that exist
is termed the unfunded actuarial liability.
However, actuarial examination usually also requires the calculation
of the expected present value of additional benefits expected to be
accrued in the future. The sum of past and expected future accruals is
termed the actuarial present value of future benefits. Firms are often
legally required to erase the difference between this latter value and the
value of reserve assets over a period of time through the adoption of a
schedule of contributions that fund the benefits of the plan.10
It has been argued, however, that these actuarial concepts must be
used with caution when applied to the government's fiscal accounts.
First, as the government's power to levy taxes or create debt instruments
to finance expenditure implies that it does not face the same solvency
constraints as the private sector, actuarial techniques designed to mea-
sure solvency may be less relevant for an analysis of fiscal policy.11
Moreover, Selling and Stickney (1986) note that recent accounting stan-
dards set for the U.S. private sector require pension liability to be
calculated on the basis of accumulated benefit and projected benefit
obligations, neither of which include consideration of the impact of
expected future service on benefit obligations.12 This similarly suggests
that actuarial methodologies and standards, as applied to fiscal account-
ing, should be amended so as to place greater emphasis on expected
future obligations, rather than on accrued obligations to date.
A frequently used criterion for measuring the degree to which a pro-
gram is funded, especially in the case of social insurance programs, is
that of "actuarial balance." It differs from the previous definition, in
that account is taken of the net expected benefits of expected future, as
well as current, participants. A contingency program would be said to be
actuarially balanced if the expected value of future payouts to all current
and future participants equaled the expected present value of the inflows
from all current and future participants, plus the value of any reserve
fund. Since there is not necessarily a balance between current partici-
pants' expected future contributions and benefits, even if a reserve fund
did exist, it would not in general be equivalent to the program's accrued
liability.
10
11
For a further discussion, see McGinn (1980).
Nonetheless, while governments' ability to issue nominal domestic currency
debt is not limited, it has become apparent in recent years that governments can
be insolvent, especially with regard to foreign currency or their "real" obliga-
tions.
12
See von Furstenberg (1979, Chap. I) for a discussion of these issues.
The accumulated benefit is calculated on the basis of current salary and
accumulated service to date; the projected benefit is calculated on the basis of
expected future salaries but accumulated service to date.

Nonetheless, when a program is actuarially balanced, government tax

revenue will not be required to meet its obligations. For example, a
growing population or inflation could provide the resources to finance a
social security system for which current participants' expected benefits
exceed expected contributions. Moreover, while an actuarially balanced
program may suffer a temporary negative cash flow, in turn requiring
financing from the government's general revenues, actuarial balance
implies that these negative cash flows will be transient and will be offset
by future inflows, so that deficits may be offset by borrowing, with
repayments fully covered by the fund's own resources.
A stricter criterion for measuring the degree to which government
programs are funded is actuarial fairness. In this case, a contingency
program would be termed funded if the expected present value of future
payouts to each of a program's current participants equaled the expected
present value of any current and future payments by current participants
(for example, fees or contributions) to the program in addition to the
value of any reserve assets. This calculation would also include consider-
ation of benefits accrued in the future. If this criterion were met, and the
revenues of the program are not added to the general revenues of the
government, a fund would exist that matched the expected value of
participants' accrued benefits, and there should be no need for govern-
ment tax revenue to be raised to meet the program's expenditure.
Unfunded programs will not meet one or both the actuarial balance of
actuarial fairness criteria, and expected receipts may fall short of ex-
pected payouts. The simplest example is that of pay-as-you-go (PAYG)
schemes, in which contribution rates are adjusted periodically as neces-
sary to meet cash outlays.
III. Contingent Liabilities in Conventional Deficit Measures
The treatment of contingencies in conventional fiscal accounting sys-

tems implicitly relies on a view that the macroeconomic impact of such
programs occurs primarily at the future date when the contingent claim
is realized and the cash flow is generated, rather than when the liability is
issued. The implicit assumption is that consumers are myopic with re-
gard to the future benefits that such schemes imply, or face liquidity
constraints that inhibit the adjustment of current consumption to ex-
pected future benefits. In the latter case, an increase in expected future
benefits, while increasing consumers' net wealth, will not permit an

increase in current consumption because capital market imperfections

restrict consumers from borrowing against future income.13
The two most prevalent deficit measures—those prescribed by the
United Nations in A System of National Accounts (SNA) or the Interna-
tional Monetary Fund in A Manual on Government Finance Statistics
(GFS)—index the current period's excess of governments' expenditures
over revenues on an accrual or cash basis, respectively. Both systems
focus on current flows of goods and services rather than on current
policy commitments that may imply future transfers between the public
and private sector. GFS recommends the exclusion of imputed or ac-
crued transactions in accounting for government activity (GFS, p. 2), in
favor of an accounting method based solely on cash transactions. While
conceding the importance of accrued assets and liabilities (for example,
uncollected tax revenues), GFS excludes them, owing to the difficulty in
accounting for them accurately and in a fashion consistent with other
macroeconomic statistics (GFS, p. 108). Thus, with regard to loan guar-
antees, only payments in the event of default are included as an expendi-
ture item. In particular, if these transactions give rise to a claim on the
borrower, payment of principal and interest in the event of default is
classified as net lending; receipts in repayment of defaulted amounts are
included as repayment of a loan to the private sector (GFS, p. 105).14
Similarly, only the cash flows related to social security programs enter
into the government's accounts under the GFS system, and the accrued
liability is ignored (GFS, p. 16).15
SNA's budget methodology differs from that prescribed in GFS, with
implications for the treatment of contingencies. In particular, govern-
ment transactions are included in the fiscal accounts on the basis of
changes in ownership of goods and services and accrued tax liabilities;
sales of assets and net lending are treated as financing items. Therefore,
the recommended treatment of social security flows "in the case of cur-
rent transfers which represent obligations to, or commitments of, organs
of general government is to record the transfer as of the date when they
13
See Bernheim (1989) for an interesting discussion and survey of the litera-
ture regarding the effect of liquidity constraints on the relationship between
deficits
14
and aggregate demand.
If no claim is created against the defaulter, the payment of the interest
obligation is treated as an interest expenditure while repayment of principal is
considered
15
negative financing (GFS, p. 179).
The exceptions to consolidation include provident funds, government em-
ployee funds, and local and regional funds, which act more as savings instru-
ments and more closely resemble their private sector analogues.

are due without penalty" (SNA, p. 128), rather than when the cash flow
is generated. This results, however, in only a modest difference in tim-
ing compared with the methodology described in GFS. Similarly, SNA
recommends accounting for payments upon default of government-
guaranteed loans on an accruals basis—again when the liability is due
with certainty. Since SNA focuses on changes in the government's net
indebtedness (rather than on the policy intent of its transactions), the
settlement of such an obligation would be classified as a financing item.
IV. Contingencies and Alternate Budget Definitions
Conventional deficit measures are likely to be deficient indicators of

fiscal impact except under extreme assumptions regarding private sector
behavior—for example, myopia regarding the future implications of cur-
rent government policy or liquidity constraints. However, if these as-
sumptions are relaxed, and if contingency programs are not operated
subject to the same incentives as the private sector, such programs may
imply a net transfer to participants and may affect economic activity in
advance of any cash outlay. For example, the announcement of ex-
panded future social security benefits could increase current aggregate
demand if some of the tax burden of the future benefits is expected to be
borne by future generations. As households expect that their future
benefits will exceed their future tax liability, their expected net wealth
increases, thereby increasing their demand for current consumption.
Therefore, the macroeconomic impact of both social security and
nonsocial security contingency programs is likely to be closely related to
the degree to which programs are funded. For example, as mentioned
above, if the contingency program operates as a forced savings vehicle,
requiring contributions from participants that exactly match their ex-
pected payouts, the program will be fully funded, and it is unlikely that
the program will have a macroeconomic impact. Similarly, a PAYG
scheme, or one that is less than fully funded, may imply a net transfer
from future generations to current consumers, and therefore induce an
increase in aggregate demand.16 Two extended measures of the fiscal
deficit are described below that address this possibility.
16
Note that GFS also seems to make this distinction in the case of social
security funds, for which benefits are not directly related to contributions and
which are assumed included as part of general government, whereas provi-
dent funds, which maintain the financial integrity of deposits, are excluded
(GFS, p. 15).

The Economic Deficit
As noted above, it is often argued that private sector behavior results

from economic agents' allocation of net wealth over their life cycle. In
this context, it has been suggested that since social welfare programs
offer a well-defined right to (possibly uncertain) benefits in the future,
economic agents may view the payment of social security taxes as the
purchase of an annuity or bond, rather than a compulsory tax, and
may similarly view the payout as repayment of principal and interest
(Kotlikoff (1984, 1986, and 1988); see also the discussion in Mackenzie
(1989)).
In this circumstance, the prescription is to describe the government's
fiscal stance in terms of an "economic deficit," in which social security
tax receipts are reclassified as a financing item, while a portion of bene-
fits are similarly included below the line as a loan repayment. Only the
excess of benefits over payments to each individual is treated as an
interest expenditure.17 Similarly, most expenditures associated with loan
and other guarantees would remain above the line—reclassified as inter-
est payments—while any premia or fees would be placed below the line
and classified as financing items.18
Clearly, however, the extent to which social security, or any other
contingency program, can be viewed as having the same fiscal impact as
any other government financing instrument will depend on the degree to
which the contingent claim represents a tax or a subsidy. For example,
for social security payments to have an economic impact equivalent to
the sale of other financing items, the rate of return on contributions
would have to be the same as on other private sector savings instruments,
adjusted for risk and other relevant factors. To the extent that the return
is less (more) than on market instruments, households must be coerced
into participation, and the program implies a tax (subsidy) in addition to
a loan. A prescription, therefore, would be to include the (possibly
noncash) transfer element of the contingency program in the fiscal ac-
counts when the government issues the contingent claim; this measure
17
As compared to the budget methodologies proposed by both GFS and SNA,
for the purpose of defining a deficit, this system would place below the line all
receipts and payments except those in excess of participants' prior payments.
Note that proponents of this system do not address the issue of decomposing
benefit payments into "principal" and "interest." Presumably, actuarial criteria
could
18
be applied.
Adjustment to the SNA system would be more dramatic since payments on
default of guarantees are already treated as a financing item.

would approximate an index of actuarial fairness described above (see

also Section V).
Proponents of the economic deficit concept have argued that the gov-
ernment's fiscal impact will also depend on the extent to which resources
are transferred across generations. The impact of contingency programs
would then depend on the degree to which current participants perceive
that programs will be financed by future generations. Thus, calculation
of the economic deficit would require detailed estimates of the intergen-
erational incidence of all fiscal activity, a task whose complexity limits its
practical significance.
Government Net Worth

Alternatively, it has been argued that private sector macroeconomic
behavior is best described in terms of a Ricardian consumer—an eco-
nomic agent whose consumption and savings behavior is based on an
extremely long-term assessment of household net wealth. In this case, a
key variable that affects household net wealth, and therefore household
behavior, is the government's net wealth position—the expected present
value of current and future tax revenue less its current net liability to the
private sector. Government policies, such as a planned tax increase, that
increase its net wealth permit greater government consumption of goods
and services while constraining the private sector's ability to finance its
own expenditure. In its most general formulation, government net
wealth will equal the expected present value of all taxes, including the
seigniorage on its nominal debt, plus the net value of current assets,
including natural resources and fixed capital, less the current value of
current liabilities.19
The government's provision of noncash contingency programs—such
as social insurance and loan guarantees—also implies future subsidies or
transfers to the private sector, and, unlike conventional deficit mea-
sures, affects the government's net wealth and therefore the index of
fiscal impact. The government's issuance of loan guarantees will imply
an expectation of a future cash payout (given a probability of default)
and a reduction in the government's net wealth (that is, the government
will be unable to finance the same level of future consumption). Simi-
larly, the expectation of a future transfer will increase household net
wealth. The provision of social security-related insurance (or other such
contingent payment systems) and the expectation of future transfers
19
See Buiter (1983) for a comprehensive discussion of this formulation.

would also reduce the government's net wealth. In either case, the
increase in the government's contingent liability will provide a stimulus
to current private sector consumption.
However, underlying this concept is the view that the infinitely lived
household is the relevant economic unit. If the economy is better de-
scribed by heterogeneous (that is, by age and wealth) households with
limited horizons, then changes in the government's net wealth position
may not be the most appropriate measure of fiscal impact.20 In addition,
it has proven difficult to apply this concept; for example, valuation of
such assets as natural resources, future seigniorage, and future tax rev-
enue is extremely subjective.21 Further, it is likely that the private sec-
tor's expectations regarding its future tax liability is not limited to exist-
ing tax regimes, but includes consideration of government reaction to
future financing requirements.22
V. Measurement of Government Contingent Liabilities
The preceding discussion suggests that it may be impractical to aban-

don conventional deficit measures in favor of comprehensive indices of
fiscal impact that encompass contingent liabilities. Nonetheless, indices
of governments' contingent liability are important. Since current contin-
gent claims may imply future financing requirements, an index of the
government's net liability will allow for proper long-term fiscal budget-
ing. Similarly, since contingency programs may imply the use of scarce
government resources, albeit at a future date, the appropriate design of
public policy will depend on some measure of the tax and subsidy ele-
ment embodied in such programs. Applications of these concepts are
discussed below.
20
See Auerbach and Kotlikoff (1987) for an interesting discussion and appli-
cation of these life-cycle concerns to the dynamics of fiscal policy. The empirical
evidence regarding the economic impact of social security is ambiguous (see
Atkinson (1987)), in some cases rejecting the pure Ricardian prediction that the
private sector would react to the institution of a social welfare program by simply
reducing savings, and in other cases, rejecting the alternate, life-cycle, hy-
pothesis that savings would be only partially depressed, causing consumption to
increase.
21
Examples of applications to specific countries include Boskin (1988) and
Eisner and Pieper (1984) for the United States, and Hills (1984) for the United
Kingdom. None of these studies was able to include consideration of expected
revenues, seigniorage, and net investment, nor was the impact of government
contingencies
22
considered.
Eisner (1984, 1986) is considerably more sanguine regarding the impact of
contingent government liabilities on private sector activity, arguing that if such
unfunded obligations are expected to be met through increased taxes, the net
impact on government net wealth will be zero.

Actuarial Balance
This measure of a government's net contingent liability is closely allied

with the net wealth concept of the overall deficit described above. As
mentioned above, it offers a useful gauge of the future tax liability or the
required reduction in government consumption implied by a given pro-
gram, information that is important both for budget management and
for economic analysis of fiscal impact. To the extent that a program is
unfunded from the perspective of its actuarial balance, the fiscal authori-
ties are provided with an indicator of the inadequacy of contribution
rates or benefits.
The actuarial balance can be calculated by comparing the net present
value of a program's assets to its liabilities:
actuarial balance = E[PF(inflows — outflows — operating expenses)]
+ reserve,
where £[•] is the expectations operator; PV(-) is the present value
operator; and inflows, outflows, expenses, and the reserve are defined in
real terms. Since the underlying rationale for the above exercise is to
examine the net worth of the program from the government's perspec-
tive, it is usually assumed that the appropriate discount rate is the gov-
ernment's opportunity cost—the real after-tax interest rate on govern-
ment bonds.23
An advantage of using the actuarial balance as a measure of the
government's net liability is that, especially in the case of social security
and annuity programs, there are well-established accounting and actuar-
ial practices to facilitate the calculation, especially with regard to the
appropriate expectations regarding, for example, mortality and fecun-
dity.24 However, care must be taken with the application of these actuar-
ial methodologies to the public sector accounts. For example, accounting
practices for private sector pension are usually based on one of two
alternate methodologies: the accumulated-benefit obligation; or the
projected-benefit obligation. The former defines the pension obligation
as based on the current salary and accumulated service to date of current
plan participants, whereas the latter is based on expected future salaries
and accumulated service to date (Selling and Stickney (1986)). Both are
23
See Buiter (1984, p. 32) for a discussion. In addition, although the formula
above is defined in real terms, an alternate specification—using nominal magni-
tudes for cash flows and interest rates—could, and often is, defined to yield the
appropriate nominal magnitude. The nominal formulation does not index the
real balance of the program in question and may therefore be of limited useful-
ness
24
as a measure of true opportunity cost.
For example, see U.S. Department of Health and Human Services (1989)
and the discussion in Ebrill (1990).

myopic with regard to the future service of current and prospective

participants and exclude consideration of new entrants and their pay-
ment of premia or receipt of benefits. For the purpose of determining the
net balance of government contingent liabilities, particularly public sec-
tor social security programs, a more economically relevant measure is
appropriate. Since participation in such programs is mandatory, the
potential future funding requirement of such programs should be de-
fined with reference to the inflow and outgo associated not only with
current but also prospective enrollees (for example, with appropriate
assumptions regarding population and income growth).
However, for other, nonannuity, contingency programs, where partic-
ipation is not universal or mandatory, long-term projections of participa-
tion would be difficult to incorporate in measures of actuarial balance.
For example, the calculation of the balance for deposit insurance pro-
grams would require assumptions regarding the growth of deposits and
new entrants, as well as an estimate of defaults. This difficulty is com-
pounded for those programs in which participation by the private sector
is optional (that is, loan guarantee and trade financing programs), where
an assumption would be required regarding public sector participation
and government policy toward their provision. In these instances, pro-
jections could be made based either on a shorter horizon, or simply with
respect to current participants (see discussion below).
A final issue concerns cash flow deficits. A business strategy for pri-
vate sector insurance companies may involve adjusting the level of re-
serves or premia, for example, to reduce the "probability of ruin" (that
is, the probability that reserves will be more than depleted in any given
period; see Buhlmann (1970)) to some minimum acceptable level. An
actuarially balanced contingency program may face either the risk or the
certainty of a negative cash flow in the near term, which would more
than eliminate any reserve. At such a point, to maintain its commitments
the program would require support from the general government bud-
get. This could present significant difficulties, especially in the case of
contingencies denominated in foreign currency, or for countries with
limited access to domestic capital markets or limited tax bases. Thus,
rather than targeting so that contingencies are actuarially balanced on an
expected value basis, a concern for the risks associated with the program
may dictate a more conservative strategy.
Market Value or Subsidy Measures

The actuarial balance of a program may be inadequate to gauge the
subsidy or transfer to current participants of contingency programs, and
thus of their short-run fiscal impact. Four theoretically equivalent in-

dices that address this concern, and which correspond more closely to
the paradigm underlying the concept of the economic deficit discussed in
Section IV, are described below.
Actuarial Fairness
A simple alternative to the net worth or actuarial balance definition
discussed above would be to calculate a program's index of actuarial
fairness. This would require the derivation of an index similar to that for
a program's actuarial balance, except that it would only include consid-
eration of the net transfer to current participants—the expected present
value of their current and future contributions less their current and
future benefits, plus the current value of any reserve fund.
Since the index is intended to measure the change in current partici-
pants' net wealth, rather than that of the government, the appropriate
discount rate may differ from that used to calculate net wealth from the
government's perspective, owing to tax distortions, the existence of ex-
ternalities (for example, the private sector may undervalue the impact of
its investment on future generations), or risk. In principle, the appropri-
ate discount rate would be the private sector's opportunity cost of the
cash flow associated with the contingency.25
The calculation of such an index would be relatively simple if proce-
dures were already in place for calculation of the actuarial balance (for
example, as is done for the U.S. social security system). However, if the
actuarial exercise is not already routinely performed, the cost of de-
veloping the actuarial model—that is, the assumptions regarding the
probabilities of default or illness, for example, as well as future eco-
nomic scenarios—and of performing the calculations may be prohibitive.
"Market Value" Measures

An alternative approach is to use current market information to derive
a market value of the contingency. For example, in the context of a loan
guarantee, the market value of a program would simply be the difference
between the rate that the borrower pays versus that rate paid in the
absence of the guarantee.26 A nominal measure of the subsidy per period
would be (ip — /g)L, where ip and ig are the annual rates of nominal
interest on private sector unguaranteed and guaranteed loans, respec-
tively, and L is the loan principal. The implicit subsidy, or the value of
25
The choice of the appropriate discount rate is discussed in the context of
cost-benefit
26
analysis by Boadway and Wildasin (1984).
For a brief description of the application to credit subsidies and guarantees,
see Wattleworth (1988) and U.S. Congress (1989).

the guarantee (G), over the life of the loan («), is simply the net present
value of associated cash flows:
where the subsidy value would be reduced by the amount of initiation

fees or other charges that the recipient is required to pay.27 The value of
a government guarantee of an annuity contract, in which interest and
principal are repaid by means of a fixed cash payment over the contract
period, can be derived in a similar fashion. For example, suppose private
sector credit is available such that a contract for an n-year loan of L
dollars requires an annual payment of cp dollars. Given the effective
interest rate of ip, the annual payment will satisfy
The analogous rate and debt-service payments (i8 and c8, respectively)
on a government guaranteed loan for the same initial amount L satisfy
The periodic subsidy to the holder of the guaranteed loan can be mea-
sured by the difference between the annual cash flows between the two
loan contracts:
where
The full value of the guarantee is the present discounted value of the
cash flow defined above, which in this case can be shown to equal
While the foregoing discussion has been in terms of loan guarantees,

the same methodology could be applied to the full range of government
contingent liabilities. For example, the subsidy associated with social
security could be defined as the difference between the premia paid to
27
Note that it is assumed for the sake of simplicity that the appropriate dis-
count rate, from the perspective of the recipient of the guarantee, is the yield in
the private and unguaranteed loan market; that is, the rate that is the opportu-
nity cost to the recipient.

the government program and the amount that the private sector would
have charged for the same or similar insurance contract. The net present
value of any subsidy could be calculated accordingly. Exchange rate
guarantees would be similarly valued as the cost of purchasing a forward
exchange contract with the same features as provided by the government.
In some cases, however, it may be difficult to determine the rate of
interest that would have been paid in the absence of the guarantee. A
representative sample of a guaranteed contract with similar risk charac-
teristics may not be available, especially if the program in question has
been initiated to resolve a market failure, or it has supplanted a private
sector market. Moreover, even if a market rate is observable, it may
understate the value of the guarantee, since the exit of the guaranteed
borrowers may reduce the pressure on private sector rates. Finally, by
guaranteeing private sector credit, the government also assigns many of
the characteristics of government debt, both with regard to risk and
transactions costs (since the guarantee may be associated with govern-
ment management of the secondary market for the guaranteed debt). If
the government is faced with anything but a perfectly elastic demand for
its debt, the effect may be to increase its own cost of borrowing, in turn
implying an additional indirect cost of the contingency program that
would be difficult to quantify.
Option-Pricing Approach
An equivalent approach to measuring the subsidy associated with
contingent claims is suggested by recent advances in option-pricing
theory. In financial markets, a call (put) option is defined as the right to
buy (sell) a prespecified quantity of a financial instrument (or commod-
ity) at a prespecified price on or before a prespecified date.28 The pur-
chaser of an option will exercise the right to purchase the underlying
instrument if the market price exceeds the specified exercise price.
Merton (1977) has observed that government provision of a loan guaran-
tee or deposit insurance may also be viewed as the provision of an
option. If the underlying value of the credit instrument falls below a
given level, the borrower is in default and will exercise the option for the
government guarantee. With modern option-pricing theory, an exact
pricing formula (initially derived by Black and Scholes (1972)) for the
implicit market value of such guarantees can be derived as a function of
readily available market data.
For example, consider the government guarantee of a loan B. If, at the
28
A European put option may only be exercised at the expiration date; an
American put option may be exercised at any time up to the expiration date.

maturity date of the loan the value of the assets of borrower V on which
the guarantee was issued exceeds B, then there is no default. However, if
the value of the assets is less than the value of the loan, then the guaran-
tee implies that the government must pay B — V; that is, the difference
between the loan value and the surrender value of the assets. The im-
plicit value of the guarantee at the maturity date of the loan (time T) is
Under the assumption of frictionless markets and that the return

process to the underlying assets evolves according to a continuous
stochastic process, Black and Scholes demonstrate the exact pricing
formula for the option.29 For example, the value of the loan guarantee
(at the initial date t) is
where
and <J> is the cumulative normal density function, o~2 is the variance rate of
the logarithmic changes in the value of V, and r is the "risk-free" rate of
interest.
The formula above is not general; it must be amended according to the
terms of the guarantee arrangement and the underlying assumptions
regarding the market structure and the type of contingent claim.30 While
this approach could be applied with some modification to the measure-
ment of the value of social security programs, it is most relevant to
financial market insurance programs. However, it is not clear that this
measure would be any simpler to calculate than those described above.
Moreover, the underlying assumption of the option-pricing model—that
of frictionless markets in which asset prices, including those on options,
adjust so as to eliminate the risk of the market portfolio—may invalidate
its use in many cases, especially those in which markets are insufficiently
deep to permit the creation of the perfect hedge assumed.
29
The frictionless market assumption requires no transaction costs, continuous
trading, unrestricted borrowing and lending at identical rates, and unrestricted
short
30
sales.
For example, Jones and Mason (1981) extend these results by examining a
richer array of guarantee arrangements. By relaxing certain assumptions regard-
ing the payment of interest, Merton (1977) derives a similar formula for the case
of deposit insurance, which is extended by Pennacchi (1987). Borensztein and
Pennacchi (1990) estimate the value of interest guarantees on developing coun-
try debt.

Welfare Measures
The above subsidy measures are explicitly based on the calculation of
the financial worth of the contingency program. An alternative is to
consider the welfare implications of the extension of a contingent claim
to the private sector.31 This would involve an examination of the change
to the consumer and producer surpluses, as defined by the areas under
the private sector's demand and supply curves for the "insurance" in
question that resulted from the government's intervention.32 However,
this approach is poorly suited to a consideration of the second-round
effects on the demand for and supply of credit of the government policy,
which might result from substitutions from other markets or from the
dynamic consequences of default on welfare.33 Moreover, although this
type of approach may be useful for gauging the cost or benefit of specific
government programs, it is less relevant for government budgeting. This
is especially apparent given the obvious difficulties associated with mea-
suring the parameters underlying the demand for and supply of credit, as
well as the well-known methodological problems associated with mea-
suring consumer surplus (see Auerbach (1985) for a discussion).
VI. Budgetary Controls
As noted above, conventional budget methodologies and deficit mea-

sures tend to ignore the future cash flow implications of government
contingent liabilities. As a result, the cost of providing contingencies
may be inadequately accounted for at the time of their provision. More-
over, the inadequacy of conventional budget methodologies may imply
an incentive for the fiscal authorities to substitute away from cash to
noncash activities as a means of circumventing constraints on the overall
cash-based deficit or expenditure.34 While it may not be feasible to adopt
the comprehensive deficit measures discussed in Section IV, a more
piecemeal approach, incorporating the concepts described in the previ-
ous section, may provide the necessary budgetary control.
31
32
See Wattleworth (1988) for an application to explicit credit subsidies.
The area under the demand curve above the price paid is termed the con-
sumer surplus; that is, the difference between what the consumer would have
been able and willing to pay and the price paid. Under certain restrictions (see
Auerbach (1985) for a discussion), this surplus can be thought of as the monetary
equivalent of the consumers' utility or welfare.
^3For an examination of these issues in the context of an overlapping genera-
tions
34
growth model, see Towe (1989).
This point has been made in the context of the United States in U.S.
Congress (1989).

Ad Hoc Constraints
One approach is to place an ad hoc limit on the level of the govern-

ment's liability. For example, preparation of the government's annual
budget could include a forecast of the desired increase in total loans on
which a government guarantee applies and the imposition of explicit
constraints on agencies' ability to exceed these limits during the year.
This method may be useful in the case of programs where measurement
of the subsidy component is difficult, or where technical expertise is
unavailable. Thus, this approach is most often used with loan and other
guarantee programs, rather than for social insurance programs where
there exist relatively well-established actuarial methodologies. More-
over, such constraints ignore the value of contingencies and do not per-
mit proper accounting of their cost or the implied subsidy; therefore, a
comparison with other programs is impossible.35
Given that the fiscal authorities' resources are limited, proper bud-
getary planning will require a measure of the subsidy transfer embodied
in contingency programs. Three possibilities are discussed below.
Divestiture
A straightforward method of establishing the current cost from the

provision of contingent liabilities is for the government to divest itself of
the liability. This goal may be accomplished by purchasing offsetting
insurance from the private sector, or by a voucher system in which
borrowers are provided vouchers to purchase either explicit insurance or
to provide lenders compensation for default risk.36 On the date the
contingent claim is offered, a cash expenditure would therefore be re-
quired by the government representing the explicit subsidy or transfer.
As a result, the future contingent cash obligation would be eliminated in
favor of a current cash flow, which should approximate the actuarial
fairness index described above.37
35
36
These points are also made in U.S. Congress (1989).
These possibilities have both been suggested for the U.S. Congress (see U.S.
Congress
37
(1989) for a discussion).
Note that this type of scheme is substantially different from placing the cash
flow associated with contingencies on an extrabudgetary basis. This latter ap-
proach, while avoiding funding of other expenditure from temporary cash sur-
pluses from contingencies, does not address the issue of how future deficits are to
be funded. Moreover, the liabilities of the extrabudgetary agency administer-
ing the contingency program will certainly be implicitly guaranteed by the cen-
tral government. Finally, since such extrabudgetary agencies are performing
governmental functions, GFS recommends consolidation with the government's
accounts.

Although this proposal has been primarily associated with loan guar-
antees, it could be applied to social welfare and other government insur-
ance programs. Its advantage is its transparency; the cost of the contin-
gency program is made explicit at the point the contingent claim is
issued. Moreover, by eliminating the ongoing administrative burden of
such programs, a net savings to the budget may arise. Finally, the advan-
tage of this approach is that the valuation of the contingent claim is
provided by the market. However, as a result, it may only be relevant to
those types of contingencies for which a private market would normally
be viable. First, it would be difficult, for example, to apply a voucher
system when the government's intervention was originally intended to
alleviate the market's inability to provide the socially desirable level of
insurance. This will be especially true in the case of social welfare sys-
tems. Second, it is possible that by pooling risks, the public sector would
be the least-cost provider of insurance. By shifting the risk to the private
sector, budgetary costs may be increased. Thus, while this approach has
merit, it may be less useful in cases where the government's original
purpose in providing the contingency was to correct a perceived in-
adequacy of the private sector, on either economic efficiency or equity
grounds.
Redefining the Deficit

A proposal closely allied with the concept of the economic deficit
discussed above is to redefine government expenditure and revenue
concepts to take explicit account of the subsidy component of contin-
gency programs. Accordingly, the cash flows associated with the admin-
istration of contingency programs would be dichotomized; instead of
including all cash receipts and disbursements as revenue and expendi-
ture, respectively, only the net implicit subsidy resulting from the change
in contingent liabilities outstanding would be included (as an expendi-
ture). The difference between the calculated net subsidy and the pro-
gram's total net receipts would be reclassified as a financing item.38
Annual subsidy outlays would be calculated with reference to the mar-
ket value of contingencies issued over a given period.39 At the beginning
of a given budgetary period, agencies responsible for issuing contingen-
cies would be assigned a constraint on their ability to issue additional
38
This is essentially the recent proposal of the U.S. Congressional Budget
Office
39
to address the issue of credit reform (U.S. Congress (1989)).
See U.S. Congress (1989, p. 43) for an example of the derivation of this
amount.

contingencies, based on their subsidy value rather than the additional

gross liability of the government. In this manner, the budget authorities
would be provided with the appropriate trade-off between expenditure
allocations, since the longer-run financing implications of contingencies
would be accounted for. Moreover, as the budget deficit would be calcu-
lated on the basis of the subsidy component rather than the current cash
flows associated with contingencies, it could more accurately represent
the government's fiscal impact on the macroeconomy.
Note that the subsidy under this scheme would be (broadly speaking)
defined in terms of the expected net present value transferred to the
current recipients of the newly issued contingent claim—that is, the
amount that would render the contingency actuarially fair. Moreover, it
would be implicitly assumed that the current fiscal impact of any current
policy that implied a subsidy to future participants would be nil. There-
fore, the resultant measure of the overall fiscal deficit would be implicitly
similar to the economic deficit discussed above.
This approach would likely be most effective when applied to pro-
grams for which administering agencies have significant discretion re-
garding the amount and value of claims issued. In such cases, a con-
straint on the subsidy value of contingent claims issued during a given
budgetary period may ensure that the issuing agency either restricts the
issue of claims or adjusts the terms under which the claim is issued to
reduce the subsidy component. However, the relevance of this method
for imposing fiscal discipline in the case of contingency programs for
which participation is mandatory and/or the terms of issuance are not
under the administering agency's control—for example, social insurance
programs—is less clear.
Funding
An alternative approach is to require full funding of the subsidy com-
ponent of contingencies—that is, agencies issuing contingent liabilities
would be required to purchase assets equal to the value of the contingent
liabilities issued in each budgetary period. Thus, instead of the some-
what artificial constraint that would be imposed under the scheme de-
scribed above, issuing agencies would face a cash constraint. With this
method, an actual cash expenditure is created that matches the change in
the government's liability, so that the overall deficit immediately in-
cludes the impact of contingencies.
However, in this case the type of asset used to match the government's
liability will be of concern. If the agency and its fund are consolidated

with the government, and government instruments are the investment

vehicles for the funding exercise, then the government's overall deficit,
and its net liability position, may continue to be understated.40 The
funding requirement may provide the necessary institutional constraints
to monitor and control the issue of contingencies appropriately, but the
government's future financing requirement will be unaffected, and the
overall fiscal deficit will continue to account inadequately for the trans-
fers associated with contingency programs.
The alternative (which has been proposed for the U.S. social security
surplus) would be to make government debt instruments ineligible for
funding of contingencies. Other things being equal, this will result in the
government's purchase of private sector debt or equity in an amount
equal to the value of the contingencies issued each budgetary period. If
and when contingent obligation became due, these assets would be sold
to finance the required expenditure. Thus, under the GFS and similar
systems, where such purchases are treated as an expenditure rather than
financing, the government's overall deficit would be increased at the
date of issuance of any contingency, but would be unaffected when the
contingency became due.
However, while constraining contingency programs in this fashion
may be necessary where budgetary resolve and control are weak, it may
be argued that the fiscal authorities' ability to finance their other activi-
ties would be unnecessarily constrained. Just as it will be occasionally
desirable to finance a government's purchases of goods and services
through domestic borrowing, it also may be desirable to issue contingent
liabilities without cash expenditure until a future date. This will be
especially relevant for countries whose capital markets are less well
developed. In such cases, the government's ability to finance the initial
cash outlay may be limited. Moreover, it is possible that this activity
could have undesirable macroeconomic consequences by altering signifi-
cantly the private sector's portfolio share of private and public debt
instruments. Finally, requiring the purchase of nongovernment assets for
investment of reserve funds would not by itself induce the desired bud-
getary discipline. It will be just as important to adopt measures that
prevent excessive issuance of liabilities.
Thus, the issue of the composition of governments' reserve or contin-
gency funds seems to hinge more on budgetary control of both the
issuance of contingent liabilities and other expenditure, rather than ap-
40
Similarly, GFS distinguishes between "funded government employee pen-
sion plans invested in the capital market or in loans and securities other than
those of the employing government" (p. 16) and other funded plans.

propriate funding techniques. If budgetary control is weak, the addi-

tional resources made available from contingency programs may provide
an incentive to expand expenditure in excess of prudent levels. How-
ever, if budgetary control is strong, and the fiscal authorities are pro-
vided with the appropriate measure of the noncash outlays implied by
contingency programs, then the content of the reserve fund will be less
relevant.
VII. Concluding Remarks
It is argued above that conventional budget methodologies ignore the

issue of contingent liabilities, except when a cash flow is created. Thus,
while conventional deficit measures will (subject to a number of caveats)
accurately describe the change in the government's nominal liabilities
resulting from the need to finance cash expenditures, the change in its
liability from noncash policies—that is, the extension of contingent
claims—will generally be ignored. As a result, fiscal accounting systems
will provide insufficient data for adequate budgetary control over such
policies. Moreover, constraints on conventionally defined levels of ex-
penditures and the deficit may have the unintended effect of creating the
incentive to substitute noncash expenditure through the issue of loan
guarantees or other means. As a result, conventional budget methodolo-
gies may lead to improper analysis of the trade-offs between current cash
expenditure policies and the issuance of contingencies.
In addition, the design of macroeconomic policy will depend on an
appropriate measure of the macroeconomic impact of the government's
fiscal activities. While there is substantial controversy regarding the
impact of such contingency programs as social welfare, deposit in-
surance, and loan guarantees, it is likely to be significant. Thus, insofar
as conventional measures of fiscal impact ignore these noncash fiscal
activities, they may underrepresent the government's effect on the
macroeconomy.
Extended deficit measures can be defined to gauge the government's
fiscal impact. Deficits can be defined that measure the intergenerational
transfers implied by contingency programs (and other government poli-
cies), or that sum government activity, including contingencies, over an
infinite horizon. The choice will depend on the relevant planning hori-
zon of the budget authority and the private sector. However, either alter-
native would likely be impractical given the data requirements, as well as
requiring a choice between polar views regarding the determinants and
the horizon relevant for private sector consumption and savings.

Nonetheless, the value of government contingencies should be mea-

sured, especially so that adequate budgetary controls may be applied.
Two alternatives were proposed here that correspond closely to the
underlying focus of the extended deficit measures: actuarial balance,
which would represent the liability from the government's (long-term)
perspective; and actuarial fairness or a contingency's subsidy value,
which measures the transfer to current participants. In the latter case, a
number of alternative measurement strategies can be defined, with the
choice depending on the type of contingency in question and the data
available. These measures can be used to form the basis of the appropri-
ate budgetary control over the government's provision of contingencies,
as well as the analytic device for gauging their impact.
REFERENCES
Atkinson, Anthony B., "Income Maintenance and Social Insurance," in Hand-

book of Public Economics, ed. by Alan J. Auerbach and Martin Feldstein
(New York: North-Holland, 1987).
Auerbach, Alan J., "The Theory of Excess Burden and Optimal Taxation,"
Chap. 2 in Handbook of Public Economics, ed. by Alan J. Auerbach and
Martin Feldstein (Amsterdam: North-Holland, 1985).
, and Laurence J. Kotlikoff, Dynamic Fiscal Policy (New York: Cam-
bridge University Press, 1987).
Bernheim, B. Douglas, "A Neoclassical Perspective on Budget Deficits," Jour-
nal of Economic Perspectives, Vol. 3 (Spring 1989), pp. 55-72.
Black, Fischer, and Myron Scholes, "The Valuation of Option Contracts and a
Test of Market Efficiency," Journal of Finance, Vol. 27 (May 1972).
Boadway, Robin W., and David E. Wildasin, Public Sector Economics (Boston:
Little, Brown, 1984).
Borensztein, Eduardo, and George Pennacchi, "Valuation of Interest Payment
Guarantees on Developing Country Debt," Staff Papers, International
Monetary Fund, Vol. 37 (December 1990), pp. 806-24.
Boskin, Michael J., "Concepts and Measures of Federal Deficits and Debt and
Their Impact on Economic Activity," in The Economics of Public Debt, ed.
by Kenneth J. Arrow and Michael J. Boskin (New York: St. Martin's Press,
1988).
Bosworth, Barry P., Andrew S. Carron, and Elisabeth H. Rhyne, The Econom-
ics of Federal Credit Programs (Washington: The Brookings Institution,
1987).
Brau, Eduard H., and Chanpen Puckahtikom, Export Credit Cover Policies and
Payments Difficulties, Occasional Paper 37 (Washington: International
Monetary Fund, 1985).
Buhlmann, Horst, Mathematical Methods in Risk Theory (Heidelberg: Springer-
Verlag-Berlin, 1970).

Buiter, Willem H., "Measurement of the Public Sector Deficit and Its Implica-
tions for Policy Evaluation and Design," Staff Papers, International Mone-
tary Fund, Vol. 30 (June 1983), pp. 306-49.
, "Measuring Aspects of Fiscal and Financial Policy," NBER Working
Paper 1332 (Cambridge, Massachusetts: National Bureau of Economic Re-
search, 1984).
Chancy, Paul K., and Anjan V. Thakor, "Incentive Effects of Benevolent Inter-
vention: The Case of Government Loan Guarantees," Journal of Public
Economics, Vol. 26 (March 1985), pp. 169-89.
Diamond, Jack, and Christian Schiller, "Government Arrears in Fiscal Adjust-
ment Programs," in Measurement of Fiscal Impact: Methodological Issues,
Occasional Paper 59 (Washington: International Monetary Fund, 1988).
Ebrill, Liam P., "Social Security, Demographic Trends, and the Federal Bud-
get," IMF Working Paper 90/14 (Washington: International Monetary Fund,
1990).
Eisner, Robert, "Which Budget Deficit? Some Issues of Measurement and Their
Implications," American Economic Review, Vol. 74 (May 1984), pp.
138-43.
, How Real is the Federal Deficit? (New York: The Free Press, 1986).
, and Paul J. Pieper, "A New View of the Federal Debt and Budget
Deficits," American Economic Review, Vol. 74 (March 1984), pp. 11-22.
Heller, Peter S., Richard D. Haas, and Ahsan S. Mansur, A Review of the Fiscal
Impulse Measure, Occasional Paper 44 (Washington: International Mone-
tary Fund, 1986).
Hills, John, "Public Assets and Liabilities and the Presentation of Budgetary
Policy," in Public Finance in Perspective, Report Series No. 8 (London:
Institute for Fiscal Studies, 1984).
International Monetary Fund, A Manual on Government Finance Statistics
(Washington: International Monetary Fund, 1986).
Ippolito, Dennis S., Hidden Spending: The Politics of Federal Credit Programs
(Chapel Hill: University of North Carolina Press, 1984).
Jones, E.P., and S.P. Mason, "Valuation of Loan Guarantees," in U.S. Con-
gressional Budget Office, Conference on the Economics of Federal Credit
Activity, Part II: Papers (Washington: Government Printing Office, Sep-
tember 1981).
Kotlikoff, Laurence J., "Economic Impact of Deficit Financing," Staff Papers,
International Monetary Fund, Vol. 31 (September 1984), pp. 549-82.
, "Deficit Delusion," The Public Interest (Summer 1986), pp. 53-65.
, "The Deficit Is not a Well-Defined Measure of Fiscal Policy," Science,
Vol. 241 (August 1988), pp. 791-95.
Levitsky, Jacob, and Ranga N. Prasad, Credit Guarantee Schemes for Small and
Medium Enterprises, World Bank Technical Paper 58 (Washington: The
World Bank, 1985).
Mackenzie, George A., "Are All Summary Indicators of the Stance of Fiscal
Policy Misleading?" Staff Papers, International Monetary Fund, Vol. 36
(December 1989), pp. 743-70.

Mayshar, Joran, "Should Government Subsidize Risky Private Projects?"

American Economic Review, Vol. 67 (March 1984), pp. 20-28.
McGinn, Daniel F., Pension Funding: Actuarial Primer for Corporate Manage-
ment (Chicago: Charles D. Spencer & Associates, 1980).
Merton, Robert C., "An Analytic Derivation of the Cost of Deposit Insurance
and Loan Guarantees," Journal of Banking and Finance, Vol. 1 (June
1977), pp. 3-11.
Pennacchi, George C., "A Reexamination of the Over- (or Under-) Pricing of
Deposit Insurance," Journal of Money, Credit, and Banking, Vol. 19 (Au-
gust 1987), pp. 340-60.
Selling, Thomas I., and Clyde P. Stickney, "Pension Accounting and Future
Cash Flows," Journal of Accounting and Public Policy, Vol. 5 (Winter
1986), pp. 267-85.
Tanzi, Vito, Mario I. Blejer, and Mario O. Teijeiro, "The Effects of Inflation on
the Measurement of Fiscal Deficits," in Measurement of Fiscal Impact:
Methodological Issues, Occasional Paper 59 (Washington: International
Monetary Fund, 1988).
Towe, Christopher M., "Optimal Fiscal Policy and Government Provision of
Contingent Liabilities: The Example of Government Loan and Deposit
Guarantees," IMF Working Paper 89/84 (Washington: International Mone-
tary Fund, 1989).
United Nations, A System of National Accounts (New York: United Nations,
1968).
U.S. Congress, Congressional Budget Office, Credit Reform: Comparable Bud-
get Costs for Cash and Credit (Washington: Government Printing Office,
1989).
U.S. Department of Health and Human Services, "Social Security Programs
Throughout the World—1987," Research Report No. 61 (Washington:
Government Printing Office, 1987).
U.S. Office of Management and Budget, Special Analyses: Budget of the United
States Government, Fiscal Year 1990 (Washington: Government Printing
Office, 1989).
U.S. Social Security Administration, "Short-Range Actuarial Projections of the
Old-Age, Survivors, and Disability Insurance Program, 1988," Actuarial
Study No. 103 (Baltimore: Department of Health and Human Resources,
1989).
von Furstenberg, George M., ed., Social Security versus Private Saving (Cam-
bridge, Massachusetts: Ballinger, 1979).
Wattleworth, Michael, "Credit Subsidies in Budgetary Lending: Computation,
Effects, and Fiscal Implications," in Measurement of Fiscal Impact: Method-
ological Issues, Occasional Paper 59 (Washington: International Monetary
Fund, 1988).

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
An Inflation-Proof Tax System?

Some Lessons from Israel
EFRAIM SADKA*
When the effect of high inflation on the tax system is taken into account,
the overall revenues from inflationary finance may well be negative. The
strength of this contention is weighed against measures taken in Israel in
an attempt to construct an inflation-proof tax system. The paper concludes
that, despite these measures, the Israeli experience suggests that it is more
appropriate to talk about the f< inflation subsidy," rather than the " infla-
tion tax." [3EL 130, 320]
ECOMISTS HAVE long been interested in the so-called inflation tax

or the real depreciation of money holdings. In a seminal paper,
Bailey (1956) studied the welfare cost of inflationary finance, and Fried-
man (1969) investigated the optimal inflation tax and concluded that it
should be negative. These early studies tended to ignore the fact that in
the real world the alternatives to an inflation tax are not nondistortion-
ary lump-sum taxes, but rather some other distortionary taxes. There-
fore, the inflation tax should be considered in a second-best framework.
Later, Phelps (1973) and Helpman and Sadka (1979) investigated the
optimality of inflationary finance in a second-best context, employing an
optimal taxation approach.
These studies considered only one aspect of inflationary finance: the
real depreciation of money holdings or the revenues from printing
money (seigniorage). These revenues tend to be relatively small as a
percentage of gross domestic product (GDP). For instance, in Israel such
revenues have averaged about 2 percent of GDP, with an inflation rate
* Efraim Sadka is a Professor of Economics at Tel Aviv University, Tel Aviv,

Israel. This paper was written while the author was a Visiting Scholar in the
Fiscal Affairs Department of the Fund. The author has benefited from discus-
sions with Vito Tanzi and Milka Casanegra de Jantscher.
135

136 EFRAIM SADKA
ranging widely between 40 percent and 500 percent a year (see, for in-
stance, Eckstein and Leiderman (1989)).1 Furthermore, as the (perfectly
foreseen) inflation rate rises, real money holdings decline; and with a
very high inflation rate a country may well find itself on "the wrong side
of the Laffer curve."
Tanzi (1977,1978) identified another much more important and prac-
tical aspect of inflationary finance: the effect of inflation on the tax
system.2 According to Tanzi, real tax revenue is eroded by inflation
owing to the collection lag, which is "the time that elapses between a
taxable event (that is, earning of income, sales of a commodity) and the
time when the tax payment related to that taxable event is received by
the government" (Tanzi (1978, p. 419, fn 7)). In fact, the overall revenue
from inflationary finance may well be negative.
The collection lag can be shortened to lessen the effect of inflation on
the tax system, but such measures are not themselves without costs. For
example, the filing period for the value-added tax (VAT) in Israel was
shortened from three months to one month when inflation reached the
triple-digit plateau. Accordingly, during the period 1979-85, businesses
had to file a VAT return and pay the tax collected to the Government
every month, thereby increasing both their bookkeeping costs and the
Government's collection costs. This is another real cost of inflation that
should be added to the lists compiled by Fischer and Modigliani (1978)
and Fischer (1981).
This paper emphasizes yet another aspect of the effect of inflation on
the tax system: the definition of income. The traditional approach fol-
lowed in macroeconomic textbooks is that with a progressive individual
income tax (that is, the average tax rate increases with income), nominal
income, which rises in proportion to the rate of inflation, causes nominal
tax liability to rise more than in proportion to nominal income and
inflation. Hence, real tax revenues increase and the progressive individ-
ual income tax serves as an automatic stabilizer. However, with full, and
almost instantaneous, indexation of the income tax brackets, individual
real tax liabilities no longer rise with inflation. Therefore, the effect of
inflation rests primarily on the definition of business income. Since
business income is defined according to nominal accounting standards,
high inflation virtually destroys the income tax base in the business
sector.
Many attempts have been made in the past to correct tax laws for the
1
See also Fischer (1982) for a study of other countries where the inflation tax
revenues seem a bit higher.
2
Tanzi (1977) credits Oliyera (1967) for first contemplating the possibility that
inflation may lead to a decline in real tax revenues.

INFLATION-PROOF TAX SYSTEM? 137
effects of inflation so as to re-establish a valid income tax base in the

business sector during periods of inflation.3 As the early (1975-82) Is-
raeli experience suggests, these attempts were partial and not even-
handed. Lawmakers were under pressure, first of all, to remove those
effects of inflation that hurt taxpayers (for example, the taxation of infla-
tionary capital gains). Only much later did they come to deal with the
effects of inflation that played into the hands of taxpayers (for example,
the tax deductibility of nominal interest payments). These "one-side-of-
the-balance-sheet" adjustments tended to worsen the detrimental effect
of inflation on the tax system (see also Kay (1977)).
It took several years of high inflation before a law providing compre-
hensive adjustment to the effects of inflation on business income was
enacted in Israel in 1982. Like any tax law, it contained several serious
loopholes, some of which were later closed. Also, it applied mainly to
corporations and neglected most proprietorships. In the event, the effec-
tiveness of the "Israeli solution" was never put to the test because in
mid-1985, the Government introduced its stabilization program, which
brought inflation down very rapidly (to a 15-20 percent range).
It is nevertheless true that during the period of high inflation (1981-
85), the tax on wage earners accounted for an unusually high fraction
(about 65 percent) of all income tax revenues. Today, it accounts for only
40 percent. Yet, when the rapid deceleration in inflation in 1985 suc-
ceeded in erasing the Government's budget deficit, a significant contri-
bution came from the automatic increase in real tax revenues caused by
the sharp decline in inflation. Should not economists therefore abandon
the traditional concept of the "inflation tax" in favor of the more realistic
"inflation subsidy"?
The paper is organized as follows. Section I describes the main effects
of inflation on business income. Section II summarizes early partial
adjustments for inflation in the tax laws. Section III analyzes the main
elements of the 1982 tax law that was supposed to provide a comprehen-
sive adjustment for inflation in the definition of business income, and
Section IV describes the major loopholes and exceptions in that law.
Section V discusses the importance of withholding when income tax
brackets are indexed and change within the same tax year. Section VI
discusses some alternatives to income taxation (for example, consump-
tion or cash flow taxation) during inflationary periods. The Appendix
provides an alternative to the derivation of real income as presented in
Section III.
Throughout this paper, inflation is assumed to take the form of an
3
See Casanegra de Jantscher (1976) for a description of the early Latin Ameri-
can experience.

138 EFRAIM SADKA
equiproportional increase in all prices, so that relative prices do not

change. This assumption makes it possible to abstract from the question
of which price index to employ in converting nominal values into real
values.
I. Effects of Inflation on Taxable Business Income
Taxable income in the business sector is calculated according to stan-

dard accounting procedures, which are nominal in nature. In other
words, 1 sheqel (the Israeli currency unit) is treated as 1 sheqel regard-
less of the date on which it was paid or received. Nominal business
income (or profit) so calculated, which is the difference between reve-
nues (or sales) and costs, is calculated by adding together sheqalim
received at different dates (and having different real values) and sub-
tracting from them sheqalim paid at different dates and having different
real values. When the inflation rates are in the range of 100-500 percent
a year, a beginning-of-the-year sheqel may be worth, in real terms, as
much as 2 to 6 end-of-the-year sheqalim. As a result, nominal income
cannot even serve as an approximation of the true, real income of a
business firm in periods of high inflation rates, such as those Israel
experienced during the late 1970s and the first half of the 1980s.
Inflation creates several deviations of the nominal income from true,
real income. Some of these deviations or biases are negative and some
are positive, but they do not offset each other. Furthermore, as I shall
explain below, their incidence and magnitude are not independent of the
behavior of the taxpayer. In other words, the taxpayer may take certain
actions that reduce the calculated nominal income even though the real
income does not change. In such a case, a higher inflation rate reduces
rather than increases real tax revenues; and the tax system fails to serve
as an automatic stabilizer.
The deviations or biases of real income from nominal income that are
caused by inflation may be classified into five main categories as follows.
(1) Nominal capital gains on an asset have two components: an artifi-
cial or inflationary component that merely reflects an increase in the
general price level of all goods and services; and a true real component
that reflects the portion of the appreciation in the value of the asset that
is over and above the increase in the general price level. Thus, nominal
income overstates real income by the sum of the inflationary component
of capital gains. For later reference, it is worth pointing out that capital
gains are normally taxed upon realization, rather than on an accrual
basis, so that the inflationary component of the capital gains is taxed
only when the asset is sold or otherwise disposed of.

(2) Analogous to the distinction between the inflationary and real

components of nominal capital gains is the distinction between the infla-
tionary and real components of the interest rate. Thus, allowing deduct-
ibility of nominal interest accumulations causes nominal income to un-
derstate real income by the sum of the inflationary component of the
interest accumulations.
At first glance, one may argue that (1) and (2) above offset each other.
On the one hand, inflationary capital gains on an asset are included in
taxable income but, on the other hand, the inflationary interest charges
incurred for the purpose of acquiring the asset are tax deductible. This
argument is invalid on two grounds. First, the purchase of an asset may
be financed by equity rather than by debt. Second, capital gains are
taxed upon realization, whereas interest is deductible on an accrual
basis. Suppose, for instance, that a firm takes out a fully indexed loan of
100 sheqalim to be repaid after five years in order to purchase a certain
asset. If the annual inflation rate is 100 percent, then the firm will be
allowed to deduct from taxable income an indexation differential (that
is, the inflationary component of the interest rate) of 100 sheqalim in the
first year, 200 sheqalim in the second year, 400 sheqalim in the third
year, and so on, even though these differentials were not actually paid
before the end of the fifth year.4 The inflationary capital gains on the
asset purchased, however, will not be taxed until the asset is sold.
(3) The depreciation allowance on a physical asset is calculated on the
basis of the nominal (historic) cost of the asset. In this respect, nominal
income overstates real income.
The above three sources for the deviation of real income from nominal
income are well known and have received considerable attention. In fact,
many economists have asserted that these are the only significant effects
of inflation on real tax liabilities (see, for instance, Halperin and Steuerle
(1988)). Many believed that in order to eliminate the effect of inflation
on real tax liability in practice, it would suffice to exempt inflationary
capital gains from tax, disallow tax deductibility of inflationary inter-
est charges, and allow replacement cost depreciation (or more simply,
indexation of historic cost depreciation).
This simple prescription for dealing with the effect of inflation on
taxable income might well be adequate for relatively low rates of infla-
tion, say up to 10-15 percent a year. But when the annual rate of
inflation reaches the three-digit range, other, less obvious, factors come
4
These indexation differentials, which are tax deductible for the borrower,
would, in principle, be taxable income for the lender. However, in the Israeli
case, where major segments of the capital market are effectively nationalized,
the lender is very often the Government itself, so there would be no taxable
lender that would pay the tax that the borrower saved.

140 EFRAIM SADKA
into play. The above suggested solution fails to recognize these factors
and is therefore inadequate for dealing with the effects of inflation on
real tax liabilities in the business sector.
When the inflation rate reaches the double-digit range on a monthly
basis, two additional major factors cause nominal income to deviate
significantly from real income. These factors, unlike the first three,
pertain primarily to the determination of real operating income (that is,
income before capital gains and long-term financing costs are taken into
account). They relate to the nature of the production process, which
takes place over time.
(4) Given that the production process takes place over time, output is
usually sold at the end of this process, while the costs of labor and other
inputs and raw materials are incurred earlier. Thus, output is sold at high
(inflated) nominal prices, relative to the low nominal prices of the inputs.
As a result, the nominal operating income overstates the real operating
income. Naturally, the time length of the production process is short
relative to the length of life of fixed assets and long-term loans. Thus, the
bias in the operating income, unlike the biases caused by capital gains
and interest payments, is significant only when the inflation rate is rela-
tively high (for instance, when the annual rate climbs into the triple-digit
range).
A special case of the preceding bias applies in particular to retail firms.
(5) A retail firm normally buys and pays for its merchandise before it
sells the merchandise. Thus, the sales price is inflated relative to the
purchase price. As a result, the nominal profit from sales includes an
inflationary appreciation in the value of the merchandise. The firm is
thus taxed on the inflationary appreciation of the merchandise it sells.
II. Partial Adjustments for Inflation: 1975-81
One might conclude from the preceding section that since the various
deviations of nominal income from real income are not all of the same
sign, then the effect of inflation on nominal taxable income, vis-a-vis real
income, is ambiguous. However, such a conclusion ignores the long-run
behavior response of the taxpaying firm to the effect of inflation on
nominal income. In the short run, when inflation unexpectedly picks up,
firms are caught by surprise and may either lose or gain from inflation.
For instance, those firms that have invested heavily in fixed assets fi-
nanced by debt usually gain (because, as will be recalled, capital gains
are taxed upon realization, whereas interest charges are deducted on an
accrual basis); those firms that have used equity capital to finance their

production process usually lose. But in the longer run, firms will have
taken various tax avoidance measures in order to reduce nominal taxable
income. For instance, they will use less and less equity capital and invest
more and more in buildings and real estate. Such a tax avoidance ac-
tivity is further fueled, as will be seen below, by changes that are made
in the tax laws in the wake of inflation—changes that are partial and
unbalanced.
As inflation persists, lawmakers start introducing provisions in the tax
laws aimed at eliminating the effect of inflation on real tax liabilities.
However, the Israeli experience suggests that these provisions are gen-
erally introduced piecemeal and tend to be unbalanced. Lawmakers first
yield to the public outcry of those who are hurt by inflation and grant
relief against taxation of inflationary (artificial) income. Only much later
do they close up the loopholes that enable taxpayers to reduce nominal
taxable income much below real income.
One of the earliest provisions introduced in 1975 in the wake of the
stubbornly persistent inflation was a reduction in the tax rate on the
inflationary component of capital gains to 10 percent, compared to the
61 percent rate on ordinary corporate income.5 Similarly, the holders of
indexed government bonds and some other bonds were exempted from
tax on the inflationary component of the interest (the indexation differ-
ential) earned. Owners of bank saving deposits were also granted an
exemption on the inflationary component of earned interest. Another ad
hoc relief measure allowed firms to take a deduction based on the size of
their (finished or unfinished) inventories. The rationale for this deduc-
tion was the need to offset artificial inflationary elements in the operat-
ing income of the firm that were related to the temporal nature of the
production process.
The form all of these provisions took was what I have called "one-side-
of-the-balance-sheet"—that is, they all pertained to the asset side. The
liability side was initially ignored. Only later on did it become evident
that it made no sense, for instance, to exempt inflationary capital gains
or to tax them at a low rate, while still allowing taxpayers to deduct from
ordinary income the total inflationary component of the interest in-
curred by them. As a result of these lopsided measures, firms increased
their borrowing in order to invest in buildings, equipment, machinery,
inventories, stocks, and indexed government bonds. Attempts to restrict
5
It is worth noting that the Israeli tax laws are generally very generous with
respect to capital gains accruing to individuals, even when they are real gains.
For instance, securities traded on the stock exchange are exempted, as is residen-
tial housing (even if not owner-occupied) under some (not significantly restric-
tive) conditions.

142 EFRAIM SADKA
the tax deductibility of inflationary interest charges were partial and

clumsy and they failed.6
III. The Comprehensive Approach: 1982-Present
The partial adjustment measures discussed in the previous section

proved to be inadequate, and after some seven years, with inflation
reaching an annual rate of 140 percent, a new law was introduced in
1982. The aim was to remove in a comprehensive manner all effects of
inflation on real tax liabilities. In principle, the law should have been
applied to all businesses. However, since the various provisions of the
law were based on the balance sheet of the firm, it was effectively
confined to "big" businesses, which are usually required to provide
balance sheets: namely, to corporations and some other proprietary
firms (usually above a certain size).
From a theoretical point of view, the most logical method for eliminat-
ing the effects of inflation on taxable income is to evaluate each transac-
tion in units of some stable currency (say, the European Currency Unit
(ECU)), instead of nominal Israeli sheqalim. Thus, if income is Calcu-
lated by subtracting costs of labor and other inputs, finance costs, and
depreciation, from receipts from sales, all evaluated in ECUs, the result
will indeed reflect the true, real income of the firm. However, imple-
menting this method would be difficult and costly. As long as the Israeli
sheqel remains the only legal medium of exchange, and transactions are
consequently made in sheqalim, this method would require a record to
be maintained of the exact date of each transaction, so that the nominal
sheqalim involved in each transaction could be translated into ECUs at
the current rate of exchange between the sheqel and the ECU, which
varies daily in periods of high inflation.
A further drawback of this method is that standard accounting proce-
dures calculate income not on a cash flow basis, but on an accrual basis.
However, transactions take place over time; that is, some time elapses
from the date on which a sale of a good or a purchase of an input is made,
and the date on which cash is actually received or paid. This time lapse
complicates the process of translating the sheqel value of a transaction
into an ECU value. For instance, suppose that firm A sells some mer-
chandise to firm B for 100 sheqalim. Suppose, further, that when the
merchandise is shipped and an invoice is issued, the rate of exchange
between the sheqel and the ECU is NIS 1 = ECU 1. Hence, at this date
6
These restrictions applied to interest charges that could be attributed to the
financing of tax-exempt government bonds and other securities.

firm A records a sale of 100 ECUs and debits the account of firm B by
100 ECUs. Suppose also that the terms of the sale allow firm B to pay for
the merchandise within 30 days, a common trade practice. Suppose,
further, that after 30 days, when the payment of 100 sheqalim is ac-
tually made, the rate of exchange between the sheqel and the ECU is
NIS 1.25 = ECU 1. Thus, when firm B pays its bill, its account with firm
A is credited for only 80 ECUs with a balance due of 20 ECUs, which is,
of course, incorrect. Hence, an additional entry in firm A's books is
required: the value of sales should be lowered by 20 ECUs and firm B's
account should be credited with the same amount.7 (An analogous entry
is also needed in firm B's books.)
Recall that the period in question was the late 1970s up to the early
1980s when the use of personal computers was not widespread, and many
small businesses still did their accounting manually. It was felt then that
the method of dealing with the effects of inflation on taxable income by
translating the sheqel value of every transaction into units of some stable
currency would be too complicated to implement in practice. Hence, an
alternative, much simpler, but indirect, method was adopted.
The main features of the new comprehensive law that was enacted in
1982 are quite simple: first, calculate income in nominal terms according
to standard accounting procedures; then for each of the five effects of
inflation enumerated in Section I, an adjustment is introduced that
either directly removes that effect or ensures that it is offset by another
one or more of these effects. Income, after these adjustments, will then
reflect the true, real income of the firm, evaluated at end-of-year prices.
The following describes in more detail the adjustments that are
needed. Consider first, items (4) and (5) of Section I. They relate to the
fact that costs (of labor inputs, raw materials, and merchandise) are paid
for some time before sales receipts are cashed in. These effects inflate
nominal income because revenues are evaluated at prices that are in-
flated relative to the prices at which costs are evaluated. But there are
now two possibilities: if the costs were financed by debt, then these two
effects are offset by item (2) (that is, the deductibility of inflationary
interest charges); if the costs were financed by equity, then these two
effects are corrected for by the following adjustment, which puts equity
on a par with debt.
(i) Allow a deduction equal to an imputed inflationary interest on
equity (that is, a deduction that is equal to the amount of equity times
the inflation rate).
Next, consider item (1) from Section I, the inclusion of inflationary
7
This example could be further complicated by supposing that firm B's pay-
ment is made by check, which takes a few more days to clear.

144 EFRAIM SADKA
capital gains in taxable income. When inflationary interest charges—

both the genuine interest on debt and an imputed interest on equity—are
tax deductible on an accrual basis, then it is correct to include inflation-
ary capital gains in taxable income. Furthermore, these inflationary gains
should be included on an accrual basis. Thus, the following adjustment
has to be made.
(ii) Add inflationary capital gains accruing (even if not yet realized)
on all fixed and other nonmonetary assets (that is, on all assets that
appreciate in nominal terms during inflation). This means adding to
nominal taxable income an amount that is equal to the book value of
these assets times the inflation rate; upon realization of the capital gains,
only their real component is taxed.8
Notice that once inflationary interest charges on both debt and equity
are deducted from taxable income, then adjustment (ii) above should be
applied also to business inventories. Specifically, recall that the cost of
sales is defined as beginning-of-the-year inventory, plus new purchases
during the year, minus end-of-year inventory. The latter should be
evaluated at end-of-year prices, so as to include in taxable income the
inflationary capital gains accruing to it.
Adjustments (i) and (ii) above fully correct for items (1), (2), (4), and
(5) of Section I. The following adjustment must be made to correct for
item (3), which is the historic cost depreciation.
(iii) Allow a depreciation that is evaluated at end-of-year prices.
These three adjustments to nominal income make it a true representa-
tion of real income, evaluated at end-of-year prices. In practice, (i) and
(ii) were combined. Since (i) calls for a deduction equaling equity times
the inflation rate, while (ii) calls for an addition to income, which is
equal to fixed (and some other) assets times the inflation rate, then the
net effect of (i) and (ii) is to allow a net deduction that is equal to
(equity - fixed assets) x inflation rate.
Notice that this net deduction may well be negative. It is referred to as
"the deduction for the preservation of equity," since it could be inter-
preted as a deduction aimed at protecting that part of the equity not
invested in "inflation-proof" assets.
An alternative way to derive this formula is to employ the definition of
real income of the firm as the difference between the firm's net worth at
the end of the year and its net worth at the beginning of the year, when
both are evaluated at the same prices, say, end-of-year prices (see the
Appendix).
8
Notice that in view of the adjustments to income discussed in this section,
dividends represent real income.

IV. Loopholes and Exceptions
The preceding section described in a schematic way the basic features

of the 1982 law. However, the actual implementation of these features
was complicated by the emergence of some serious practical consider-
ations, which are described below.
Equity may vary within any one tax year, since new equity may be
issued and some old equity may be retired by paying dividends. Thus,
according to adjustment (i) described in Section III, the firm has to keep
track of the movements of equity within the tax year in order to be able
to calculate the tax deduction to which it is entitled. Beginning-of-the-
year equity, for instance, will be entitled to a deduction based on the
annual rate of inflation (that is, from the beginning to the end of the
year), whereas a new issuance of equity will be entitled to a deduction
that is based on the rate of inflation only from the date of issuance of the
new equity to the end of the year.
A similar caveat applies to fixed assets (as described in adjustment (ii)
in Section III). Since one usually encounters only a few changes in equity
or in the stock of fixed assets within a relatively short period of one year,
the calculation of the deductions from and additions to income is fairly
manageable. However, this is not usually the case with business invento-
ries, which are normally fast moving and typically include an extremely
large number of items. Hence, calculating the inflationary capital gains
accruing to end-of-year inventories is not feasible and has therefore not
been put into practice, even though it was one of the principles em-
bodied in the 1982 law. The effect of this deviation from the rule has
been to postpone the tax on inflationary capital gains on end-of-year
inventories to the next year.9 A public committee recommended in 1985
that some accounting formula could be used to calculate the average
holding period of inventories and, accordingly, adjust the value of the
end-of-year inventories, but this recommendation was never adopted.
Industrial Equipment and Machinery
A second exception was granted to industrial equipment and machin-

ery. Industry in Israel has traditionally been accorded favorable treat-
ment via the tax-subsidy system, in the belief that such assistance is
9
It should be pointed out that using the last-in-first-out (LIFO) method
(rather than the first-in-first-out (FIFO) method) for evaluating end-of-year
inventories would only increase the disparity, because the LIFO method deflates
rather than inflates the monetary value of end-of-year inventories.

146 EFRAIM SADKA
essential for long-term, export-led growth.10 Until 1986, for instance, the
corporate income tax rate on industrial firms was substantially lower
than on nonindustrial firms. Similarly, industrial firms are exempted
from a general payroll tax.
Following this tradition, the 1982 law provided for the inflationary
appreciation of industrial equipment and machinery to be exempted
from tax until the date of realization, so as to encourage investments in
these capital assets. As a partial offset for this tax relief, depreciation
allowances for industrial equipment and machinery were not indexed.
The tax relief reduced the effective tax rates on income from industrial
equipment and machinery and generated an overinvestment in them.
The problem was that the magnitude of the decline in the effective tax
rates depended on the rate of inflation: the higher is the rate of inflation,
the lower are the effective tax rates. A method for determining the mag-
nitude of the reduction was suggested by Sadka and Zigelman (1989),
who adapted standard effective tax rate formulas (see, for instance,
Auerbach (1983, 1987)) to the 1982 law and concluded that the effective
tax rate (tt) on the income generated by asset i is given by
where MPKi is governed by the profit-maximization condition:
and where
MPKi = marginal product of asset i
t = statutory corporate tax rate
DI = physical depreciation of asset i
r = real rate of return required by equity holders
Zi = real present value of the depreciation allowances for asset i
Ut = real present value of the "equity preservation deduction"
(see preceding section).
The interpretation of the above formulas is straightforward. The effec-
tive tax rate (tt) is the rate by which the before-tax marginal return to
capital (MPKi ~~ A) exceeds the real rate of return required by equity
holders (r). The firm invests up to the point where the net-of-tax mar-
ginal product of capital equals the tax-adjusted cost of capital (A + r).
10
Of course, most academic economists in Israel do not approve of favorable
treatment for one sector of the economy.

The tax-adjustment parameters are related to the depreciation al-

lowances (Z,) and the equity preservation deduction (£//).
Employing the above formulas, Table 1 presents the effect of inflation
on the effective tax rates on income from industrial equipment and
machinery (for t = 52 percent, which until 1986 was the statutory tax
rate, and r = 4 percent). As the annual inflation rate rises from zero to
400 percent, the effective tax rates fall by about 12-15 percentage points.
A reduction in effective tax rates as a response to a higher inflation
rate introduces a built-in automatic destabilizer into the tax system,
against the conventional wisdom of all public finance textbooks, which
advocate fiscal automatic stabilizers. Indeed, the tax relief that was
granted in 1982 to industrial equipment and machinery was abolished in
1985, when inflation in Israel reached its peak. Inflationary capital gains
accruing to industrial equipment and machinery became taxable, and the
depreciation allowances were indexed.
Proprietorships and Self-Employed Individuals

The 1982 law covered all corporations, but largely ignored the in-
comes of most proprietorships and self-employed persons (for example,
small businesses, brokers, law firms, plumbers, accountants, and clin-
ics), giving rise to the third loophole. Proprietorships were only some-
what restricted with respect to the amount of interest deductions they
could claim. As a result, the 1982 law essentially created two tax sectors
within the business sector: one to which the law did apply, which I shall
call the indexed sector (mostly corporations); and one that escaped the
provisions of the law, which I shall call the nonindexed sector (mostly
proprietorships and self-employed individuals).
The nonindexed sector can maneuver the timing of its cash receipts
and payments so as to deflate its taxable income and reduce its real tax
Table 1. Effect of Inflation on Effective Tax Rates on Industrial

Equipment and Machinery
(t = 52 percent, r = 4 percent)
Inflation Rate
Asset 0 percent 400 percent
General industrial equipment 28.9 16.5
Tools 27.3 15.4
Special industrial equipment 26.4 14.8
Construction equipment 34.7 20.5

148 EFRAIM SADKA
liability. By advancing the date of the cash receipt for a certain real
revenue, one can deflate nominal revenues. Similarly, by postponing the
date of the cash payment for a certain real expense, one can inflate real
tax liability.
A simple example can serve to illustrate this argument. Consider an
individual whose real revenues, expenses, and net income, when mea-
sured in terms of a stable currency (say, the ECU), are as follows:
Revenues ECU 250,000
Expenses - ECU 150,000
Net income ECU 100,000
Suppose that the annual inflation rate is 100 percent, so that prices
double from the beginning to the end of the year. Thus, if the rate of
exchange is NIS 1 = ECU 1 at the beginning of the year, it will be
NIS 2 = ECU 1 at the end of the year. Suppose further that the individ-
ual is able to advance the receipt of revenues to the beginning of the year
and postpone the payment of expenses until the end of the year. The
statement of nominal income will then show the following entries:
Revenues NIS 250,000
Expenses - NIS 300,000
Net income -NIS 50,000
Thus, through a few simple maneuvers, real income can be turned into a
loss for tax purposes.
Notice that the validity of the above example rests on the ability of the
individual to advance cash receipts or postpone cash payments or both.
However, a receipt for one agent is also payment of another agent.
Therefore, there should be other agents in the above example for whom
payments of expenses were advanced and receipts of revenues were post-
poned. Would not the incomes of these agents be inflated and their real
tax liabilities increased? The answer is not necessarily, since these agents
could belong to the indexed sector and their real tax liabilities would not
then be affected by the manipulations described in the example. Alter-
natively they could belong to the nontaxable, nonprofit sector, or the
public sector, or they could be foreign residents, or final consumers—
in any case, their liabilities would not be affected.
V. Wage Taxation: The Role of Withholding
Unlike taxpayers in the business sector, wage earners cannot maneu-

ver with the indexed sector in order to reduce their real tax burden
because of the withholding system. This system ensures that any manip-

ulation of the timing of cash receipts for wages earned will have little, if
any, effect on real tax payment.
A simple example will serve to illustrate this point. Consider, for the
sake of simplifying the arithmetic, that the tax year consists of just two
months. Suppose that in the first month the income tax schedule is
Month 1
Income Bracket Marginal Tax Rate
0-l,000sheqalim 0 percent
Over 1, 000 sheqalim 30 percent
Suppose further that prices double between the first and the second
month. With full indexation of the income tax brackets (as is the case in
Israel), the income tax schedule in the second month would be
Month 2
0-2,000 sheqalim 0 percent
Over 2,000 sheqalim 30 percent
The income tax schedule is calculated for the annual tax returns by
adding up the brackets for the various months. Thus, the annual tax
schedule will be
Annual Schedule
0-3,000 sheqalim 0 percent
Over 3, 000 sheqalim 30 percent
Consider an individual who earns a steady wage income of ECU 1,500
a month. Suppose also that the rate of exchange is NIS 1 = ECU 1 in the
first month, and, consequently, NIS 2 = ECU 1 in the second month.
Thus, the individual earns NIS 1,500 in the first month and NIS 3,000
in the second month. In the first month, the individual will be subject to
a withholding tax of (NIS 1,500 - NIS 1,000) x 30 percent = NIS 150,
which is worth ECU 150. In the second month, she will be subject to a
withholding tax of (NIS 3,000-NIS 2,000) x 30 percent = NIS 300,
which is worth ECU 150. Altogether, an amount NIS 450, or ECU 300,
is withheld at the source. When this individual files a tax return at the
end of the year, she will report an annual income of NIS 4,500, on which
the tax liability is (NIS 4,500 - NIS 3,000) x 30 percent = NIS 450. The
latter is also the amount that was withheld, and hence she will pay no
further taxes. Real tax payment is therefore ECU 300.
Now, suppose the individual advances the receipt of her wage for the
second month to the first month; that is, she receives her total annual
wage of ECU 3,000, or NIS 3,000, in the first month. The amount of tax

150 EFRAIM SADKA
withheld will then be (NIS 3,000 - NIS 1,000) x 30 percent = NIS 600,
which is worth ECU 600. When she files a tax return at the end of the
year, she reports an annual income of NIS 3,000, on which the tax liability
is zero. Therefore, she receives a refund of NIS 600, which is now worth
only ECU 300. Hence, her real tax payment is ECU 600 - ECU 300
ECU 300. This is also exactly what she paid when her income was spread
evenly over the two months.
In the above example, the individual gained nothing by maneuvering
the timing of her wage receipts. Although one may be able to devise an
example in which some gain could occur, nevertheless, this example
serves to show how withholding substantially curtails the gains from
advancing wage receipts. When a wage receipt is advanced, the tax is
withheld, thereby limiting the real gain that can be realized from such a
maneuver.
VI. Consumption Tax Versus Income Tax with Inflation
The income (or direct) versus consumption (or indirect) tax contro-
versy has been discussed at length in the literature (see, for example,
Atkinson (1977)). Although this controversy can be addressed from
several angles,111 shall confine the discussion here to the relative perfor-
mance of these two taxes in the presence of inflation.
As was seen above, the presence of inflation poses some serious
complications in the definition of business income, but the consumption
tax appears to be unaffected by these difficulties. This is why many
economists and policymakers argue in favor of a consumption tax in a
period of high inflation. Indeed, a consumption-type VAT performed
remarkably well in Israel, even during the peak inflation period of
1984-85.
However, a consumption tax is usually levied at a flat rate. In a life-
cycle model, or in a Ricardian world, the present value of consumption is
equal to the present value of wages. Hence, a proportional consumption
tax has the same equity implications as a proportional wage tax (that is, it
is not progressive). In order to make the consumption tax more progres-
sive, one could exempt from tax some necessities (such as food products,
for example) and impose a higher tax rate on luxuries.12 This is indeed
11
Atkinson and Sandmo (1980), for example, point out that the income tax
causes both intra- and intertemporal distortions, whereas the consumption tax
causes only intertemporal distortions—the so-called double-taxation-of-savings
argument.
12
For the theoretical foundation of this result see Deaton (1977) and Balcer
and Sadka (1981).

the practice followed in many European Community countries with re-

spect to the VAT. Yet, it is highly questionable how much progression
one can achieve through a three-tier VAT system (that is, a zero rate, a
standard rate, and a luxury rate).
In order to strengthen the progression capacity of the consumption
tax, the tax rate has to be tied to the total consumption of the individual
(in the same way as the income tax rate depends on the total income of
the individual). In other words, there should be a consumption tax
schedule that is applied to the total consumption of each individual or
household. Each individual would have to report her total consumption,
which could only be done accurately by subtracting real personal saving
from real personal income. For a typical wage earner, it presumably
would not be difficult to calculate real personal income—real interest
income, dividends, and real capital gains would be added to the wage
income to yield real personal income. The corporate income tax would
no longer be needed and would be abolished. But a self-employed in-
dividual or an owner of a small business (unincorporated) would still
need to calculate real business income. Thus with this group of tax-
payers, we are back to square one: how to define real business income in
a period of high inflation. In addition, the problem arises of how to
calculate real personal savings (that is, the real increase in net (of debt)
wealth). A progressive consumption tax therefore is not necessarily
an easy alternative to the progressive income tax in a period of high
inflation.
Another alternative to the business income tax is a business cash flow
tax (see King (1987)). This is a tax on the net cash flow a company
receives from its real economic activities.13 It differs from the income tax,
in that it grants immediate expensing (100 percent first-year depreciation
allowances) to all forms of investments. This feature gives the cash flow
tax an administrative advantage, since its implementation does not re-
quire a calculation of "true" or "economic" depreciation upon which to
base depreciation allowances. After analyzing the positive implications
of the cash flow tax for the efficiency of resource allocation, King (1987,
p. 379) concluded: "It is attractive for a further reason, namely that the
base of the tax requires no adjustment for inflation, and hence that the
complicated indexation provisions for depreciation, for example, re-
13
In order for the cash flow tax to have the same effect as the consumption tax,
interest payments should still qualify as a deduction if interest income is taxed at
the individual level; otherwise, the firm and the individual will be using different
rates of discount—the firm's rate of discount will be the pretax rate of interest,
while the individual's discount rate will be the after-tax rate of interest. Such a
divergence between the firm's rate and the individual's rate causes an intertem-
poral inefficiency of resource allocation.

152 EFRAIM SADKA
quired under alternative corporate tax systems are unnecessary with a

cash flow tax The tax eliminates the necessity of calculating 'eco-
nomic profit.' Hence, there is no need to construct a true measure of
depreciation or to make any adjustment for the effects of inflation."
Obviously, with a cash flow tax that grants immediate expensing, the
need for indexation provisions for depreciation vanishes. However, with
a high inflation rate, other adjustments for inflation still need to be made
in the tax base. When the inflation rate is sufficiently high (say, in the
triple-digit range), the cumulative price increase within any single tax
year, from the first months to the last months, is quite substantial. In
such a case, one cannot simply subtract cash outflows from cash inflows
in order to calculate the annual net cash flow of the firm, if the cash
inflows were received at different points in time than the cash outflows
were paid, even though both flows occurred within the same tax year.
Thus, cash flows have to be indexed in calculating the annual net cash
flow of the firm. Alternatively, the tax period can be shortened from one
year to one month. (In fact, the idea of a cash flow tax on a monthly basis
was briefly considered in Israel in 1984, but the time was not yet ripe for
such a tax "revolution.") A big advantage of the cash flow tax in this
respect is that, unlike the income tax, it does not require the calculation
of depreciation allowances or a complicated evaluation of business in-
ventories (especially inventories of unfinished goods in the production
process). Hence, a monthly cash flow tax would not be excessively costly
to administer.
However, the monthly net cash flow of the firm varies considerably
over time; often, it may be negative. (For example, one would certainly
expect a negative net cash flow in a month in which the firm makes a
major investment.) Therefore, it is essential for the smooth functioning
of the monthly cash flow tax either to grant a full tax rebate in case the
net cash flow is negative or to allow net negative cash flows to be carried
forward with full indexation and real interest.
The cash flow tax deserves serious consideration as an alternative to
the business income tax. In addition to its "fiscal neutrality" advantage
over the standard income tax, it may also perform more effectively in a
period of high inflation.
APPENDIX
Alternate Derivation of Real Income

In Section III the real income of the firm, evaluated at end-of-year prices, was
shown to be equal to nominal income adjusted by the formula given on p. 144,
"the deduction for the preservation of equity," and by an indexation differential

on depreciation. This Appendix provides an alternative, but equivalent, defini-

tion of real income via the balance sheet of the firm.
If no new equity is issued within the tax year and no dividends are distributed,
then the real change in the firm's net worth is equal to its real income. Thus, real
income, evaluated at end-of-year prices, is equal to end-of-year net worth, eval-
uated at end-of-year prices, less beginning-of-the-year net worth, also evaluated
at end-of-year prices.
If the firm neither purchases nor sells any fixed asset during the year, then its
nominal balance sheets at the beginning and the end of the year will typically
appear as follows (a "0" subscript stands for the beginning of the year and a "1"
subscript stands for the end of the year).
Beginning of the Year
Assets Equity and Liabilities

FA E
NNA0 Lo
End of the Year
Assets Equity and Liabilities

FA-D E
NNA1 NI
L1
where
FA — fixed assets at historic (that is, beginning-of-the-year) prices
NNA = net nominal (nonindexed) assets. These may include, for example,
cash, checking accounts, and balances due from clients (less bal-
ances due to suppliers)
E = equity at historic prices
L = long-term indexed liabilities at current prices
NI = nominal income
D = depreciation at historic prices.
Notice that
and
Suppose that the price level rises from the beginning to the end of the year at
the rate TT. Then, the beginning-of-the-year net worth of the firm, evaluated at
end-of-year prices is
Similarly, the end-of-the-year net worth of the firm, evaluated at end-of-the-year

prices is

154 EFRAIM SADKA
Hence, the real income of the firm, evaluated at end-of-year prices is obtained by
subtracting (3) from (4):
real income
Employing (1) and (2), equation (5) reduces to

real income
Rearranging terms, equation (6) becomes
real income
Notice that (E - FA)tr is the deduction for the preservation of equity that is
given by the formula in Section III. The term DTT is the indexation differential on
depreciation. Thus, equation (7) suggests that real income is indeed equal to
nominal income, adjusted by the deduction for the preservation of equity and by
an indexation differential on depreciation. A similar formula is proposed by
Harberger (1988), but he understated the difficulties involved in the evaluation
of business inventories, especially unfinished goods in the production process
(see also equation (9) in Tanzi (1981)).
REFERENCES
Atkinson, Anthony B., "Optimal Taxation and the Direct versus Indirect Tax
Controversy," Canadian Journal of Economics, Vol. 10 (November 1977),
pp. 590-606.
Atkinson, Anthony B., and Agnar Sandmo, "Welfare Implications of the Taxa-
tion of Savings," Economic Journal, Vol. 90 (September 1980), pp. 529-49.
Auerbach, Alan J., "Taxation, Corporate Financial Policy and the Cost of
Capital," Journal of Economic Literature, Vol. 21 (September 1983), pp.
905-40.
, "Corporate Taxation in the United States," in Economic Policy in
Theory and Practice, ed. by Assaf Razin and Efraim Sadka (London:
Macmillan, 1987).
Bailey, Martin J., "The Welfare Cost of Inflationary Finance," Journal of Politi-
cal Economy, Vol. 64 (April 1956), pp. 93-110.
Balcer, Yves, and Efraim Sadka, "Budget Share and Optimal Commodity Taxes
Without Computation," Economics Letters, Vol. 7, No. 3 (1981), pp.
265-71.
Casanegra de Jantscher, Milka, "Taxing Business Profits during Inflation: The
Latin American Experience," International Tax Journal (1976), pp. 128-46.
Deaton, Angus S., "Equity, Efficiency, and the Structure of Indirect Taxation,"
Journal of Public Economics, Vol. 8 (December 1977), pp. 299-312.
Eckstein, Zvi, and Leonardo Leiderman, "Estimating an Intertemporal Model

of Consumption, Money Demand, and Seigniorage" (unpublished; Foerder

Institute for Economic Research, Tel Aviv University, 1989).
Fischer, Stanley, "Towards an Understanding of the Costs of Inflation: II,"
Carnegie-Rochester Series on Public Policy, Vol. 15 (1981), pp. 5-42.
Fischer, Stanley, "Seigniorage and the Case for a National Money," Journal of
Political Economy, Vol. 90 (April 1982), pp. 295-313.
Fischer, Stanley, and Franco Modigliani, "Towards an Understanding of the
Real Effects and Costs of Inflation," Weltwirtschaftliches Arciv, 114, No. 4
(1978), pp. 810-33.
Friedman, Milton, "The Optimal Quantity of Money," in his The Optimum
Quantity of Money: and Other Essays (Chicago: Aldine Publishing Co.,
1969).
Halperin, Daniel, and Eugene Steuerle, "Indexing the Tax System for Infla-
tion," in Uneasy Compromise: Problems of a Hybrid Income-Consumption
Tax, ed. by Henry J. Aaron, Harvey Galper, and Joseph A. Pechman
(Washington: The Brookings Institution, 1988).
Harberger, Arnold C., "Comments on Indexing the Tax System for Inflation,"
in Uneasy Compromise: Problems of a Hybrid Income-Consumption Tax,
ed. by Henry J. Aaron, Harvey Galper, and Joseph A. Pechman (Washing-
ton: The Brookings Institution, 1988), pp. 380-83.
Helpman, Elhanan, and Efraim Sadka, "Optimal Financing of the Govern-
ment's Budget: Taxes, Bonds or Money," American Economic Review, Vol.
69(1) (March 1979), pp. 152-60.
Kay, J.A., "Inflation Accounting—A Review Article," The Economic Journal,
Vol. 87 (June 1977), pp. 300-11.
King, Mervyn A., "The Cash Flow Corporate Income Tax," in The Effects of
Taxation on Capital Accumulation, ed. by Martin S. Feldstein, University of
Chicago Press, 1987.
Olivera, Julio H.G., "Money, Prices and Fiscal Lags: A Note on the Dynamics
of Inflation," Quarterly Review, Banca Nazionale del Lavoro, Vol. 20 (Sep-
tember 1967), pp. 258-67.
Phelps, Edmund S., "Inflation in the Theory of Public Finance," Swedish Jour-
nal of Economics, Vol. 75 (March 1973), pp. 67-82.
Sadka, Efraim, and Avi Zigelman, "Income Taxation in the Industrial Sector,"
Economic Quarterly (1990).
, "Inflation and Effective Tax Rates in the Industrial Sector," Bank of
Israel Economic Review, Vol. 64 (1990), pp. 71-82.
Tanzi, Vito, "Inflation, Lags in Collection, and the Real Value of Tax Revenue,"
Staff Papers, International Monetary Fund, Vol. 24 (March 1977), pp.
154-67.
, "Inflation, Real Tax Revenue, and the Case for Inflationary Finance:
Theory with an Application to Argentina," International Monetary Fund,
Staff Papers, Vol. 25 (September 1978), pp. 417-51.
, "Inflation Accounting and the Taxation of Capital Gains of Business
Enterprises," in Reforms of Tax Systems: Proceedings of the 35th Congress
of the International Institute of Public Finance, ed. by Karl W. Roskamp and
Francesco Forte (Detroit: Wayne State University Press, 1981).

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
Exchange Rate Expectations

A Survey of Survey Studies
SHINJI TAKAGI*
The empirical literature on survey-based exchange rate expectations is

briefly surveyed. The literature in general supports the presence of a non-
zero risk premium and rejects the hypothesis of rational expectations. The
crucial result is that, whereas short-run expectations tend to move away
from some long-run "normal" values, long-run expectations tend to move
back toward them. If this behavior of short-run expectations increases the
volatility of exchange rate movements, there may be a basis for an official
measure to minimize short-run exchange rate movements. [JEL 431]
LUGH IT is firmly established that expectations play a central

A LTHO role in the determination of exchange rates, little is known about
the exact nature of those expectations. Of course, the problem with
empirically testing hypotheses about exchange rate expectations is that
expectations are unobservable. In the past, a popular way to get around
this problem was to use either the forward exchange rate or the ex post
spot exchange rate as a proxy for the expected exchange rate.
There is an obvious drawback to this approach. First, the use of the
forward exchange rate presupposes that there is no risk premium, but
the absence of a risk premium itself is a hypothesis of interest.1 Second,
* Shinji Takagi, an Economist in the Research Department of the Fund when

this paper was written, is an Associate Professor of Economics at the University
of Osaka. He holds a Ph.D. in economics from the University of Rochester. The
author thanks Charles Adams, James Boughton, Michael Dooley, and Masayuki
Ohtaki for useful comments, and the Japan Center for International Finance for
providing survey data.
ôr example, if there is no risk premium in the exchange rate of any two
currencies, assets denominated in one currency are perfect substitutes for those
denominated in the other, making sterilized foreign exchange market interven-
tion ineffective.
156

EXCHANGE RATE EXPECTATIONS 157
the use of the ex post exchange rate imposes rationality of expectations,

but the nature of expectations can only be determined empirically. Most
empirical tests involving exchange rate expectations are thus joint tests
of a hypothesis about the degree of risk aversion (or a more structural
model of exchange rate determination) and a hypothesis about the na-
ture of expectations.
In order to avoid the joint nature of conventional hypothesis testing,
an increasing number of researchers have recently begun to use survey
data in tests involving exchange rate expectations. The use of observable
survey expectations allows separate testing of an underlying model of
exchange rate determination and a hypothesis about expectations. There
is strong professional resistance to the use of nonmarket data, perhaps
for good reasons. For one thing, there is no assurance that economic
agents have enough incentive to disclose their truthful expectations. For
another, even if they did, no precise link seems to exist between average
(or individual) expectations and the actual exchange rates that are in fact
marginal prices in the foreign exchange market. However, in the absence
of better alternatives, an empirical literature based on survey data of
exchange rate expectations has been expanding in recent years. This
paper presents a brief survey of this growing literature.
The paper is organized as follows. Section I summarizes the features
of five major sets of survey data used in the literature. Section II presents
a few characteristics of survey exchange rate expectations. Sections
III-VI survey, respectively, major empirical results on forward discounts
and risk premia, the rationality of expectations, the mechanism of ex-
pectations formation, and the relationship between short-run and long-
run expectations. The concluding section presents a summary and a few
policy implications.
I. Data Sets
Five sets of survey data have been used in the literature (Table 1). The
oldest data set comes from the (roughly) annual surveys of the six-month
and twelve-month expectations of over 250 monetary officials and other
financial market experts, conducted by American Express Banking Cor-
poration (Amex) of London for the period 1976-85. The exchange rates
in the surveys were the U.S. dollar rates of five major currencies,
namely, the U.K. pound, deutsche mark, Japanese yen, Swiss franc, and
French franc. A problem with the Amex data is that the surveys were
irregular and conducted by mail, making the timing of surveys somewhat
imprecise. The Amex survey was discontinued in 1985.

158 SHINJI TAKAGI
Table 1. Major Surveys of Exchange Rate Expectations

Survey Description
American Express Banking January 1976-July 1985; roughly annually;
Corporation (Amex), London a sample of 250-300 financial market ex-
perts; U.K. pound, deutsche mark,
Japanese yen, Swiss franc, and French
franc against the U.S. dollar; six-month
and twelve-month expectations.
Economist Financial Report June 1981-present; every six weeks; a
(Economist), London sample of 14 international banks; U.K.
pound, deutsche mark, Japanese yen,
Swiss franc, and French franc against the
U.S. dollar; three-month, six-month, and
twelve-month expectations.
Money Market Services October 1984-present; weekly; a sample of
(MMS), New York about 30 professional traders; U.K. pound,
and Londona deutsche mark, Japanese yen, Swiss franc
against the U.S. dollar; one-week and
one-month expectations.
Godwins, London January 1981-present; monthly; a sample
of over 50 investment managers; U.K.
pound's effective exchange rate and bilat-
eral rate against the U.S. dollar; twelve-
month expectations.
Japan Center for International May 1985-present; semimonthly; a sample
Finance (JCIF), Tokyo of 44 market participants; Japanese yen
against U.S. dollar; one-month, three-
month, and six-month expectations.
a
For the period between November 1982 and October 1984, the surveys were
biweekly and included only two-week and three-month expectations.
Second, the Economist Financial Report (Economist) has been pub-

lishing telephone surveys of the three-month, six-month, and twelve-
month expectations of 14 leading international banks every six weeks
since 1981. The exchange rates in the surveys are the same as the Amex
data, comprising the U.S. dollar rates of the five major currencies.
Third, a similar data set comes from weekly telephone surveys on the
one-week and one-month exchange rate expectations conducted by
Money Market Services (MMS) of California. Weekly surveys have been
conducted, since 1984 in both New York and London, and each sample
includes about 30 professional traders. Between November 1982 and
October 1984, the surveys were biweekly and consisted of two-week and
three-month expectations. The exchange rates in the MMS data are the

U.S. dollar rates of four major currencies, namely, the U.K. pound,
deutsche mark, Japanese yen, and Swiss franc.
Fourth, since January 1981, Godwins of London has conducted
monthly surveys of 50 leading investment managers. The Godwins data
contain only the twelve-month qualitative expectations of the effective
and the U.S. dollar exchange rates of the pound. That is to say, the
survey asks the respondents only their opinions of the direction of ex-
pected future change; that is, whether the exchange rate would go up, go
down, or remain the same. Taylor (1989), the only researcher to have
used the Godwins data, used a subjective probability method to derive a
quantitative series of mean expectations in his study.
Finally, semimonthly telephone surveys have been conducted since
May 1985 by the Japan Center for International Finance (JCIF), a private
institution affiliated with the Japanese Ministry of Finance. The JCIF
data contain the one-month, three-month, and six-month expectations
of the U.S. dollar exchange rate of the Japanese yen held by 44 market
participants in Tokyo. The surveys have consistently included the same
respondents classified into six industry groups, consisting of banks and
brokers, securities companies, general trading companies, insurance
companies, importers, and exporters; the data can thus be used both as
panel data and as industry data.
II. Some Characteristics of Survey Data
Before reviewing the empirical literature on survey-based exchange

rate expectations, one should take note of general qualitative character-
istics of the survey data. Among many such characteristics, this section
will summarize three, under the headings of heterogeneity, underpredic-
tion, and "twist," as a background against which major empirical studies
can be reviewed.
Heterogeneity
Although in what follows the convention of treating expectations as if

they were homogeneous is adopted, the limitation of this standard prac-
tice should be noted at the outset. In reality, any survey data will imme-
diately reveal that expectations have a distribution. If the exchange rate
of the Japanese yen against the U.S. dollar is taken as an example, the
standard deviations in recent years have ranged roughly between 2 yen
and 5 yen per dollar (1 percent and 3.5 percent) for one-month expecta-
tions, between 3 yen and 8 yen (1.5 percent and 5.5 percent) for three-

160 SHINJITAKAGI
month expectations, and between 3 yen and 11 yen (1.5 percent and 7.5
percent) for six-month expectations. As expected, there is a tendency
for dispersion (as measured by standard deviations) to increase for
longer-term expectations, although the size of increase appears to be far
less than would be the case if the variance were to remain constant over
expectations horizons (Figure 1). Casual observation also indicates that
from the time immediately preceding the Louvre accord in February
1987, the degree of dispersion declined significantly but it began to
increase again in the first part of 1989.
This heterogeneity in expectations may reflect, in addition to the usual
distributional factors, systematic individual or group effects. On the basis
of disaggregated JCIF data, for example, Wakita (1989) and Ito (1990)
found significant industry-specific bias in expectations. According to
their findings, exporters had expectations of greater yen depreciation (or
smaller yen appreciation) and importers expressed exactly the opposite
expectations. This systematic expectational bias in favor of one's interest
may indicate either wishful thinking or strategic behavior (to influence
the movement of the exchange rate in a desired direction). Wakita (1989)
suggests a possibility that such industry-specific bias may reflect private
information, while Ito (1990) argues that, to the extent that individuals
are not likely to possess private information, the presence of individual
effects must reflect the failure of the rational expectations hypothesis.
Underprediction
Another important characteristic of survey data is a general tendency
for the expected future exchange rate of all time horizons to follow
closely the current spot exchange rate. This means that expected
changes in exchange rates as reported in survey data tend to underpre-
dict consistently the extent of actual exchange rate movements; this is
demonstrated by Figure 2, which depicts the actual and expected ex-
change rates of the U.S. dollar against the Japanese yen during the
recent period of sharp dollar depreciation. This is another way of saying
that much of actual exchange rate change is unexpected and is consistent
with a similar conclusion based on forward exchange rates as a measure
of expected exchange rates (Mussa (1979)).
For the earlier period of dollar appreciation (that is, from 1981
through early 1985), survey data indicated a persistent underprediction
of the extent of the actual dollar appreciation. In fact, the market partic-
ipants surveyed in general expected the major currencies to appreciate
against the U.S. dollar (Table 2). In contrast, during the period of dollar

Figure 1. Standard Deviations of Expected Exchange Rates,

May 1985 to December 1989
(yen/dollar)
Source: Japan Center for International Finance (JCIF).

Note: Monthly time series of one-month, three-month, and six-month ex-
pected exchange rates of the Japanese yen against the U.S. dollar.

162 SHINJI TAKAGI
Figure 2. Actual and Expected Exchange Rates, May 1985 to April 1986
(yen/dollar)

Note: Semimonthly time series of the spot and one-month, three-month, and
six-month expected exchange rates of the Japanese yen against the U.S. dollar.

Table 2. Actual and Expected U.S. Dollar Depreciations

in Selected Sample Periods
Data Dates Currency Horizon Actual Expected
MMSa 1/83-10/84 pound 2-week -16.15 -2.66
3-month -13.92 4.46
mark 2-week -15.19 5.09
3-month -13.92 0.37
yen 2-week -4.23 8.40
3-month -2.90 8.68
10/84-2/86 pound 1-week 14.96 -12.84
1 -month 10.13 -11.91
mark 1-week 21.36 2.84
1-month 23.82 -2.26
yen 1-week 24.39 5.40
1-month 27.55 2.99
Economist3 6/81-12/85 pound 6-month -6.79 4.19
12-month -9.47 3.38
mark 6-month -0.96 12.39
12-month -5.60 10.67
JCIF 5/85-1/87 yen 1-month 25.20 16.80
3-month 30.32 5.00
6-month 26.16 -0.40
2/87-7/89 yen 1 -month 6.60 11.88
3-month 2.24 9.20
6-month 1.48 5.20
Note: Average annualized logarithmic change in basis points (roughly equal to
percentage change); an increase in value means a depreciation of the U.S. dollar
against
a
the U.K. pound, deutsche mark, or Japanese yen.
Frankel and Froot (1987a).
depreciation (that is, from late 1985 through early 1987), the market
participants expected a much more moderate depreciation of the dollar;
they even expected a sizable appreciation of the dollar against the
pound, when the dollar in the event depreciated in subsequent months.
This characteristic of expectations is important to bear in mind when the
results of rational expectations tests are interpreted.
Twist
The final important characteristic of survey data is a tendency for

longer-run expectations to reverse the direction of the short-run expecta-
tions. That is to say, a depreciation tends to be followed by expectations
of further depreciations in the short run, but by expectations of moder-

164 SHINJI TAKAGI
ate reversals (or appreciations) in the long run. This characteristic has
been called in the literature a "twist" in expectations.
This tendency becomes conspicuous during periods of sharp exchange
rate movement, as in the recent period of dollar depreciation against the
major currencies. For example, when we look at the movements of the
expected exchange rates of the U.S. dollar against the Japanese yen
during the six-month period between October 1985 and April 1986, we
note that the market participants surveyed expected the dollar to con-
tinue to depreciate over the period of one month but to appreciate over
the period of six months (Figure 3). Some have interpreted these results
as reflecting the view of market participants that exchange rates are de-
termined by "momentum" models in the short run but return to histori-
cal norms over longer periods (see Section VI for a further discussion).
III. Forward Discounts and Risk Premia
The presence or absence of a risk premium has been of considerable

interest to economists and policymakers because it has far-reaching
implications for the substitutability of assets denominated in different
currencies and, hence, for the efficacy of sterilized foreign exchange
market intervention. The use of survey data allows the direct measure-
ment of a risk premium from the observation of the forward discount
(fd\ which can be decomposed into the expected currency depreciation
and a risk premium:
where tf is the log of the forward exchange rate set in period t> Et is an
expectations operator2 based on the set of information available in pe-
riod t, st(st+j) is the log of the spot exchange rate in period t(t + /), and rpt
is a risk premium; an increase in the exchange rate is defined as a
depreciation of the domestic currency (the U.S. dollar).
Table 3 presents the decomposition of the forward discount of the
U.S. dollar into expected depreciation (as reported in survey data) and
the risk premium in selected sample periods. For the most part, the
dollar was expected to depreciate against the deutsche mark and the
Japanese yen by an amount greater than the size of the forward discount;
in contrast, the dollar was expected to appreciate against the U.K.
2
Note that expectations (expressed by the operator £) may or may not be
mathematical expectations.

Figure 3. Twists in Exchange Rate Expectations,

October 1985 to September 1986
(yen/dollar)

Note: Semimonthly time series of the spot and one-month, three-month, and
six-month exchange rates of the Japanese yen against the U.S. dollar. The
expected exchange rates are dated according to the future periods for which the
expectations are formed.

166 SHINJI TAKAGI
Table 3. Decomposition of the Forward Discount on the U.S. Dollar

in Selected Sample Periods
Forward Expected Risk
Data Currency Horizon Discount Depreciation Premium
MMSa
10/84-2/86 pound 1-month -3.85 -11.91 8.06
mark 3.23 -2.26 5.49
yen 1.68 2.99 -1.31
1/83-10/84 pound 3-month 0.37 4.46 -4.09
mark 4.68 8.33 -3.65
yen 3.85 8.68 -4.83
Economist3
6/81-12/85 pound 3-month -0.06 3.66 -3.72
6-month 0.14 4.19 -4.05
12-month 0.36 3.38 -3.02
mark 3-month 4.36 11.84 -7.48
6-month 4.35 12.39 -8.04
12-month 4.24 10.67 -6.43
yen 3-month 4.67 12.66 -7.99
6-month 4.74 12.94 -8.20
12-month 4.66 10.67 -6.01
Godwinb
1/81-7/85 pound 12-month -0.79 -6.41 5,62
JCIF
5/85-2/87 yen 3-month 1.52 5.00 -3.48
6-month 1.64 -0.40 2.04
2/87-7/89 yen 3-month 3.64 9.20 -5.56
6-month 3.64 5.20 -1.54
Note: Average annualized logarithmic changes in basis points (close to per-
centage
a
change).
b
Frankel and Froot (1987a).
Taylor (1989).
pound by an amount greater than the size of the forward premium. In

general, exchange rate expectations were positively correlated with the
forward discount (that is, the currencies that were expected to depreciate
were at a forward discount).
Although descriptive statistics suggest the presence of a risk premium,
it is also important to know if the risk premium is significant (in a
statistical sense) and if it is correlated with the forward discount. This
question is important because it sheds light on the source of the forward
discount bias. The forward rate can fail to be an unbiased predictor of

the future exchange rate either because of the failure of rational expecta-
tions (to be treated in the next section) or because of a risk premium that
is time-varying.
The null hypothesis that the correlation of the risk premium with the
forward discount is zero can be tested by running the following regres-
sion:
where e is a random error term. The null hypothesis is that a2 = 1.

According to the results of Froot and Frankel (1989) as selectively
summarized in Table 4, the hypothesis of a2 = 1 was rejected for one-
month expectations but could not be rejected for expectations of three
months or longer (see Sections V and VI for the difference between
short-run and long-run expectations). There is thus little evidence that a
risk premium is correlated with the forward discount, at least for longer-
run expectations. However, the F-statistics rejected the joint hypothesis
of fli = 0 and a2 = 1, suggesting the presence of a nonzero risk premium
(this is consistent with the results in Table 3). The failure to reject the
hypothesis of a2 = 1 may be interpreted as indicating evidence of the
perfect substitutability of assets denominated in different currencies, to
the extent that a change in expected exchange rates is fully reflected in a
one-to-one change in forward exchange rates, at least for changes in
fundamentals observed during the sample period.
Table 4. Forward Discounts: Summary Results

Data Currency3 Horizon a2b F-statisticc
MMS, 11/82-1/88 all 1-month 3.07 (0.47)** 36.34**
Economist, 6/81-12/85 all 3-month 1.30(0.26) 16.55**
6-month 1.03 (0.17) 52.06**
12-month 0.93 (0.15) 65.82**
Amex, 1/76-7/85 all 6-month 1.21 (0.21) 6.32**
12-month 0.88 (0.28) 8.10**
Source: Froot and Frankel (1989).
Note:
a
See equation (2).
All against the U.S. dollar. See Table 1 for the currencies included in each
data
b
set.
Standard errors in parentheses; (**) indicates that the hypothesis of a2 = 1 is
rejected
c
at the 1 percent level of significance.
F-statistics on the joint hypothesis of a\ = 0 and a2 = 1; (**) indicates that the
hypothesis is rejected at the 1 percent level of significance.

168 SHINJITAKAGI
IV. Rationality of Expectations
The overwhelming majority of the empirical literature is concerned

with the rationality of survey exchange rate expectations. Normally, ra-
tionality is defined in terms of two criteria: (1) whether the expected
exchange rate is an unbiased predictor of the future spot exchange rate
(unbiasedness); and (2) whether the expected exchange rate fully incor-
porates all available information (orthogonality). The tests of rational
expectations reported in the literature also correspond to these two
types.
Unbiasedness
Unbiasedness is an important aspect of the rationality of exchange

rate expectations. The use of survey data allows direct testing of the
hypothesis that the expected spot exchange rate for period t + / (formed
in period t) is an unbiased predictor of the future spot rate (in period
t + j):
where the survey expectations Etst+j is free from the presence of a risk
premium, and u is a random error term. Tests of the unbiasedness of ex-
change rate expectations would involve tests of the hypothesis of b\ = 0
and &2= 1> when equation (3) is estimated, usually in first-difference
form.
Dominguez (1986) and Ito (1990) regressed actual depreciation on
expected depreciation using MMS and JCIF data, respectively, for dif-
ferent time horizons and for different dollar exchange rates (Table 5).
For the earlier period (1983-85), Dominguez almost unanimously re-
jected the joint hypothesis of bi = 0 and b2 = 1 for one-week, one-
month, and three-month expectations for all currencies.3 The negative
estimates of b2 for some exchange rates suggest that the forecasts missed
the direction of exchange rate movements. Moreover, the estimate of b2
was below unity in many cases, implying the tendency of forecasters to
overpredict the size of future dollar depreciations.
For the later period (1985-87), however, Ito (1990) could not reject the
joint hypothesis except for the six-month expectation. The difference
between the two studies may reflect the extraordinary nature of the ear-
lier sample period. As noted earlier, the period studied by Dominguez
3
Her study also included the Swiss franc and two-week expectations.

Table 5. Unbiasedness: Some Regression Estimates

Study Currency3 Horizon bi b2 X2-statisticb
Dominguez (1986) pound 1-week 0.001 -0.171** 44.3**
MMS, 1983-85 (0.003) (0.181)
1-month -0.001 -0.505** 21.4**
(0.006) (0.329)
3-month -0.029** -0.450** 39.2**
(0.015) (0.395)
mark 1-week 0.002 0.049** 48.8**
(0.003) (0.137)
1-month 0.014 -0.248** 15.1**
(0.007) (0.392)
3-month -0.043** 0.412 23.3**
(0.016) (0.529)
yen 1-week 0.003 0.502** 12.8**
(0.002) (0.146)
1-month 0.015 0.404* 6.2*
(0.008) (0.291)
3-month 0.003 -0.457** 9.4**
(0.016) (0.626)
Ito (1990) yenc 1-month -0.028 -0.485 2.6
JCIF, 1985-87 (0.017) (0.969)
3-month -0.043 1.167 5.2
(0.034) (1.167)
6-month -0.119** 0.908 10.1**
(0.041) (0.741)
Note: See equation (3). Standard errors are in parentheses; (*) indicates
rejection of the unbiasedness hypothesis at the 5 percent level of significance,
anda
(**), rejection at the 1 percent level.
Against
b 2
the U.S. dollar.
c
X -statistic on the joint hypothesis that b\ = 0 and b2=l.
Average data.
was one in which the U.S. dollar continued to appreciate on a sustained

basis despite expectations to the contrary. Given the extremely low
values of R2 in all of these studies (not reported in the table), only a small
portion of actual exchange rate changes were predicted in practice. The
exact outcome of empirical tests of the unbiasedness hypothesis is thus
likely to depend on the sample.
Orthogonality
Orthogonality is another important aspect of the rationality of ex-
change rate expectations. If expectations are to be efficient (in the sense

170 SHINJITAKAGI
that they incorporate all available information), their predictable power

cannot be improved by inclusion of any variable that is already in the set
of information available at the time when the expectations are formed.
That is to say, prediction errors must be uncorrelated with any variable
in the set of known information. This orthogonality condition can be
formally tested by running the following regression:
where the left-hand-side variable is a prediction error, Xt is a set of

information known in period t, and v is a random error term; popular
candidate variables for Xt have included forward discounts (or nominal
interest rate differentials) and lagged exchange rates. The orthogonality
hypothesis is that c\ = c2 = 0.
Table 6 summarizes some of the regression results, along with the
choice of Xt variables and samples, reported in the literature. While the
summary is by no means exhaustive, it gives an indication of the range
of results that have been obtained from survey studies. As a general
rule, given the large standard errors, Mests often failed to reject the
separate null (orthogonality) hypothesis of c\ = 0 or c2 = 0. However, x2-
or F-statistics almost unanimously rejected the joint orthogonality hy-
pothesis of Ci = c2 = 0, particularly for time horizons longer than three
months. These results, taken together, seem to suggest that the expected
exchange rates as reported in the survey data did not fully incorporate all
available information.
V. Mechanisms of Expectations Formation
Regardless of whether or not expectations are rational, it is of interest

to investigate how they are formed. Survey data have been used to test
three broad types of expectations and their formation mechanisms, ac-
cording to the classification popularized by Frankel and Froot (1987a):
extrapolative, adaptive, and regressive. It should be noted in this exer-
cise that no attempt is being made to determine which of the three
expectations mechanisms is correct or even closest to the actual process.
Extrapolative Expectations
The first mechanism is called extrapolative expectations:

or
where h = (1 + g). This mechanism is called extrapolative for the obvious

reason that the expected currency movement for the next period is given
by the past currency movement for the most recent period, as indicated
in equation (5a). Equation (5a), however, can be equivalently expressed
as equation (5b), which indicates that an extrapolated expectation of the
spot exchange rate for period t + / is given by a weighted average of the
current spot exchange rate and the lagged exchange rate for period t — j.
Of crucial interest is the sign of the coefficient g: g < 0 is the case of
distributed lag (where past currency movement is followed by an expec-
tation of currency movement in the opposite direction); g - 0 is the case
of static expectations (where currency movement is expected to follow a
random walk), and g > 0 is the case of bandwagon expectations (where
past currency movement is followed by an expectation of currency move-
ment in the same direction). In the case of bandwagon expectations,
g > 1 indicates that expectations are explosive.
Table 7 summarizes some of the major regression results on the signs
and magnitudes of g reported in the literature. First, in almost all cases
the f-statistics rejected the hypothesis of static expectations (that is,
g = 0); this means that, despite the fact that empirical exchange rates
followed a process that is closely approximated by a random walk
(Mussa (1979) and Takagi (1988)), market participants did not expect
the expected future exchange rates to follow the same process.
Second, the sign of g is positive (that is, bandwagon expectations) for
the short-run horizons of one or two weeks and one month; the sign of g
is negative (distributed-lag expectations) for the long-run horizons of six
and twelve months; and the sign of g was mixed for the three-month
expectations. This behavioral difference between short-run and long-run
expectations is a general characteristic of survey expectations that will
show up repeatedly throughout this section. For the magnitude of the
estimates, the regression results indicated that the absolute values of g
were less than unity in all cases. This means that, for short-run expecta-
tions of the bandwagon type, expectations were stabilizing.
Adaptive Expectations
The second mechanism is called adaptive expectations, in which ex-
pected currency movement is determined as a fraction of the current
prediction error:

Table 6. Orthogonality: Some Regression Results
X2-statistic/
3
Study Xt Currency Horizon Ci C2 F-statisticb
Dominguez (1986)
MMS, 1983-85 Forward pound 1 -month 0.012 -1.26 5.9
discount (0.017) (1.56)
3-month -0.039** -7.89** 128.8**
(0.006) (1.03)
mark 1 -month 0.019 -1.14 6.4*
(0.008) (1.18)
3-month 0.003 -4.96 36.9**
(0.053) (4.33)
yen 1 -month 0.015 -0.36 3.1
(0.008) (0.78)
3-month 0.003 -3.26 6.9*
(0.035) (3.59)
Froot and Frankel (1989)
Economist, 1981-85 Forward all0 3-month — 2.51 1.3
discount (1.29)
6-month — 2.99 1.5
(1.60)
12-month — 0.52 6.0**
(1.23)

Amex, 1976-85 6-month — 3.63** 3.3**
(1.34)
12-month — 3.11* 1.5
(1.30)
MMS, 1982-88 1-month — 4.81** 8.7**
(1.22)
Ito (1990)
JCIF, 1985-87 Lagged yen 1-month 0.004 0.17 3.9
exchange (0.010) (0.20)
rate change 3-month 0.042 0.31 9.5**
(0.025) (0.23)
6-month 0.114** 0.23 18.9**
(0.036) (0.36)
MacDonald and
Torrance (1989)
MMS (UK), 1982-87 Nominal pound 1 -month -0.014** 8.20** 19.6**
interest (0.004) (1.55)
differential mark 1 -month 0.033** 12.31** 14.7**
(0.011) (3.34)
yen 1 -month 0.018* 7.41* 4.6
(0.010) (3.48)
Note: See equation (4). Standard errors are in parentheses; (*) indicates rejection of the orthogonality hypothesis at the
5 percent
a
level of significance, and (**), rejection at the 1 percent level.
Against the U.S. dollar.
b 2
0
X -statistics on the joint hypothesis that Ci = c2 = 0; for Froot and Frankel (1989) only, F-statistics are reported.
All the currencies in the sample are pooled.

174 SHINJI TAKAGI
Table 7. Extrapolative Expectations: Some Regression Results

Study Currency3 Horizon g
Frankel and Froot (1987a)
MMS, 1983-84 allb 3-month 0.039* (0.017)
Amex, 1976-85 6-month -0.299** (0.049)
12-month -0.380** (0.080)
Frankel and Froot (1987b)
MMS, 1983-84 yen 2-week 0.114* (0.055)
3-month 0.069** (0.023)
MMS, 1984-86 1-week 0.239** (0.049)
1-month 0.121** (0.044)
Economist, 1981-85 3-month -0.143** (0.052)
6-month -0.178** (0.050)
12-month -0.342** (0.066)
Amex, 1976-85 6-month -0.291* (0.112)
12-month -0.441* (0.170)
Bank of Japan (1989)c
JCIF, 1985-89 yen 1 -month 0.029 (0.022)
3-month -0.122** (0.024)
6-month -0.249** (0.032)
MMS, 1984-88 allb 1-week 0.13** (0.04)
1 -month 0.08 (0.07)
Economist, 1981-88 3-month -0.08** (0.03)
6-month -0.17** (0.03)
12-month -0.33** (0.05)
Note: See equation (5a). Standard errors are in parentheses; (*) indicates that
the coefficient is significant at the 5 percent level of significance, and (**), at the
1 percent
a
level.
b
c
The right-hand-side variable is always (st-st- 2) in biweekly data, regard-
less of the choice of;.
or
As equation (6b), an alternative expression, indicates, the adaptively

formed expected exchange rate is given by a weighted average of the
actual current exchange rate and the expected current exchange rate.
Adaptive expectations, however, have not been very popular in em-
pirical investigations. Table 8 summarizes some of the empirical results

Table 8. Adaptive Expectations: Some Regression Results

Study Currency3 Horizon k
Economist, 1981-85 allb 3-month 0.080** (0.020)
6-month 0.052** (0.016)
12-month -0.009 (0.024)
MMS, 1983-84 3-month -0.027 (0.022)
Amex, 1976-85 6-month -0.070 (0.120)
12-month 0.095** (0.021)
MMS, 1983-84 yen 2-week -0.074 (0.049)
3-month -0.054 (0.037)
MMS, 1984-86 yen 1-week -0.129* (0.053)
1 -month -0.091 (0.046)
Economist, 1981-85 yen 3-month 0.142* (0.054)
6-month 0.121** (0.038)
12-month 0.150* (0.054)
1 percent
a
level.
b
on the sign and magnitude of k that are reported in two studies by

Frankel and Froot (1987a, 1987b). First, the signs of k again tend to
differ for different time horizons, although the distinction between
short-run and long-run expectations is less clear-cut than in the case of
extrapolative expectations. For the MSS data containing the expec-
tations over one-week to three-month horizons, the sign of k was unani-
mously negative, indicating that unanticipated appreciation for the cur-
rent period (that is, an increase in value of the right-hand-side
expression in equation (6a)) leads to continued expected appreciation
(that is, a fall in value of the left-hand-side expression).
On the other hand, for the long-run expectations of three to twelve
months, the sign of k was generally positive; the negative estimates in
Frankel and Froot (1987a) were statistically insignificant. Although
these results are not conclusive, they are consistent with an expectations
mechanism under which unanticipated appreciation leads to expected
depreciation in the longer run. This may reflect the belief of market
participants that there is a long-run "normal" level for the exchange
rate. In all cases, the absolute values of all the coefficients were less than
unity, suggesting that expectations were stabilizing.

176 SHINJITAKAGI
Regressive Expectations
The last expectations mechanism discussed in the literature is called
regressive expectations, and has the general form:
or
where s, is the long-run equilibrium exchange rate. This mechanism is

called regressive for the obvious reason that the actual exchange rate is
assumed to regress toward the equilibrium exchange rate (at the speed of
adjustment given by q)\ in this formulation, the expected exchange rate
can also be expressed as a weighted average of the current exchange rate
and the long-run equilibrium exchange rate.
Obviously, the estimate of q would depend on which model is used to
specify the equilibrium exchange rate (s,); popular candidates in the
literature have been constants, moving averages, and purchasing power
parity (PPP) exchange rates. If the sign of q is found to be positive, on
the one hand, the expectation is regressive, such that the exchange rate
is expected to move in the direction of the specified long-run equilibrium
rate; on the other hand, a negative sign for q would indicate an expecta-
tions mechanism under which the exchange rate is expected to deviate
from the long-run equilibrium rate.
Table 9 summarizes some of the representative regression results re-
ported in the literature. In general, the sign of q was found to be negative
for the short-run expectations of one week to one month; the sign of q
was positive for the long-run expectations of six and twelve months (ex-
cept when the coefficient was statistically insignificant); and for the
three-month expectations, the signs were ambiguous. The negative sign
of q for the short-run expectations suggests that exchange rate expecta-
tions can be destabilizing in the short run. Once again, the same striking
contrast in behavior is evident between short-run and long-run expecta-
tions as in the case of extrapolative and adaptive expectations.
VI. Short-Run and Long-Run Expectations
Regardless of which formation mechanism is assumed, short-run

(generally shorter than one month) and long-run (generally longer than
three months) expectations have been shown to display strikingly differ-
ent behavior. There is a tendency for short-run expectations to respond
to lagged exchange rate movements in the same direction or move away

Table 9. Regressive Expectations: Some Regression Results

Study Currency3 s Horizon 4
MMS, 1983-84 allb constant 3-month 0.010 (0.016)
Economist, 1981-85 3-month 0.036** (0.010)
6-month 0.076** (0.013)
12-month 0.172** (0.016)
Amex, 1976-85 6-month -0.000 (0.024)
12-month 0.079* (0.035)
MMS, 1983-84 PPP 3-month -0.021 (0.015)
Amex, 1976-85 6-month 0.032 (0.020)
12-month 0.124** (0.028)
MMS, 1983-84 yen PPP 2-week -0.094** (0.020)
3-month -0.167** (0.037)
MMS, 1984-86 1-week -0.042** (0.012)
1 -month -0.075** (0.019)
Economist, 1981-85 3-month -0.024 (0.057)
6-month -0.036 (0,057)
12-month 0.080 (0.086)
Amex, 1976-85 6-month 0.022 (0.048)
12-month 0.119** (0.059)
Bank of Japan (1989)
JCIF, 1985-89 yen 6-month 1 -month -0.085** (0.014)
moving 3-month 0.088** (0.026)
average 6-month 0.374** (0.031)
Economist, 1981-88 allb PPP 3-month 0.010 (0.014)
6-month 0.048** (0.018)
12-month 0.143** (0.030)
1 percent
a
level.
b
from some long-run "normal" values, while long-run expectations tend

to respond to lagged movements in an opposite direction or move to-
ward the long-run normal values. This behavioral difference has gener-
ated considerable interest in recent years in terms of how to characterize
such behavior ("consistency") and how to explain it ("chartists and
fundamentalists").

178 SHINJITAKAGI
Consistency
As a way to capture the apparent behavioral difference between short-

run and long-run expectations, Froot and Ito (1989) recently proposed
an analytical concept called consistency, which is a weaker condition of
rationality; it is weaker because it does not require that expectations
have a certain relationship with the actual exchange rate process. It is
important to note, however, that consistency is a model-based concept,
such that rejection of the consistency hypothesis is a rejection of the
joint hypothesis of consistency and a particular expectations formation
process.4
To simplify exposition, assume that the expected one-period deprecia-
tion from t to t + 1 is determined by the lagged one-period depreciation
from t — 1 to r, as follows:
Equation (8) expresses the formation mechanism of short-run expecta-

tions. By updating equation (8), the expected one-period depreciation
from t + k-ltot + k can be expressed as
This means that the expected fc-period depreciation from t to t + k can

be expressed in terms of the one-period lagged depreciation from t — 1 to
t, as follows:
Equation (10) is the expected ^-period expectation from t to fc, as

obtained from sequentially updating the expected one-period expecta-
tion by k times. Now, assume that the expected fc-period depreciation
from t to k can also be directly determined by the one-period lagged
depreciation from t — 1 to t, as follows:
Equation (11) is the formation mechanism of long-run expectations,

4
Boughton (1988) suggests an interesting possibility that short-run and long-
run markets are segregated in terms of market participants. According to this
interpretation, the expectations of foreign exchange participants in each market
can be rational even it they are not consistent with those in the other market.

analogous to equation (8) for short-run expectations. Consistency, de-

fined by Froot and Ito (1989), requires that the A>period expectation
obtained iteratively from short-run expectations (equation (10)) equal
the same A>period expectation obtained directly (equation (11)). In
terms of this simple model, consistency thus imposes the following cross-
equation restriction when equations (8) and (11) are estimated:
The restriction (12) indicates that, as long as the absolute value of Wi is

less than unity, the sign of wk must be the same as the sign of H^,
indicating that the "twist" observed between short-run and long-run
expectations in the previous section cannot be consistent. However, such
a conclusion is valid only if one accepts an autoregressive process of
order one as the correct expectations formation process. For an al-
ternative time-series process, it can be shown that consistent expecta-
tions can generate a twist. Tests of consistency are thus conditional upon
the hypothesis about the correct expectations formation process.
Froot and Ito (1989) estimated a system of either two or three equa-
tions subject to the cross-equation consistency restrictions; they used
both first-order and second-order autoregressions to describe the expec-
tations formation process. According to their estimation results (Table
10), consistency could not be rejected for the immediate horizon of
one-week to one-month expectations, indicating that these very short-
run expectations behave in a similar fashion. In contrast, for the horizon
encompassing three, six, and twelve months, one to three months, or
one to six months, the cross-equation consistency restrictions were
unanimously rejected for the expectations of different horizons at least
two months apart. Froot and Ito attributed this finding to the observa-
tion that short-run expectations tend to overreact relative to longer-run
expectations when the exchange rate changes.
Chartists and Fundamentalists

As a way to explain the difference between short-term and long-term
expectations, Froot and Frankel (1990) suggested that participants in the
foreign exchange market may be using two types of forecasting tech-
niques (see also Frankel and Froot (1986, 1988)). It may be that, for the
short-run forecasts, the predominant method is "chartist" or technical
analysis; and that, for the longer-run forecasts, the common method is
fundamental analysis based on such "fundamental" variables as PPP.

180 SHINJI TAKAGI
Table 10. Consistency: Summary Results

Data Set Horizon Currency8 AR order x 2b
Economist, 1981-87 3, 6, 12 months pound 1 12.7*
2 19.5*
mark 1 32.3*
2 60.1*
yen 1 73.6*
2 142.9*
MMS, 1984-87 1 week, 1 month pound 1 0.8
2 1.4
mark 1 1.8
2 11.0**
yen 1 1.9
2 11.2**
MMS (UK), 1984-87 1 week, 1 month pound 1 1.6
2 1.4
mark 1 0.1
2 2.1
yen 1 0.1
2 2.3
JCIF, 1985-87 1,3 months yen 1 546.1**
2 832.1**
3, 6 months yen 1 1483.3**
2 1135.9**
Source:
a
Froot and Ito (1989).
b
Wald statistics on the cross-equation restrictions imposed by consistency;
(**) indicates that the restrictions are rejected at the 1 percent level of signifi-
cance.
They present evidence based on annual Euromoney surveys that the

weight of chartists in the market has been increasing in recent years. To
the extent that, at least in the short run, the market is dominated by
chartists who concentrate on the recent pattern of price movements,
their increasing influence may have been a factor in the increased volatil-
ity of exchange rates in recent years.
It is, however, too early to give a final verdict on the cause of exchange
rate volatility. A recent study by Allen and Taylor (1989), based on a
survey of over 200 foreign exchange market practitioners in London,
questions the notion that chartism necessarily increases volatility. Ac-
cording to this study, Allen and Taylor find that, while chartism is the
most actively used forecasting method over short-time horizons (intra-
day to one week), it is by no means used exclusively. Both the chartist
and fundamentalist methods are often used in a complementary manner,

with the latter assuming greater weight at longer horizons. Moreover,

there was no evidence that chartist expectations overreacted to changes
in the current spot exchange rate and were thus destabilizing.
VII. Conclusions and Some Policy Implications
This paper has presented a brief survey of the empirical literature on

survey-based exchange rate expectations. After summarizing the fea-
tures of survey data sets and the broad descriptive characteristics of
survey exchange rate expectations, the paper reviewed the empirical
literature under the headings of forward discounts and risk premia,
rationality of expectations, mechanisms of expectations formation, and
short-run and long-run expectations.
Three broad generalizations emerged from the survey of the empirical
literature. First, survey data generally indicated the presence of a
nonzero risk premium, which appeared to be stable and uncorrelated
with the forward discount. This means that changes in the forward dis-
count fully reflect changes in the expected exchange rate, so that assets
denominated in different major currencies can be considered to be per-
fect substitutes, at least for changes in fundamentals observed during the
sample period.
Second, empirical tests were generally unfavorable to the hypothesis
that exchange rate expectations are rational in terms of both unbiased-
ness and orthogonality. Except for certain time periods and horizons,
survey expectations were shown to be biased predictors of future ex-
change rates, and the forecast errors were correlated with some variables
that are known to be in the set of information available when the
expectations are formed. Given the extraordinary nature of some sam-
ple periods, however, the rejection of the rationality hypothesis may be
saying more about the peculiarity of actual exchange rate movements
than the nature of exchange rate expectations.5 Lewis (1989), for exam-
ple, suggests that systematic forecast errors can be consistent with the
behavior of rational agents who are learning about the new process
governing fundamental economic variables.
Third, the most substantive conclusion that emerges from the lit-
erature concerns the consistently observed behavioral difference be-
5
The rejection may also reflect the so-called peso problem which is a finite
sample bias attributable to the failure of correct expectations to materialize
during the sample period. In the earlier period of dollar appreciation, for exam-
ple, the consistent bias in expectations might have reflected the expectations of
rational agents who correctly perceived the dollar to be "too high."

182 SHINJITAKAGI
tween short-run (generally shorter than one month) and long-run (gen-
erally longer than three months) expectations. Short-run expectations
tend to respond to lagged exchange rate movements in the same direc-
tion and move away from some long-run "normal" values, while long-
run expectations tend to respond to lagged movements in an opposite
direction and move toward the long-run normal values. This suggests the
possibility that the foreign exchange market in the short run reflects an
element of "noise trading," trading that is based on factors other than
"fundamentals."
The conventional wisdom among economists has long held that such
noise traders, who base their trading on factors other than market funda-
mentals, would on average buy high and sell low, and thus would be
driven out of the market. De Long and others (1987), however, have
recently proposed a model in which such noise traders may survive in the
long run, even if they do buy high and sell low on average. In this model,
noise traders are rewarded with a higher return for the greater risk they
assume; moreover, the greater risk introduced by noise trading would
lead rational investors to demand a higher return on risky assets, such
that the asset prices can deviate from their fundamental values. Noise
trading could also cause a greater volatility of price movements relative
to what is warranted by the movements of long-run fundamentals.
If it is indeed the case that a major portion of the short-run volatility of
exchange rate movements is attributable to the chartist nature of short-
run exchange rate expectations, the literature may provide justification
for some type of policy measure to intervene in the foreign exchange
market in the short run. Such a measure might be a fixed transactions tax
in the foreign exchange market, which would increase the cost of short-
run trading relative to long-run trading. A majority of economists, how-
ever, would probably remain skeptical of such a policy recommendation
until they could be convinced of a firm theoretical link relating average
expectations to marginal prices in the foreign exchange market.
REFERENCES
Allen, Helen, and Mark P. Taylor, "Charts, Noise and Fundamentals: A Study
of the London Foreign Exchange Market," CEPR Discussion Paper No. 341
(London: Centre for Economic Policy Research, September 1989).
Bank of Japan, Research and Statistics Department, "External Balance Adjust-
ment and Monetary Policy Management under International Private Capital
Flows" (Tokyo: Bank of Japan, November 1989).
Boughton, James M., "Exchange Rates and the Term Structure of Interest
Rates," Staff Papers, International Monetary Fund, Vol. 35 (March 1988,
pp. 36-62).
De Long, J. Bradford, Andrei Shleifer, Lawrence H. Summers, and Robert J.

Waldmann, "The Economic Consequences of Noise Traders," NBER

Working Paper 2395 (Cambridge, Massachusetts: National Bureau of Eco-
nomic Research, October 1987).
Dominguez, Kathryn M., "Are Foreign Exchange Forecasts Rational? New
Evidence from Survey Data," Economics Letters, Vol. 21, No. 3 (1986), pp.
277-81.
Frankel, Jeffrey A., and Kenneth A. Froot, "Understanding the Dollar in the
Eighties: The Expectations of Chartists and Fundamentalists," Economic
Record, Supplement (1986), pp. 24-38.
(1987a), "Using Survey Data to Test Propositions Regarding Exchange
Rate Expectations," American Economic Review, Vol. 77 (March), pp.
133-53.
(1987b), "Short-Term and Long-Term Expectations of the Yen/Dollar
Exchange Rate: Evidence from Survey Data," Journal of the Japanese and
International Economies, Vol. 3 (September), pp. 249-74.
-, "Chartists, Fundamentalists and the Demand for Dollars," Greek Eco-
nomic Review, Vol. 10 (June 1988), pp. 49-102.
Froot, Kenneth A., and Jeffrey A. Frankel, "Forward Discount Bias: Is It An
Exchange Risk Premium?" Quarterly Journal of Economics, Vol. 104
(February 1989), pp. 139-61.
, "Exchange Rate Forecasting Techniques, Survey Data, and Implications
for the Foreign Exchange Market," IMF Working Paper 90/43 (Washington:
International Monetary Fund, 1990).
Froot, Kenneth A., and Takatoshi Ito, "On the Consistency of Short-Run and
Long-Run Exchange Rate Expectations," Journal of International Money
and Finance, Vol. 8 (December 1989), pp. 487-510.
Ito,Takatoshi, "Foreign Exchange Rate Expectations: Micro Survey Data,"
American Economic Review, Vol. 80 (June 1990), pp. 434-49.
Lewis, Karen K., "Changing Beliefs and Systematic Rational Forecast Errors
with Evidence from Foreign Exchange," American Economic Review, Vol.
79 (September 1989), pp. 621-36.
MacDonald, Ronald, and Thomas S. Torrance, "Some Survey-Based Tests of
Uncovered Interest Parity," in Exchange Rates and Open Economy Macro-
economics, ed. by Ronald MacDonald and Mark P. Taylor (Oxford and
New York: B. Blackwell, 1989).
Mussa, Michael, "Empirical Regularities in the Behavior of Exchange Rates and
Theories of the Foreign Exchange Market," Carnegie-Rochester Confer-
ence Series on Public Policy, Vol. 11 (1979), pp. 9-57.
Takagi, Shinji, "On the Statistical Properties of Floating Exchange Rates: A
Reassessment of Recent Experience and Literature," Bank of Japan Mone-
tary and Economic Studies, Vol. 6 (May 1988), pp. 61-91.
Taylor, Mark P., "Expectations, Risk and Uncertainty in the Foreign Exchange
Market: Some Results Based on Survey Data," Manchester School of Eco-
nomic and Social Studies, Vol. 57 (June 1989), pp. 142-53.
Wakita, Shigeru, "Are Survey Forecasts Trusted? American Trade Account
Deficit and Yen/Dollar Rate," Economics Letters, Vol. 29, No. 4 (1989), pp.
339-44.

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
Currency Bands, Target Zones,

and Price Flexibility
MARCUS MILLER and PAUL WELLER*
Exchange rate behavior is analyzed in the context of a stochastic rational

expectations model in which there are random shocks to the price-setting
mechanism and in which the authorities choose to impose either nominal
or real exchange rate bands. The effects of rules for realignment of the
band are also examined. Results are compared with those that emerge from
a simple monetary model subject to velocity shocks. [JEL 431, 432]
S conducted
INCE 1979 macroeconomic policy in many European countries has been
within the constraints imposed by the exchange rate com-
mitments of the European Monetary System (EMS). For member coun-
tries like France and Italy these arrangements have provided a useful
anti-inflationary anchor in the form of an exchange rate link against a
hard currency—the German mark; and the surrender of national mone-
tary autonomy has been more obvious as recourse to realignments has
been reduced and the pace of capital market integration deliberately
accelerated. (In place of the current asymmetric arrangement, the De-
* Marcus Miller is Professor of Economics at the University of Warwick and

co-director of the International Macroeconomics Program at the Center for
Economic Policy Research, London. He holds a Ph.D. from Yale University and
has taught at the London School of Economics and Political Science and
Manchester University. This paper was written while he was a Visiting Scholar in
the European Department of the Fund.
Paul Weller, who was also a Visiting Scholar in the European Department
when this paper was written, is Associate Professor in the Department of Fi-
nance at the University of Iowa and a Research Fellow at the Center for Eco-
nomic Policy Research. He holds a Ph.D. from Essex University and has taught
at Warwick University and Cornell University.
The authors are grateful to Massimo Russo, Robert Flood, Peter Garber, and
Alan Sutherland for useful discussions.
184

CURRENCY BANDS AND PRICE FLEXIBILITY 185
lors Report1 recommended the creation of a Central Bank for Europe to

conduct monetary policy for the EMS.) At the global level, the Group of
Seven industrial countries have been trying since 1987, with more or less
success, to keep their currencies within informal bands against the U.S.
dollar—although the extent of policy coordination involved has (like the
bands themselves) been much less explicit.
In this paper we discuss what defending a currency band implies for
monetary policy (at the edge of the band) and for the exchange rate
within the band, using a version of the popular Dornbusch model where
there are stochastic inflation shocks. We look both at nominal bands,
which may be fully credible or subject to known rules for realignment,
and at real currency bands (or target zones) as advocated by John
Williamson since 1983.
The techniques used for the purpose were first developed by Paul
Krugman in the context of a full employment, flexible price model where
the shocks were those affecting the velocity of money. The explicit
closed-form solutions that he and others have obtained to analyze the
consequences of exchange rate commitments do not, however, carry over
to the case where there is price inertia. Qualitative techniques (or nu-
merical solutions) have to be used instead. But the principles underlying
these qualitative solutions are essentially the same as for the monetary
model.
To provide both an accessible introduction to the method of analysis
and a benchmark for comparison, we begin with an outline of the
methodology used and key results obtained for currency bands in the
flex-price model. (Real bands cannot sensibly be discussed in this model,
which assumes the real rate is constant.) We also discuss the "regime
switching" that arises when reserves are limited.
The stochastic model of price inertia is presented in Section II, and the
solutions associated with it are described (with the relevant analytical
background supplied in the Appendices). We look first at the stabilizing
effect of currency bands on the exchange rate. Although the S-shaped
curve obtained resembles that for the monetary model, we note that it is
associated with discrete monetary intervention (and argue that it is at-
tributable to a locally reversible regime shift rather than to the regula-
tion of fundamentals). Then the effect of a known realignment rule in
"undoing" the stabilizing influence of the currency band is discussed, as
is its influence on the real exchange rate. Lastly, the model is trans-
formed into real terms, so as to consider the consequences of implement-
ing Williamson's target zones.
1
The report presented to the European Council in April 1989 by the Commit-
tee for the Study of Economic and Monetary Union.

186 MARCUS MILLER and PAUL WELLER
I. The Flex-Price Monetary Model
Perhaps because of its analytical tractability, the monetary model has

been widely used to study the behavior of exchange rates in a stochastic
environment where there are state-dependent changes in the conduct of
monetary policy; see, for example, Krugman (1988), Froot and Obstfeld
(1989), Flood and Garber (1989), and Bertola and Caballero (1990). In
what follows we describe both the general solution to the model and the
way in which different regimes imply specific boundary conditions to
identify the particular solution. Closed-form results are reported, but we
concentrate on a qualitative account of solutions and boundary condi-
tions, to facilitate comparison with our treatment of the Dornbusch
model in the next section.
The essential equations are as follows:
Equation (1) states the condition for equilibrium in the money market.
On the left is the domestic money supply measured in logs and denoted
m; on the right are the determinants of demand, where p is the log of the
price level, y is the log of full employment output, v measures cumulative
shocks to velocity, and E(ds)/dt is the expected rate of change of the
exchange rate, with s denoting the log of the domestic currency value of
foreign currency. (This definition of the exchange rate is the inverse of
standard usage, but it has the useful feature that, under a neutral mone-
tary expansion, the exchange rate rises along with other prices.)
In Krugman's model of currency substitution, the term E(ds)/dt, the
expected capital gain on foreign exchange, is a direct measure of the
opportunity cost of holding domestic currency. But models with liquidity
preference, where the opportunity cost of holding money is given by
the (short) rate of interest on bonds, lead to the same solution paths for
the exchange rate, as Froot and Obstfeld (1989) have shown. Changes in
the term E(ds)ldt will also measure changes in the incentive to hold
money as long as international interest differentials are based on cur-
rency arbitrage, so that i = i* + E(ds)ldt, where i and i* are domestic
and foreign interest rates, respectively (and, for convenience, i* is taken
to be constant).
Equation (2) states that purchasing power parity (PPP) always holds,
so (in logs) the domestic currency price of any good equals its foreign
price plus the price of foreign currency. (The asterisk denotes the vari-

able in the foreign country.) Equation (3) indicates that the velocity
variable, v, follows a Wiener process with variance a2. Thus, velocity
follows the continuous time equivalent of a random walk.
Domestic output is exogenous, as are all the foreign variables, so they
may be set to zero (or subsumed into the velocity variable) to yield
where k = m + v denotes "economic fundamentals." Thus, s — k is

positively correlated with E(ds)ldt in such monetary models.
Krugman (1988) and Froot and Obstfeld (1989) describe a two-step
procedure, which may be followed to solve explicitly for exchange rate
behavior. First, find the family of functions of the form s =f(k) that
satisfies equation (4) when velocity evolves according to equation (3),
with the money supply held constant. Second, select whichever one of
these solutions satisfies the boundary conditions appropriate to the cur-
rency regime in force. The general solution is derived algebraically in
Appendix I; and the qualitative nature of the family of solution paths can
be studied with the aid of Figures 1 and 2.
For a currency band (s;, J) symmetrically placed around zero, so that
5 = —s, the solution can be written in the form
where p = V(2/Xcr2), and the parameter A is determined by the boundary

conditions; the variety of solutions obtained for a constant money supply
is shown by the three paths drawn passing through the origin of Figure 1.
(Initially we assume that m = 0, so k = v; the effects that subsequent
intervention may have on the money stock are considered later.) The
line with a slope of 45 degrees is the free-float solution (FF), where, in
the absence of bubbles, s = k and E(ds)/dt - 0. Thus, if the money
supply is fixed and no regime change is expected, the exchange rate
follows the same continuous-time random walk as velocity.
Since PPP is always preserved, the general price level must move in
line with the exchange rate, so it is evident that constancy of the money
supply is no guarantee of price stability. Despite its name, the monetary
model does not, in these applications, lead to "monetarist" conclusions.
On the contrary, the assumed behavior of velocity implies that active
intervention, not a fixed rule for money, is needed to secure price stabil-
ity; and the other solutions reflect the expectations of such intervention.
Those solutions shown as R'OR and B'OB in Figure 1 are nonlinear
and diverge progressively from the 45-degree line for values of fun-
damentals increasingly far from the origin. Formally, this divergence
reflects the nonlinear terms in equation (5), but it is not difficult to see

Figure 1. A Free Float, a Currency Band, and Realignment Prospects
how the curvature of these two solutions follows from the behavioral
assumptions already made.
Consider, for example, points N and B, on and below the free float
solution, respectively. Given that m = 0, at both points the velocity vari-
able is v = fc>0. At N, on the 45-degree line, the total increase of
velocity above zero is exactly matched by the increase in the exchange
rate (and in prices); money market equilibrium is thus preserved at N
without any change to E(ds). At point By however, the exchange rate
(and/?) has risen by less than v, so some extra inducement is required to
ensure that the demand for money matches the given supply. The con-
cavity of the function / at B, together with the diffusion of the velocity
variablev, implies that E(ds)/dtis negative (in fact, E(ds)ldt = ^f"(K)l2\
and this—the anticipated capital loss on holdings of foreign currency—
supplies the necessary inducement. (In models without direct currency
substitution but with financial arbitrage, E(ds)/dt<0 leads to a fall in

the international interest differential, i — /*, which increases liquidity

preference indirectly.) Conversely, the convexity of the function at A
ensures that the incentive to hold domestic money will be reduced as
necessary to preserve money market equilibrium when real balances are
lower than at N.
Which of these solutions is relevant for the exchange rate in a currency
band depends on what rules are being used to implement it. In what
follows the boundary conditions associated with marginal and intramar-
ginal intervention are both described with reference to Figure 1, as is the
effect of anticipated realignments.
A Nominal Currency Band with Marginal Intervention

It is clear that the exchange rate could be perfectly stabilized in this
model if the money stock were to be continuously adjusted so as to offset
all velocity shocks (dm = —dv = —crdz). But what happens if there is no
accommodation as long as the exchange rate is within the band (s,s), so
that the money stock is adjusted only when the exchange rate hits the
edges of the band?
For the case where the band is symmetrical around zero (so that
5 = —s) and where the intervention is infinitesimal, Krugman (1988)
showed that the relevant solution should be tangent to the chosen bands;
that is, the S-shaped curve, B'OB in Figure 1.
The intuitive justification for this tangency condition is that any other
solution is inconsistent with money market equilibrium at the edge of the
band.2 Consider point R, for example, where the exchange rate has just
reached its upper limit, but the solution cuts the upper edge, as shown. If
v were to increase by a small amount, s would—in the absence of inter-
vention—be expected to rise above s, and this expectation is the only one
consistent with equation (4), as already discussed. Consequently, inter-
vention to hold s at s if v increases (but not to prevent a decline if it
decreased) will reduce E(ds)ldt at R and violate the conditions for equi-
2
A more formal justification of the "smooth-pasting" boundary condition
applied here is provided in the papers by Froot and Obstfeld (1989) and Flood
and Garber (1989). They observe that the problem studied by Krugman is
equivalent to that of regulating a Brownian motion process. They appeal to the
results of Harrison (1985) and the simplified discrete-time argument of Dixit
(1988). The interpretation in terms of regulated Brownian motion is of consider-
able theoretical interest but is conceptually quite distinct from the notion of
"locally reversible" regime switches to which we appeal to justify the smooth
pasting in the Dornbusch model. Our logic is much closer to that applied by
Krugman and Rotemberg (1988) for analyzing "switches" to periods of tempo-
rary floating, see below.

librium. Only at point B where the exchange rate path is smoothly

tangent to the^ edge of the band will intervention (to prevent k from
rising above k} leave E(ds)ldt unchanged as is required to preserve
money market equilibrium.
Note that the amount of time that the exchange rate is expected to
remain on the edge of the band must be infinitesimally short. To sup
pose otherwise means that over some finite period of time E(ds)ldt = 0;
but this is only true for paths along the 45-degree line FF (wher^
E(ds)/dt = (o2/2)f" = ty. It also follows that the adjustment to the
money stock necessary to defend the band itself must be infinitesimal, so
as not to disturb E(ds)ldt. As a consequence, the money stock will be
reduced infinitesimally each time the currency reaches the top of the
band (and increased when it reaches the bottom). Inside the band,
however, the money stock will remain constant, so the solution given in
equation (5) continues to apply. (The implications of these policies for
the behavior of interest rates inside a currency band in the monetary
model with liquidity preference are analyzed in Svensson (1989).)
A Nominal Currency Band with Intramarginal Intervention

Flood and Garber (1989) show how a different set of boundary condi-
tions, not involving smooth pasting, can sustain the same currency band.
They suppose that the monetary authorities announce a "discrete" inter-
vention rule, which specifies both the upper and lower limits of the
fundamental k (denoted U, L, respectively) at which intervention will
occur and the magnitude of the intervention at each limit. If, for exam-
ple, the points chosen are equidistant from the origin and the rule is
to accommodate exactly the accumulated velocity shock, then these
boundary conditions select the solution path LB'OBU, as shown in
Figure 1.
Assuming, as before, that initially m = 0, so k = v, then when velocity
variable v reaches its upper limit, [/, the authorities must immediately
offset it by reducing the money supply, setting m = — U. Since this inter-
vention is fully anticipated, there must be no discontinuous jump in the
exchange rate, and the solution paths for the system immediately after
the intervention must be those consistent with a lower money supply;
graphically, fundamentals must jump from U to O, and the exchange
rate will continue to satisfy the same solution path as before (shown as
LB'OBU) with fundamentals now measured by k = v - U until the
next intervention.
Note that any given currency band can be defended by any one of an
infinite number of (fully credible) intervention rules. The same path, for

example, could be selected by smaller discrete interventions at points

closer to the edges of the band. In the limit, as the size of interventions
tends to zero, the solution reverts to the S-curve identified by Krugman
(where fundamentals do not jump when intervention takes place, since
they are simply checked or "regulated" at k and k).
It is evident that the discrete intervention described by Flood and
Garber occurs at a point when the exchange rate lies strictly within the
interior of the band it is designed to support; that is, it involves intramar-
ginal intervention. Moreover, it is not locally reversible; thus, if an
intervention is provoked by a positive velocity shock, then an interven-
tion of the same size will not occur if there is an immediately subsequent
negative shock.
Realignment Prospects and the Inverse S-curve

Another example of an adjustment rule that is not locally reversible is
described by Bertola and Caballero (1990). They show how expectations
of realignment can reverse Krugman's S-shaped curve, selecting instead
a curve such as R'OR.
Their argument runs as follows. Assume that the authorities only
intervene when fundamentals reach preannounced points ct — b,ct + b,
where ct denotes the central parity applying at time t. Assume further
that, when fundamentals reach either of these limits, there will be dis-
crete intervention of Am = ±b either to take the rate back to its existing
central parity, or to take it to the center of a realigned band of the same
width, with a central parity
or
depending on whether the upper or lower boundary is reached; and that

the probabilities of either defense or realignment are known.
Given an initial central parity at the origin, and an upper intervention
point at b in Figure 1, then the curve shown as R'OR will arise, given a
sufficiently high probability IT of realignment, so that the expected capi-
tal gain on foreign currency in the case of realignment (a shift from R to
point O') is matched by the expected capital loss in the case of defense (a
shift from R back to the origin O). Since the increase in s above s on
realignment is less than the fall below s on defense, it is evident that the
assumed realignment probability along R'OR must be greater than one
half. It is important to emphasize that for a given value of b, s, which

identifies the width of the band, changes as IT is changed (alternatively, if

s is fixed while IT changes, b must change).
If the probability of realignment were zero, then, as the authors point
out, their rule would be one of the intramarginal intervention rules
considered by Flood and Garber. But as IT rises, the S-shaped curve
becomes steeper, coinciding with the 45-degree line when IT = |, and
reversing its curvature when fn>\. Note that for sufficiently large IT, and
certainly for TT > |, all interventions take place at the edges of the cur-
rency band. Bertola and Caballero say that their interest in such high
realignment probabilities is motivated by "the qualitative consistency of
an inverted S-shape with EMS evidence on exchange rate behaviour
within bands" (p. 10). In particular, they note that the inverted S-curve
produces a different asymptotic distribution for the exchange rate with
more weight on the middle of the band.
Currency Bands with Limited Reserves

In a recent paper examining how limited reserves will affect the sus-
tainability of a currency band, Krugman and Rotemberg (1990) show
how a "smooth-pasting solution" at the edge of the band can arise from
discrete adjustments to the money stock caused by a "locally reversible"
switch from a currency band to (temporarily) floating exchange rates.
The setting is the model of currency substitution where increases in v
represent increased demand for foreign currency and where any mone-
tary intervention designed to meet this demand involves depleting the
limited stock of official foreign currency reserves. If it is the case that the
money supply only varies when reserves do (so m = \n(R + C), where R
is the quantity of reserves and C is a constant) then one can draw as in
Figure 2 the locus that would correspond to a free float with zero re-
serves, shown as F'F'. Note that in this figure it is the velocity variable
itself that is measured on the axis, not the fundamental k = m + v, as in
Figure 1. For m = 0 (so that R = 1 - C), the solution for the exchange
rate is that already identified as B'OB\ but as reserves are depleted by
marginal intervention, the S-shaped curve will shift to the right. Krugman
and Rotemberg ask what would happen if the authorities were forced to
suspend sales of foreign currency at s when reserves finally run out, but
remained willing to purchase foreign currency at s. For such a locally
reversible regime switch between a currency band and a float, they
conclude that the smooth-pasting solution tangent to the end of the band
is appropriate; so the solution to equation (4) for the exchange rate
outside the band takes the form

Figure 2. A Currency Band, Limited Reserves, and Temporary Floating
where k = ln(C) + v, p = V(2/Xa2), and A2 is determined by the bound-

ary condition that f ( k ) = 0 when s =s. This is shown in Figure 2 as the
exponential curve AE, which asymptotically approaches the free-float
locus F'F' as v tends to infinity.
Note that at point A, reserves have already been depleted by interven-
tions that have shifted the S-shaped curve from tangency at B to tangency
at A. Even though reserves are still positive, any further increase in
velocity will take the rate outside the band, since it will trigger a stock
shift out of domestic money (a speculative attack), which will reduce
reserves to zero. The stock shift arises because of the change of curva-
ture from the concavity of the S-shaped curve A 'A (where E(ds) < 0) to
the convexity of AE (where E(ds) > 0), which enhances the attractive-
ness of foreign currency and depletes reserves to zero. The willingness of
the authorities to continue to purchase foreign currency at s means that a
subsequent decline in v, which drives the exchange rate back to point A,

will now cause a sudden speculative inflow into the domestic currency—
the exact reversal of the attack that caused the collapse.
What the various cases shown in Figures 1 and 2 have in common is
that the appropriate (rational expectations) path for the exchange rate is
tied down by the specific rules for financial policy believed to be in force.
These supply the boundary conditions needed to identify the particular
solution appropriate to the regime in question. Firm expectations of
intervention to defend a band, be it marginal or discrete, bend the rate
toward the center of the band, for example; whereas prospects of re-
alignment tend to do the opposite. Of particular interest is the observa-
tion that the smooth-pasting condition arises not only when there is
marginal intervention at the edge of the band but also when there is a
locally reversible shift of regime, as instanced by a temporary float.
The model with price inertia we now discuss is sbmewhat more compli-
cated, because the solution path for the exchange rate directly affects
fundamentals even within the band. But the same principles can be
applied to determine the impact on the exchange rate of expected de-
fense or possible realignment where the source of disturbance lies not in
the velocity of money but in supply-side shocks to inflation.
II. A Model with Price Inertia
Although financial markets are taken to be forward-looking through-

out our analysis, we focus on the simple case where the process of price
adjustment is not. Specifically, we use a stochastic version of the popu-
lar Dornbusch model (1976) where the process of price setting is a sim-
ple Phillips curve. (The procedures we use can be extended to include
forward-looking labor contracts, but we do not pursue this here.)
The following equations (7) through (10), to be used in this section,
are expressions of, respectively, equilibrium in the money market and
the goods market, currency arbitrage, and price adjustment.
where the variables are as defined in the last section.

Two of these behavioral equations are much as before, namely equa-
tion (7), which is the LM curve defining equilibrium in the money mar-
ket, and equation (9), the arbitrage condition for the foreign exchange

market. Note that disturbances to the velocity of money are omitted

here, however. Instead, the price adjustment equation—where prices
rise if gross national product (GNP) is high, and vice versa—is disturbed
by white noise shocks. With prices evolving in this way, there is no
guarantee that output will remain at any given level, nor that the ex-
change rate will remain at PPP. The log of the level of output, y, is
therefore taken to be demand determined, where the level of demand
depends on the real exchange rate (s - p) and on the interest rate, i
(equation (8)).3
The dynamics of the system can be summarized as
where
Alternatively, by redefining the variables s and p to denote deviations

from the long-run equilibrium, this may be simplified to
where A is as identified above. So long as the boundary conditions are in-

dependent of time, these stochastic differential equations may be solved
in two stages. First one determines a family of functions of the time-
independent form s =f(p)', and then one imposes the relevant bound-
ary conditions, in the form of currency bands around the equilibrium
exchange rate, for example.
Beginning at the first stage involves deriving a differential equation
defining the required family of functions, and examining its solutions
using a qualitative approach. The equation is obtained by applying Ito's
lemma to f(p) to yield
3
The choice of nominal interest rate here as the influence on output is for
simplicity only. Nothing of substance in our analysis changes if we work with the
real interest rate.

Taking expectations, conditional on information at time t, we obtain
and then substituting for the expected changes in s and p from equation
(12), we obtain the desired result:
where a^ denotes the appropriate element of the matrix A.

For specific parameter values, numerical methods may be used to
solve this nonlinear differential equation; but there is no general closed-
form solution. In this respect it differs from the analogous equation of
the monetary model, which has solutions of exponential form. The ab-
sence of closed forms is not a serious drawback, however, because for
stochastic saddlepoint systems there exists a reasonably straightforward
qualitative description of the family of solutions. Not surprisingly, their
properties are intimately related to key features of deterministic solu-
tions obtained when there is no stochastic disturbance—that is, when
a = 0—which for the "overshooting" Dornbusch model take the familiar
saddlepoint form shown in Figure 3 (see Appendix II for a more formal
treatment).
The nature of symmetric solutions for the system once it is disturbed
by stochastic shocks is shown in Figure 4 (see Appendix III for an analy-
sis of the relation between these stochastic solutions and the determinis-
tic phase paths). All the solutions shown pass through the origin,4 but
there are only two linear solutions, which are labeled 55 and UU, since
they correspond exactly to the stable and unstable eigenvectors of the
deterministic system. There are in addition an infinity of nonlinear solu-
tions, all of which are technically antisymmetric; that is,/(/?) = -/(-p),
and have a point of inflection at the origin.
Solutions with a slope at the origin less than that of the stable eigen-
vector diverge further and further below 55 for increasing values of p
(see, for example, the path ZZ). Those, such as WW, whose slope at the
origin lies between the two eigenvectors, initially exhibit the same
smooth divergence, but go through a point of inflection, after which they
approach UU asymptotically. Paths whose slope at the origin are steeper
than that of the unstable eigenvector are strictly concave to the right
4
The reason is that we intend to consider only currency bands that are sym-
metric around the equilibrium of the origin. For currency bands that are not so
placed, one would need to consider solutions that did not pass through the
origin.

Figure 3. Phase Diagram for Non-Stochastic Case
until they go through a point of inflection and approach UU asymptoti-

cally from above.5
It may be useful at this point to give an intuitive idea of how the
presence of "noise" in the inflation rate can generate so many more
solution paths linked to the origin and why they bend away from the
stable eigenvector, as shown. For concreteness, consider the three points
shown as M, A, and B, and the solutions passing through them. The first
point, M, lies on the stable manifold s = 65p, where, on differentiating
and applying the expectations operator, we find that E(ds) = QsE(dp).
5
Theie patterns of concavity and convexity shown by these solutions in a neigh-
borhoodDa of the
trie origin may be readily confirmed by the formal argument given in
Appendix III. Properties of the solutions in the large are less easy to prove, but
their pattern can be deduced relative to the saddlepath solutions shown in Figure
3 by rearranging equation (13), as is shown in Appendix III.

Figure 4. Stochastic Solutions Through the Origin
Since 6S < 0 in this case, and the price level is expected to fall to equi-
librium at the origin, an appreciation of the price of foreign currency is
implied; that is, E(ds) > 0; thus, given arbitrage, domestic interest rates
exceed those on foreign currency by the extent of the expected apprecia-
tion of the latter.
The second point, A, lies on YY, which represents price stability for
the deterministic model and expected price stability in the stochastic
case. At point A the interest differential is wider than at M, so arbitrage
requires a larger anticipated change in the exchange rate.6 Since A lies on
the line of expected stationarity (where E(dp) = 0, and s =/(/?)), there
5
It is evident from equation (11) that E(ds) is increasing in s, given p.

would be no expected change in s in the absence of inflation shocks. But

with noise in the inflation rate, the price level will depart from its current
value over time, with the upward curvature of the function / leading to
expectations of a higher exchange rate. (Formally, the upward bias re-
quired to match the interest differential is E(ds)ldt = (a2/2)/" = i — /*. )
At point B below the stable eigenvector, the opposite curvature is
required to satisfy arbitrage. There the interest rate differential is less
than at N, but both E(dp) and/' have increased in absolute value;7 so
E(ds) would rise were it not for the concavity of the solution (which
together with the noise reduces E(ds), as required for arbitrage).
A Nominal Currency Band
Having characterized the various solutions to the differential equa-

tion, we can proceed to the second stage and ask which of these is
relevant for any particular currency regime (that is, we need to impose
the boundary conditions associated with that regime). We consider first
the monetary policy adjustments required (and the boundary conditions
implied) by a commitment to keeping the exchange rate within a nomi-
nal band (5, s). Then we turn to the (very different) boundary conditions
implied by a realignment rule.8
As a preliminary, we note that to keep the exchange rate pegged does
not, as in the monetary model, involve the "perfect accommodation" of
the stochastic disturbances; what is needed instead is to offset, by mone-
tary policy, any effects that shocks to the price level might have on
interest rates. From equation (11), one finds that the money supply rule
necessary to achieve this is
where A = K*/ + X, and s is the level of the peg. Thus, in the overshooting
case (where 1 > KTI > 0), the money supply must increase with the price
level but less than in proportion; whereas in the undershooting case the
money supply has to be cut when the price level rises.
What, then, are the boundary conditions implied by adjusting mone-
7
Since, from equation (11), E(dp) is increasing in s, given p, the rate of
disinflation
8
must increase between M and B.
In both cases, we consider rules that are symmetric around the equilibrium;
otherwise, we would have to consider a wider set of solutions to the fundamental
equations (namely, all those that do not pass through the origin). The same
principles may, of course, be applied to asymmetric cases, such as one-sided
currency bands.

tary policy at the edges so as to maintain a currency band (s,s)? Once

again, the answer is that the stochastic solutions will be tangent to the
edges of the band. The result, shown as C'B1 OBC in Figure 5, looks
much like the S-shaped curve of the monetary model; but this similarity
conceals two important differences. First, the fundamental/? is not "reg-
ulated" when the edge of the band is reached. Consequently, the edges
of the band C'B' and BC also constitute part of the solution path (and
the exchange rate can spend finite periods of time at the top or bottom
edge of the band). Second, this smooth-pasting solution calls for discrete
changes to the money stock at the edge of the band.
One could indeed think of the proposed solution as consisting of three
segments or regimes (being pegged at s, being pegged at s, and floating
in between), with the regime switches being induced by stochastic shocks
to the price level. Therefore, as was argued in Miller and Weller (1990),
the reason for smooth pasting is not that there is a Brownian motion
process being regulated by marginal intervention, but that there are
locally reversible regime switches9 (compare Krugman and Rotemberg's
argument for smooth pasting in the monetary model discussed above).
As in their case, the money supply makes a discrete adjustment each time
a switch is triggered.
A solution is implemented as follows. The monetary authorities an-
nounce that when s hits s (or s), the money supply will be instantly
adjusted so as to set domestic interest rates equal to world levels. Let us
suppose that this occurs when/? is above its long-run equilibrium value,
say at /?i > 0 in Figure 5. So long as price shocks hold p above pi, the
money supply is adjusted according to the rule in equation (15), but once
p hits the value /?i, the money supply is again immediately returned to its
original level, and is then held constant so long as s < s < s. The required
monetary adjustments are shown in the lower panel of Figure 5.
It may not be immediately obvious why the smooth-pasting solution is
singled out by the intervention rule described above. There are, as we
have seen, many other solutions to the stochastic system that cut the
edge of the band, and might be considered possible candidates for a
solution. How can they be ruled out? Consider, for example, the stable
manifold itself. Suppose OA in Figure 5 were part of a solution path, and
that the exchange rate has just entered the support regime at A. This
means that the money supply has been adjusted so that i = i*. But over
the next short interval of time dt, p is more likely to move back toward
the origin than it is to move further away, lip falls, there will be a change
of order dt in the value of s (as the rate moves up OA), whereas ifp rises,
9
See Whittle (1983) for an account of matching conditions for diffusion pro-
cesses at prescribed boundaries.

Figure 5. A Nominal Currency Band with Discrete Monetary Changes

s stays fixed at s. This means that s is expected, on average, to rise. But

this is inconsistent with arbitrage, since / = i*. The same applies to all
other solutions that cut the edge of the band. Only if there is a tangency
at the edge of the band will the arbitrage condition be satisfied.10
The behavior of interest rates is significantly different from that im-
plied by the monetary model. In the case of exchange rate overshooting
being considered here, the higher the price level is, the higher interest
rates are—until the point at which the exchange rate enters the support
regime at the edge of the band, when there is a discrete downward jump
in interest rates to the world level i*. The length of time the rate spends
at the edge is random but could be quite short, since the price level
process is still mean reverting.
The effects described above depend crucially on the assumption that
the band is fully credible, so a return inside the band is confidently ex-
pected. Reaching the edge of the band may, however, trigger prospects
of realignment. We therefore consider next how various realignment
rules can affect the rate within the band, leading, for example, to a
reversal of the usual smooth-pasting result.11
Discrete Realignment of a Currency Band

Suppose that the authorities announce a rule of the following form:
whenever the exchange rate hits the top of the band, the band will be
shifted upwards by an amount equal to half the total width of the band;
and this realignment will be validated by a change in money supply
designed to shift the long-run equilibrium exchange rate to the center of
the new band. In this case, the exchange rate will follow the path A'OA,
snaking around the 45-degree line in Figure 6, since any other trajec-
tories would be found to imply fully anticipated gains or losses at the
edge of the band.
In the original band, the exchange rate will follow the solid line, but
when it hits the top edge of the band at point A, an upward realignment
is triggered. By construction, A becomes the long-run equilibrium for
the realigned band, and the rate now moves along the dotted trajectory
10
Note that for this argument to work it is essential that the switch from a
floating to a fixed rate regime be instantaneously reversible. If this is not the
case, then it is certainly possible that OA in Figure 5 could be part of the solution
path. For, once the switch occurred, price shocks in either direction would then
produce monetary adjustments to hold s at s, and there would be no violation of
the11arbitrage condition.
It is possible, however, to have stochastic prospects for realignment at the
edge of the band, which prevent interest rates from moving to world levels but do
not change the smooth-pasting condition; see Miller and Weller (1990).

Figure 6. Realignment Prospects and the Inverse-^ Solution
centered at A. If subsequent price shocks drive the system back to point

O, the realignment is reversed and the money stock is returned to its
original level. Note that there is no smooth pasting here in the transition
from one regime to another because of the locally irreversible nature of
the regime shift.12
12
A case of both mathematical and practical interest arises when one considers
the effect of a realignment rule where the top of the old band becomes the
bottom of the new one. This means that there is a reversible switch of regime at
the edge of the band, and so smooth pasting should apply. With the larger
realignment, the dotted reverse-S solution shown in Figure 6 will not overlap the
original solution, but will slide further up the 45-degree line and have its lower
end point at A', thus the smooth-pasting condition for the two solutions will be
satisfied. It is not smooth pasting on to the edge of the band, however, because
under the realignment rule considered there is no longer a transition to a (tempo-
rarily) fixed rate regime at the edge of the band.

Though these conclusions apply equally in the overshooting case dis-

cussed earlier, we have chosen to illustrate the impact of the realignment
rule in the undershooting case where KTI > 1 and the stable manifold has
a positive slope. The contrast between the S-shaped curve B'OB defined
by smooth-pasting conditions attached to a credible band, and the in-
verse S-shape induced by expectations of realignment is immediately
apparent; in Miller and Weller (1989), we noted that the various inter-
mediate cases lying between the smooth-pasting solution and the inverse
S-shape curve—including the stable manifold itself—can be obtained
when the realignment rule discussed above is applied probabilistically.
Bertola and Caballero (1990), who obtained similar results for the mone-
tary model, have stressed the need to incorporate the effects of realign-
ments in order to reconcile stochastic theories of currency bands with
observed reality (at least for EMS currencies up until 1987).
The realignment rules discussed here involve not just a shift in the
currency band but a discrete adjustment of monetary policy associated
with it. To the extent that they are triggered by price increases they imply
that local nonaccommodation gives way to global price accommodation.
Another proposal that shares these features is monetary control inside
real exchange rate bands, which will be examined next, after the model
has been recast in real terms. Reformulating the model in this way makes
clear that while real bands involve marginal monetary intervention, re-
alignments correspond to discrete monetary intervention along the lines
discussed by Flood and Garber (1989).
Target Zones
Williamson (1985) has argued forcefully that a sharp distinction
should be drawn between nominal and real currency bands; and his target
zone proposal is explicitly couched in terms of real currency bands. The
monetary model cannot be used here, since, under the assumption of
PPP, the real exchange rate is effectively constant. This is not the case in
a model with price inertia, however, and a number of important differ-
ences between a nominal currency band and a target zone emerge in that
context.
It is convenient to redefine variables in the model spelled out in equa-
tions (7)-(10), introducing the real exchange rate, c = s — p, and real
balances, / = ra -/?. The transformed model can then be written as

Where, again, the variables are expressed as deviations from long-run

equilibrium. The arbitrage condition now imposes the requirement that
the real interest rate differential be equal to the expected depreciation of
the real exchange rate.
The system is formally identical to that in equation (12), except that
the matrix A' differs from A in an obvious fashion. The only significant
qualitative difference is that the stable saddlepath now always has a
positive slope, as shown in Figure 7.
In order to enforce a target zone, the authorities must now concen-
trate on the relationship between real balances and the real exchange
rate. We assume, as before, that m is held fixed in the interior of the
zone. But a willingness to adjust the stance of monetary policy at the
edge of the band (by offsetting any price shocks that would take real
balances and the real exchange rate further from the center of the band)
means that policy is essentially one of regulating the fundamental—here,
real balances. (Compare this with the regulation of velocity-adjusted
money in the monetary model.) What this means is that infinitesimal
intervention at the limits to a target zone will lead to a solution path such
as A 'A for the real exchange rate as depicted in Figure 7. The real ex-
change rate will never remain for more than an instant at the edge of the
target zone, but will be reflected at the limits of its range of variation.13
In order to show what effect regulating real balances has on nominal
variables, we need to consider the target zone illustrated in Figure 8. Its
limits run parallel to the PPP line, and the path for the nominal exchange
rate within the zone will drift up and down as the money stock is adjusted
periodically to defend the zone. It is immediately clear that the behavior
of the nominal exchange rate is strikingly different within a target zone
involving perfect accommodation of shocks to prices at the edges, than it
is within a nominal currency band, where defense must involve reversi-
ble switches of regime. Figure 8 reveals that the impact of a target zone is
exactly the same as a nominal currency band whose limits are indexed to
changes in the money stock, when those occur each time the exchange
rate hits the edge of the band.
In the long run, both the money stock and the price level will follow a
random walk under a target zone arrangement. However, both will
display short-run stability. This is in contrast to what happens to money
and prices in the case of a nominal currency band, where both variables
13
Compare this with the discrete monetary intervention under the realignment
rules examined in the last section. Under those rules real variables can be shown
to follow a path shaped like the trajectory F'OF in Figure 7, with discrete
intervention shifting real balances directly from F to O (compare Flood and
Garber's (1989) analysis of discrete intervention).

Figure 7. Target Zone for the Real Exchange Rate
display a global stability, in the sense that they are always expected to
adjust in the direction of long-run equilibrium.
This discussion is subject to one important qualification. In the case of
a target zone, depending upon parameter values, there may exist a
second smooth-pasting solution, shown as B'B in Figure 7. This solution
is unstable and raises a potentially serious problem. If the intervention
rule is formed in terms of what will happen when the exchange rate,
rather than the fundamental, reaches a particular level, then the market
will have no means of distinguishing between the two possible trajecto-

Figure 8. Target Zone as an Indexed Currency Band
ries. This suggests the need for policy to be directed toward deterring the
market from embarking upon the unstable path.14
It has been shown that a target zone for the real exchange rate may be
implemented by infinitesimal adjustments to real balances. We note in
14
The idea that this second solution could be ruled out by a threat to adjust
monetary policy within the band, if ever the market were to move in the perverse
fashion that this solution implies, is discussed in Miller and Weller (1990). A
more ambitious scheme, to use fiscal policy to help stabilize nominal income, was
outlined in a blueprint for policy coordination (Williamson and Miller (1987)).

conclusion that the argument advanced by Flood and Garber (1989) that
such regulated solutions can alternatively be sustained by discrete inter-
ventions applies here too. An illustration is provided by the realignment
rule for a nominal band studied in the previous section, under which
discrete adjustments to the nominal money supply (and so to real bal-
ances) were triggered for certain at the edges of the band. It will be
found that the path followed by the real exchange rate under such a rule
is precisely of the form FA 'OAF shown in Figure 7, and that the shift in
real balances on realignment corresponds to the shift from F (or F') to
the origin O. This makes the important point that periodic realignment
of a nominal band is, for certain realignment rules, equivalent to de-
fending a fixed real band by means of discrete intervention. Other re-
alignment rules which involve larger or smaller discrete changes to real
balances are analyzed in Miller and Weller (1989). As Figure 8 suggests,
the smooth-pasting solution emerges in the limit where the change to
real balances on realignment shrinks to zero.
III. Summary and Conclusions
Policy intervention to stabilize exchange rates has led to the macro-

economic application of techniques more commonly found in finance
theory. In their study of the behavior of sterling before its widely antici-
pated return to the gold standard in 1925, for example, Flood and
Garber (1983) used the techniques of stochastic process switching. Ex-
tensive further use of these techniques has recently been made to study
the operation of currency bands, with a focus on the experience of the
EMS. This paper is designed to show that, although the technical details
may seem esoteric, the main results are reasonably accessible.
As is to be expected, the policy implications depend to a great extent
on the model used and the way in which shocks are introduced. We have
looked in some detail at both a monetary model with velocity shocks
and a Dornbusch-style model with shocks to the process by which prices
are set.
The monetary model has attracted a good deal of attention, not least
because it can be solved explicitly—and hence, subjected more easily to
empirical testing. Given cumulative shocks to velocity, a policy of con-
trolling the money supply cannot ensure price stability; and it is only by
intervening at some point to offset these velocity shocks that finite limits
can be set to these fluctuations in nominal values. But the credible ex-
pectation of such intervention can induce a good deal of stability before

that, as Krugman (1988) demonstrated with the S-shaped solution for

the rate inside a currency band.
Evidence from the EMS, however, strongly suggests that expectations
of exchange rate realignment have played an important offsetting role—
at least up until 1987—which can reverse this stabilizing effect. How
important is the limitation of official reserves needed to defend currency
bands in the EMS is a debatable issue (given the arrangements for
extending credit). But in any case, introducing reserve limits suggests the
need for considering locally reversible regime switches, even in the
monetary model.
Much of the policy concern with exchange rate fluctuations, however,
has been over the consequences for real exchange rates and international
competitiveness, issues that cannot be addressed in the monetary model.
In the main body of the paper, therefore, we have shown how the same
techniques can be used to study the operation of currency bands (and
their realignment) in a model with price inertia. Though the model is
different (fixing the money supply will, for example, ensure long-run
price stability), and the results cannot be expressed in closed-form solu-
tions, nevertheless one finds similar tendencies for currency bands to
stabilize the rate, and for realignment rules to undo this stabilization.
Nominal bands were found to require discrete intervention at the edge,
but this was not true of Williamson's target zones where marginal inter-
vention is sufficient. (The possibility of multiple equilibria inside the
target zone implied that intramarginal intervention might be required, if
only as a threat.)
The analysis in this paper can be considerably enhanced by allowing
for time dependence in the solutions. This is appropriate when the
market is expecting a policy shift at a reasonably predictable date, as was
true in the interval before the United Kingdom's entry into the exchange
rate mechanism of the EMS in October 1990. Fortunately, this is straight-
forward in principle (see for example, Ichikawa, Miller, and Sutherland
(1990)). As Svensson (1989) has shown, it is also possible to include
systematic realignment risk inside the band without much difficulty.
To assess the impact of currency stabilization measures in recent
years, it would be desirable to go further than the current literature, so
as to include multiple sources of disturbances and a more realistic ac-
count of wage price setting. Another important task for future research
is to introduce explicitly elements of market inefficiency into a model of
currency bands, since such inefficiencies are likely to be a major reason
for imposing a band in the first place. (For recent evidence on this issue
see Frankel and Froot (1987) and Cutler, Porterba, and Summers

(1990).) Development of a convincing theoretical framework within

which one can analyze the welfare implications of asset market ineffi-
ciencies and strategies to mitigate their effects is still at an early stage,
but recent work on noise trading holds much promise.
APPENDIX I
Stationary Solutions for the Monetary Model

To obtain the desired family of functions, one can express E(ds)ldt in terms of
the function f(k) and its first two derivatives, since by Ito's lemma
when the money stock is kept constant.

Substitution into equation (4) yields the equation
One of the great attractions of working with the monetary model is that this
differential equation has an explicit solution in terms of exponential functions,
namely:
where p = V(2/Xa2) andAi,A2 are constants of integration (see Krugman (1988)

and Froot and Obstfeld (1989), who show the effect of including a trend in
velocity).
APPENDIX II
Properties of the Saddlepoint Phase Diagram

In the absence of noise in the inflation process, equation (12) can be rewritten
as
where A has roots of opposite sign. The integral curves (phase curves) for such
a deterministic system shown in Figure 3 form a saddlepoint phase diagram.
Some of the properties of this phase diagram are useful for characterizing solu-
tions for the stochastic differential equations that result when noise is present.
For convenience, these are listed below.

Eigenvectors and Eigenvalues (Roots)

Each of the eigenvectors (shown as 55, UU in Figure 3) satisfies the condition
that
where 0 represents the slope of the eigenvector, and p denotes the eigenvalue,
one of the roots of characteristic equation [A — p/| = 0. It follows, by substitu-
tion, that these slope coefficients can also be obtained as the roots of the
quadratic equation
Notice that, from the signs of the coefficients appearing in A, the slope of the
unstable eigenvector must always be positive:
but the sign of 65 depends on specific parameter values:
as
When KTI < 1, the exchange rate overshoots the long-run equilibrium in response
to a change in the money stock; therefore, the stable eigenvector slopes down to
the right as in Figure 3. (The undershooting case, 05 > 0, is considered in the
text.)
Integral or Phase Curves
The solutions to the differential equation (18) are represented by the so-called
integral curves shown in Figure 3. The slope of these curves at any point,
denoted g(p,s) is given by the ratio of the two equations of motion:
See, for example, Petrovskii (1969) for further discussion.

These integral curves have turning points along the lines of stationarity shown
as YY (where dp = 0) and // (where ds = 0) in Figure 3. Notice that the slope of
YY (given by -an/a12 = (<$>1 + c|)X'n)/<t)X7i) is greater than unity, but less than 6M.
The slope of //, the locus of stationary points for the exchange rate, has the same
sign as 65, and is at least as large in absolute terms.

APPENDIX III
The Curvature of the Stochastic Solutions

The pattern of convexity and concavity off(p) near the origin can be examined
by expressing the curvature as a quadratic function of its slope at the origin, as
follows:
where
and
Note that #(6) is the quadratic expression already encountered in Appendix I,

with its roots 6M and 6, (the slopes of the eigenvectors); it is positive for
Qs < 6 < 6U, and negative otherwise.
The curvature of stochastic solutions elsewhere may be studied using algebraic
methods (as in Miller and Weller (1988)), but a simple geometric argument
follows directly from rewriting the differential equation (13) in the form
and substituting to obtain
Note that
where g(p,s) is the slope of the (deterministic) integral curve. Thus, there will be
curvature in the function /whenever its slope/' differs from that of the determin-
istic phase curve at the same point. Specifically sgn(/") = sgn (g — /') for
E(dp) > 0; while for E(dp) < 0, sgn (/") = sgn(f' - g).
To show this relationship, in Figure 9 we superimpose the integral curves from
the deterministic case on the stochastic solutions. Because of symmetry one
needs to consider only half the plane.
In the half plane on the right-hand side, consider first those paths where
E(dp)<Qm, that is, those lying below YY, the line of expected stationarity for
prices. In region B (beneath the line 55), g >/', so/" < 0, and the curves diverge
from 55 asp increases; whereas in region A (between 55 and YY), where/' > g,
the curvature is reversed.
Above the line of stationarity, however, points of inflection appear near the
unstable eigenvector. In region A' (above UU), /' falls as p increases, until
eventually it is tangent to an integral curve at point Ta. But the solution for the
tochastic system through Ta must always lie below the phase curve for the de-

Figure 9. Curvature of Stochastic Solutions
terministic system through the same point. It must therefore approach UU

asymptotically, as shown. A similar argument can be applied to the solution path
through the tangency point Tb.
Although the qualitative properties of the stochastic solutions are readily
apparent, obtaining exact solutions will typically involve numerical methods (of
shooting or power-series expansion), since the required integrals are not tabu-
lated. An exception is the case where there is simple mean reversion in the
fundamentals, so that dp = anpdt + vdz\ that is, a\2 - 0. In this case, where YY
and UU coincide with the vertical axis, the solutions can be found by using
tabulated values of the confluent hypergeometric function, as noted by Froot
and Obstfeld (1989) and Delgado and Dumas (1990).

REFERENCES
Bertola, Giuseppe, and Ricardo Caballero, "Target Zones and Realignments,"
CEPR Discussion Paper 3986 (London: Centre for Economic Policy Re-
search, March 1990).
Cutler, David M., James M. Poterba, and Lawrence H. Summers, "Speculative
Dynamics," NBER Working Paper 3242 (Cambridge, Massachusetts: Na-
tional Bureau of Economic Research, January 1990).
Delgado, F, and B. Dumas, "Target Zones Big and Small" (unpublished;
November 1990). Forthcoming in Currency Bands and Exchange Rate
Targets, ed. by P. Krugman and M. Miller (Cambridge: Cambridge Univer-
sity Press).
Dixit, Avinash, "A Simplified Exposition of Some Results Concerning Regu-
lated Brownian Motion" (unpublished; Princeton University, August 1988).
Dornbusch, R., "Expectations and Exchange Rate Dynamics," Journal of Polit-
ical Economy, Vol. 84 (December 1976), pp. 1116-76.
Flood, Robert P., and Peter M. Garber, "A Model of Stochastic Process Switch-
ing," Econometrica, Vol. 51 (May 1983), pp. 537-51.
, "The Linkage between Speculative Attack and Target Zone Models of
Exchange Rates," NBER Working Paper 2918 (Cambridge, Massachusetts:
National Bureau of Economic Research, April 1989).
Frankel, J., and K. Froot, "Using Survey Data to Test Standard Propositions
Regarding Exchange Rate Expectations," American Economic Review,
Vol. 77 (March 1987), pp. 133-53.
Froot, Kenneth, and Maurice Obstfeld, "Exchange Rate Dynamics Under
Stochastic Regime Shifts: A Unified Approach," Harvard Institute of Eco-
nomic Research Discussion Paper 1451 (September 1989).
Harrison, J.M., Brownian Motion and Stochastic Flow Systems (New York: John
Wiley and Sons, 1985).
Ichikawa, M., M. Miller, and A. Sutherland, "Entering a Preannounced Cur-
rency Band," Economics Letters, Vol. 34 (December 1990), pp. 263-368.
Krugman, Paul R., "Target Zones and Exchange Rate Dynamics," NBER
Working Paper 2481 (Cambridge, Massachusetts: National Bureau of Eco
nomic Research, January 1988).
, and Julio Rotemberg, "Target Zones with Limited Reserves," NBER
Working Paper 3418 (Cambridge, Massachusetts: National Bureau of Eco-
nomic Research, July, 1990).
Miller, Marcus, and Paul Weller, "Solving Stochastic Saddlepoint Systems: A
Qualitative Treatment with Economic Applications," Warwick Economic
Research Paper 309 (Coventry: University of Warwick (December 1988)).
, "Exchange Rate Bands and Realignments in a Stationary Stochastic
Setting," in Blueprints for Exchange Rate Management, ed. by M. Miller,
B. Eichengreen, and R. Portes (London; San Diego, California: Academic
Press, 1989).
-, "Exchange Rate Bands with Price Inertia," CEPR Discussion Paper 421
(Centre for Economic Policy Research, June 1990).

Petrovskii, G., "Ordinary Differential Equations," in Mathematics: Its Content,

Methods and Meaning, ed. by A.D. Aleksandrov and others, Vol. 1 (Cam-
bridge, Massachusetts: MIT Press, 1969).
Svensson, Lars, "Target Zones and Interest Rate Variability," CEPR Discussion
Paper 372 December 1989. NBER Working Paper 3218 (Cambridge, Mas-
sachusetts: National Bureau of Economic Research, December 1989).
Whittle, P., Optimization Over Time: Dynamic Programming and Stochastic
Control, Vol. II (New York: John Wiley, 1983).
Williamson, John, The Exchange Rate System, 2d ed. (Washington: Institute for
International Economics, 1985).
, and Marcus H. Miller, Targets and Indicators: A Blueprint for the Inter-
national Coordination of Economic Policy (Washington: Institute for Inter-
national Economics, 1987).

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
Shorter Papers and Comments

The 1990 Reform of U.K. Local
Authority Finance
TAMIM BAYOUMI*
In April 1990 the U.K. Government implemented the largest change to

local authority finance in England and Wales in the postwar period. It
involves the introduction of a poll tax on domestic residents, the central-
ization of local business taxes, and an overhaul of central government
grants. This paper analyzes the effect of these reforms on local govern-
ment and the wider economy. [JEL 324, 325]
NAPRIL 1990 the U.K. Government implemented the most significant

I change to local authority finance in England and Wales in the postwar
period. It involved the introduction of a completely new tax on domestic
householders (the community charge, a poll tax), the centralization of
the current local taxation of business, and an overhaul of the method by
which the Government allocates grants to local authorities.1
The most controversial of these proposals is the imposition of a
poll tax. The most famous attempt to impose such a tax in England (in
1381) ended in its withdrawal following the peasants' revolt. Historically,
poll taxes appear to have been highly unpopular, even by the standards
of taxes in general, and the current proposal has proved to be no excep-
* Tamim Bayoumi is an Economist in the Maritime Division of the European

Department. He is a graduate of Cambridge and Stanford Universities. The
author thanks Gordon Hughes, David King, Steven Smith, Jim Gordon, Graham
Haache, and Alan Tait for guidance and advice.
lr
The discussion here focuses on the arrangements for England. Those for
Scotland and Wales vary in some minor respects. In Scotland the most important
reform, the community charge, was introduced in April 1989. The discussion
also concentrates on the effects of the new system when it is fully operational,
rather than the implications of the transitional arrangements.
216

REFORM OF U.K. LOCAL AUTHORITY FINANCE 217
tion. The new tax has produced widespread opposition, the most dra-
matic manifestations being the largest riots in London in this century and
the biggest swing against a government in a parliamentary by-election
since 1935.
The controversy over the poll tax has overshadowed other aspects of
the reforms, such as the changes in the system of government grants and
centralization of local business taxation, which themselves represent a
significant change in the relationship between central and local govern-
ment. In this paper the effects of all aspects of the reforms are analyzed.2
I. The Objectives of the Government

and the Reform Proposals
Local authorities represent a major sector of the U.K. economy, with
spending representing 10 percent of gross domestic product (GDP) in
1988/89. In financing expenditure prior to April 1990, approximately
equal amounts were derived from local governments' own resources and
from central government grants. The major source from local resources
was a property tax, known as the rates, levied on all buildings, both
domestic (residential) and nondomestic (business), which was the sole
local tax. Central government grants had two functions: to equate tax
levels across different authorities for a given level of services (negating
the effects of differences in the local tax base) and to influence the level
of expenditure through variations in grant amounts.
The broad features of the reforms are contained in a consultative
document (Green Paper), Paying for Local Government (HMSO
(1986)). The objectives of the Government since 1979 have been to
contain local expenditure at affordable levels; to encourage local author-
ities to carry out services more efficiently; and to reduce detailed con-
trols over local government. In the view of the Government, local gov-
ernment spending was excessive due to a lack of accountability to the
local taxpayers. This lack of accountability was endemic to the system of
local authority finances, affecting all three principal forms of revenue, in
particular:
• Nondomestic rates (those rates paid by businesses) were paid by
organizations to which the local authority was not directly answerable.
• Domestic rates, which do affect voters, were directly paid by only a
2
This is particularly true given the recent announcement of a further review of
local authority finance, which will almost certainly result in a radical change to
the community charge.

218 TAMIM BAYOUMI
minority of the local electorate and varied in a way that had little regard
to the use made of local services.
• Central government grants were calculated in a complex manner,
which made it difficult for the local electorate to evaluate the true cost of
services being provided.
The reforms changed the way in which all three sources of revenue
operate. Domestic rates have been replaced by a new tax, called the
community charge (a poll tax), which is levied at an equal rate for each
adult, with rebates for the poorer members of society. Like the rate
rebates they replace, they cover up to 80 percent of the charge.
Nondomestic rates continue to be levied, but the poundage (the term
for the tax rate) is now set by the Central Government. Although the
revenues continue to be collected locally, they are paid into a central
government pool and then reallocated on a per adult basis. Hence, non-
domestic rates have changed from being a local tax to being, effectively,
a central government tax levied specifically to finance local authorities.
The system of government grants was changed and simplified. It now
aims to ensure that a given level of services provided implies an equiva-
lent community charge in different local authorities, with the level of the
grant being independent of actual expenditure. Hence, local authorities
must finance all marginal expenditure increases from the community
charge.
In order to understand the effects of these changes, consider two
authorities under both the old and new systems. One authority, labeled
N (for needy), has low property values and high per capita service needs.
The other, labeled P (for prosperous), has high property values and
relatively low needs. The situation under the rates is shown in the upper
panel of Figure 1 (based on King (1988b)). The vertical axis shows the
per capita level of services, and the horizontal axis represents the rate of
tax (poundage).
In the case of authority N, property values are low, and hence the lines
Tn and Dn, which represent the revenue from the all rates and domestic
rates, respectively, are less steep than the corresponding lines Tp and Dp
for the more prosperous authority. The lines In and Ip show income after
the general grants. The difference between line T (total local tax rev-
enue) and line / (total income) represents the level of government
grants. At a given level of tax (Tg), which is the same for each authority,
they pass through a fixed level of services, Sg, defined by the Govern-
ment as the grant-related expenditure assessment (GRE). Since N re-
quires more services, this level is higher for N than for P; the assessment
equates tax rates among different authorities for the level of services
given by the GRE.

Figure 1. The Two Systems of Local Finance
Note: Based on King (1988b).

220 TAMIM BAYOUMI
The income lines (In and Ip) are less steep than the respective tax lines
(Tn and Tp). This is because the Government sought to equate the
marginal increase in the tax rate for additional expenditure. Each extra
pound of expenditure per capita involved a rise of 1.1 pence in the
poundage of the local authority, which generally brought in over £1 in
revenue per capita. Hence, at higher poundages, the grant from the
Central Government was reduced, and the tax and income lines thus
move closer together.3 The income lines, In and Ip, also have a kink at a
level 10 percent above the Government's defined level of services (the
GRE), Sg. At this point, the marginal tax rate faced by the authorities
rises from 1.1 pence per pound of extra expenditure per capita to 1.5
pence per pound.
The lower panel of Figure 1 shows the arrangements under the new
system. The vertical axis still represents the level of expenditure and
revenue per capita, but the horizontal axis now represents the level of
the community charge. The lines Cn and Cp, representing the revenues
from the community charge, are parallel, since it is a poll tax. The lines
Tn and Tp, as in the upper panel, reflect total tax receipts; that is, the
sum of the community charge and business rates. Since receipts from
business rates are no longer under the control of local authorities, the
line Tn (or Tp) is parallel to the line Cn (or Cp). Lines I'n and Fp
reflect total income including central government grants. These income
lines pass through the projected expenditure points, Sg, at the same tax
rate, Cg. Since neither nondomestic rates nor government grants are
related to actual expenditures, this line is also parallel to the community
charge lines, Cn and Cp.
Comparing the top and bottom panels, several features emerge.
Under the old system, grants were calculated to equate poundages,
while the new system equates levels of the community charge. Hence, a
low-resource authority, like N, will have to raise a larger amount of
money from its tax base. Also, the marginal tax price of the community
charge on the domestic sector (the slope of lines Cn and Cp) is between
that of domestic and total rates (the slopes of the D and T lines, respec-
tively). The tax price rises for domestic rate payers and falls to zero for
nondomestic rate payers.
The reforms also change the tax paid by individuals within a given
authority. The move from a property-based tax to a per capita charge
involves a shift in the tax burden; those with high-value properties pay
less, and those with low-value properties pay more. The view of the gov-
3
The income lines, Ip and In, are parallel, since the Government ensured that
equal increases in the tax rate provided the same increase in expenditure per
capita across authorities.

ernment is that this change will eliminate the ability of councils to levy
high taxation on "prosperous" households in order to cross-subsidize
households paying relatively small amounts of rates.4
II. Analysis of the New System
This section focuses on the implications of the reforms for local

accountability and equity, business taxation, and the wider economy.
Accountability
The Government's argument is that the new system will improve the
accountability of local government. It will raise the number of taxpayers,
broaden the tax base, and reduce the number of people who receive
services without paying for them. In terms of the number of taxpayers,
there were over 35 million electors in England in 1985, while the number
of householders was under 18 million. Since only one person per
dwelling was formally a ratepayer, this implies that only half the elec-
torate actually paid rates. However, the effective incidence was certainly
much higher. For example, if spouses are added to the total, the number
of ratepayers would rise to 30 million people, making the rates a fairly
broad-based tax. Thus, although the community charge will have both a
higher actual and effective incidence than the rates, the differences are
probably not large.
A second issue relates to the marginal tax rate faced by domestic
taxpayers. On average, the cost of an extra pound of expenditure in
England (excluding London) under the old system was 69 pence if the
authority is to the left of the kink in Figure 1 (King (1988a, p. 142)).
Under the rates, a rise in spending across different local authorities
meant an equal increase in the property tax rate; hence, the marginal tax
price faced by the domestic sector was proportional to the domestic
ratable values. It is difficult to see an efficiency (as opposed to equity)
argument for such variations in the local marginal tax price. Under the
new system, the marginal tax price will be unified at 1. For many au-
thorities, particularly those with low property values, the new system
implies a substantial rise in the marginal cost of extra spending falling on
domestic taxpayers.
The significance of this change depends on the degree to which the
4
Interestingly, the tax price faced by the median voter in any given authority is
similar under the two systems (see Table 1 below).

222 TAMIM BAYOUMI
domestic tax price affects spending behavior. Under the rates, since the
domestic tax price was proportional to domestic ratable value per head,
the correlation between domestic ratable value and expenditures mea-
sures the importance of this connection. Figure 2 shows scatter plots for
four types of English local authorities: inner London boroughs; outer
London boroughs; metropolitan districts; and nonmetropolitan coun-
ties. The vertical axis shows budgeted spending in excess of GRE per
head in 1989/90, and the horizontal axis shows the domestic ratable value
per head. All four sets of authorities show the expected negative correla-
tion. Regressing excess spending against ratable values plus dummy
variables representing the different types of local authorities yields the
following results:
excess spending = 55.9 — 0.64 ratable value + 106.7 inner L

(11.3) (0.12) (15.2)
+ 53.0 outer L + 28.0 metropolitan
(11.1) (8.7)
R2 = 0.36 SE = 36.7 No. of observations = 104,
where excess spending is spending over GRE per head in pounds ster-
ling; ratable value is domestic ratable value per head; inner L, outer L,
and metropolitan are dummy variables representing inner London bor-
oughs, outer London boroughs, and metropolitan districts, respectively;
R2 is the proportion of variance explained by the equation; and SE is
the standard error of the regression (standard errors for coefficients are
reported in parentheses).
The results indicate that marginal domestic tax prices have a signifi-
cant effect upon spending decisions. The coefficient on ratable value per
head is negative, as expected, and is significantly different from zero.
These results imply that by raising the tax price to unity, the local
authority reforms will lower local spending in England (outside London)
by about 2Vi percent of total expenditure.
A further issue is whether making the new tax a per capita charge will
strengthen the connection between taxation and services. In the words of
the green paper "a community charge would provide a closer reflection
of the benefit from a modern people-based service than a property tax"
(HMSO (1986, p. 25)). Although it is true that much spending by local
authorities is oriented toward people, the large component is education,
where cost is not related to charge-paying adults. Recent evidence on the
provision of benefits by Cheshire County Council, reported in Bramley,
LeGrand, and Low (1989), indicates that benefits from local government

Figure 2. Relationship Between Domestic Ratable Value and Spending

Table 1. Percentage of Net Household Income Paid Under the Rates and the Community Charge
(Ranges of equivalent household income; in pounds per week)
Under 50- 75- 100- ISO- 200- 250- 300- 350- 400- 500 All
Item 50 75 100 150 ZOO 250 300 350 400 500 Plus Households
Rates 2.1 3.3 3.7 3.2 2.8 2.6 2.6 2.5 2.5 2.3 1.8 3.0
Community
charge 1.9 3.3 3.9 3.3 2.8 2.4 2.1 1.9 1.6 1.4 1.0 3.0
Difference3 0.2 — -0.2 -0.1 — 0.2 0.5 0.6 0.9 0.9 0.8 —
Percentage of
households in
income band 8.2 23.9 39.9 20.3 7.9 100.0
Source: HMSO (1986, Annex F).
Note: Equivalent household income measures relative standard of living by adjusting actual income for differences in household
composition.
a
A positive number indicates a gain under the community charge.

services rise with both income and socioeconomic class,5 implying that a
poll tax is not a satisfactory benefit charge. However, when rebates are
taken into account, the poll tax is quite a good reflection of benefits
across income bands, even if it is still regressive across socioeconomic
classes (Bramley, Le Grand, and Low (1989, Figures 6 and 7)).
Cost and compliance represent another issue. The rates were cheap
and easy to collect. The community charge will probably double admin-
istrative costs, due to its wide coverage and the need to keep up-to-date
information on residences. Evidence from Scotland indicates that there
may also be a considerable problem with compliance; Strathclyde has
initiated legal action against 350,000 people for nonpayment of the com-
munity charge, over 1 in 6 taxpayers. The evidence from Scotland also
indicates that local authorities may have used the confusion surrounding
the introduction of the new system to raise revenues significantly more
than expenditure (Hughes (1989)), presumably in order to lower future
increases in taxes for which the local authority will be more directly
accountable.
Equity
Any poll tax, even one with rebates for the less well-off, is likely to be
regressive. The Green Paper (HMSO (1986)) concedes that, as well as a
beneficial principle, the ideal local tax would have a redistributive princi-
ple. However, it concludes that "no tax could satisfy both aims simulta-
neously" (p. 24, emphasis in original), but that the rates failed both tests,
being badly correlated with both income and services.
Table 1 shows a comparison of payments of the rates and the commu-
nity charge among different income bands for the same total tax revenue.
In comparison with the rates, which were themselves a fairly regressive
tax, under the community charge the well-off (top three deciles) and very
poor (bottom decile) gain at the expense of those in the middle. Even as
a percentage of income, however, the gains to the well-off are much
larger than to the very poor. For example, the top income decile gains
about 0.8 percent of income on average, the bottom decile, 0.2 percent.6
The community charge is clearly a very regressive tax.
5
This picture may not be accurate for some urban councils where social expen-
ditures are considerably higher.
6
The reason the very poor gain is that many of them are pensioners and single
parent families, who are the group who gain most from the change. HMSO
(1986, Annex I) indicates that pensioners and single families generally gain,
other single adults lose, while couples are broadly unaffected.

226 TAMIM BAYOUMI
Nondomestic Taxation
Business rates under the old system made up over half of rates re-
ceived by local authorities, with each £1 of extra expenditure raising
local business taxes by an average of 74 pence. Differences in local rates
of taxation imply differences in costs for similar business. In general,
however, the rates were a relatively small part of aggregate costs, repre-
senting some 1.9 percent of net output in 1982. Overall, Bennett (1988)
concludes "rates do not have a general effect on business decisions, but
can be of major significance in the case of a restricted number of high tax
locations" (p. 159, emphasis in original). Clearly, the uniform business
rate will solve such problems, although at the cost of a considerable loss
of local fiscal autonomy.
Macroeconomic Effects
The most significant effect of the reforms on the macroeconomy have
to do with the housing market. The rates were the only important tax on
housing, and their abolition will raise housing consumption, further
distorting behavior in an already distorted market.
The community charge lowers the marginal cost of living in a more
expensive house, since the local tax bill is no longer affected by property
values. Calculations of the cost of trading up on a house indicate that the
rates represented one fourth of the rise in costs, and the implied fall in
the cost of housing could raise house prices in the long run by 15 to 20
percent (Spencer (1988) and Hughes (1988a, 1988b)). These effects may
have been one element in the sharp increase in housing values in the
United Kingdom in 1988 and 1989.
The supply response to this increase in demand will be delayed because
of the costs of moving and lags in the system. The current rate of increase
of the housing stock of about 1 percent a year could well double, al-
though some of this increase will come through improvements to the
current stock of housing rather than new houses. This implies that the
construction industry (in its broadest sense) will experience a significant
increase in demand over the medium term.
The increase in house prices also has implications for the macroecon-
omy, particularly consumption. Both the permanent income and life-
cycle theories of consumption imply an important role for wealth in
consumption decisions. The boom in house prices has probably been a

major factor in the fall in the U.K. household saving rate over the late
1980s (Muellbauer (1989) and Muellbauer and Murphy (1989)).
International Comparisons
There are few examples of poll taxes in the modern world; Japan has a
prefectural and municipal inhabitant tax, which forms an important part
of the local taxation system, raising 48 percent of income. However, the
rate of tax is progressive with respect to income. Tanzania introduced a
poll tax in 1984/85, with a flat rate charge of 200 shillings per adult in Dar
es Saalam. The following year this was changed to a progressive charge,
going a long way up the income scale. Overall, the international evi-
dence does not provide examples of an important residence tax unrelated
to income.
III. Conclusions
The reform of U.K. local authority finance was both far-reaching and
radical. The most controversial part of the reforms involves the replace-
ment of the existing tax on property with the community charge, a poll
tax. This tax is generally more regressive than its predecessor and sig-
nificantly more costly to administer. Although it might be expected that
such a lump-sum tax would improve efficiency, its main effect is to
further distort the housing market. For businesses, the switch to a uni-
form national rate of property tax ends distortions caused by large differ-
ences in the rate of tax, but the effect of this change on behavior is
unlikely to be large. For local government outside London, the reforms
raise the cost to electors of £1 of local services from an average of 70
pence to a uniform rate of £1, which may lower spending by some 2Vi
percent in the long run.
Overall, the reforms simplify and improve many areas of the relation-
ship between central and local government. Nevertheless, outside com-
mentators have been almost uniformly critical of the reforms. There
appear to be four reasons for this. First, many people believe that the
imposition of a poll tax represents a significant backward step in the tax
system. Second, the reforms, by lowering the tax base, reduce local
autonomy and limit choice. Third, attempts to avoid the tax will lead to
disenfranchisement of voters. Finally, given the scope of the reforms and
the unsatisfactory nature of the existing system, many believe an oppor-

228 TAMIM BAYOUMI
tunity for a radical improvement of the system was lost to the objective
of containing expenditure.
REFERENCES
Bayoumi, Tamim, "The 1990 Reform of United Kingdom Local Authority
Finance," IMF Working Paper 90/58 (Washington: International Monetary
Fund, July 1990).
Bennett, Robert, "Non-Domestic Rates and Local Taxation of Business," in The
Reform of Local Government Finance in Britain, ed. by S.J. Bailey and
R. Paddison (London; New York: Routledge, 1988).
Bramley, G., J. Le Grand, and W. Low, "How Far is the Poll Tax a 'Community
Charge'? The Implications of Service Usage Evidence," Discussion Paper
WSP/42 (London: The Welfare State Programme, London School of Eco-
nomics, April 1989).
Her Majesty's Stationery Office, Paying for Local Government, Cmnd. 9714
(London: HMSO, 1986).
Hughes, Gordon (1988a), "Rates Reform and the Housing Market," in The
Reform of Local Government Finance in Britain, ed. by S.J. Bailey and
R. Paddison (London; New York: Routledge).
(1988b), "Rates, Community Charge and the Housing Market," The
Housing Research Foundation (December).
-, "The Switch from Domestic Rates to the Community Charge in Scot-
land," Fiscal Studies, Vol. 10 (August 1989).
King, D. (1988a), "The Future Role of Grants in Local Government Finance,"
in The Reform of Local Government Finance in Britain, ed. by S.J. Bailey
and R. Paddison (London and New York: Routledge).
(1988b), "Accountability and Equity in British Local Finance—The Poll
Tax," University of Sterling Discussion Paper in Economics, Finance, and
Investment No. 145 (Sterling: University of Sterling, March).
Muellbauer, J., "Some Macroeconomic Causes and Consequences of U.K.
Housing Market Developments," unpublished (Nuffield College, October
1989).
, and A. Murphy, "Why Has U.K. Personal Saving Collapsed?" Credit
Suisse First Boston Research Report (London: Credit Suisse First Boston,
July 1989).
Spencer, P., "The Community Charge and Its Likely Effects on the UK Econ-
omy," Credit Suisse First Boston Report (June 1988).

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
Issues in Interest Rate Management

and Liberalization
Comment on Leite and Sundararajan
STEPHEN H. AXILROD*
I survey
WOULD LIKE to comment on one part of the excellent and concise
1
by Leite and Sundararajan of issues in interest rate manage-
ment. Perhaps the comment—which relates to the section on open mar-
ket policies—may seem excessively technical or involve the splitting of
hairs. However, the article appears designed to provide persuasive policy
advice, and I believe the monetary policy discussion runs the risk of
misleading policymakers in certain important respects.
The authors clearly and correctly state the conditions under which a
monetary aggregate or an interest rate should be the preferred target for
monetary authorities. They also make it clear that because of multiple
disturbances—some stemming from conditions in the financial sector,
others from conditions affecting the demand for goods and services—and
imperfect information, the policy authorities can adopt a strategy of
continuously reviewing their target settings.
The authors then go on to state the following:
Although it would still be necessary to choose between interest rate or money
supply targets at any point in time, this choice would generally be subsidiary
to the more important task of setting the consistent target levels for these
variables. Thus, even in a liberalized interest rate regime, the authorities
must constantly hold a view of the appropriate level of the interest rate and
strive to achieve it (pp. 750-51).
There are a number of problems with those two sentences, with very
practical implications for how monetary policy is conducted.
* Stephen H. Axilrod is Vice-Chairman of Nikko Securities Company Interna-
tional. He was formerly Staff Director for Monetary and Financial Policy for the
Board of Governors of the Federal Reserve System.
1
Sergio Pereira Leite and V. Sundararajan, "Issues in Interest Rate Manage-
ment and Liberalization," Staff Papers, International Monetary Fund, Vol. 37
(December 1990), pp. 735-52.
229

230 STEPHEN H. AXILROD
First, it is not necessary to choose between an interest rate or money

supply target during any reasonable operating period, even one as short
as two weeks. Monetary policy can (and in most practical cases probably
should) function with a range of tolerance for both interest rate and
money supply targets (or outcomes). In a very short period, the author-
ities can achieve a specified target for the interest rate of choice, but that
target can be changed within the operating period, depending on evolving
behavior of the money supply. It is not, incidentally, practically possible
to achieve a money supply target in a short operating period (which is my
interpretation of the authors' somewhat ambiguous phrase, "at any point
in time"), but that of course is no argument against setting and being
guided by one.
In any event, while an interest rate may be chosen as a daily operating
guide, that decision can be viewed as a matter of operational convenience
and need not imply that the monetary authorities necessarily must make
a choice in any fundamental sense about a particular level of that interest
rate or of the money supply. If they believe they have made such a choice,
they will—given human nature and the soul-searching that will have gone
into the choice—be slow, possibly too slow, to change the target as
economic circumstances alter.
Second, even if it were necessary to decide between an interest rate or
money supply target "at any point in time," I would not agree the choice
is generally subsidiary to the task of setting consistent target levels for
both. In my experience, it is not given to economists in practice to be able
to find mutual consistency between the money supply and interest rates.
Even if they could do so on the basis of particular economic and financial
assumptions, that would not be of great help to policymakers. Policymak-
ers would still have to make the judgment about whether to put more
stress on money supply or interest rates once the economy and/or finan-
cial structure start varying (as they most certainly will) from the initial
assumptions that lay behind the targets.
The problem for policymakers is not to choose among a number of
alternative, and presumably internally consistent (at the time made),
relationships between money supply and interest rates, but to choose
which to follow or emphasize when one or the other starts going off
course. Policymakers may even wish to choose what might appear in
advance to be an inconsistent set of relationships, depending on how
strongly they wish to give operating weight to a measure of money supply
or to an interest rate.
While it is crucial and also evidence of objectivity in staff analysis for
policymakers to be given technically consistent money and interest rate
relationships for consideration, policymakers in an obviously inflationary

INTEREST RATE MANAGEMENT: COMMENT 231
period may wish to "err" on the side of money growth restraint. In

a clearly recessionary period, they may wish to "err" by adopting a
relatively low interest rate target.
Finally, I do not believe the authorities must constantly hold a view of
the interest rate level and strive to achieve it. The authorities may say to
themselves, as the Federal Reserve did from late 1979 to late 1982, that,
under the particular circumstances of the time, they are not prepared to
make a judgment about what interest rate level is desirable (within a very
broad range of tolerance) and that they will control only the level of the
depository system's reserves (nonborrowed reserves in the case of the Fed
at that time).
There can be circumstances when it is wise to make such a judgment.
One would be when control of inflation is the prime objective and when
in particularly volatile market and economic conditions neither inflation
expectations nor the real return on capital is knowable with any reason-
able degree of certainty. It would be practically impossible for policy to
construct, and thus by definition, to achieve, an appropriate interest rate
level.

IMF Staff Papers
Vol. 38, No. 1 (March 1991)
IMF Working Papers

Staff Papers draws on IMF Working Papers, which are research studies by
members of the Fund's staff. A list of Working Papers issued in 1990:4
follows; additions will be noted in future issues.
"Tax Policy Options for a United Germany," by Bernd Genser [90/89]

"Alternative Dual Exchange Market Regimes: Some Steady State Compari-
sons," by Saul Lizondo [90/90]
"Fiscal Policy, the Real Exchange Rate, and Commodity Prices," by Carmen M.
Reinhart [90/91]
"Fiscal Policy During the Demise of Central Planning: The Transition to a
Market Economy," by Andrew Feltenstein [90/92]
"Industrial Relations and Macroeconomic Performance: An Application to
Spain," by Fabrizio Coricelli [90/93]
"Tax Efficiency in an Open Economy," by W.R.M. Perraudin and T. Pujol
[90/94]
"Financial Innovation and Consumption in the United Kingdom," by Tamim
Bayoumi [90/95]
"European Fiscal Harmonization and the French Economy," by W.R.M. Per-
raudin and T. Pujol [90/96]
"A Note on Saving-Investment Correlations in the EMS," by Jagdeep S. Bhan-
dari and Thomas H. Mayer [90/97]
"The Fiscal and Economic Effects of Federal Credit Assistance Programs," by
Stephen M. Fries [90/98]
"Investment in Housing in the United States: A Portfolio Approach—The Possi-
ble Effects of Changes in Tax Policy," by Krister Andersson [90/99]
"Trade Liberalization in Developing Countries: Initial Trade Distortions and
Imported Intermediate Inputs," by Jonathan D. Ostry [90/100]
"Why Is Unemployment So High at Full Capacity? The Persistence of Unem-
ployment, the Natural Rate, and Potential Output in the Federal Republic
of Germany," by David T. Coe and Thomas Krueger [90/101]
"Structural Models of the Dollar," by Charles Adams and Bankim Chadha
[90/102]
"Corporate Income Tax Harmonization and Capital Allocation in the European
Community," by Angel de la Fuente and Edward Gardner [90/103]
"European Policy Convergence and the EMS," by Ronald MacDonald and
Mark P. Taylor [90/104]
"Analytical and Methodological Issues in the Measurement of Fiscal Deficits,"
by Mario I. Blejer and Adrienne Cheasty [90/105]
232
IMF WORKING PAPERS 233
"The Simplest Test of Target Zone Credibility," by Lars E.G. Svensson [90/106]
"The Changing Structure of Japanese Trade Flows," by Robert Corker [90/107]
"Exchange Rate Bands with Point Process Fundamentals," by W.R.M. Per-
raudin [90/108]
"The Macroeconomic Effects of Capital Controls and the Stabilization of the
Balance of Trade," by Enrique G. Mendoza [90/109]
"Credibility and the Dynamics of Stabilization Policy: A Basic Framework," by
Guillermo A. Calvo and Carlos A. Vegh [90/110]
"Real Interest Rate Targeting: An Example from Brazil," by Eric V. Clifton
[90/111]
"Financial Market Volatility and the Implications for Market Regulation: A
Survey," by Louis O. Scott [90/112]
"Output, Employment, and Financial Sanctions in South Africa," by Tamim
Bayoumi [90/113]
"Parallel Currency Markets in Developing Countries: Theory, Evidence, and
Policy Implications," by Pierre-Richard Agenor [90/114]
"Management of the Nominal Public Debt: Theory and Applications," by
Guillermo A. Calvo and Pablo E. Guidotti [90/115]
"Time Varying Risk Premia in Futures Markets," by Graciela Kaminsky and
Manmohan S. Kumar [90/116]
"Capital Mobility in Developing Countries—Some Empirical Tests," by
Nadeem U. Haque and Peter Montiel [90/117]
"Long-Run Purchasing Power Parity and the Dollar-Sterling Exchange Rate in
the 1920s," by Mark P. Taylor [90/118]
"Money Supply and Interest Rate Policy in a New-Keynesian Framework," by
Guillermo A. Calvo and Carlos A. Vegh [90/119]
"Financial Sector Reform and Central Banking in Centrally Planned
Economies," by V. Sundararajan [90/120]
"Financial Market Constraints and Private Investment in a Developing Coun-
try," by E.G. Johnson [90/121]
"Export Processing Zones for Growth and Development: The Mauritian Exam-
ple," by Rolf G. Alter [90/122]
"Taxation and the Cost of Capital in Hungary and Poland: A Comparison with
Selected European Countries," by Krister Andersson [90/123]
A limited supply of IMF Working Papers is available free of charge to interested

scholars and research institutions. In requesting a particular paper (or papers),
please indicate the Working Paper number (for example, WP/90/1), title, and
author(s). Requests should be made to:
International Monetary Fund

Publication Services
Washington, D.C. 20431, U.S.A.
Telephone: (202) 623-7430
Telefax: (202) 623-7201

In the current issue of
THE WORLD BANK
ECONOMIC REVIEW
Volume 5 January 1991 Number 1
THE MEASUREMENT OF POVERTY

Comparing Poverty Rates Internationally: Lessons
from Recent Studies in Developed Countries
Anthony B. Atkinson
Poverty in Palanpur
Peter Lanjouw and Nicholas Stern
Measuring Changes in Poverty: A Methodological Case Study

of Indonesia during an Adjustment Period
Martin Ravallion and Monika Huppi
TOPICS IN LAND RIGHTS AND LAND TAXATION

Introduction: Agricultural Taxation and Land Rights Systems
Land Taxes, Output Taxes, and Sharecropping: Was Henry George Right?
Karla Hoff
If Agricultural Taxation Is So Efficient, Why Is It So Rarely Used?

Jonathan Skinner
Land Tenure and Property Rights: Theory and Implications

for Development Policy
Gershon Feder and David Feeny
Indigenous Land Rights Systems in Sub-Saharan Africa:

A Constraint on Productivity?
Shem Migot-Adholla and others
For information on how to order a subscription to The World Bank Economic

Review, please contact World Bank Publications, 300 Raritan Center Parkway,
Edison, NJ 08818-7816 U.S.A. Telephone: 201-225-2165.

View publication stats

PrivateInvestment PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PrivateInvestment PDF

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Private Investment in Developing Countries: An Empirical Analysis

Article in IMF Staff Papers · March 1991

Joshua Greene Delano Villanueva

SEE PROFILE SEE PROFILE

Economic Adjustment and Growth: Theory and Practice View project

The user has requested enhancement of the downloaded file.

Vol. 38 No. 1 MARCH 1991

©International Monetary Fund. Not for Redistribution

The term "country," as used in this publication, may not refer to a

® 1991 by the International Monetary Fund

International Monetary Fund

Three no. a year, 1950-1977; four no. a year, 1978-

1. Foreign exchange—Periodicals. 2. Commerce—Periodicals.

HG3810.15 332.082 53-35483

©International Monetary Fund. Not for Redistribution

Vol. 38 No. 1 MARCH 1991

Long-Run Money Demand in Large Industrial Countries

Wages, Profitability, and Growth

The Transmission Mechanism

The Budgetary Control and Fiscal Impact

An Inflation-Proof Tax System?

Exchange Rate Expectations:

Currency Bands, Target Zones, and Price Flexibility

©International Monetary Fund. Not for Redistribution

Shorter Papers and Comments

©International Monetary Fund. Not for Redistribution

Long-Run Money Demand

The long-run properties of money demand functions in the large industrial

HE REPUTATION of the aggregate demand function for money bal-

©International Monetary Fund. Not for Redistribution

In spite of these difficulties, a renaissance seems to be underway, or at

I. State of the Art

©International Monetary Fund. Not for Redistribution

effects of financial innovation and deregulation on the nature and stabil-

where m, y, and p are the logarithms of money, real gross national

©International Monetary Fund. Not for Redistribution

©International Monetary Fund. Not for Redistribution

where rl is a long-term interest rate.8 As with most earlier models, this

©International Monetary Fund. Not for Redistribution

When the demand function is estimated with all arguments entering

This equation is completely standard except for the inclusion of both

©International Monetary Fund. Not for Redistribution

which is very close to Hacche's much earlier estimate (equation (3),

This sketchy review of recent developments suggests a few conclusions

©International Monetary Fund. Not for Redistribution

II. Cointegration Analysis

One of the main sources of confusion regarding the parameterization

Engle and Granger (1987) propose a two-step procedure under which

©International Monetary Fund. Not for Redistribution

©International Monetary Fund. Not for Redistribution

The difficulty of applying the Engle-Granger procedure may be illus-

Solving for the static long-run solution gives

Here, the coefficients move somewhat closer to their expected values,

©International Monetary Fund. Not for Redistribution

but it still seems doubtful that equation (9) is a valid characterization of a

and the dynamic error-correction process be

where D denotes a general differencing operator. In principle, these

where L is a general lag operator on the levels of y and x. Depending on

Multiple Cointegrating Vectors: The Johansen Procedure

The initial development of cointegration analysis was aimed at bivari-

©International Monetary Fund. Not for Redistribution

©International Monetary Fund. Not for Redistribution