INTERNATIONAL MONETARY FUND...economists, and by others concerned with monetary and financial problems. Much of what is now presented is quite provisional. On some international monetary

INTERNATIONAL MONETARY FUND

S T A F FP A P E R S

C O N T E N T S

Interest Rate Determination in Developing Countries:A Conceptual Framework

SEBASTIAN EDWARDS and MOHSIN S. KHAN • 311

Exchange Rate Changes and Exportsof Selected Japanese Industries

DANIEL CITRIN • 404

Debt-Equity Ratios of Firms and Interest Rate Policy:Macroeconomic Effects of High Leverage

in Developing CountriesV. SUNDARARAJAN • 43

Effects of Exchange Rate Volatility on Trade:Some Further Evidence

PADMA GOTUR • 415

Real Exchange Rates, Import Penetration,and Protectionism in Industrial Countries

ERIC V. CLIFTON • 513

Summaries • 537 Resumes • 541 Resumenes • 546

SEPTEMBER 1985V O L 3 2 N O . 3 ©International Monetary Fund. Not for Redistribution

STAFF PAPERS

NORMAN K. HUMPHREYS, EditorJAMES MCEUENAssistant Editor

Editorial Committee

Norman K. Humphreys, Chairman

Bijan B. Aghevli Anthony Lanyi

Gerard Belanger M. Ranji P. Salgado

Morris Goldstein Joanne Salop

Ernesto Hernandez-Cata V. Sundararajan

Ronald A. Krieger Alan A. Tait

From the Foreword to the first issue:

Among the responsibilities of the International Monetary Fund, as setforth in the Articles of Agreement, is the obligation to "act as a center for thecollection and exchange of information on monetary and financial prob-lems," and thereby to facilitate "the preparation of studies designed to assistmembers in developing policies which further the purposes of the Fund."The publications of the Fund are one way in which this responsibility isdischarged.

Through the publication of Staff Papers, the Fund is making availablesome of the work of members of its staff. The Fund believes that thesepapers will be found helpful by government officials, by professionaleconomists, and by others concerned with monetary and financial problems.Much of what is now presented is quite provisional. On some internationalmonetary problems, final and definitive views are scarcely to be expected inthe near future, and several alternative, or even conflicting, approachesmay profitably be explored. The views presented in these papers are not,therefore, to be interpreted as necessarily indicating the position of theExecutive Board or of the officials of the Fund.

The authors of the papers in this issue have received considerable assistancefrom their colleagues on the staff of the Fund. This general statement ofindebtedness may be accepted in place of a detailed list of acknowledgments.

Subscription: US$15.00 a volume or the approximate equivalent in thecurrencies of most countries. Four numbers constitute a volume. Single copiesmay be purchased at $4.00. Special rates to university libraries, faculty members,and students: $7.50 a volume; $4.00 a single copy. Subscriptions and ordersshould be sent to:

Publications UnitInternational Monetary FundWashington, D.C. 20431, U.S.A.

Telephone number: (202) 473-7430Cable address: Interfund

©International Monetary Fund. Not for Redistribution

INTERNATIONAL MONETARY FUND

S T A F FP A P E R S

Vol. 32 No. 3 SEPTEMBER 1985


EDITOR'S NOTE

The Editor invites from contributors outside the Fundbrief comments (not more than 1,000 words) on pub-lished articles in Staff Papers. These comments shouldbe addressed to the Editor, who will forward them to theauthor of the original article for reply. Both the com-ments and the reply will be published in the same issue ofStaff Papers.

The term "country," as used in this publication, may not refer to aterritorial entity that is a state as understood by international law andpractice; the term may also cover some territorial entities that are notstates but for which statistical data are maintained and providedinternationally on a separate and independent basis.

® 1985 by the International Monetary FundInternational Standard Serial Number: ISSN 0020-8027

The Library of Congress has cataloged this serial publication as follows:

International Monetary FundStaff 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-Text in English with summaries in English, French, and Spanish.Indexes:

Vols. 1-27, 1950-80. 1 v.ISSN 0020-8027 = Staff papers — International Monetary Fund.

1. Foreign exchange—Periodicals. 2. Commerce—Periodicals. 3. Curren-cy question—Periodicals.

HG3810.15 332.082 53-35483


C O N T E N T S

Vol. 32 No. 3 SEPTEMBER 1985

Interest Rate Determination in Developing Countries:A Conceptual Framework

SEBASTIAN EDWARDS and MOHSIN S. KHAN • 377


DANIEL CITRIN • 404

Debt-Equity Ratios of Firms and Interest Rate Policy:Macroeconomic Effects of High Leverage

in Developing Countries

V. SUNDARARAJAN • 430

Effects of Exchange Rate Volatility on Trade:Some Further Evidence

PADMA GOTUR • 475

Real Exchange Rates, Import Penetration,and Protectionism in Industrial Countries

ERIC V. CLIFTON • 513

Summaries • 537 Resumes • 541 Resumenes • 546

in


This page intentionally left blank


Interest Rate Determinationin Developing Countries

A Conceptual Framework

SEBASTIAN EDWARDS and MOHSIN S. KHAN*

URING THE PAST DECADE or so economists have emphasizedthe critical role that interest rate policies play in the devel-

opment process. The growing literature on financial "reforms"and financial "liberalization" in developing countries has dealtwith a variety of issues, such as the relation between financialintermediation and economic growth, the sensitivity of the vol-ume of savings to changes in real interest rates, and the relationbetween investment and interest rates. Generally speaking, theempirical evidence indicates that there is indeed a positive associ-ation between the degree of development of the financial sector,including in particular freer interest rates, and economic per-formance in developing countries.1 This finding has undoubtedlyprompted the authorities in a number of such countries to pursuepolicies to remove controls on interest rates and to allow marketforces to play a relatively greater role in the determination ofinterest rates.

Now that the process of financial liberalization is well underway, however, economists and policymakers are faced with a dif-ferent set of issues relating to interest rates in developing coun-tries. The focus has begun to shift away from investigating the

* Professor Edwards was a consultant with the Research Department whenthis paper was written. He is currently with the University of California, LosAngeles, and with the National Bureau of Economic Research. He is a graduateof the Universidad Cat61ica de Chile and of the University of Chicago.

Mr. Khan, Advisor in the Research Department, is a graduate of ColumbiaUniversity and of the London School of Economics and Political Science.

1 See, for example, McKinnon (1973), Fry (1982), Lanyi and Saracoglu (1983),Mathieson (1983), andTownsend (1983); for a contrary view, see van Wijnbergen(1983).

377

D


378 SEBASTIAN EDWARDS and MOHSIN S. KHAN

effects of freeing interest rates to examining how interest rates arein fact determined once the domestic financial market has beenliberalized. The interest in this particular issue has been height-ened by two factors. The first is the recent experiences of thecountries of the Southern Cone of Latin America—Argentina,Chile, and Uruguay—where domestic interest rates rose to extra-ordinarily high levels following the implementation of financialreform policies.2 The second is the evidence that has accumulatedsuggesting that the high and volatile world interest rates in recentyears were at least partially transmitted into developing countries.Both these factors have been a cause of concern to policymakersand have generated some fundamental questions about the behav-ior of interest rates in developing countries—in particular, aboutwhat should be expected when controls on interest rates are elim-inated. At present, however, there are few studies dealing withthis general issue, and even fewer specifically examining the re-spective influences of foreign factors and domestic monetary con-ditions as they affect interest rates in developing countries.3

It is obvious that the process of determination of interest rateswill be significantly different under alternative degrees of open-ness of the capital account of the balance of payments. For exam-ple, in the case of a fully open capital account some form ofinterest arbitrage will hold, with domestic interest rates de-pending on world interest rates, expected devaluation, and per-haps some risk factors. In contrast, in countries with a completelyclosed economy (closed capital and current accounts) open econ-omy factors will obviously play no role, and the nominal interestrate will be determined by conditions prevailing in the domesticmoney market and by expected inflation. Most developing coun-tries, however, do not fall in either of these two extreme cate-gories, so that interest rates will in general depend on domesticmoney market conditions, as well as on the expected rate ofdevaluation and world interest rates.4 From a policy perspective it

2This subject has been addressed by, among others, Diaz-Alejandro (1981),Edwards (1985b), Hanson and de Melo (1985), Harberger (1982), Sjaastad(1983), and Zahler (1983).

3 The only studies we are aware of that include both open economy and domes-tic monetary factors in the analysis of interest rates are Mathieson (1982,1983),on Argentina and Chile respectively; Blejer and Gil Diaz (1985) and Hanson andde Melo (1985) on Uruguay; and Edwards (1985a) on Colombia.

4 Even if the capital account of the balance of payments is closed but there issome trade with me rest of the world, open economy factors can still indirectlyaffect domestic interest rates. For example, a terms of trade shock can producechanges in real income and prices that will affect the domestic demand for creditand, thus, equilibrium interest rates.


INTEREST RATES IN DEVELOPING COUNTRIES 379

is important to determine the way in which these different factorsactually affect interest rates. For example, how expected deval-uations or changes in domestic monetary conditions or both affectinterest rates in developing countries is crucial for assessing thesignificance of one of the possible mechanisms through whichstabilization policies will affect aggregate demand. Stabilizationprograms typically involve both exchange rate adjustments andtighter credit and monetary policies. If these policies generate anincrease in the domestic (real) interest rate, there will be anadditional channel (usually not considered in formal studies aboutstabilization programs in developing countries) through which ag-gregate demand will be affected.5

In this paper a framework is proposed for empirically analyzingthe determination of nominal interest rates in developing coun-tries. Even though the model is quite general and of relevance forany small country, the discussion is carried out with those devel-oping countries in mind that have liberalized their domestic fi-nancial sectors in the sense that controls on interest rates havebeen removed. The model, which is described in Section I, com-bines features of models for both closed and open economies, andit is shown that the relative importance of the domestic monetaryconditions and the open economy factors will depend essentiallyon the openness of the capital account. An interesting property ofthe model is that the approximate degree of openness of thefinancial sector in a particular country can be estimated from thedata. In Section II of the paper the usefulness of this frameworkfor analyzing interest rate behavior is illustrated using data forColombia and Singapore. The results obtained indicate that, asexpected, in Singapore only open economy factors appear tomatter; in Colombia, however, both domestic monetary dis-equilibria and open economy conditions have influenced nominalinterest rates during the past 15 years. Section III describes someareas in which the analysis could be extended—including, forexample, studying the behavior of real interest rates, the deter-mination of interest rates under changing degrees of openness, themodeling of the effects of expected exchange rate changes, and,

5 Until now most studies that have analyzed the effect of stabilization policieson output, prices, and the balance of payments in developing countries have notincluded the interest rate as a possible transmission mechanism. The main reasonfor this omission is that the experience with liberalized capital markets is stillrelatively recent. A theoretical discussion, however, of the effects of a stabiliza-tion program working through increases in real interest rates is contained inDornbusch (1982b).



finally, introducing the possibility of currency substitution. Theconcluding section summarizes the main points and results of theanalysis.

I. Theoretical Models of Interest Rate Determination

In this section three basic models for analyzing interest ratebehavior in developing economies are briefly presented. The firstis a simple model that assumes that the country in question iscompletely closed to the rest of the world. Under these circum-stances it is assumed that the nominal interest rate depends on thereal interest rate and on expected inflation. The second modelconsiders the other extreme, in which the capital account is com-pletely open. In this case domestic interest rates are closely linkedto world interest rates through interest arbitrage. Finally, a moregeneral model that allows both foreign and domestic factors toaffect the behavior of the nominal interest rate, and thus containsthe other two models as special cases, is presented and discussed.

Interest Rates in a Closed Economy

Following the standard Fisher approach, we can specify thenominal interest rate as equal to6

where

i = the nominal rate of interestrr = the real (ex ante) rate of interestTte = the expected rate of inflation.

The real interest rate in turn can be specified as

where p is a constant and represents the long-run equilibrium realinterest rate. The variable EMS represents the excess supply ofmoney, X is a parameter (X > 0), and co, is a random error term.According to equation (2), the real rate of interest deviates from

6 We are ignoring here, for example, the effects of taxation on the relationbetween expected inflation and the nominal interest rate. On this topic seeDarby (1975), and Tanzi (1976).



its long-run value p if there is monetary disequilibrium; and excessdemand (supply) for (of) real money balances will yield a tempo-rarily higher (lower) real interest rate. This relation has beencalled the "liquidity effect" in the literature (Mundell (1963)). Inthe long run, however, the money market would be in equi-librium, and the variable EMS would play no role in the behaviorof rrt 7 Introducing this liquidity effect into the model, contrary tomost recent empirical studies of interest rate behavior, allows thereal rate of interest to be variable in the short run.8 As such, eventhough the Fisher equation (1) is assumed to hold continuously,the possibility of slow adjustment of the real interest rate (givenby X) implicitly allows for the possibility of delayed response of thenominal interest rate to monetary changes.

The solution for the nominal interest rate in a closed economy,therefore, is

To estimate equation (3), however, some assumptions have tobe made regarding the unobserved variables, ire and EMS. Theexpected rate of inflation can be specified in a variety of ways.One is to use the traditional adaptive expectations model, inwhich the expected rate of inflation is assumed to be a (geo-metrically) distributed lag function of past rates of inflation. Anempirical generalization of this approach is to assume an auto-regressive process for the rate of inflation and to use the predictedvalues as representing the expected rate of inflation.9 Other possi-ble methods include the use of survey data (for example, theLivingston series on inflationary expectations) or of models thatallow for the influence of additional economic variables (otherthan only past rates of inflation) in the formation of expecta-tions.10 Of course, it can also be assumed that actual and expectedrates of inflation are the same—an assumption that would imply

7 Note that EMSt could also affect TT*. Furthermore, it is assumed here thatchanges in < have no direct effects on rrt. On these types of effects, see Mundell(1963).

Recent empirical studies on interest rate behavior in the United States in-clude, among others, Fama (1975), Tanzi (1980), Makin (1982), and Melvin(1983).

9 In this formulation the weights of the lag distribution are not assumed tofollow any specific pattern.

10 These would be the empirical representations of the rational expectationsmodel in which economic agents are assumed to take into account all availableinformation in forming their (conditional) expectations.



a strict form of rational expectations (that is, perfect foresight).There is no compelling theoretical reason for preferring onemethod over any other, and the choice is ultimately an empiricalone.

The excess supply of money is defined as

where m is the actual stock, and md the desired equilibrium stock,of real money balances.11 In an economy that has completed thefinancial reform process, we would expect substitution to takeplace between both money and goods, as well as between moneyand financial assets, so that the demand for money would be afunction of two opportunity-cost variables (the expected rate ofinflation and the rate of interest) along with a scale variable (realincome).12 The equilibrium demand for money can therefore bewritten as

It should be noted that long-run demand for money is assumedto be a function of the equilibrium nominal interest rate, definedas the equilibrium real interest rate (p) plus the expected rate ofinflation (77*), rather than of the current nominal interest rate.

The model can be closed by assuming that the stock of realmoney balances adjusts according to

where A is a first-difference operator, Alog mt = log mt - log m,_i,and p is the coefficient of adjustment, 0 ^ (3 ^ 1. If the nominalstock of money is exogenous, then equation (6) really describes anadjustment mechanism for domestic prices. In essence, equation(6) introduces a process by which the nominal interest rate returnseventually to its equilibrium level.

11 Note that equation (4) is only one of the alternative ways to specify excessmoney supply, or monetary disequilibrium. For example, it can be postulatedthat only money surprises will influence the real interest rate (Makin (1982)). Insuch a case EMS would have to be replaced by some proxy of unanticipatedmonetary changes in equation (2).

12 Of course, one could also introduce an "own" rate of return into the moneydemand formulation. This would certainly be advisable when dealing with broaddefinitions of money that include deposits paying positive rates of interest (seeMathieson (1982, 1983)). Because we work with narrow money (currency plusdemand deposits) throughout, in our case this omission is obviously not serious,since demand deposits typically are non-interest-bearing.



The workings of the model given by equations (3), (4), and (6)can be conveniently described within the framework of Figure 1.In the figure the initial equilibrium is point A, where the long-rundemand for real money balances is equal to the supply (EMS = 0),the nominal interest rate is at its equilibrium level (p + ire), andthe actual stock of real money balances is equal to m0. Supposenow that there is an increase in the supply of money from ms

Q tom(. This would create an excess supply of real money balances(EMS >0), and the nominal interest rate would fall below itsequilibrium value (say, to ^). The movement from A to B inessence represents the short-run liquidity effect we referred toearlier. B, however, is only a temporary equilibrium positionbecause in the next period the (unchanged) long-run demand formoney is less than the actual stock in the previous period,m?+l<mt(=ms

2}\ therefore, by equation (6) the actual stock ofreal money balances would begin to decline. In Figure 1 the m5

schedule would shift to the left until the actual money supply isonce again equal to equilibrium money demand, and conse-quently the nominal interest rate would be given by p + rf.

Figure 1. Interest Rate Determination in a Closed Economy



Equation (6) can be simplified to

combining equations (4) and (6a), we obtain

Using equations (1), (5), and (7), we can derive the reduced-formequation for the nominal interest rate:

where the composite parameters are

Once TTe is replaced by some appropriate measured variable,equation (9) can be directly estimated. In the estimation it wouldbe expected that yl > 0 and that y2 < 0; the sign of 73 would benegative or positive depending on whether X(l - P)(a2 + a3) isgreater or less than unity.

Interest Rates in a Fully Open Economy

If the economy is completely open to the rest of the world, andthere are no impediments to capital flows, domestic and foreigninterest rates will be closely linked. In particular, in a world withno transaction costs and risk-neutral agents the following un-covered interest arbitrage relation will hold:

where i* is the world interest rate for a financial asset of thesame characteristics (maturity and so on) as the domestic instru-ment, and et is the expected rate of change of the exchange rate(defined as the domestic price of foreign currency). If agents areassumed to be risk averse, however, et should be replaced by theforward premium; alternatively, a (time- varying) risk-premiumterm should be added to equation (9).13

13 Introducing the forward premium into the specification in place of the ex-pected change in the exchange rate implies, of course, that the forward premiumis a good approximation of the change in the future spot exchange rate.



The analysis of interest rate behavior in open economies usuallyhas amounted to investigating the extent to which equation (9), orsome variant of it, holds. One way of doing this is by addingtransaction costs and defining a band within which the interest-parity differential can vary, without violating the arbitrage condi-tion. Another way of testing equation (9) is through analysis of thetime-series properties of the interest-parity differential. If thesetime series are not serially correlated—that is, if they are whitenoise—it is usually concluded that the domestic interest rate de-pends only on open economy factors.14 Frenkel and Levich (1975,1977), for example, have analyzed the extent to which the coveredarbitrage condition, which replaces et by the forward premium inequation (9), held for industrialized countries during the periodafter adoption of floating rates in 1973. They showed that, oncetransaction costs are allowed into the analysis, this arbitrage con-dition has worked well for these countries. Using a similar meth-odology, Lizondo (1983), however, found evidence of large andpersistent deviations in Mexico during 1979-80. Cumby and Obst-feld (1981) adopted the second of the two approaches andanalyzed the time-series properties of the uncovered interest arbi-trage differential using weekly data for six industrialized coun-tries; they found that in five of the six cases these series exhibitedstrong serial correlation. They interpret these results as providingevidence that there exists a (time-varying) foreign exchange pre-mium for most currencies (see Levich (1985) for a review of otherstudies of related isues). The tests performed by Blejer (1982)using monthly data for Argentina for June 1977 through August1981, however, could not disprove the hypothesis that for Argen-tina during this period the uncovered interest rate differential waswhite noise.15 Broadly speaking, the evidence appears fairlymixed on the interest-parity condition in open economies.

Of course there exists the possibility that, because of frictionsarising from transactions costs, information lags, and the like,domestic interest rates respond with delay to any changes in theforeign rate of interest or in exchange rate expectations. This type

14 From a methodological point of view, even if interest-parity arbitrage differ-entials are white noise it is still possible that other variables, besides the worldinterest rate and the expected rate of devaluation, will affect the domesticinterest rate. For this reason a more appropriate procedure is to test directlywhether other variables suggested by the theory have an effect on /,.

15 In a more recent study of Uruguay, Blejer and Gil Diaz (1985) found thatthe risk premium was highly serially correlated.



of lagged response can be modeled straightforwardly in a partialadjustment framework as follows:

where 0 is the adjustment parameter, 0 < 0 < 1. If the financialmarket adjusts rapidly, this parameter 0 will tend toward unity.Conversely, a small value of 0 would imply slow adjustment of thedomestic interest rate.16 The solution of equation (10) in terms ofthe domestic interest rate is

The General Case

The preceding discussion has examined interest rate deter-mination in the two polar cases related to the degree of opennessof the economy. If, however, the economy under consideration isone that has some controls on capital movements, as most devel-oping countries do, it is possible to visualize that both open andclosed economy factors will affect the behavior of domestic inter-est rates at least in the short run. A straightforward way of con-structing a model for such an economy is to combine the closedeconomy and open economy extremes. In particular, it can beassumed that the equation for the nominal interest rate can bespecified as a weighted average, or linear combination, of theopen and closed economy expressions discussed above. Denotingthe weights by i|i and (1 - i|i) and combining equations (1) and (9)allows the following model for the nominal interest rate to bespecified:

where the parameter i|i can be interpreted as an index measuringthe degree of financial openness of the country. If t|i = l, theeconomy is fully open, and equation (12) collapses into the inter-est arbitrage condition (9). If i|i = 0, however, the capital account

16 During the period when the parity condition does not exactly hold therewould obviously be unexploited profit opportunities. The attempts by trans-actors to take advantage of these opportunities would set in motion the veryforces that would bring about equality between domestic and foreign interestrates (adjusted for expected exchange rate changes). How long this process takesis an empirical question and would have to be estimated from the data.



is closed, and equation (12) becomes equal to the Fisher closedeconomy equation (1). In the intermediate case of a semiopen(semiclosed) economy, the parameter i|i will lie between zero andunity; the closer it is to unity, the more open the economy will be.In a sense, estimating fy from the data makes it possible to deter-mine the degree of openness of the financial sector in a particularcountry. This estimated degree of openness will provide someinformation on the actual degree of integration of the domesticcapital market with the world financial market.17 To the extentthat official capital and exchange controls are not fully effective,the empirically estimated "economic" degree of openness can besignificantly higher than the "legal" degree of openness given bythe system of capital controls in the country.

If we assume slow adjustment to interest parity and thus useequation (11) instead of equation (9), the appropriate form for thegeneral case becomes

In this case full interest parity would require the conditioni|i = 6 = 1; when \\i = 0, the Fisher closed economy condition wouldemerge. It should be noted that there will be some relation be-tween the index of financial openness, i|i, and the speed of adjust-ment, 6. For example, if the domestic financial market is fullyintegrated with the international capital markets, it is also likelythat domestic interest rates would adjust quite rapidly.

Assuming that the excess money supply term is given by equa-tion (4) and that the demand for real money function is providedby equation (5),18 we obtain from equation (13) the followingexpression for the nominal interest rate:19

17 It is, of course, assumed here that the degree of openness (v|/) is constant overtime. The implications of relaxing this assumption, and the possible proceduresfor doing so, are considered in Section III.

18 Strictly speaking, in the shift from the closed economy to the open economycase the demand for money function should be generalized to allow for foreigninterest rates, the expected change in the exchange rate, or both. A suggestedprocedure for doing so is presented in Section III (under "Effects of CurrencySubstitution").

19 Note that when 6 = 1 the lagged interest rate term would drop from thespecification, so that the equilibrium model is only a restricted version of thisformulation.



where the reduced-form parameters 8, are

and e is a random error term. If we assume that the incomeelasticity of the demand for money is unity, then the model can befurther simplified. In this case 82 = — 83, and real income andlagged real money balances can be combined into one compositevariable — that is, [log yt - log mr_J.

Equation (14) is quite general because it not only incorporatesopen economy and closed economy features but also permits thepossibility of slow adjustment on both the foreign and domesticsides.20 One can see that, in the case of a completely open econ-omy with instantaneous adjustment of the domestic interest rate(that is, i|i = 0 = l), §! becomes equal to unity, and 80 = 82 =83 = 84 = 85 = 0. According to equation (14), the nominal interestrate will then be equal, in both the long and short run, to (i* + et).In the case of a completely closed economy (i|j = 0), the param-eters Si and 85 will be equal to zero, and equation (14) collapsesto the closed economy equation (8).

The preceding discussion has assumed that agents are risk neu-tral. As mentioned, if agents are risk averse, equation (14) shouldbe modified to take this fact into account. The simplest way ofdoing so is to replace the expected rate of devaluation et by theforward premium. From a practical viewpoint, however, this sub-stitution poses difficulties because there are few developing coun-tries that have forward markets for their currencies. An alterna-tive way to deal with the problem of risk aversion is to introducea risk premium explicitly into the analysis, and to make someassumptions about its statistical properties. For example, it can beassumed that the risk premium is equal to a constant plus a ran-dom term. In this case the constant part of the premium will be

20 Note that an equation of the form of equation (14) can be derived froma portfolio model with imperfect substitutaoility between domestic and for-eign assets.



added to the constant in equation (14), and the random com-ponent becomes a part of the error term. In principle it would bepossible to incorporate any number of alternative assumptionsabout the behavior of the risk premium into the empiricalanalysis.

II. Empirical Tests of the Model

To assess the ability of the general model to describe the processof interest rate determination in developing countries, it was esti-mated using quarterly data for Colombia and Singapore. Becausethese two countries are quite different, both in the developmentof their domestic financial markets and in the extent of controlsover capital flows, they should provide a fair test of the basicmodel. Since both countries vary in their openness, it would havebeen preferable to round out the picture by also including in theanalysis an example of a closed economy. For obvious reasons thiswas not possible.21

Since 1967 Colombia has followed a growth strategy based onexport promotion. During the past 15 years a crawling peg ex-change rate system has been in effect, and, at least in a segmentof the capital market, interest rates have been allowed to fluctuatefreely (Diaz-Alejandro (1976), Wiesner (1978), and Montes andCandelo (1982)). Over this period the domestic capital marketwas slowly liberalized, but some restrictions to capital movementswere maintained. For example, there were restrictions on theminimum maturity of loans obtained from abroad (usually fiveyears); the movement of capital in and out of the country requiredformal approval from a number of government agencies, includingthe Exchange Office, the Ministry of Finance, and the NationalPlanning Department; and there was a 95 percent advance pay-ment deposit on all capital outflows.22 Although there was somecapital mobility, the existence of such legal restrictions make itbest, for practical purposes, to characterize Colombia as a semi-open economy rather than a fully open one. In terms of ourmodel, therefore, we would expect to obtain a positive value forthe openness parameter i|j, and a value for 9 of less than unity.

21 First of all, there are few developing countries that can be viewed as com-pletely closed; second, those that would qualify do not have developed financialsystems with market-determined interest rates.

22 See International Monetary Fund (1984) for a detailed description of thenature and extent of capital controls in Colombia.



In contrast, the Singapore economy can be regarded as highlyopen, with virtually no restrictions on trade and capital flows(see Blejer and Khan (1983)). For example, imports are mostlyunrestricted, with a very small number subject to tariffs, and allpayments can be made freely. As far as the capital account isconcerned, the last elements of exchange controls were eliminatedin June 1978, and there are no hindrances to the movement ofcapital.23 After being pegged to the pound sterling, the Singaporedollar floated from June 1973 through late 1975. From then on thecurrency has been pegged to a trade-weighted basket of the cur-rencies of Singapore's major trading partners. The floating of theSingapore dollar led to a rapid development of the foreign ex-change market, and, although the volume of transactions is not aslarge as in the world's major financial centers, the Singaporemarket has over the years become among the largest in developingcountries. An active forward market, covering transactions ofvarious maturities, has also developed, with quotations beinggiven on a daily basis by participating banks. In general, theprogressive freeing of financial transactions, the exchange ratepolicy, and direct encouragement by the government through itsfinancial development program have combined to make Singa-pore an important financial center with close links to other majorfinancial markets. These institutional factors would suggest thatfor Singapore the openness parameter v|; would be close to unity,and that domestic interest rates would respond rapidly to foreigndevelopments (6 — 1).

Equation (14) and its equilibrium variant excluding the laggedinterest rate term were estimated by ordinary least-squares meth-ods for the two countries using quarterly data. For Colombia thedata were for the period running from the third quarter of 1968through the fourth quarter of 1982, whereas for Singapore thedata cover the period from the third quarter of 1976 through thelast quarter of 1983 (see the Appendix for data sources and defi-nitions). In the estimation equations for Colombia, the expectedrate of devaluation between periods t and t + 1, et, was replacedby the actual rate of depreciation in period t. This assumptionimplies that, during the period under consideration, the rate of

23 Even before 1978 there were no limits on residents' investments in theScheduled Territories (comprising the former Sterling Area). Because HongKong was included in the Scheduled Territories, residents could, in theory,transfer funds anywhere via the Hong Kong market, so that this restriction wasnot particularly effective.



devaluation in Colombia can be represented approximately as arandom walk process with zero drift (Edwards (1985a)). Becauseforward rates are available for the Singapore dollar, we used theforward premium to proxy the expected exchange rate change.Thus it is implicitly assumed in the analysis that the exchange raterisk premium is captured in the constant and error terms.24 Forboth countries the expected rate of inflation was calculated byfitting an autoregressive process (with seven lags) to the actualrate of inflation, then using the predicted values to represent TT* ,25

Finally, for reasons of efficiency the income elasticity for moneywas set equal to unity, and thus we were able to combine theincome and lagged money variables.26 The results for the twocountries are shown in Table 1.

Taking the case of Colombia first, we can see from Table 1 thatthe results are quite satisfactory. All the coefficients have thecorrect signs and are significant at the conventional levels.27 Inparticular, the significance of the coefficients of (if + et] and[log yt - log mr_i] clearly indicates that the nominal interest ratein Colombia has been sensitive to both foreign and domesticinfluences. If either of these factors is ignored—as is the casewhen more traditional approaches to interest rate determinationare used—important elements are left out of the story. Becausethe coefficient of the lagged interest rate is different from zero atthe 5 percent level of significance, implying that 0 is significantlydifferent from unity, to exclude this variable from the specifica-tion would obviously not be warranted. This is borne out by theresults, in which the restricted version of the equation yields apoorer fit.

We further calculated the values of what we regard as the keystructural parameters: the openness parameter (i|/) and the adjust-ment parameter for the interest rate (0). The value of \\f turns outto be 0.84 (with a t-value of 5.94), which is quite high and indi-

24 Experiments with alternative approximations for the expected exchangerate, such as the fitted values from a distributed lag function of the actualexchange rate, yielded broadly similar results. This is to be expected because inSingapore the forward rate has been a reasonably good predictor of the futurespot exchange rate. See Blejer and Khan (1983).

25 Using the actual rate of inflation (that is, the perfect foresight model) did notproduce any significant differences in the results.

26 This assumption is consistent with independent empirical evidence on thedemand for money relation for both countries—for example, Montes and Can-delo (1982) for Colombia, and Khan (1981) for Singapore.

27 Recall that the sign of the reduced-form coefficient for expected inflation(84) was ambiguous; the result in Table 1 indicates that \(1 - P)(a2 + a3) < 1.


Table 1. Results of General Interest Rate Model

Countryand Period

Colombia(1968-82)2

Singapore(1976-83)2

Note: Thestatistic; and

1 Adjusted2 From the

Constant(So)

-0.189(3.08)

-0.326(5.78)

-0.200(0.20)

-0.203(0.21)

ForeignInterestRate1

(Si)

0.353(1.98)0.786

(4.99)0.922

(23.68)0.923

(39.62)

Real Income- Lagged Real

Money Balances(82 = 83)

0.303(3.03)0.517

(5.49)0.052

(0.24)0.053

(0.25)

ExpectedInflation

(84)

0.256(2.00)0.422

(3.11)0.026

(1.40)0.026

(1.42)

LaggedInterest

Rate(85)

0.484(3.89)

0.001(0.02)

R2 DW H

0.820 — -1.29

0.768 1.41 —

0.991 — 0.39

0.991 1.83 —

values reported in parentheses are £ -ratios; R2 is the coefficient of determination; DW is the Durbin- Watson testH is the Durbin statistic for serial correlation in a model with lagged dependent variables,for expected exchange rate change,third quarter through the fourth quarter of the years indicated.



cates that the Colombian financial sector has, in practice, beenmore integrated with the rest of the world than one would havebelieved from the nature and extent of capital controls during thisperiod. According to this estimated value of v);, an increase in theforeign interest rate of 10 percentage points, for example, wouldbe translated into an increase of the domestic interest rate of over8 percentage points in the long run. Because the coefficient ofadjustment (0) is equal to 0.422 (with a t-value of 2.5), however,the average (or mean-time) lag in adjustment of the nominalinterest rate to a change in either the foreign interest rate or theexchange rate would be between three and four quarters.

The results for Singapore are quite different from those forColombia, with foreign factors clearly playing the dominant rolein the determination of the domestic interest rate. The coefficientof the foreign interest rate and expected exchange rate change, Si,is not significantly different from unity at the 5 percent level. Theremaining coefficients in the equation have the expected signs butare all statistically insignificant. This result implies that for allintents and purposes the openness parameter v|; is unity, which isa result one would have expected in the case of Singapore. Do-mestic monetary developments have no direct effect on the inter-est rate, although it is possible that they still could have indirectinfluence through their effect on the forward premium. This par-ticular channel, however, has not been considered here (see Sec-tion III). Moreover, because the value of 9 is unity, implying thatthe adjustment of the domestic interest rate is instantaneous andthat interest parity is maintained continuously, it is clearly amatter of indifference which of the two specifications for Sing-apore is considered. Both the equations—that is, with and withoutthe lagged interest rate term—appear equally well specified.

The results reported above were obtained by using the excesssupply of real money balances as the appropriate formulation forthe monetary disequilibrium term. As mentioned earlier, thereare other ways in which a monetary disequilibrium could affectnominal interest rates. For example, it has recently been argued(in Makin (1982), for example) that nominal monetary surprisescan have a temporary effect on nominal interest rates. To in-vestigate this proposition, equation (14) was re-estimated by re-placing [log mt - log mf] in equation (4) with a nominal money-surprise variable, defined as the residuals from an equation inwhich the rate of growth of nominal money was regressed on itslagged values for up to seven periods. The results for both coun-



tries with this formulation were quite similar to those reported inTable 1.

III. Limitations and Extensions

The model presented here has its limitations and can obviouslybe expanded in several directions. In this section we briefly discussfour possible extensions: (1) analysis of the determinants of realinterest rates in developing countries; (2) analysis of interest ratebehavior during the process of liberalization of the capital accountof the balance of payments; (3) explicit modeling of the expectedrate of devaluation in the context of interest rate behavior in opendeveloping countries; and (4) consideration of the effects of cur-rency substitution. This list is by no means exhaustive; specifi-cally, it does not incorporate various econometric issues thatcould arise in estimating a model of interest rate determination.Such issues would include, among others, simultaneity, specifica-tion of the underlying dynamics, and the proper treatment of theerror structure. Here we focus on what we see as the principaltheoretical extensions.

Real Interest Rates in Developing Countries

Some recent studies (for example, Cumby and Mishkin (1984))have empirically analyzed the behavior of real interest rates inindustrialized countries, placing special emphasis on whetherthese rates have tended to equalize across countries. Even if thereare no exchange controls, the capital account is fully open, and thenominal arbitrage condition holds, from a theoretical perspectivereal interest rates can still differ among countries. For example, anexpectation of a real depreciation would cause a country to havea higher real interest rate than the rest of the world.28

The framework discussed in this paper can be easily extendedto analyze the process of determination of (ex post and ex ante)real interest rates. Because the ex post real interest rate is definedas the nominal rate minus the actual rate of inflation, a simple wayof analyzing this issue is to add an explicit inflation equation to themodel. The resultant two-equation model could then be used to

28 On the relation between real exchange rates and real interest rates, seeDornbusch (1982a).

29 Note that the adjustment equation (6) in our model could be interpreted asan inflation equation, although we do not explicitly do so.



determine simultaneously the nominal interest rates and the rateof inflation, and the ex post real interest rates can then be directlyobtained from these two equations.30 Furthermore, if the inflationequation is used to determine the expected rate of inflation, thenone can calculate the ex ante real rate of interest.

To keep within the spirit of the model outlined here, the in-flation equation specified should be general enough to allow bothclosed and open economy factors to play a role. In the extremecase of a fully open economy, domestic monetary conditions willhave no direct effect, and the inflation rate will depend solely onforeign inflation and the (actual) rate of devaluation. If, in addi-tion, it is assumed that the expected real exchange rate will remainconstant, the model will predict the equality of domestic andforeign real interest rates. If the economy is completely closed,however, the domestic rate of inflation and the nominal and realinterest rates will have no relation to their world counterparts.

Interest Rates and Liberalization

One of the limitations of the model presented in this paper is thatit assumes a constant degree of openness of the financial sector inthe country under study. But several developing countries haverecently gone through liberalization processes characterized by,among other things, the relaxation or removal of existing capitalcontrols. To the extent that these liberalization processes yield ahigher degree of integration of domestic and world capital mar-kets, the assumption of a constant v|> is clearly inappropriate.31

There are several possible ways to proceed if the degree ofopenness is changing over time. The simplest way to model thisvariation would be to make the openness parameter a linear func-tion of time:

where i|i0 is the constant part of the openness parameter and t isa time trend. We would expect that i^ > 0. If the level and in-tensity of capital controls vary smoothly and gradually over the

30 Bleier and Gil Diaz (1985) specify a two-equation model for the real interestrate and inflation. Their model, of course, can be used to determine the nominalinterest rate as well.

31 Note also that \|> would depend on the interest rate chosen. For differentinterest rates one could easily have different values of i|i. We are indebted toMichael Mussa for this point.



period of study, then equation (15) would be a reasonable approx-imation. One could use equation (15) to substitute for i|/ in theinterest rate equation and then directly estimate the resultantreduced form. This simple form would obviously break down ifthe changes in capital controls were abrupt or erratic, and it wouldbe necessary to consider other methods to capture the liberal-ization process formally.

Ideally, of course, one would wish to have some type of indexthat directly measured the degree of legal capital controls. Itwould then be possible to specify openness as a function of thisindex (C):

In the estimation process several alternative functional forms canbe assumed.32 The main problem with this formulation, however,is obtaining data for the capital controls index C. One possibleway would be to construct a subjective measure from actual infor-mation on the system of capital controls in the country in ques-tion. Another approach would be to use some type of proxymeasuring the severity of capital controls, such as the black mar-ket exchange premium.33

Expected Devaluation and Interest Rate Determination

No mention has yet been made of the way in which the expectedrate of devaluation or the forward premium is determined. Forpurposes of the present exercise, these were assumed to be ex-ogenous. This is quite a restrictive assumption, and a more real-istic analysis would have to recognize that the expected exchangerate change is likely to be affected by movements in domesticinterest rates and by domestic monetary conditions in general.But recognizing this issue and actually doing something about itare quite different matters, since in practice endogenizing theexpected rate of devaluation or the forward premium has in mostcases proved to be exceedingly difficult.

32In formulating such equations, one has to recognize that the endogenousvariable (v|/) is bounded (0,1). For this constraint to be taken properly intoaccount, the precise functional forms would be more complicated than the linearones described here.

33 A problem with the black market premium is that it will tend to capture avariety of factors, including the effect of actual and expected capital controls.



The way to proceed would depend on the exchange rate systemoperative in the country in question. If the country has a floatingexchange rate, standard contemporary theories of exchange ratebehavior can perhaps be used. Even so, the task would not be easybecause these models have not been particularly successful inpredicting exchange rate movements (Levich (1985) has surveyedsuch models for the major industrial countries). Under fixed ratesthe problem becomes even more complicated because the proba-bility of an exchange rate crisis would then have to be modeledexplicitly. Some initial attempts have been made in this direction,but the modeling of exchange rate crises is still in its infancy (seeBlanco and Garber (1983) for one such attempt for Mexico). Byand large it seems that the present state of the art of exchange ratemodeling would preclude paying anything more than lip service tothis particular issue.

Effects of Currency Substitution

In combining the closed economy version of the interest ratemodel with the open economy formulation, the basic money de-mand function was left unchanged. Recall that this function allowsfor substitution to take place between money and domestic finan-cial assets and goods. This substitutability is the appropriate speci-fication for a closed economy, but it does prove to be somewhatrestrictive when the possibility of substitution between domesticand foreign money, defined in general as currency substitution,is admitted. In other words, one now has another asset in thesystem—that is, foreign money, for which the rate of return alsohas to be taken into account. Thus, in combining the two modelsone has to recognize that the money demand function in anopen economy could be different from that function for a closedeconomy.

The importance of the phenomenon of currency substitutionhas been documented in several studies (for example, Ortiz (1983)and Ramirez-Rojas (1985)). In contrast to earlier opinion, whichheld that currency substitution was relevant only in countries withdeveloped financial and capital markets, these writers have re-cently shown that currency substitution takes place frequently indeveloping countries as well. Furthermore, it has been found tooccur in countries that differ considerably in the degree of fi-nancial development and integration with the rest of the world



and in the types of exchange rate regimes and practices. Currencysubstitution clearly is a factor that should be explicitly taken intoaccount in any realistic analysis.

How one would proceed to model the effects of currency substi-tution is not, however, all that clear. The general consensus is thatthe principal determinant of currency substitution is the expectedchange in the exchange rate, although (as pointed out in thepreceding subsection) there is considerable controversy abouthow it should be measured. Other things being equal, an expecteddepreciation of the domestic currency, for whatever reason,would cause residents to switch from domestic money into foreignmoney, and vice versa. Once the difficult problems associatedwith the choice of an appropriate empirical proxy for exchangerate expectations have been surmounted, however, the rest ofthe analysis becomes relatively straightforward. The (domestic)money demand function (5) in an open economy could be re-specified as:

The last term in this modified equation would then capture theeffects of currency substitution.

This type of formulation would not be applicable in the extremecases of interest rate determination in completely closed and com-pletely open economies. In a closed economy the variable e wouldobviously not enter; in a fully open economy domestic monetarydisequilibrium (and thus the demand for money), with or withoutcurrency substitution, does not matter. Equation (5 a) would cer-tainly be relevant in the intermediate case, which of course doescorrespond to the actual case in most developing countries.

IV. Conclusions

As more developing countries proceed to liberalize their do-mestic financial systems and to remove restrictions on capitalflows, the issue of interest rate determination becomes increas-ingly important. In particular, how interest rates can be expectedto behave in the changed environment and how they will respondto foreign influences and domestic policies are questions thatpolicymakers in developing countries must consider. Only wheninterest rate behavior is well understood will it be possible to



predict confidently the effects of interest rate changes on keymacroeconomic variables such as saving, investment, the balanceof payments, and economic growth. To affect these variables ispresumably the real purpose for which the liberalization policieswere originally designed.

In this paper we have derived a theoretically consistent modelthat we believe can serve as a starting point for analyzing theprocess of interest rate determination in those developing coun-tries that have undertaken policies of financial reform. Althoughthe model has a fairly simple structure, it is nevertheless able toincorporate the principal determinants of interest rates, such asforeign interest rates, expected changes in exchange rates, anddomestic monetary developments. One of the interesting charac-teristics of the model is that it is sufficiently general to be applica-ble to a variety of developing countries that differ widely in theirfinancial openness. Indeed, through the model it is possible todetermine empirically, from data for the individual country, thedegree of financial openness (defined as both the extent to whichdomestic interest rates are linked to foreign interest rates and thespeed with which domestic rates respond to changes in worldrates). This measure of "economic" openness may differ quitesignificantly from the "official" or "legal" degree of opennessimplied by the prevailing system of capital controls.

For illustrative purposes the model was applied to two coun-tries—-Colombia and Singapore—that, because they are at quitedifferent stages of financial development, provided a useful firsttest of the general nature of the model. Colombia still maintainsrestrictions on capital movements, and only part of the financialsector can be characterized as free; Singapore, in contrast, is ahighly open economy with a dynamic and sophisticated financialmarket that has close links with the world's major financial cen-ters. The estimates from the model confirmed our prior assump-tions: we found that both foreign and domestic factors were im-portant in interest rate determination in Colombia, but that onlyforeign factors appeared to matter in Singapore. Our results alsoindicated that Colombia is more open than suggested by the actualsystem of capital controls. In conclusion, although one shouldobviously be careful in generalizing from the results for only twocountries, we nonetheless feel that this model has considerablepotential and can serve as a useful starting point for studying thebehavior of interest rates in developing countries.



APPENDIX

Data Sources and Definitions

This Appendix briefly gives the major sources of the country data and defines theprincipal variables used in the model.

Colombia (1968-82)

The basic sources for the data were Montes and Candelo (1982); DireccionNacional de Planeacion (DNP); Direccion Administrativa Nacional deEstadistica (DANE), Boletin Mensual de Estadistica (Bogota), various issues;and International Financial Statistics (IPS), International Monetary Fund (Wash-ington), various issues.

The definitions of the variables and specific sources are as follows:

e Percentage change in the official buying rate for export receipts andcapital inflows (Montes and Candelo (1982), and DNP)

i Domestic interest rate: for 1968-69, the average rate on mortgage bills,for 1970-82, the effective annual yield on three-month certificadosde abono tributario, or tax certificates (Montes and Candelo (1982)and DNP)

/* Three-month U.S. Treasury bill rate (IPS)

M Narrow (Ml) money balances: for 1968-80, data are from Montes andCandelo (1982); for 1981-82, from DNP

P Consumer price index (DANE, Boletin Mensual de Estadistica)', TT isdefined as the percentage change in this index

y Quarterly real gross domestic product (GDP; Montes and Candelo(1982)); updated through 1982.

Singapore (1976-83)

The sources of the data are IPS and the Monetary Authority of Singapore(MAS), Monthly Bulletin (Singapore), various issues.

The definitions and specific sources are as follows:

e Three-month forward premium (MAS)

/ Three-month interbank rate (MAS)

/* Three-month Eurodollar rate (IPS)

M Narrow (Ml) money balances (IPS)

P Consumer price index (IPS); the variable IT is defined as the percent-age change in this index

y Quarterly real GDP; the annual series were obtained from IPS andinterpolated to a quarterly basis using an index of manufacturingproduction, also from IPS.



REFERENCES

Blanco, Herminio, and Peter M. Garber, "Recurrent Devaluation and Specu-lative Attacks on the Mexican Peso" (unpublished; Rochester, New York:University of Rochester, December 1983).

Blejer, Mario I., "Interest Rate Differentials and Exchange Risk: Recent Ar-gentine Experience," Staff Papers, International Monetary Fund (Wash-ington), Vol. 29 (June 1982), pp. 270-80.

, and Mohsin S. Khan, "The Foreign Exchange Market in a Highly OpenDeveloping Economy: The Case of Singapore," Journal of DevelopmentEconomics (Amsterdam), Vol. 12 (February-April 1983), pp. 237-49.

-, and Jose Gil Diaz, "On the Determination of the Real Interest Rate ina Small Open Economy: Domestic Versus External Factors" (unpublished;Washington: International Monetary Fund, 1985).

Cumby, Robert E., and Maurice Obstfeld, "A Note on Exchange-Rate Ex-pectations and Nominal Interest Differentials: A Test of the Fisher Hypoth-esis," Journal of Finance (New York), Vol. 36 (June 1981), pp. 697-703.

Cumby, Robert E., and Frederic S. Mishkin, "The International Linkage ofReal Interest Rates: The European-U.S. Connection" (unpublished; NewYork: New York University, 1984).

Darby, Michael R., "The Financial and Tax Effects of Monetary Policy onInterest Rates," Economic Inquiry (Long Beach, California), Vol. 13 (June1975), pp. 266-76.

Diaz-Alejandro, Carlos F., Colombia (New York: Columbia University Pressfor the National Bureau of Economic Research, 1976).

, "Southern-Cone Stabilization Plans," in Economic Stabilization in De-veloping Countries, ed. by William R. Cline and Sidney Weintraub (Wash-ington: Brookings Institution, 1981), pp. 119-41.

Dornbusch, Rudiger (1982a), "PPP Exchange-Rate Rules and MacroeconomicStability," Journal of Political Economy (Chicago), Vol. 90 (February),pp. 158-65.

(1982b), "Stabilization Policies in Developing Countries: What Have WeLearned?" World Development (Oxford, England), Vol. 10 (September),pp. 701-08.

Edwards, Sebastian (1985a), "Money, the Rate of Devaluation, and InterestRates in a Semi-Open Economy: Colombia, 1968-1982," Journal of Money,Credit and Banking (Columbus, Ohio), Vol. 17 (February), pp. 59-68.

(1985b), "Stabilization with Liberalization: An Evaluation of Ten Yearsof Chile's Experiment with Free-Market Policies, 1973-1983," EconomicDevelopment and Cultural Change (Chicago), Vol. 32 (January),pp. 223-54.

Fama, Eugene F., "Short-Term Interest Rates as Predictors of Inflation," Amer-ican Economic Review (Nashville, Tennessee), Vol. 65 (June 1975),pp. 269-82.

Frenkel, Jacob A., and Richard M. Levich, "Covered Interest Arbitrage: Un-exploited Profits?" Journal of Political Economy (Chicago), Vol. 83 (April1975), pp. 325-38.



, "Transactions Costs and Interest Arbitrage: Tranquil Versus TurbulentPeriods," Journal of Political Economy (Chicago), Vol. 85 (December1977), pp. 1209-26.

Fry, Maxwell J., "Models of Financially Repressed Developing Economies,"World Development (Oxford, England), Vol. 10 (September 1982),pp. 731-50.

Hanson, James A., and Jaime de Melo, "External Shocks, Financial Reforms,and Stabilization Attempts in Uruguay: 1974-1983," World Development(Oxford, England), Vol. 13 (forthcoming, August 1985).

Harberger, Arnold C., "The Chilean Economy in the 1970s: Crisis, Sta-bilization, Liberalization, Reform," in Economic Policy in a World ofChange, ed. by Karl Brunner and Allan M. Meltzer, Carnegie-RochesterConference Series on Public Policy, Vol. 17 (Amsterdam: North-Holland,1982; New York: Elsevier, 1982), pp. 115-52.

International Monetary Fund, Annual Report on Exchange Arrangements andExchange Restrictions, 1984 (Washington, 1984).

Khan, Mohsin S., "The Dynamics of Money Demand and Monetary Policy inSingapore," in Monetary Authority of Singapore, Papers on Monetary Eco-nomics (Singapore: Singapore University Press for the MAS, 1981),pp. 46-76.

Lanyi, Anthony, and Rii§du Saracoglu, "The Importance of Interest Rates inDeveloping Economies," Finance & Development (Washington), Vol. 20(June 1983), pp. 20-23.

Levich, Richard M., "Empirical Studies of Exchange Rates: Price Behavior,Rate Determination and Market Efficiency," in Handbook of InternationalEconomics, ed. by Ronald W. Jones and Peter B. Kenen (Amsterdam:North-Holland, 1985; New York: Elsevier, 1985), pp. 979-1040.

Lizondo, Jose Saul, "Interest Differential and Covered Arbitrage," in FinancialPolicies and the World Capital Market: The Problems of Latin AmericanCountries, ed. by Pedro Aspe Armella, Rudiger Dornbusch, and MauriceObstfeld (Chicago: University of Chicago Press, 1983), pp. 221-40.

McKinnon, Ronald I., Money and Capital in Economic Development (Washing-ton: Brookings Institution, 1973).

Makin, John H., "Effects of Inflation Control Programs on Expected RealInterest Rates," Staff Papers, International Monetary Fund (Washington),Vol. 29 (June 1982), pp. 204-32.

Mathieson, Donald J., "Inflation, Interest Rates, and the Balance of PaymentsDuring a Financial Reform: The Case of Argentina," World Development(Oxford, England), Vol. 10 (September 1982), pp. 813-28.

, "Estimating Models of Financial Market Behavior During Periods ofExtensive Structural Reform: The Experience of Chile," Staff Papers, Inter-national Monetary Fund (Washington), Vol. 30 (June 1983), pp. 350-93.

Melvin, Michael, "The Vanishing Liquidity Effect of Money on Interest Rates:Analysis and Implications for Policy," Economic Inquiry (Long Beach,California), Vol. 21 (April 1983), pp. 188-202.



Monies, Gabriel, and Ricardo Candelo, "El enfoque monetario de la balanza depagos: El caso de Colombia, 1968-1980," Revista de Planeacion y De-sarrollo (Bogota), Vol. 14 (May-August 1982), pp. 11-40.

Mundell, Robert A., "Inflation and Real Interest," Journal of Political Econ-omy (Chicago), Vol. 71 (June 1963), pp. 280-83.

Ortiz, Guillermo, "Currency Substitution in Mexico: The Dollarization Prob-lem," Journal of Money, Credit and Banking (Columbus, Ohio), Vol. 15(May 1983), pp. 174-85.

Ramirez-Rojas, C. Luis, "Currency Substitution in Argentina, Mexico, andUruguay" (unpublished; Washington: International Monetary Fund, 1985).

Sjaastad, Larry A., "Failure of Economic Liberalism in the Cone of LatinAmerica," World Economy (Oxford, England), Vol. 6 (March 1983),pp. 5-26.

Tanzi, Vito, "Inflation, Indexation and Interest Income Taxation," QuarterlyReview, Banca Nazionale del Lavoro (Rome), No. 116 (March 1976),pp. 64-76.

, "Inflationary Expectations, Economic Activity, Taxes, and InterestRates," American Economic Review (Nashville, Tennessee), Vol. 70(March 1980), pp. 12-21.

Townsend, Robert M., "Financial Structure and Economic Activity," AmericanEconomic Review (Nashville, Tennessee), Vol. 73 (December 1983),pp. 895-911.

van Wijnbergen, Sweder, "Interest Rate Management in LDCs," Journal ofMonetary Economics (Amsterdam), Vol. 12 (September 1983), pp. 433-52.

Wiesner, Eduardo, "Devaluation y mecanismo de ajuste en Colombia," inPolitica economica externa Colombiana, ed. by Eduardo Wiesner (Bogota:Asociacion Bancaria de Colombia, 1978).

Zahler, Roberto, "Recent Southern Cone Liberalization Reforms and Sta-bilization Policies: The Chilean Case, 1974-1982," Journal of Inter american(EMS >0), and the nominal interest rate would fall below its1983), pp. 509-62.



DANIEL CITRIN*

PAPER PRESENTS the results of an empirical investigationof the impact of exchange rate variation on the exports of

three major Japanese industries—motor vehicles, consumer elec-tronics, and iron and steel. The study is a disaggregated one, withthe following five major Japanese exports chosen for analysis:subcompact passenger cars, color television sets, galvanized steelsheet, heavy steel plate, and tin plate.

Determining the extent to which the exchange rate influencestrade flows is especially relevant for Japan. One motivation for theadoption of protectionist measures by some of Japan's majortrading partners may be doubts as to the efficiency of the yenexchange rate for adjusting Japan's trade account. Between 1976and 1978, in particular, while Japan's real effective exchange rateappreciated by 24 percent, the surplus on merchandise trade grewfrom US$2.4 billion in 1976 to US$9.7 billion in 1977 and US$18.2billion in 1978.1 In 1979 Japan's trade balance finally deteriorated,registering a deficit of US$7.6 billion, but its real effective ex-change rate depreciated in that year by 22 percent.

With comparable rates of technological progress and of foreignand domestic inflation assumed, the magnitude of the impact ofthe exchange rate on exports depends on how far an exchange ratechange is passed through to foreign currency prices of exports andhow far export volumes then respond to such price changes. Inrecent years increasing attention has been devoted to estimating

*Mr. Citrin, currently the Fund's resident representative in Jamaica, holdsdegrees from the University of California, Berkeley, and from the University ofMichigan. This paper was prepared while he was a member of the AsianDepartment.

International Monetary Fund (1982) and Japan, Ministry of Finance (1978,1980). The real effective exchange rate utilized here is based on wholesale priceindices.

404

'HIST


EXCHANGE RATES AND JAPANESE EXPORTS 405

the adjustment of export prices and volumes over the short run aswell as over the long run. Volumes have in general been found toadjust less rapidly than prices, leading to perverse short-runmovements in merchandise trade balances following an exchangerate change, or "J-curves" (Branson (1972), and Magee (1973)).Thus, this analysis places emphasis on the explicit measurement ofthe time lags involved in the response of export prices and vol-umes to exchange rate variation.

Among others, researchers at Japan's Economic PlanningAgency (EPA) have devoted considerable attention to Japan'strade adjustment problems during the late 1970s (Komine andothers (1978) and EPA (1978)). Based on the Agency's quarterlymacroeconomic model, with revised export and import equationsto account for lags in adjustment, various simulation exerciseswere conducted within a general equilibrium framework to esti-mate the effects of exchange rate changes on various aspects of theJapanese economy. The results of these simulations indicatedperverse short-run effects of changes in the yen exchange rate. Forexample, in fiscal year 1977/78 (April 1977 through March 1978),during which Japan's nominal effective exchange rate appreciatedby an estimated 22 percent and its real effective exchange rateappreciated by an estimated 13 percent (according to the Inter-national Financial Statistics of the International Monetary Fund),J-curve effects were estimated to have yielded a US$3.2 billionincrease in Japan's trade balance (EPA (1979, p. 393)). The esti-mated positive impact amounts to roughly one third of the US$9.7billion increase in Japan's trade surplus in 1977/78 over that of theprevious year.

Whereas the EPA's analysis incorporated lags in the adjust-ment of Japan's exports and imports to exchange rate changes,Wilson and Takacs (1980) took the analysis one step further byestimating the impact of exchange rate expectations, or leads inadjustment, on the behavior of Japanese trade flows. Specifically,they estimated the extent to which expected appreciations of theyen exchange rate may have led to an export acceleration andimport deceleration that exacerbated the perverse movement ofthe Japanese trade balance during the late 1970s. Their resultswere rather striking: the estimated J-curve for Japan's trade bal-ance continued to move in a perverse direction for several quar-ters subsequent to an exchange rate change and, for most of theadjustment period, implied a considerably larger perverse move-ment than that implied by a J-curve that incorporated only adjust-


406 DANIEL CITRIN

ment lags. Such a result suggests that, between 1977 and 1979,traders' expectations of a further yen appreciation gave rise toperverse movements in the trade balance that overwhelmed theadjustment of trade flows to past exchange rate changes. Follow-ing the approach of Wilson and Takacs, this analysis also incorpo-rates the impact of expectations of relative export prices on thedemand for Japan's exports.

As mentioned previously, this paper examines the response ofJapanese exports to changes in the yen exchange rate at a highlydisaggregated level, looking at specific major export products.From a theoretical viewpoint, such a disaggregated analysis avoidsthe aggregation problem of biased elasticity estimates made fa-mous by Orcutt (see Orcutt (1950) and Learner and Stern (1970)).One such problem arises because the response of export prices toexchange rate changes varies across industries, since suppliersdiffer in the extent to which they maintain prices in line with thoseof foreign competitors; an aggregate demand equation thereforeimplies biased estimates. Another reason for the disaggregation isthat tensions between Japan and her trading partners have tendedto focus on specific industries, such as steel in the late 1970s andautomobiles, electronic equipment, and machine tools in the1980s. Exports of the industries sampled—that is, motor vehicles,consumer electronics, and iron and steel goods—accounted for 34percent of the U.S. dollar value of Japan's total exports during the1975-79 period and have, in addition, all been involved in recentdisputes between Japan and her trading partners.

Simultaneous-equations models of the supply of and demandfor both exports and domestic sales of each product consideredwere estimated. The theoretical model specifies the following dy-namic elements in the adjustment process: the extent to which anexchange rate change is passed through to foreign currency exportprices over time; leads and lags in the response of export ordersto relative export price changes; and delivery lags between ordersand shipments. Although the analysis does take into account theindirect impact of exchange rate shifts on export prices throughchanges in prices of imported raw materials, the study is partialequilibrium in nature in the sense that foreign prices, wages,inventory accumulation, and the exchange rate are treated asexogenous. Given the unique characteristics of the Japanese labormarket, the tendency for wage increases to be linked to profit-ability of firms, and the observed flexibility in real wages in Japan,the assumption of exogenous wages may not be overly restrictive.



(See Komiya and Suzuki (1977) and Shinkai (1982).) In addition,wages account for a small proportion of total variable costs for allof the sample products, as indicated by input-output coefficients(see the Appendix).

I. The Model

This section presents the basic theoretical model of demand forand supply of exports and domestic sales underlying the empiricalanalysis presented in the paper. The additional product-specificvariables included in the estimation are reviewed in the discussionof the estimation results.

Demand Side

The demand side of the model consists of four equations. Forboth exports and domestic sales, the model specifies a demandequation that determines factory orders and an equation thatspecifies actual shipments as a function of current and past orders.

Export orders are specified in real terms as a log-linear functionof foreign income, the relative price of exports, and a non-price-rationing variable. (Demand theory provides little guidance onthe appropriate functional form. The choice is thus essentially anempirical one, and the log-linear formulation is adopted on thebasis of work by Khan and Ross (1977).) Nonprice rationing isincorporated on the assumption that, since quoted prices may berelatively sticky in the short run, goods markets are also clearedby variables such as changes in delivery dates, length of queues,or credit terms. Because nonprice rationing is essentially cyclical,it is represented by the ratio of inventories to the volume of total(export and domestic) orders.2 Past values of foreign income andthe relative export price are included because recognition andother lags may delay the response of export orders to changes inthese factors, and expected relative export prices are included toaccount for the impact of price and exchange rate expectations on

2 See Gregory (1971) for an extensive treatment of the impact of nonpriceelements on trade flows. Gregory considers only the effect of domestic nonpricerationing on import demand, but the same arguments may be used for includinga domestic cyclical variable in the export demand equation; see also Ahluwaliaand Hernandez-Cata (1975) and Hooper (1976).


408 DANIEL CITRIN

current export demand. The volume of export orders (XO) attime t is therefore specified as follows:

where YF is foreign activity (in real terms); PX is the contractexport price in domestic currency; ^? is the yen-per- foreign-currency effective exchange rate, weighted according to exportmarket shares; PF is the weighted average foreign price, withweights as for the exchange rate (/?); and IO is the ratio of inven-tories to total volume of orders.

Equation (1) takes account of adjustment lags in explainingexport demand. Actual export shipments, however, represent aflow of goods generated by orders placed at some time in the past.With a constant delivery-lag structure, the volume of export ship-ments (QX) at time t is specified according to the following linearrelation:

where S6, equals 1, and bt is the share of export orders at time(t — i) shipped at time t. This equation should not be specified inlog-linear form since, as orders fluctuate over time, the relativeshares of current and past orders in current shipments will alsovary. In addition, the vector of delivery-lag coefficients (the bt)should vary over time, as a function (for instance) of the ratio ofunfilled orders to current production. Such variations are notconsidered here; see Hooper (1976), Ahluwalia and Hernandez-Cata (1975) and Artus (1973, 1974) for a similar treatment ofdelivery lags in the determination of shipments and unit values oftrade flows.

Domestic orders (DO) and shipments (QD) are derivedanalogously and are specified (with time subscripts deleted) asfollows:

and


where Qv is the output from capital of vintage v; Kv is the amountof capital of vintage v in operation; Lv is the amount of laboremployed by capital of vintage v; RMV is the amount of rawmaterials employed by capital of vintage v; and b and c are thefixed coefficients of production.

Labor and raw materials are combined by a Cobb-Douglasprocess into a composite variable input, with the assumption ofconstant returns to scale, so that a\ + a2 equals 1. Embodied tech-

The supply side of the model is developed within the context ofa firm that discriminates between its export and domestic markets.The derivation proceeds as follows for both export and domesticsales. First, the optimal short-run price is derived from the theoryof a firm that maximizes profits subject to a short-run productionfunction; because the model considers adjustment in the short tomedium term, the stock of capital is regarded as fixed. Second,the optimal adjustment path for contract prices is determined byincorporating certain longer-run considerations in addition toshort-run profits. The above steps are then integrated to yield anequation for the actual contract price.

Maximizing total profits from both exports and domestic salessubject to a production function yields the familiar first-ordercondition that sets the optimal short-run export price (PX*) equalto a markup over marginal cost, as follows:

where NX is the price elasticity of export demand, and MC is themarginal cost of production.

The vintage capital production function of de Menil (1974) isused to determine short-run marginal cost. Under this model,machines of different vintages are each assumed to have a fixed-coefficient production function given by:

where ^dt equals 1; YD is domestic activity (in real terms); andRPD is the relative domestic price, defined as the domestic priceof the good relative to competing domestic prices. Import pricesare not included in the equation because imports to Japan of theproducts considered in this paper are small.

Supply Side



410 DANIEL CITRIN

nical change raises the efficiency of new machines at the rate of r1?

and disembodied technical change raises the efficiency of all ma-chines at the rate of r2.

Short-run marginal cost is derived by noting that marginal costfor the firm is equal to the marginal cost of production on theoldest machine in operation and is given by the following expres-sion:3

where

The second stage of the derivation of the export price involvesthe determination of its optimal adjustment path in domestic cur-rency. The contract export price may deviate from the short-runoptimum because of several factors. Prices in domestic currencymay be sticky because of administrative costs of frequent pricechanges or general uncertainty about the reactions of either buy-ers or competitors to price changes. Firms exporting goods incompetitive markets or those with a strong preference for main-taining or expanding market shares will be inclined, if necessary,to sacrifice short-term profits to stay in line with competitors'prices. In addition, firms may be willing to cut prices to smoothout fluctuations in levels of inventories or in rates of capacityutilization; this applies especially to firms with a relatively highproportion of fixed costs.

The contract export price is assumed to be set to achieve thebest possible compromise between the short-run profit maximiza-tion target and the additional considerations noted above. The

3 See de Menil (1974, p. 132) for a full derivation of the marginal cost function.

where W is the wage rate; PRM is the price of raw materials; andv' is the vintage of the oldest machine in operation. Replacing v'by (t - U)9 where £7 is the age of the oldest machine in operation,and substituting equation (7) into equation (5) allows the optimalshort-run export price to be written according to the followinglog-linear expression:



contract export price (PX) at time t is determined by minimizingthe total cost from not meeting all targets simultaneously:4

where L is the total loss subjectively perceived by suppliers; /i isthe loss coefficient associated with deviating from the short-runoptimum, and 12 is that associated with not maintaining pricestability. The loss related to the price-competitiveness target issplit into two elements—/3 reflects the loss from deviating fromthe competing foreign price in terms of its level, and the I4i are theloss coefficients on price changes over various time horizons thatdiffer from those of competitors. The first coefficient measuresthe desire of firms to stay in line with competing prices in the longrun; the second set of coefficients determines losses associatedwith short-run relative price changes, and k is thus finite andassumed to be small. Finally, /5 determines the loss resulting fromfluctuation in real activity, which is represented by the ratio ofinventories to total orders.

Minimizing the loss function (10) with respect to PX, the con-tract export price, while assuming a constant short-run exportprice elasticity of the inventory-orders ratio (NIOi PX), one obtains:

where ml equals IJS\ m2 equals /2/5; m3 equals 13IS\ m4i equals IJS\m5 equals —15 NIO, px/S\ and 5 is equal to /i + /2 + /a + 2/4,.

The equation determining the contract export price is now ob-

4 Although this approach is rather ad hoc, a rigorous formulation of the adjust-ment path of the contract price may be extremely complex because of theconflicting nature of the various goals and the fact that the additional longer-rungoals are motivated by entrepreneurial preferences not easily measured by tradi-tional economic variables. The approach is motivated by similar formulationsfound in Artus (1974) and Ahluwalia and Hernandez-Cata (1975).


412 DANIEL CITRIN

tained by substituting equation (9) into equation (11) and com-bining terms:

The domestic contract price equation is similarly derived. Indetermining the adjustment path of the domestic price, it is as-sumed that firms on the domestic market achieve the best possiblecompromise between short-run profits, price stability, and "cy-clical smoothness." Given the limited extent of import pene-tration into the Japanese markets of the products sampled in thispaper, the maintenance of competitiveness relative to imports isnot considered. The domestic contract price (PD) is thus specifiedas follows:

The theoretical model utilized as a basis for the empirical analy-sis consists, therefore, of equations for the volumes of exportorders and shipments, equations (1) and (2); the volumes of do-mestic orders and shipments, equations (3) and (4); the contractexport price, equation (12); and the contract domestic price,equation (13).

II. Estimation Results

This section presents the results obtained by applying the theo-retical model presented above to the estimation of prices andquantities (domestic and export) of Japanese subcompact pas-senger cars, color television sets, galvanized steel sheet, heavysteel plate, and tin plate (see the Appendix for a detailed descrip-tion of the data). Separate equations for orders and shipmentswere estimated only for heavy steel and tin plate, owing to a lackof adequate data for the other three products. In addition to thevariables specified in the basic model, explanatory variables rele-vant to specific products were also included. The price of gasolinerelative to other consumer items (in the United States) wasincluded in the demand equation for subcompact automobile ex-ports, to take account of the dramatic shift in consumer prefer-



ences toward Japanese cars, particularly in the United States, asa result of the oil price increases during the 1970s. Dummy vari-ables were employed to account for export restraint agreementsduring the sample period with respect to color television sets andsteel products. The proxies for expected relative export priceswere created by utilizing ARIMA time-series techniques to gen-erate separate forecasts for export prices, foreign prices, and ex-change rates.

The structural model of simultaneous supply and demand, alsoincluding lagged endogenous variables, was estimated accordingto an iterative estimation procedure appropriate for such sys-tems.5 Statistical estimation was based on monthly data over the1970-79 period; thus, whereas the theoretical relations of themodel would be appropriate in any current applications, the esti-mates of its behavioral parameters are strictly applicable to onlythe sample period.

Regression results for the export equations are presented inTable 1. (Because of the relatively complex estimation method-ology, preliminary regressions were run using single-equationtechniques to determine lag lengths and to ascertain whether toinclude certain variables in the final estimation.) The model per-forms quite well, as indicated by the R2 calculated with respect tothe original data (before transformation of the data to account forserial correlation), except for the equations for shipments andorders of tin plate. In addition, the estimated coefficients of theexplanatory variables included in the simultaneous-equations esti-mation procedure all have the expected signs.

Looking first at the export demand equations, one finds that thedistributed lag structure on foreign income is statistically signifi-cant at the 5 percent level for only two of the five productsconsidered—heavy steel plate and galvanized steel sheet. Furtheranalysis suggests that the insignificance of the distributed lags for

5 The presence of lagged endogenous variables coupled with the expectation ofserial correlation, reflecting estimation over monthly time series, implies that theusual two-stage or three-stage least-squares estimation techniques are inconsis-tent. Estimation was therefore conducted by applying an iterative modificationof a two-step procedure developed for such models by Hatanaka (1976). Distrib-uted lag structures were incorporated according to the method derived by Shiller(1973).

The ^-statistics cannot be interpreted in the usual way for the distributed lagcoefficients estimated according to the Shiller lag method (1973), a Bayesianprocedure that imposes prior constraints on the shape of lag structures. Thus thestatistical significance of the various lag structures was tested by calculatingappropriate F-statistics for their group influence.


414 DANIEL CITRIN

Table 1. Regression Estimates for Export Volume and Price Equations

VariableSubcompact Color Galvanized Heavy

Passenger Cars Television Sets Steel Sheet Steel PlateTin

Plate

Export shipments(QX)

XO [lag length]1

Export demand(In XO)

1[5]

0.729

1[5]

0.560

In (PGAS)

In (10)

In (YF) [laggedterms]

In (RPX)e 2

In (RPX) [laggedterms]

R2

Export price(In PX)

In (VCOST)

In (7O)

\n(PF-R)44In (PX/PF •/?)_/In (PX/PF -R)_2

4

In (PX)_,

R2

1.32(5.81)0.04

(0.77)0.97

[2]0.10***3

-2.80***3

[18]0.936

0.15(3.95)

-0.016(-3.79)

0.43***0.23***0.71***0.42

(6.32)0.990

0.16(1.78)0.79

[2]-0.78**3

-2.18**3

[13]0.915

0.14(3.61)

0.66***0.38***0.22***0.20

(3.77)0.980

0.22(3.57)1.73***

[2]0.001

-1.06[10]

0.861

0.070(1.80)

-0.027(-2.50)

0.47***0.28***0.17***0.46

(4.49)0.990

0.61(3.08)0.94***[2]

-0.13-0.65

[4]0.823

0.46(4.88)

0.54***0.29***0.13***

0.987

0.94(3.34)0.36[2]1.15

-2.65[15]0.466

0.023(1.23)

0.47***0.29***0.077***0.50

(5.78)0.990

Note: QXis export shipments (real); XO is export orders (real); PGAS is therelative price of gasoline (in U.S. dollars); IO is the inventory-orders orinventory-sales ratio; YFis foreign income; RPXe is the expected relative exportprice; RPX is the relative export price (PX/PF - R)', PX is the contract exportprice (in yen); VCOST is variable input cost; and PFis the foreign price. Coef-ficient estimates for the constant and various dummy variable terms are omitted.For variables with lagged terms, the sums of estimated coefficients are pres-ented; the number of lagged terms is indicated within brackets. Estimationprocedure was corrected for first-order serial correlation; t-statistics are shownin parentheses; ** denotes F-statistics for distributed lag structure significant atthe 5 percent level; *** denotes F-statistics significant at the 1 percent level; R2

(coefficient of determination) values were calculated on the original data; thatis, before transformation to correct for serial correlation.

*In months.2 The time horizon for relative export price expectations was assumed to be

three months for all products.3 Significance level of F-statistics for distributed lag structure including the

expected relative price terms.4 Significance level of F-statistics for distributed lag structure related to foreign

competing prices.

r2



the other three products reflects an even quicker response ofexport demand with respect to changes in foreign income thanspecified. Regressions that include only current foreign incomeand foreign income lagged one month show a significant estimatedcoefficient on foreign income for subcompact cars and tin plate,and on lagged foreign income for color television sets.

Relative export prices are significant at the 5 percent level onlyin the export demand equations for subcompact cars and colortelevision sets; they are significant, however, at the 15 percentlevel for export demand for heavy steel plate and tin plate. Thenon-price-rationing variable (the inventory-sales ratio) is signifi-cant for all but one product, subcompact passenger cars; it isparticularly significant with a large estimated coefficient for thethree steel products, suggesting the initiation of "export drives"during times of weak domestic demand.

Table 2 presents the structural equation effects of various fac-tors on export volumes and prices in the long run. (These effectsare based on single-equation results and do not take account ofthe simultaneity of the model. Because of the nonlinear nature ofthe specified model, in-sample simulation of the estimated modelis required to calculate the simultaneous effects of changes inexogenous variables. The presence of the inventory sales ratio asan explanatory variable makes the model nonlinear in logarithmsand prevents a reduced-form solution. Thus, reduced-form elas-ticities cannot be calculated.) Export demand is estimated to bequite responsive to relative price changes for subcompact cars andcolor television sets, but it is relatively price inelastic for the threesteel products. The results shown in the first and second parts ofTable 2 indicate that export shipments are estimated to adjust toa relative price change with a lag ranging from 9 months for heavysteel plate to 20 months for tin plate.

Relative price expectations are estimated to have a strong im-pact on the export demand for tin plate and for color televisionsets. The elasticity of tin plate export orders with respect to ex-pected relative export prices (the sum of the estimated coefficientson the variables for relative price expectations) is estimated toamount to 1.15. This implies a considerable initial increase (de-crease) in export demand for tin plate subsequent to a rise (fall)in relative export prices, owing to anticipations of further priceincreases (decreases) in the near future. For color television sets,however, the estimated price expectations elasticity of export de-mand is equal to -0.78; this indicates an initial fall (rise) in


416 DANIEL CITRIN

Table 2. Long-Run Effects of Selected Factors on Japanese Export Volumesand Prices: Estimates of Structural Equations

Long-Run AdjustmentFactor Elasticity Period (In months)

Export shipments equationsExport orders

Heavy steel plate l.OO1 5Tin plate l.OO1 5

Export demand equationsRelative export price

(including expectations)Subcompact passenger cars -2.702 18Color television sets -2.962 13Galvanized steel sheet -1.062 10Heavy steel plate -0.782 4Tin plate -1.512 15

Contract export price equationsForeign competing prices

Subcompact passenger cars 0.47 4Color television sets 0.30 26Galvanized steel sheet 0.21 363

Heavy steel plate 0.20 6Tin plate 0.85 231 The cumulative elasticity of shipments with respect to orders was constrained

to be equal to 1.2 Sum of the structural coefficients of relative export prices.3 Cumulative elasticity estimate within 1 percent of long-run solution.

foreign demand following an effective appreciation (depreci-ation), reflecting expectations of a reversal in relative prices in thefuture. During the 1970-79 sample period, when there was a trendappreciation of the yen exchange rate, one would expect thatexpectations would have had a temporary positive effect on exportdemand; such an impact was found, by Wilson and Takacs (1980),to have been an important effect of exchange rate changes on totalJapanese exports during 1972-78. The opposite result, obtained inthe case of exports of color television sets, could reflect the com-petitive nature of the North American market during the 1970s,where customers expected that exchange rate changes soon wouldbe offset by cost-cutting or quality-improving technological ad-vances by Japanese suppliers.

With regard to the export price equations, the lag structuresrelated to competing foreign prices, the lagged export price (rep-



resenting the price-stability target), and the price of variable in-puts are all significant at the 5 percent level, with the exception ofthe lagged export price for heavy steel plate (insignificant and notincluded) and the price of variable inputs for tin plate (significantat the 11 percent level). (Wages and raw material prices are com-bined into a single-variable cost term according to input-outputweights in order to overcome multicollinearity problems encoun-tered in the estimation process. See the Appendix.) The exportprice seems to respond to cyclical conditions in the case of sub-compact cars and galvanized steel sheet; the vintage of the capitalstock is not estimated to affect the export price of any productsignificantly.

Reflecting the significance of the competing-price and laggedexport price variables, the estimation results indicate that contractexport prices can deviate significantly from levels based purely onprofit-maximizing considerations. In the short run, export pricesin domestic currency are estimated to respond to changes in for-eign prices for all products; the implication is that exchange ratechanges are not fully passed through to foreign currency prices.The results also suggest that exchange rate changes may not befully passed through in the long run. The estimated long-runforeign price (or exchange rate) elasticity of the contract exportprice in yen ranges from 0.85 (for tin plate) to 0.20 (for heavyplate); that is, the pass-through estimates range, correspondingly,between 15 percent and 80 percent. The long-run elasticity esti-mates are all significantly different from zero except in the case ofgalvanized steel sheet.7 The adjustment period is estimated to berelatively short for subcompact cars and heavy steel plate and tobe quite long for color television sets, galvanized steel sheet, andtin plate, although a substantial portion of the adjustment for thelatter group is estimated to occur within 12 months.

III. Simulation Results

To account for simultaneity, the impact of a hypothetical10 percent appreciation of the yen was simulated for each productover the period 1977-79. Simulations were run on the estimatedmodels according to two constant exchange rates during 1977-79,

7 The estimated variances of the long-run elasticity estimates were calculatedaccording to Kmenta (1971, equation (11.40), p. 444).


418 DANIEL CITRIN

one at the December 1976 level and the other at a level 10 percenthigher, and the results were then compared to determine theimpact of appreciation. The simulation exercise also incorporatedestimates of the indirect impact of appreciation on contract exportprices due to changes in raw material costs.8 The simulated ex-change rate effects over time are presented in Table 3. The adjust-ment of prices and volumes may be viewed as virtually completewithin a three-year period subsequent to the hypothetical 10 per-cent appreciation in January 1977.

The results of full model simulation indicate once again thatexchange rate changes are partially offset by movements in exportprices in domestic currency terms and, therefore, are not fullyreflected in prices in foreign markets. The extent to which thedomestic currency price is estimated to decline during the three-year period following a 10 percent appreciation ranges from a highof 8.0 percent for tin plate to 3.4 percent for heavy steel plate. Thesimulated decreases in export prices (in yen) are greater thanthose based on the structural coefficient estimates implied by thethird part of Table 2; the difference can be attributed almosttotally to the fall in raw material prices following an appreciation.For subcompact cars and galvanized steel sheet, the significantcyclical variable in the export price equation has an additional,albeit marginal, negative effect. The offsetting effect of lower rawmaterial costs on export prices is most significant for heavy steelplate and for galvanized steel sheet, reflecting the importance ofimported inputs in Japan's steel industry coupled with a largeestimated coefficient on the variable input cost term in the exportprice equation.

A relatively rapid adjustment of the export price toward itslong-run position is simulated for all products; the adjustmentoccurs almost immediately for subcompact cars and heavy steelplate. The products are split into two groups, however, with re-spect to whether the domestic-currency price response increasesor decreases over time. For color television sets, heavy steel plate,and galvanized steel sheet, the fall in the contract export price in

8 Using coefficients from Japan's 1975 input-output table (Japan, Adminis-trative Management Agency (1978)), the EPA estimated the exchange rateelasticity of raw material prices at a sectoral level; the relevant estimates aretransport equipment, 0.09; electrical equipment, 0.12; and iron and steel prod-ucts, 0.23 (see EPA (1978)). The variable input cost terms are adjusted accord-ing to these estimates, with the adjustment assumed to take place smoothly overa three-month period.



Table 3. Simulated Cumulative Impact of 10 Percent Appreciation of the Yenin January 1977 on Prices and Volumes of Japanese Exports, 1977-79

(In percent)

-4.1-4.7-4.8-4.8

-4.9-5.0-5.0-5.0

-5.1-5.1-5.1-5.2-5.2

0.6-0.2-2.1-5.3

-8.8-11.1-11.6-11.4

-11.1-11.0-10.9-10.8-10.8

1977JanuaryAprilJulyOctober


1979JanuaryAprilJulyOctoberDecember

Color television sets1977

JanuaryAprilJulyOctober



-6.1-5.3-4.6-4.1

-3.9-3.7-3.6-3.6

-3.6-3.5-3.5-3.5-3.5


-4.4-4.8-4.6-4.5

Galvanized steel sheet

-0.6-2.3-3.7-4.4

-5 .-4.8-8.1

-10.8

-12.7-13.5-13.8-14.2

-14.4-14.6-14.4-14.5-14.7

contractExport Prices Export

Monthly Period (In yen) Export Orders Shipments

sUBCOMPACT PASSENGER CARS


420 DANIEL CITRIN

Table 3 (concluded).

ContractExport Prices Export

Monthly Period (In yen) Export Orders Shipments









-4.4-4.3-4.2-4.2

-4.1-4.1-4.1-4.1-4.1

-5.3-3.8-3.4-3.4

-3.4-3.4-3.4-3.4

-3.4-3.4-3.4-3.4-3.4

-4.4-6.1-6.8-7.3

-7.6-7.7-7.9-7.9

-8.0-8.0-8.0-8.0-8.0

-1.6-3.8-4.4-4.9

-5.1-5.2-5.4-5.5

-5.7-5.8-6.1-6.2-6.4

5.50.6

-1.7-3.6

-4.2-3.4-2.7-2.3

-2.1-1.9-1.9-1.7-1.7

-4.8-4.8-4.9-5.0

-5.2-5.1-5.2-5.2-5.1

-0.1-1.9-3.9-4.6

-5.0-5.2-5.3-5.4

-5.6-5.7-5.9-6.1-6.2

0.71.5

-0.1-2.1

-3.6-3.8-3.2-2.7

-2.3-2.1-1.9-1.8-1.8

Tin plate

Heavy steel plate




yen terms decreases over time, reflecting increasing pressure tomove back toward a profit-maximizing price. For subcompact carsand, in particular, tin plate, however, the price declines continu-ously with time. In the case of subcompact cars, the slight furtherincrease in the price response is attributable largely to the laggedresponse of lower raw material prices to appreciation. The adjust-ment path for tin plate reflects slow adjustment from the pre-appreciation price because of a large estimated coefficient on thelagged export price; this price stickiness in turn may be due tostrong supplier preference for minimizing price fluctuation so asto reduce administrative costs or to prevent the loss of goodwillamong customers.

A 10 percent appreciation causes considerable declines in ex-port volumes of subcompact cars (10.8 percent) and of color tele-vision sets (14.7 percent). Exports of the three steel products, andparticularly those of tin plate, however, are quite inelastic withrespect to the exchange rate; the estimated decline in volumeamounts to 6 percent or less. For all items, the exchange rateresponsiveness of export volumes is substantially lower than thatindicated by the estimated structural relative price elasticities,primarily because the appreciation is not fully passed through toexport prices in foreign currency.

The simulations indicate that export volumes adjust to appre-ciation with a considerable lag. Nevertheless, over three fourths ofthe response of shipments occurs during the first year for all fiveproducts. With the exception of tin plate exports, volumes arelargely observed to decline gradually toward long-run levels. Fortin plate, however, the results indicate a considerable positiveresponse of volumes in the short run; this unusual movement re-flects the estimated strong positive impact of exchange rate ex-pectations on export orders as demand rises initially due toexpectations of further appreciation in the near future.

A stronger yen is estimated to have a dampening long-run effecton the value of exports in foreign currency for two of the fiveproducts in the sample, subcompact cars and color television sets(Table 4). For these two items, export value is estimated to adjustdownward as the decline in volume in the long run becomes largeenough to exceed the increase in the foreign currency price. Withrespect to the three steel products, however, yen appreciation isestimated to cause small positive movements of export value in thelong run; this counterintuitive result reflects inelastic price re-sponsiveness of demand.


422 DANIEL CITRIN

Table 4. Simulated Long-Run Impact of 10 Percent Appreciation of the Yenon Export Shipment Values in Foreign Currency

(In percent)

Product Response of Export Value

Subcompact passenger cars -6.0

Color television sets -8.2

Galvanized steel sheet 0.8

Heavy steel plate 0.4

Tin plate 0.2

Note: The long-run adjustment of export prices and volumes is defined to beequal to that obtained at the end of the three-year simulation period, with theunit value of exports equal to the contract export price.

The dynamic response of export values over the adjustmentperiod may not be precisely calculated for all products, since thelack of data about orders for some items prevented estimation ofthe dynamic response of unit values. (Export unit value at time ofdelivery is a function of the order-delivery lag structure, the cur-rency denomination of trade contracts, and exchange ratechanges between order and delivery; see Artus (1974), for in-stance.) Nevertheless, the results indicate the presence of signifi-cant J-curve effects. In addition to an increase in the value ofexports of the three steel products throughout the adjustmentperiod, the slow response of export volumes while contract exportprices increase in foreign currency terms suggests a considerablerise in the export value of subcompact cars during most of the firstyear after yen appreciation.

IV. Conclusions

Notwithstanding that the simulation results are strictly applica-ble to only the 1977-79 period,9 several implications seem well

9 The exchange rate effect on export volumes would be less during periodswhen exports accounted for a higher share of total sales, since the offsettinginfluence of the non-price-rationing variable would be greater.



established. For all five Japanese export items studied, the impactof exchange rate changes on exports is subject to lags on both thesupply side and the demand side, with the adjustment of demandbeing particularly slow. Furthermore, competitors' prices appearto play an important role in the pricing decisions of suppliers;thus, exchange rate changes are not fully passed through to exportprices (in foreign currency), either in the short run or in the longrun. (The estimated lack of full pass-through in the long run mayreflect suppliers' expectations of relative productivity gains, lowerreal wages, or offsetting future exchange rate movements.)

The analysis clearly indicates that exchange rate effects overtime differ by product. The exchange rate does seem to work asa tool for adjusting Japan's subcompact passenger car and colortelevision exports, even though adjustment is particularly slow inthe case of automobiles. (This result, of course, is based on the1970-79 sample period and should be interpreted with caution inthe presence of voluntary export restraints.) The exchange ratedoes not seem to have as great an impact on the various exportsof Japan's iron and steel industry; indeed, export values in foreigncurrency of these products are estimated to increase after appre-ciation. Even though such products are in general considered tobe homogeneous, Japanese suppliers seem to face relatively price-inelastic foreign demand for the three steel products in the sam-ple; this price inelasticity of demand suggests that quality may bea significant factor in world demand for Japanese iron and steelproducts. Moreover, the relative importance of the non-price-rationing and cyclical variables in the demand equations suggeststhat suppliers in the steel industry place a premium on the main-tenance of stable production levels and prices.

In addition to the partial equilibrium nature of some aspects ofthe model, the divergence of estimated effects of an exchange ratechange by product cautions one against generalizing the resultsobtained to an aggregate level. Although the results suggest thatthe full response of aggregate exports will entail a lag of at leastone year and that exchange rate changes may not be fully passedthrough to foreign currency prices in the aggregate, additionalestimation is required that would incorporate other major Jap-anese export industries. Further analysis would preferably be con-ducted on a disaggregated basis, since changes in commodity com-position will alter the adjustment of Japan's aggregate exports toshifts in exchange rates.


424 DANIEL CITRIN

APPENDIX

Data Sources and Methodology

The data employed in the empirical analysis were obtained from a number ofsources and were constructed as described below.

Export Shipments and Orders

Volumes of Japanese export shipments were obtained from the Japan TariffAssociation, Japan Exports and Imports (Tokyo: JTA, various issues). Theexport commodities covered by the study are: (1) subcompact passenger cars, orpassenger cars with engines of a piston displacement of not more than 2,000cubic centimeters; (2) color television broadcast receivers, including chassis andkits; (3) heavy steel plate, or sheets and plates of iron and steel not less than6 millimeters in thickness; (4) tin plate, or trimmed sheets, plates, hoops, andstrips of iron and steel; and (5) galvanized steel sheets, plates, hoops, and stripsof iron and steel.

Export order volumes of heavy steel plate and tin plate were obtained from theJapan Iron and Steel Federation (JISF).

Domestic Shipments and Orders

Volumes of domestic shipments and orders were obtained from Japan's Min-istry of International Trade and Industry (MITI) and JISF sources, respectively,with commodity coverage equivalent to that defined for export volumes above.

Foreign and Domestic Activity

Except in the case of heavy steel plate, foreign activity (YF) in the exportdemand equations was proxied by total world imports in real terms less (real)imports by Japan, obtained from the International Financial Statistics of theInternational Monetary Fund (various issues). In the case of heavy steel plate,foreign activity was defined in terms of the volume of foreign shipping underconstruction, since the major end-user of Japan's heavy plate exports duringthe sample period was the foreign ship construction industry. Quarterly data onthe gross tonnage of foreign shipping under construction were obtained from theUnited Nations, Monthly Bulletin of Statistics (New York, various issues);monthly estimates used in estimation were generated on the basis of thesequarterly data through the TROLL software package's spline function routine.

Domestic activity was represented by three different indices, depending on theproduct under consideration. In the case of subcompact cars and color televisionsets, both consumer items, domestic activity was represented by an index of realdisposable income (YD) constructed as follows:

where YDEF is the average disposable income of Japanese employee house-holds; JPOP is the Japanese population; ASIZE is the average family size ofJapanese employee households; and JCPI is Japan's overall consumer priceindex. The data used in constructing YD were obtained from the Japanese



Economic Planning Agency (EPA). For the three steel products, it was assumedthat domestic activity would be more appropriately represented by an indexbased on activity in the manufacturing and construction sectors. Regressionswere run on the basis of two alternative indicators: (1) an index of manufacturingoutput, and (2) a composite index of manufacturing and construction activity.The composite index was calculated as:

where QI is the index of manufacturing output; QC is the value of constructionworks executed, deflated by the construction cost index; Wl is the value of totalmanufacturing output, adjusted annually; and W2 is the value of constructionwork completed, adjusted annually. The index of manufacturing output wasobtained from MITI sources, and the construction data were taken from Japan,Ministry of Construction, Construction Statistics Monthly (Tokyo: JMC, variousissues).

Prices

Japanese prices were obtained on a monthly basis from the Bank of Japan,Price Indices Annual (Tokyo: BOJ, various issues). Contract export price indices(based on f.o.b. value) were available corresponding to the five products coveredby the study; domestic prices were represented by the wholesale price indices forthe various commodities. For each product considered, the relative domesticprice was defined as the ratio of its domestic price to the wholesale price indexfor all other commodities.

Raw material price indices were not available at the relevant product levelsand were therefore proxied by indices constructed at appropriate industry levelsof aggregation; that is, automobiles (subcompact passenger cars), consumerelectrical appliances (color television sets), hot-rolled steel products (heavy steelplate), and cold-rolled steel products (tin plate and galvanized steel sheet). Theproxy indices were calculated as weighted-average wholesale price indices of themajor raw material inputs of each of the above industries, with weights accordingto input coefficients given by the 1970 input-output table for Japan (Japan,Administrative Management Agency (1973)).

For subcompact cars and color television sets, foreign prices in yen (R • PF)were calculated as weighted averages of foreign wholesale prices converted intoyen, with the weights based on average foreign market shares of Japanese ex-ports for each category during 1971-75. Among the major markets, monthlyprice series for subcompact passenger cars and color television sets were avail-able for the United States, the United Kingdom, and Canada, which togetheraccounted for about three fourths of Japan's total exports for both productsduring the above period. The sources were U.S. Department of Labor, Bureauof Labor Statistics, Producer Prices and Price Indexes (Washington: GovernmentPrinting Office, various issues); United Kingdom, Central Statistical Office,Price Index Numbers for Current Cost Accounting (London: H. M. StationeryOffice, various issues); and Canada, Ministry of Industry, Trade, and Com-merce, Industry Selling Prices (Ottawa: Canadian Government PublishingCentre, various issues).

In the case of the three steel products, disaggregated prices in the relevantexport markets were readily available only for the United States. The competingforeign price was thus defined as follows. For galvanized steel sheet, the foreign


426 DANIEL CITRIN

price was represented solely by the U.S. wholesale price of galvanized steelsheet; the North American market accounted for about one half of total Jap-anese exports of galvanized steel sheet during 1971-75. For heavy steel plate andtin plate, foreign prices were calculated as weighted averages of the U.S. whole-sale price of the relevant product and overall wholesale price indices in othermajor export markets, of which the latter were obtained from the Fund's Inter-national Financial Statistics. This treatment thus captured broad price trends inexport markets outside the United States and also yen exchange rate changes inrelation to those markets. The countries covered accounted for 51 percent and54 percent, respectively, of Japanese exports of heavy steel and tin plate.

The proxies for expected relative export prices were generated by forecastingone to three months forward on the basis of ARIMA models estimated sepa-rately for the various foreign prices, exchange rates, and contract export pricesused in the analysis. (The estimated ARIMA models utilized are available fromthe author upon request.)

The relative price of gasoline in the United States and in Japan, includedrespectively in the export and domestic demand equations for subcompact pas-senger cars, was defined in both cases as the consumer price of gasoline relativeto that for all other items included in the consumer price index.

Wages

The monthly wage rate (W) was defined as equal to total regular wages(excluding bonuses) divided by total regular hours worked (excluding overtime).Because data were not available at the relevant product level of disaggregation,wage costs were assumed to be equal to those in the automobile, consumerelectrical appliance, and rolled steel product industries, respectively, for sub-compact cars, color television sets, and the three steel products. Data wereobtained from Japan, Ministry of Labor, Monthly Labor Statistics (Tokyo: JML,various issues).

Variable Costs

Because of multicollinearity problems encountered in preliminary regressions,a variable cost variable, defined as a geometric weighted average of raw materialand wage costs, was employed in the final regressions. The weights were deter-mined by average input-output weights for 1970-72 obtained from MITI, Censusof Manufactures (Tokyo, various issues); the weights and the respective levels ofdisaggregation available in the Census for each product considered are listed inTable 5.

Non-Price-Rationing Variable

The non-price-rationing variable was represented by either the inventory-salesor the inventory-orders ratio. Inventory data were obtained from MITI sources;total sales were defined as the sum of export and domestic shipments, and totalorders as the sum of export and domestic orders.

Vintage of Capital Stock

Because data on the age of the oldest machine in operation were not available,such age was approximated as a linear function of the average age of the capital



Table 5. Input Variable Weights Used in Final Regression Analyses

Weights

Product

Subcompact passenger cars

Color television sets


Heavy steel plate

Tin plate

Industry

Automobiles

Televisions and radios


Hot-rolled steel products

Galvanized steel

Wages

0.081

0.095

0.086

0.098

0.229

Rawmaterialprices

0.919

0.905

0.9144

0.902

0.771

Source: Japan, Ministry of International Trade and Industry, Census of Man-ufactures (Tokyo: MITI, issues for 1970-72).

stock in place and the rate of capacity utilization. (This approach was motivatedby de Menil (1974).) For subcompact cars and color television sets, the averageage of machinery in place was estimated from firm-level balance-sheet data onplant and equipment stocks and plant and equipment investment and depreci-ation obtained from the Japan Industrial Development Bank. Data on machin-ery vintage for the steel industry were provided by Sumitomo Metal Industries.Because the balance-sheet data were available only on an annual basis, monthlyestimates were generated by the TROLL software package's spline functionroutine. Capacity utilization rates were available for automobiles and colortelevision sets from MITI and were estimated for each of the three steel productsfrom output data according to the methodology employed in the construction ofthe Wharton index of capacity utilization (Klein and Summers (1966)).

REFERENCES

Ahluwalia, Isher J., and Ernesto Hernandez-Cata, "An Econometric Model ofU.S. Merchandise Imports Under Fixed and Fluctuating Exchange Rates,1959-73," Staff Papers, International Monetary Fund (Washington), Vol. 22(November 1975), pp. 791-824.

Artus, Jacques R., "The Short-Run Effects of Domestic Demand Pressure onExport Delivery Delays for Machinery," Journal of International Eco-nomics (Amsterdam), Vol. 3 (February 1973), pp. 21-36.

, "The Behavior of Export Prices for Manufactures," in The Effects ofExchange Rate Adjustments, ed. by P. Clark (Washington: Department ofthe Treasury, April 1974).

Branson, William H., "The Trade Effects of the 1971 Currency Realignments,"Brookings Papers on Economic Activity: 1 (1972), The Brookings Institu-tion (Washington), pp. 15-58.


428 DANIEL CITRIN

de Menil, George, "Aggregate Price Dynamics," The Review of Economics andStatistics (Cambridge, Massachusetts), Vol. 56 (May 1974), pp. 129-40.

Gregory, R.G., "United States Imports and Internal Pressure of Demand:1948-68," American Economic Review (Nashville, Tennessee), Vol. 61(March 1971), pp. 28-47.

Hatanaka, Michio, "Several Efficient Two-Step Estimators for the DynamicSimultaneous Equations Model with Autoregressive Disturbances," Jour-nal of Econometrics (Amsterdam), Vol. 4 (1976), pp. 189-204.

Hooper, Peter, "Forecasting U.S. Export and Import Prices and Volumes in aChanging World Economy," International Finance Discussion Paper No. 99(Washington: Board of Governors of the Federal Reserve System, Decem-ber 1976).

International Monetary Fund, International Financial Statistics Yearbook 1982,Vol. 35 (Washington, 1982).

Japan, Administrative Management Agency, 1970 Input-Output Table (Tokyo:JAMA, 1973).

, 7975 Input-Output Table (Tokyo: JAMA, 1978).Japan, Economic Planning Agency, 1978 Economic White Paper (Tokyo: EPA,

August 1978)., 7979 Economic White Paper (Tokyo: EPA, August 1979).

Japan, Ministry of Finance, Balance of Payments Statistics (Tokyo: JMF, 1978,1980).

Khan, Mohsin S., and Knud Z. Ross, "The Functional Form of the AggregateImport Demand Equation," Journal of International Economics (Am-sterdam), Vol. 7 (May 1977), pp. 149-60.

Klein, Lawrence, and Robert Summers, The Wharton Index of Capacity Utiliza-tion, Studies in Quantitative Economics No. 1 (Philadelphia: University ofPennsylvania, 1966).

Kmenta, Jan, Elements of Econometrics (New York: Macmillan, 1971).Komine, Takao, and others, "Exchange Rate Fluctuations and Macro-

economics," in Economics, Society, Policy (Tokyo: Economic PlanningAgency, June 1978).

Komiya, Ryutaro, and Y. Suzuki, "Inflation in Japan," in Worldwide Inflation:Theory and Recent Experience, ed. by Lawrence B. Krause and Walter S.Salant (Washington: The Brookings Institution, 1977).

Learner, Edward E., and Robert Mitchell Stern, Quantitative International Eco-nomics (Boston: Allyn and Bacon, 1970).

Magee, Stephen P., "Currency Contracts, Pass-Through, and Devaluation,"Brookings Papers on Economic Activity: 1 (1973), The Brookings Institu-tion (Washington), pp. 303-23.

Orcutt, Guy H., "Measurement of Price Elasticities in International Trade,"Review of Economics and Statistics (Cambridge, Massachusetts), Vol. 32(May 1950), pp. 117-32.



Shiller, Robert J., A Distributed Lag Estimator Derived from SmoothnessPriors," Econometrica (Evanston, Illinois), Vol. 41 (July 1973), pp. 775-88.

Shinkai, Yoichi, "Terms of Trade, Wages, and Exchange Rates in Japan," Dis-cussion Paper No. 112 (Osaka: Institute of Social and Economic Research,Osaka University, July 1982).

Wilson, John F., and Wendy E. Takacs, "Expectations and the Adjustmentof Trade Flows Under Floating Exchange Rates: Leads, Lags, and theJ-Curve," International Finance Discussion Paper No. 160 (Washington:Board of Governors of the Federal Reserve System, April 1980).


Debt-Equity Ratios of Firmsand Interest Rate Policy

Macroeconomic Effects of High Leveragein Developing Countries

V. SUNDARARAJAN*

HE MAIN PURPOSE of this paper is to demonstrate that both thechoice between debt and equity by firms and the institutional

circumstances governing this choice have a crucial bearing on theeffect of interest rate policy on saving and investment in devel-oping countries. The reliance on debt finance has been quitesubstantial in some developing economies because loans from thebanking system have constituted substitutes for stock issue, andthe flow of foreign saving has been mainly in the form of debtrather than equity. In effect the banking system and, in somecases, the curb markets have together assumed the risk of bank-ruptcy of firms, and the equity instruments have remained under-developed.

The economy of the Republic of Korea provides an interestingcase study of rapid economic growth with heavy reliance on debtfinance. The average debt-equity ratio (that is, the ratio of totalliabilities to net worth) of firms in the industrial sector in Koreahas grown from about 100 percent in the early 1960s to about500 percent in recent years. This sharp rise is due mainly to therapid growth of the Korean banking system that occurred after theinterest rate reform in 1965 and to the large use of foreign borrow-ing. Other factors that contributed to the rise include the inade-quacy of business saving in relation to investment needs and thebiases in the tax system that have favored debt finance. Policy-makers in Korea have in general held the view that the resultant

* Mr. Sundararajan, Advisor in the Central Banking Department, is a gradu-ate of the Indian Statistical Institute and Harvard University.

430

t


DEBT-EQUITY RATIOS OF FIRMS 431

overleveraged financial structure restricts their macroeconomicpolicy options and have, on various occasions, adopted measuresto reduce the debt-equity ratio of firms as part of financial reform.(Sakong II (1977) has given a historical account of measures takenby the Korean authorities to improve the corporate financialstructure.)

On the basis of an analysis of corporate financial structure inJapan—another example of rapid growth achieved predominantlythrough debt finance, with interesting parallels to the Koreansituation—Patrick (1972) concluded that underdeveloped capitalmarkets have not had any adverse effect on saving and realizedinvestment.

Whereas rapid economic growth in Japan and Korea may sug-gest such a conclusion, a closer analysis of the situation in manydeveloping countries reveals that the prevalence of high corporatedebt-equity ratios is detrimental to macroeconomic stability, andthat the effect of interest rate policy on saving and investment issignificantly altered by the size of the ratio. Of interest is thatwhen the debt-equity ratio exceeds a critical limit, even the direc-tion of the effect of financial policies is changed, and stabilizationpolicies involve very high costs in growth forgone. These macro-economic consequences of the financial structure of firms willbecome apparent when the role of interest rate policy is examinedfrom the viewpoint of its effects on the cost of capital to investors,an aspect that is ignored in much of the debate on interest ratepolicy in developing countries.

The analysis of interest rate policy in developing countries hasevolved along two distinct lines. The analytical framework pio-neered by Shaw (1973) and McKinnon (1973) considers dis-equilibrium systems in which investment opportunities abound,but actual investment is constrained by available saving, in partbecause high inflation and controls on the monetary system fosterfinancial repression. Because of controls on interest rates, short-run monetary equilibrium is achieved mainly through variations inthe rate of inflation. The role of interest rate policy in this frame-work is to increase saving, improve allocative efficiency, spur thedemand for financial assets, and facilitate stabilization.

An alternative line of analysis, developed in van Wijnbergen(1983) and Taylor (1983), focuses more closely on the specificcharacteristics of the financial markets in many developing coun-tries. It is argued that active curb markets, or deregulated seg-ments of the organized financial markets ("free markets" for


432 V. SUNDARARAJAN

brevity), exist in many countries and that private loans in thesefree markets often constitute an important share in the portfoliosof savers. Therefore, the interest rate in the free markets can beexpected to play a role in equilibrating demand and supply ofcredit. In this structuralist framework, both the administered in-terest rate and the curb market rate (or the free rate) influencesaving, investment, portfolio choice, working capital costs, andinflation. Whereas the Shaw-McKinnon analysis deals with onlytwo types of assets in savers' portfolios—monetary assets andinflation hedges—the structuralist model introduces a third asset,private loans in the free market. This extension of the asset menusignificantly alters the implications of interest rate policy.

In this paper the structuralist analysis of interest rate policy isextended by formulating an appropriate definition of the real costof capital to investors in developing countries that are character-ized by segmented financial markets, controls on the banking sys-tem, and substantial reliance on debt, including foreign-currencydebt. The earlier models ignored the important issue of how thereal cost of capital to investors is influenced by interest rate policyand the financial structure. Although this neglect is understand-able in the Shaw-McKinnon framework, in which the emphasis ison saving and not on investment, it is not valid in the structuralistmodel, in which both investment and saving respond to interestrates. Even in the Shaw-McKinnon framework, the appropriateformulation of the real cost of capital is relevant because it is animportant component of the rental-wage ratio that influencesfactor allocation and the efficiency of capital use. Such aspectsof efficiency are highlighted in many models that are based onthe Shaw-McKinnon tradition (for example, Sundararajan andThakur (1980) and Fry (1982)).

The relation between the cost of capital, the interest rate, andthe debt ratio is a subject with a long history and a voluminousliterature in the field of finance.1 (Throughout the paper, debtratio (a) refers to the ratio of total liabilities to total assets, andthe term debt-equity ratio (e) refers to the ratio of total liabilitiesto net worth; the two terms will be used interchangeably in viewof the one-to-one correspondence between the two ratios, givenby e = a/1 - a.) The discussion below will focus on those aspectswhich appear relevant to developing countries, with a view to

ôr a survey of this literature, see Nickell (1978) and Beranek (1981); thebasic reference on the subject is Modigliani and Miller (1963).



providing a heuristic explanation of why debt ratios matter inunderstanding the effects of interest rate policy.

In its simplest formulation, the cost of capital, defined as theminimum required return on investment, can be expressed as aweighted average of the cost of equity and the cost of debt, withthe weights representing the marginal shares in total assets ofequity and of debt. Thus, the larger is the debt ratio, the greateris the effect of changes in the cost of debt on the overall cost ofcapital. If foreign currency debt is ignored, the cost of debt inmost developing countries is simply the administratively con-trolled loan rate in the banking system. The cost of equity, how-ever, cannot be readily identified in developing countries withunderdeveloped and fragmented financial markets; it is the op-portunity cost of equity funds or, equivalently, the rate of discountused by businessmen in capitalizing the net income stream fromprojects. By its nature, the cost of equity is likely to vary with thestructure of the financial system and with the extent of financialrepression. For example, in a heavily repressed financial system,the major perceived alternative to using funds for fixed in-vestment could be the acquisition of inflation hedges such as goldor inventories. If so, the expected rate of change in the price ofgold or the general rate of inflation would be the relevant oppor-tunity cost of equity. In general, the average rate of return on arepresentative portfolio of savings instruments—the curb marketloans, inflation hedges, foreign-currency assets, and bankdeposits—is likely to be the appropriate opportunity cost of equityfunds. In general, however, the cost of equity is higher than thecost of debt, due in part to a risk premium. The gap between thetwo is particularly large in developing countries because of therepression of interest rates through administrative controls.

Against this background, it is clear that the ultimate impact onthe cost of capital of an increase in the administered interest ratedepends on how this increase affects the cost of equity and theshare of debt, both of which also influence the cost of capital.Indeed, a change in interest rate can either reduce or increase thecost of capital and saving, depending on the initial size of the debtratio and on the induced adjustments in the cost of equity and inthe share of debt.

In other words, the financial structure of firms—or more broad-ly, the institutional framework of the financial system that under-lies such structure—has significant implications for interest ratepolicy. This point can be illustrated by considering a dual financial


434 V. SUNDARARAJAN

structure, consisting of a controlled banking system and an un-fettered curb market, where the rate in the curb market could beregarded as the relevant opportunity cost of equity. An upwardadjustment in the administered interest rate would initially raisethe cost of capital and lower investment demand. The reductionin investment demand would be larger, the greater the debt ratio,because as indicated the increase in the cost of capital from a risein the interest rate grows with the debt ratio. With a high enoughdebt ratio, the reduction in investment would be sharp enough todepress the demand for funds in the curb market and therebylower the curb market rate. (The effects on the supply of curbmarket funds, or of equity funds in general, arising from portfolioadjustments is ignored here for illustrative purposes; such effectsare taken into account in the next section, where the completemodel is presented.)

If, now, saving depends positively on real returns to availableassets, then the negative effect on saving caused by the fall in thecurb market rate would counter the positive effect on savingcaused by the increase in the bank interest rate. The overall im-pact on saving would be negative, or would be weakened substan-tially, if the fall in the curb market rate is large because of a highdebt ratio. Thus the debt ratio used by firms can significantlyinfluence the effectiveness of interest rate policies. This funda-mental result remains valid when the analysis incorporates bothportfolio adjustments and adjustments in the debt ratio in re-sponse to interest rates and inflation.

The paper is organized as follows. Section I presents the modeldetermining saving, investment, the debt ratio, the cost of capital,and portfolio adjustments. The model emphasizes the linkagebetween debt and investment. Such linkage is in general ignoredin the theory of investment where the debt ratio is assumed to befixed; in the theory of corporate financial behavior, the rate ofinvestment is taken as exogenous. (For a recent analysis of theinterdependence between investment and financing, see Kite(1977).) In Section II, the Fisher effect and the effects of interestrate policy are analyzed under alternative assumptions about thedeterminants of the debt ratios of firms. The first subsection ofSection II ("Flexible Amortization, Exogenously Given TargetDebt Ratio, and No External Debt") demonstrates that a largedebt ratio can lead to macroeconomic instability and can generateperverse effects from monetary policies.

The second subsection of Section II ("Financial Repression and



Supply-Determined Debt Ratio") considers a financially re-pressed environment in which considerable scope exists for raisingthe share of financial saving in total saving and, in this context,analyzes the links between interest rates, financial saving, and thecost of capital, thereby elucidating the relation between theanalysis in this paper and the analytical framework of McKinnon(1973).

To the extent that an increase in the debt ratio raises the risk-iness of net returns from investment, firms might adjust their debtratio optimally to balance the benefits of additional subsidizedcredit from banks with the associated costs from the increasedriskiness of investment. The implications of such optimal debtbehavior for stability and interest rate policy are analyzed in thethird subsection of Section II ("Optimal Choice of Debt Ratio").

The fourth subsection of Section II ("Predetermined Amortiza-tion Schedule") deals with an aspect of debt policy that in generalhas been ignored in the literature: the effect of the maturity struc-ture of debt—the rate of amortization—on investment incentives.When the gap between the cost of equity and the cost of debt islarge, as in most developing countries, it can readily be shown thatthe choice of the maturity pattern of debt will significantly influ-ence the present value of the project and, hence, investmentincentives (on the effect of such maturity decisions, see Morris(1976)). Moreover, the rate of amortization has an importantbearing on how the debt ratio evolves over time. Therefore, be-havior regarding amortization can significantly influence the ef-fect of interest rates on investment and saving.

The effect of foreign-currency debt on the cost of capital isanalyzed in the fifth and final subsection of Section II ("Foreign-Currency Debt with Predetermined Target Debt Ratio and FixedAmortization Rate") in view of the importance of such debt infinancing investment in many developing countries. Section IIIcontains conclusions from the analysis and highlights its policyimplications. The algebraic details of the analysis are given in thetwo appendices.

I. A Model of Saving, Investment, and Debt

The model specifies the determinants of saving, investment, thecost of capital, the financial structure of firms, and the assetportfolios of savers in a dual financial system characterized by a


436 V. SUNDARARAJAN

controlled banking sector and a free financial market, such as thecurb market. (The case of a curb market is used in this paper forillustrative purposes only: the model can readily be adapted forcases in which other substitutes for bank credit exist, such as creditfrom nonbank financial intermediaries or equity instrumentswhose yields are market determined and are not controlled by thegovernment; the market-determined free rates can be substitutedfor the curb market rate, and the analysis can be appropriatelymodified.) The purpose of the model is to highlight the linkagebetween the debt behavior of firms and incentives for saving andinvestment. Special attention is paid to the determinants of thecost of capital to investors because the links between the financialstructure of firms and investment incentives arise in part from theeffects of financial policies on the cost of capital.

Saving and the Debt Ratio

It is assumed that aggregate real saving S depends positively onthe real returns obtainable in the controlled banking system andon returns in the free market:

S = S(p-ir,fl-ir), (1)

where p is the average nominal rate of return in the free market,or the opportunity cost of equity funds; R is the interest rate onbank deposits; and TT is the fully anticipated rate of inflation.Other variables that influence saving, such as real wealth andtransitory income, are assumed to be fixed and hence are sup-pressed for simplicity.

The interest sensitivity of aggregate saving is influenced by thedistribution of saving among government, corporations, andhouseholds. For the purposes of this paper, real government sav-ing is assumed to remain unchanged during the time span relevantfor the analysis. Although government saving will change becauseof the differential response of receipts and expenditures tochanges in inflation, these considerations will be left out for sim-plicity, and the focus will be on private saving, which is more likelyto be sensitive to interest rates.

Private savers fall into two distinct categories: those who areunaware of the full spectrum of financial alternatives and relymainly on banks for the placement of financial savings, and thosewho exhibit sophisticated portfolio behavior by diversifying their



savings among available assets. The saving function specifiedabove is consistent with these two behaviors if the rate of returnp is interpreted not simply as the return in the curb market, butas the average return on the optimal portfolios of sophisticatedsavers.

In developing countries, such optimal portfolios will typicallyconsist of (1) deposits in the financial system that yield risk-freereturns (determined by government policy); (2) loans supplied tothe unorganized money market, or to deregulated segments of theorganized markets, that offer risky returns; (3) equity holdingsthat also offer risky returns; and (4) holdings of physical assetsthat embody saving in the form of producer durables (saving in theform of consumer durables is treated as consumption). In a worldof diversification and risk aversion, an optimal portfolio and thereturn on it can be derived from a mean-variance framework (fora simple exposition of such a framework, see Rubinstein (1976)).

In this framework the share of various assets in the portfolios ofprivate savers, and the mean return on those portfolios, will de-pend, among other things, on the variances and covariances of thereturns to various assets and on the debt ratio of firms. An in-crease in the debt ratio, to the extent that it raises the riskiness ofthe equity streams in the portfolio, may require a higher return onthe optimal portfolio to compensate for the additional risk. Giventhis typical assumption found in the literature, the debt ratio isseen to influence saving through its effect on the return on savers'asset portfolios.

The interest sensitivity of saving is also influenced by the debtratio: if the level of corporate debt is relatively large, an increasein interest rates will transfer significant amounts of resources fromcorporations to households (usually after a time lag), and thistransfer may eventually depress aggregate private saving becausecorporate (and government) saving falls by the full amount ofadditional interest costs, whereas household saving rises by lessthan the increase in interest incomes. Therefore, the interest sen-sitivity of saving is likely to be inversely related to the debt ratio.

Investment and the Real Cost of Capital

Desired real investment depends on the real cost of capital, thereal wage rate, output expectations, and the size and character-istics of the existing stock of capital. All factors other than the real


438 V. SUNDARARAJAN

cost of capital are assumed to be fixed, so that the analysis mayfocus on the short-run interactions. The real cost of capital, de-fined as the minimum acceptable return on investment, can bederived by assuming that firms choose the level of investment tominimize their total cost of producing the desired output, includ-ing the acquisition cost of capital, and their debt service costs.

Let Q* denote planned output, K real capital stock, and Lemployed labor; given the production function, and the nominalwage rate W, total labor cost C can be written as

where the subscript t denotes time. The present value of total costsTC is given by

where p is the opportunity cost of equity; / is real gross in-vestment; P is the price level; am is the proportion of investmentfinanced by debt (that is, the marginal debt ratio); G is totaldomestic-currency debt outstanding; Fis total external (foreign-currency) debt outstanding; E is the exchange rate, measured innumber of domestic-currency units per unit of foreign currency;Rd is the domestic interest rate; Rf is the foreign interest rate; andad, af are the amortization rates on domestic and foreign loans,respectively. At each point in time, total cost — cash outflows fromthe point of view of the owners of the firm — consists of labor costsC (<2*, K), funds supplied by the owners to acquire and installnew plant and equipment (1 - am) IP, and the debt service pay-ments on domestic and external debt (Rd + ad) G and (Rf + af) FE.

The present value of these cash outflows is obtained by applyingthe rate of discount p, which is the opportunity cost of equityfunds given by, say, the curb market rate, or the rate of return onthe optimal portfolio of sophisticated savers. (In line with thisassumption, loans in the curb market are treated as equity financeand are excluded from the computation of am.) In some econo-mies, the interest rate in the deregulated segment of the organizedfinancial sector may serve as the opportunity cost of equity. In therest of the discussion, discount rate will refer to the cost of equity,



which can clearly take on a variety of forms depending on thestructure of the financial system.

The task is to minimize the total cost—the present value of allcash outflows given by equation (3)—with respect to the controlvariable /, subject to the constraints:

where 8 is the rate of economic depreciation, amd is the proportionof investment financed by domestic-currency loans, and am/is theproportion of investment financed by foreign-currency loans. Themarginal debt ratio am is the sum of amd and am/. The dot abovea variable denotes a time derivative.

Equation (4) states that the change in capital stock, K, equalsgross investment / minus depreciation. The rate of economic de-preciation, stated as proportion 8 of existing capital, is assumed toremain unchanged over time.

Equations (5) and (6) describe the time path of loans out-standing, both domestic and external. They state that the changein debt outstanding—the net inflow of loans—equals total newloans minus the amortization of existing loans. Equation (6) refersto foreign loans measured in foreign-currency units. Thus, the in-vestor's external debt obligations are all denominated in foreign-currency units, and the investor bears the full exchange risk. Thisis the typical situation in developing countries. The amortizationpayments on both domestic and external loans are assumed to beproportional to the stock of loans outstanding, whereas new loansare obtained only for financing fixed investment. Loans to financeworking capital requirements can be readily incorporated, butsuch loans are ignored for simplicity (as is borrowing for thepurposes of dividend distribution and the maintenance of cashreserves).

The problem of minimizing equation (3) subject to the con-straints (4), (5), and (6) is a well-defined control problem that canbe solved to characterize the path of optimal capital accumu-lation. The first-order conditions (see Appendix I) for the cost-minimizing investment path simply reduce to the familiar rule thatstates that, at each point in time, investment should be expandeduntil the present value of cost reductions from a change in in-


440 V. SUNDARARAJAN

vestment minus the present value of debt service incurred in fi-nancing the investment equals the amount of equity finance sup-plied by the owners (both new and old). To highlight the expres-sion for the cost of capital, this rule can be restated as follows:

where -dC/dK is the reduction in nominal labor costs from unitaddition to capital stock, exp (IT* ) is the price level at time t, IT isthe rate of inflation, and rb is the real cost of capital given by

where x is the expected rate of change in the nominal exchangerate. (For an alternative derivation of equation (8) that is based onproposition I of Modigliani and Miller (1963), see Appendix I.)The expression in large brackets on the right side of equation (8)is the present value of debt service payments on amd of domesticloans and on amf of foreign loans.

Differentiating equation (2) with respect to capital stock, andusing equations (7) and (8), allows investment to be expressed as

Assuming that real wage W/P is constant and that the desiredoutput is predetermined, and suppressing Kt in order to focus onthe short run, allows the investment function to be written as

An examination of the formula for the cost of capital (equation(8)) underscores the importance of explicit consideration of thedebt policies of firms in developing countries. In the special case —when domestic capital markets are perfect, default risk is absent,capital is fully mobile internationally, and the exchange rate isexpected to remain unchanged — all interest rates are equalized(Rd = p = Rf ; domestic and foreign rates of inflation are assumedto be identical for the time being, so that the expected change inthe exchange rate is zero), and the cost of capital is given by

Thus, only under these special assumptions, the cost of capital forinvestment purposes is independent of amortization rates as well



as debt ratios. Clearly these assumptions do not adequately char-acterize developing economies. If for any reason the foreign inter-est rate differs from the domestic rate, then the share of externaldebt in investment finance will enter the calculations of the costof capital. If, in addition, the interest rate on domestic debt devi-ates significantly from the discount rate, then debt policy assumeseven greater significance. Therefore, the next section will exam-ine the determinants of the debt ratio.

Determinants of the Debt Ratio

The marginal debt ratio that entered the calculations of the costof capital is determined in part by the institutional environment—including in that term the stance of credit policy—and in part bythe long-run average debt ratio aa that firms regard as prudent.(The average debt ratio refers to the share of debt in total assets,whereas the marginal debt ratio am refers to the share of debt infinancing additions to total assets.) If, for prudential reasons,financial institutions adhere to some predetermined debt-equitynorms (see Madan (1978) for a discussion of the use of such normsin India), or if the rigidities and imperfections in the financialsystem lock firms into some historically determined debt-equityratios, then it is best to regard the marginal debt ratio as aninstitutionally determined parameter.

In contrast, in the early stages of financial evolution when thescope for raising the share of financial saving in total saving islarge, the debt-equity mix is likely to be determined mainly by theavailability of financial savings, which, in turn, could be influ-enced by interest rate policy.2 Quite often, however, firms indeveloping countries do have some flexibility in controlling thedebt-equity mix. Firms can vary the debt-equity mix by adjustingthe policy toward retention of earnings for reinvestment purposes,by varying the extent of use of foreign-currency debt, and byaccessing domestic finance from informal and equity markets.

If firms are able to adjust their financing mix, then it is reason-able to assume, on the basis of available empirical evidence, that

2 The assumption that the debt ratio would rise with larger financial savingswould be reasonable mainly when debt ratios are initially small because of thelow level of financial saving and high level of self-finance. This is the case offinancial repression. The focus of this paper, however, is on situations in whichdebt ratios are relatively high.


442 V. SUNDARARAJAN

debt decisions by firms will be guided by the long-run averagedebt ratio that they strive to achieve. (For empirical evidence thatfirms try to achieve a target average debt ratio, see Ang (1976)and Marsh (1982).) This behavior is likely because the value of thefirm, or equivalently the discount rate required by the owners ofthe firm, is likely to depend on the average debt ratio. It is theaverage debt ratio — not the marginal debt ratio — that is relevantfor assessing the riskiness of equity streams of firms that arisesfrom constraints on future investment options and from the like-lihood of bankruptcy. Therefore, the target value for the averagedebt ratio is likely to be an important determinant of the marginaldebt ratio. This consideration can be formalized by solving thedifferential equation (5) to obtain

where aad is the long-run target value of the average domestic debtratio, amd is the marginal domestic debt ratio used by the firms,and g is the expected rate of growth of real capital stock.3

Equation (10) states that the marginal share of debt in financinginvestment will depend not only on the target value chosen for theaverage debt ratio, but also on the expected rate of growth ofcapital stock, its rate of depreciation, the rate of amortization ofdebt, and the rate of inflation. For example, if inflation risesbecause of expansionary credit policies, firms will choose to raisethe marginal debt ratio and will be able to do so. Although thisdecision would raise the average debt ratio temporarily, thehigher inflation would eventually reduce the average debt ratio toits target level by raising the value of assets in relation to debtoutstanding. Another implication of equation (10) is that when-ever firms are able to adjust the rate of amortization — say,through frequent funding operations — to match the maturitystructure of assets and the projected inflation (so that ad = 8 - TT),then the average and marginal debt ratios will be identical. (Wheninflation is high, the equation ad = 8 - IT implies a negative rate ofamortization, and equation (10) implies a large marginal debtratio that could even exceed unity; these situations occur when

3 See Appendix II for the derivation of equation (10). The particular functionalform has been used for analytical convenience only, despite its limitation that themarginal debt ratio can exceed unity when inflation is large. A more satisfactoryspecification, linking the marginal ratio to the target average ratio, the rate ofinflation, and other variables, would complicate the analysis without materiallyaffecting the main results.



firms build up debt far in excess of investment needs as a conse-quence of the expected reduction in the real value of debt throughinflation.) If the rate of amortization is fixed a priori, however, thefixed marginal debt ratio will fall short of the target average debtratio when inflation is low (that is, less than 8 - ad) and will ex-ceed the average debt ratio when inflation is high (that is, exceed-ing 8 - ad). This point can be illustrated by noting that, for anygiven expected inflation, a faster rate of amortization (that is, ashortening of loan maturities) will, by reducing the outstandingdebt more rapidly than desired, induce firms to raise the marginaluse of debt.

A relation similar to equation (10) can be derived for foreigncurrency debt:

where am/is the marginal share of foreign-currency debt; afl/is thelong-run average share of foreign-currency debt; x is the expectedrate of increase in the nominal exchange rate; TTW is the foreignrate of inflation; and 9 is the expected rate of change in the realexchange rate, given by 0 = x - TT + TTW . It will be assumed thatthe rate of foreign inflation, as well as the rate of amortization offoreign loans, is fixed. Equation (11) implies that, given the long-run target value aaf of the ratio of foreign-currency debt to totalassets, a reduction in domestic inflation (with no change in therate of currency depreciation) will induce firms to lower the mar-ginal share of foreign-currency debt. But if the fall in domesticinflation is expected to be offset by changes in the exchange rate,so that the expected path of the real exchange rate is unchanged,then the marginal foreign-currency debt ratio will remain stable.Thus the share of foreign-currency debt in financing investmentwill depend on exchange rate policy, a consideration that hasimportant implications for the effect of alternative exchange rateregimes on the average cost of capital, and hence on investmentincentives.

So far the analysis of the marginal debt ratio has been based onthe assumption that the target value of the average debt ratio isgiven a priori. The next step is to specify how this target is chosenby firms. To simplify the analysis and to sidestep the difficult issueof determining the proper mix of domestic- and foreign-currencydebt, it will be assumed that foreign-currency debt constitutes a


444 V. SUNDARARAJAN

fixed share (1 - (3) of total debt. Therefore, given the averagedebt ratio afl, the average foreign-currency debt ratio is given by

and the average domestic debt ratio is given by The determinants of the target value of the average debt ratio

can now be identified by analyzing the benefits and costs of raisingthe debt ratio. The benefit is simply the additional interest subsidythat can be garnered by increasing the share of the cheaper sourceof finance — loans from the financial system available at the con-trolled interest rate. The cost of incurring larger debt (in relationto assets) derives from the increased probability of future cash-flow problems, hence of bankruptcy. This cost can be summarizedby the equity-cost function that links the discount rate and theaverage debt ratio:4

(A prime denotes the first derivative with respect to the subscriptvariable; a double prime denotes the second derivative with re-spect to the subscript variable.)

Thus the discount rate is assumed to be a strictly concave andincreasing function of the average debt ratio. The optimal value ofthis ratio can be chosen so as to minimize the overall cost of capitalby balancing at the margin the benefit of additional interest sub-sidy with the cost of increased riskiness of investment, and theoptimal debt ratio so derived will be treated as the target valuethat firms strive to achieve.5

Because the choice variable in this optimization exercise is theaverage debt ratio, it is necessary to express the formula for the

4For example, see Feldstein, Green, and Sheshinski (1978); Feldstein andGreen (1979); and Ericksson (1980). Myers (1977) and Kim (1978) containinteresting discussions of the reasons that the riskiness of returns from equity,and hence the required rate of discount, rises with increased use of debt. For abrief summary of this literature on the supply side of debt, see Modigliani (1982).

5 Ideally, the debt ratio should be treated as a control variable along with therate of investment, and the full optimal control problem of minimizing thepresent value of costs should be solved by using the appropriate constraints oncontrol variables. The problem has been simplified by assuming that the mar-ginal and the target average debt ratios are linearly related. For an analysis ofdebt policy under the optimal control framework, see Ekman (1982), which alsocontains a detailed bibliography on this area of research. In most of thesestudies, the rate of interest is assumed to vary with debt, whereas the cost ofequity is fixed. In the problem considerd here, the cost of equity varies with debt,whereas the interest rate is fixed by policy. This considerably complicates anoptimal control approach to the problem.



cost of capital, equation (8), which involves the marginal debtratios a.md and am/, in terms of the average debt ratios by usingequations (10) and (11). A complete expression for the cost ofcapital is given by

where

The optimal debt ratio can be obtained by substituting equation(12) into equation (13) and equating the first derivative of rb (withrespect to afl) to zero:

The interpretation of condition (14) is facilitated if it is assumedfor simplicity that the marginal and average debt ratios are identi-cal,6 so that the expression for the cost of capital simplifies to theweighted- aver age formula:

where a denotes the common value of the marginal and averagedebt ratios (a = afl = am).

In this special case, the first-order condition reduces to

6 Marginal and average debt ratios will be equal if ad = 6 - TT and af = 5- IT + x , so that CI(TT) = ca(irw , 6) = 1. These conditions require that firms oper-ate in a well-developed domestic financial system, with easy access to interna-tional capital markets, so that the maturity of loans can be readily adjusted in linewith inflation and the rate of depreciation of assets. Therefore, the equality ofmarginal and average debt ratios will be an unrealistic assumption for mostdeveloping countries.


446 V. SUNDARARAJAN

The equation above serves to define implicitly the optimal targetvalue of the average debt ratio. The expression on the right-handside of equation (14a) is the implicit interest subsidy, whereas theexpression on the left-hand side can be interpreted as the marginalrisk premium demanded by the firm's owners. Optimality thusrequires balancing at the margin the benefits of additional subsidywith the costs of increased risk measured by the slope of the equitycost function. This balance will clearly be disturbed wheneverdomestic interest rate policy or the foreign interest rate changes.Equally important, a change in inflation would affect the discountrate, the marginal debt ratios, and probably also the marginal riskpremium (through shifts in p«) and thereby would induce changesin the average debt ratio.

Monetary Equilibrium

The discount rate, the rate of interest, and the rate of inflationare assumed to be consistent with equilibrium in the money mar-kets. This requirement serves to capture the portfolio choices ofprivate asset holders and can be incorporated into the model byspecifying that the demand for real balances should equal supply:

where M/P is the supply of real balances and the right-hand siderepresents the demand for real balances expressed as a function ofreal output ( y ) and returns to different types of assets in theportfolio. The money demand function specified above explicitlyrecognizes that individuals hold in their portfolios not only mon-etary assets and inflation hedges, but also loans in the curb marketand claims to equity bearing similar risk.

Equilibrium in the Goods Market

Excess supply or demand in the goods market will clearly influ-ence the rate of inflation and interest rates. Equilibrium in thegoods market requires that

where / is domestic investment, S is domestic saving, and FS isforeign saving, assumed to be determined outside the model. Theequilibrium condition (15) states that investment should equal



available saving, and that such equilibrium will come aboutthrough variations in the discount rate, the rate of inflation, andthe debt ratio.

Complete Model

The model set out in Table 1 can be viewed as an adaptation ofthe standard IS-LM model, with emphasis on the determinants ofthe cost of capital and the debt ratio in developing economies with

Average debt ratioFor the case of fi-

nancial repression

For the case ofoptimal choiceof debt

Monetary equilibrium

Goods-marketequilibrium

Note: Endogenous variables are as follows: 5 is domestic saving; /is domesticinvestment; rb is the weighted-average real cost of capital; amd is the marginaldomestic-currency debt ratio; am/is the marginal foreign-currency debt ratio; afl

is the average debt ratio; IT is the rate of inflation; P is the general price level;p is the curb market rate. Exogenous variables are as follows: Rd is the domesticinterest rate; Rfis the interest rate on external debt; a/is the rate of amortizationof external debt; ad is the rate of amortization of domestic-currency debt; 8 is therate of depreciation of capital stock; p is the share of foreign-currency debt intotal debt; M is the nominal quantity of money; x is the rate of change of theexchange rate, defined as the number of domestic-currency units per unit offoreign currency; FS is foreign saving.

Table 1. A Model of Saving, Investment, and Debt

Item Equation

Saving

Investment

Cost of capital

Marginal debt ratio


448 V. SUNDARARAJAN

segmented financial markets. The model determines saving, in-vestment, the discount rate, the overall cost of capital, the debtratio, and the rate of inflation. Before a detailed analysis of themodel is undertaken, it is useful to illustrate graphically the work-ings of the model for the simple case where the debt ratio as wellas the administered interest rate are assumed to be fixed.

In Figure 1, the IS curve denotes the combinations of the dis-count rate and the rate of inflation that are consistent with goods-market equilibrium. It is upward sloping because an increase ininflation with a fixed administered rate reduces the real adminis-tered rate and stimulates investment. To elicit a matching increasein saving, the discount rate rises. The usual upward-sloping LMcurve represents money market equilibrium. The equilibriumvalues of the discount rate, and the rate of inflation, are given by(P*, IT*).

It will be shown in the next section that the magnitude as wellas the direction of the slope of the IS curve is sensitive to the sizeof the debt ratio, the response of the administered interest rate tovariations in inflation, and the conditions affecting debt service(for example, the rate of amortization). Because, as is wellknown, the stability of the system, as well as the impact of policychanges, depends on the relative slopes of the IS and LM sched-

Figure 1. Discount Rate and Rate of Inflation



ules, it follows that differences in the debt-equity ratio, the inter-est rate policy, and other conditions governing debt will shift theimpact of demand management policies. The next section illus-trates that these shifts can indeed be substantial.

II. Analysis of the Model

The model is complex, and a general analysis of its comparativestatic properties is therefore not attemped in this paper. Usefulinsights can be gained, however, by analyzing several simple spe-cial cases. For example, it is convenient to assume that the rate ofinflation is determined exogenously and is not influenced bychanges in interest rates. This assumption will be valid if thedemand for real balances is regarded as a function of the rate ofinflation alone (for any given real output), a reasonable specifica-tion in many developing economies. Unless otherwise mentioned,this assumption will be maintained to simplify the analysis andthereby to highlight the critical role of the debt-equity ratio.Modifications that result from using a more general money de-mand function—thereby allowing for variations in inflation fromchanges in the interest rate—are indicated in several places inthe text. The effects of interest rates on working capital costs,and on short-run capacity utilization, are ignored throughout forsimplicity.

The effect of financial policies on the cost of capital and onreturns to savers will be analyzed under alternative assumptionsabout debt and amortization. There are several cases of interestthat depend on whether the rate of amortization is fixed or vari-able (in response to inflation), whether the average debt ratio isfixed or is determined by loan supply or by loan demand, andwhether foreign-currency debt is significant. The choice of partic-ular combinations of assumptions about debt influences the con-struction of the cost of capital and alters the final results. Tounderscore the effects of alternative assumptions about debt be-havior, a sequence of simple models will be considered in turn.

Flexible Amortization, Exogenously GivenTarget Debt Ratio, and No External Debt

The importance of the debt ratio is best illustrated by consid-ering the simplest possible model, in which it is assumed thatforeign-currency debt is absent; the rate of amortization is variedin line with the rate of depreciation of capital assets and inflation


450 V. SUNDARARAJAN

(fld = 8-7r), so that the marginal and average debt ratios areidentical; and that the firm adheres to a target debt ratio a anddoes not optimally adjust the ratio as the environment changes.Foreign saving is ignored because there is no external debt in thismodel. Under these assumptions the model becomes

In all subsequent discussions, it will be assumed that S'R , Sp > 0unless otherwise specified.

The role of the debt ratio is brought out sharply in the specialcase in which saving is interest inelastic (S'p = SR - 0) and thecontrolled interest rate is nof adjusted in line with inflation. In thiscase,

= l + o/(l-a). (19a)

The interesting aspect of the formula is the implied magnitude ofthe Fisher effect when the debt-equity ratio (a/1 - a) is large, asin many developing countries. For example, a debt-equity ratio of3:1 implies dp/dir = 4; that is, the discount rate will increase byfour times the increase in inflation, assuming passive interest ratepolicy. This result is a special case of the more general observationthat in thin markets price fluctuations will be large; a large debt-equity ratio implies that the free market where the rate of discountis determined is quite thin.

The size of the debt ratio also governs the effect of interest ratepolicy on saving under inflationary conditions. Total saving willincrease — equivalently, the cost of capital will fall — in response toan increase in inflation if and only if

Substituting equation (19) into the above expression and sim-plifying allows the condition for improved saving to be given by:

It can be verified that when dR/dfn>l the above inequality willhold if and only if the debt-equity ratio is less than the critical limit

The effect of inflation on the discount rate is obtained by differ-entiating the equilibrium condition with respect to TT and re-grouping terms, which yields



7], given by 7] = SR/S'P . The parameter TI—the ratio of the effect onsaving of a change in the controlled interest rate to the effect onsaving of a change in the free-market discount rate (the "interest-sensitivity ratio" for brevity)—is a critical determinant of the safeupper limit for the debt-equity ratio. If the debt-equity ratio ex-ceeds 7i, then the policy of raising the administered rate by morethan the increase in inflation (dR/d^>\) will depress saving.

Even when there is no change in inflation, the ultimate effect ofinterest rate policy on saving is governed by the critical limit 7]; theeffect of the interest rate on saving (dS/dK) is given by the expres-sion within braces in equation (19b), which will be positive if andonly if the debt-equity ratio is less than the interest-sensitivityratio.

The rationale of the above results is simple. The rates of savingand investment depend on both the controlled interest rate andthe discount rate. When the controlled rate is raised, the ultimateeffect on saving or investment depends naturally on whether theresulting upward shift in the saving schedule (linking saving andthe discount rate) exceeds or falls short of the downward shift inthe corresponding investment schedule. What is interesting isthat, for the response of saving and investment to interest ratepolicy to be positive, the interest-sensitivity ratio, which measuresthe relative effect of interest rates on saving, should bear anappropriate relationship to the debt-equity ratio, which influencesthe effect of interest rate on investment.

The conditions under which the debt-equity ratio may exceedthe interest-sensitivity ratio are of interest. A high debt-equityratio may induce a significant redistribution of incomes betweenhouseholds and businesses that, as indicated in the first subsectionof Section I ("Saving and the Debt Ratio"), is likely to dampenthe size of the saving response to a change in the controlledinterest rate and thereby reduce the interest-sensitivity ratio. Inother words, a high debt-equity ratio will by itself serve to reduceiq and raise the likelihood of a perverse saving response to interestrate policy. Even if the redistributive aspect is empirically insignif-icant, the debt-equity ratio could exceed the interest-sensitivityratio if investment is financed to a substantial degree throughgovernment net lending programs.

When financial markets are segmented, the interest-sensitivityratio could be negative. For example, this is the case when theeffect of the controlled interest rate on saving is opposite in signto the effect of the discount rate, because the balance betweenincome and substitution effects of interest rates is different for


452 V. SUNDARARAJAN

different groups of savers; hence the debt-equity ratio alwaysexceeds the interest-sensitivity ratio, thereby causing perversesaving response. This result highlights the potential shortcomingof segmented financial markets—and the advantage of unifyingthese markets through financial reform.

The prevalence of high debt ratios can lead to macroeconomicinstability—for example, progressively higher inflation and realdiscount rates—and the avoidance of such instability could involvehigh costs in the form of lower investment and saving. This asser-tion can be verified by extending the model to include portfolioequilibrium and by examining the relative slopes of the IS and LMschedules in Figure 1 as the debt ratio rises. Equation (19) impliesthat, when the administered interest rate is kept unchanged asinflation accelerates (or is raised by less than the increase ininflation), the slope of the IS curve is positive and becomessteeper as the debt ratio rises. In other words, with passive inter-est rate policy, the increase in the discount rate from a change ininflation (dp/drf becomes larger as the debt ratio increases.Therefore, when the debt ratio is sufficiently large, the slope ofthe IS curve could become steeper than that of the LM curve, andthis situation leads to an unstable economic system in which mon-etary action can trigger unpredictable effects, such as acceleratinginflation or deflation. With a steeper IS curve, an increase inmonetary expansion will yield a new unstable equilibrium withlower inflation.

Such instability in a high-debt economy can be avoided throughmore active interest rate policy. For example, an increase in thereal administered interest rate will make the slope of the IS curvenegative and eliminate the source of instability. As already noted,however, an increase in the real interest rate would depress savingand investment when the debt-equity ratio exceeds a critical limitor when the interest-sensitivity ratio is negative. In other words,achievement of stability in a high-debt economy or in an economywith segmented financial markets is likely to involve high cost ingrowth forgone.

Financial Repression and Supply-DeterminedDebt Ratio

The consequences of financial repression and large reliance onself-finance, the impact of the interest rate on financial saving,and the beneficial effect of increased financial saving on in-



vestment have been topics extensively discussed in the literature,notably in McKinnon (1973) and Galbis (1977).7 These dis-cussions, however, ignore the possible beneficial effect of in-creased financial saving on the cost of capital and investmentincentives in financially repressed economies. In the model devel-oped in this paper, the effect of larger financial saving on the costof capital can be captured by assuming that the debt ratio ispositively related to the administered interest rate R and nega-tively related to the discount rate p. That is,

In other words, an increase in the administered interest rate (ora fall in the discount rate) raises the debt ratio by increasing theavailability of investment credit through the banking system,thereby reducing the share of self-financed investments. The in-creased availability of investment credit need not lead to a higherdebt ratio for individual firms if banks finance additional projectson the basis of predetermined debt-equity norms and let firmsobtain the needed equity finance from the free markets (at themarket-determined discount rate). In situations of severe fi-nancial repression, however, the debt ratio is likely to rise inresponse to an increase in the bank interest rate.

Using this assumption and differentiating, as before, the saving-investment equilibrium condition, one can readily show that inter-est rate policy would improve saving (dS/dR > 0) if and only if

The above condition for improved saving is less stringent than inthe case of a fixed debt ratio, and it may not even be binding if theinterest sensitivity of the debt ratio (a^) is sufficiently large andpositive. The less stringent results obtain because an increase inthe debt ratio from the higher administered interest rate, and theaccompanying increase in the share of cheaper source of fundsavailable through the banking system, serve to lower the cost ofcapital and to raise investment incentives and saving. Moreover,the inequality above implies that the scope for raising saving andinvestment through interest rate policy becomes greater as theinitial distortion in the interest rate becomes greater (that is,

7 McKinnon emphasizes the complementarity between financial saving andinvestment that arises from the "conduit" effect, whereas Galbis emphasizes theimproved efficiency of investment allocation that arises from increased financialintermediation.


454 V. SUNDARARAJAN

R — p is large) and the increase in the debt ratio in response tohigher interest rates and financial savings also grows larger. Oncethe banking system reaches a certain size and sophistication, how-ever, the scope for raising the debt ratio through further im-provements in financial saving would be limited, debt-equity ra-tios would be governed more by decisions of firms and by the debtnorms used by banks, and the assumption of a supply-determineddebt ratio would no longer be appropriate. (Preliminary empiricalevidence for the Republic of Korea suggests that the assumptionof a supply-determined debt ratio is not a valid description of thedebt behavior of firms.) The next section discusses the conse-quences of firms choosing the debt ratio optimally.

Optimal Choice of Debt Ratio

To highlight the effect of optimal debt decisions, foreign-currency debt will continue to be ignored. In addition, as before,the rate of amortization will be assumed to be variable so that themarginal and target average debt ratios are identical. Under theseassumptions, the model determining saving, investment, and theoptimal debt ratio can be stated as follows:

The first equation is the equilibrium condition between savingand investment. The second equation is the first-order conditionfor the optimal choice of the debt ratio and implies that an in-crease in the controlled interest rate will reduce interest subsidyand thereby cause a reduction in the optimal debt ratio (in con-trast to the increase in the debt ratio under financial repression,discussed in the previous subsection). The third is the expressionfor the real cost of capital. Noting that the discount rate is nowimplicitly a function of a and IT, one can differentiate the two-equation system obtained by substituting rb into the first equationand derive the effects of inflation on the discount rate and the debtratio. The differentiation yields


and p'̂ is the derivative of pi with respect to inflation TT.The stability condition for the model has an interesting eco-

nomic interpretation that points to the possible adverse effects oflarge interest subsidy. Stability requires that the determinant ofthe matrix on the left-hand side of equation (20) be positive. Thatis,

The first term on the right-hand side above (in braces) is positivebecause the second-order condition for the optimal choice of debtratio ensures that (1 - a) p'â - pi > 0. But the second term isnegative, and its magnitude depends on the size of the implicitinterest subsidy, p - R. The larger is the subsidy, the greater is thelikelihood that instability would obtain. The instability wouldmanifest itself in the following way. A substantial distortion in thefinancial market, causing a large implicit subsidy p-R, wouldinduce firms to raise the debt ratio so as to capture the subsidy.This behavior in turn would raise the discount rate because ofhigher risk. The increase in the discount rate would raise theimplicit subsidy, inducing firms to borrow even more. Only strin-gent credit rationing would limit the achievable debt ratio, andexcess demand for credit would continue to persist. This possi-bility, which is a realistic description of many developing econo-mies, will be ignored, and the stability condition A > 0 will beassumed to hold.

The model and the stability conditions are illustrated in Figure2. The line FF shows the combinations of p and a that ensureequilibrium between saving and investment. An increase in thedebt ratio lowers the cost of capital, raises investment, and pushesup the discount rate. Therefore the slope is positive. The line ODshows the optimal combinations of p and a that are consistent withthe minimization of the cost of capital. An increase in p raises theimplicit interest subsidy obtainable on debt and induces a largerdebt ratio. Again the slope is positive. The intersection of the twocurves indicates the equilibrium level of the discount rate and thedebt ratio (p*, a*). The slope of the OD curve should be steeperthan the FF curve if stability is to be ensured.


where


456 V. SUNDARARAJAN

Figure 2. Debt Ratio and Discount Rate

Solving equation system (20) allows the effect of inflation to bestated as follows:

The magnitude of the derivatives depends on, among otherthings, the sign and size of p^, which is the shift in risk premiumattributable to a change in inflation. If higher inflation is associ-ated with increased uncertainty, and investors therefore perceivegreater risks, then p^ > 0. Under this assumption, equations (21)



and (21a) imply that dp/d-rr<l and that da/d^<0, wheneverdR/dtT> 1. Thus an increase in the rate of inflation leads to areduction in the optimal debt ratio—a counterintuitive result—because of increased marginal risk premiums required by in-vestors.8 The decline in the debt ratio in turn implies a reductionin the share of the cheaper source of finance, hence an increase inthe overall cost of capital. This effect on the cost of capital, how-ever, is partially offset by the reduction in the real discount rate.

The consequences of optimal choice of debt ratio are best illus-trated if it is assumed that the real administered rate is keptunchanged (that is, dR/dv = 1). Under this assumption, it is seenfrom equation (19) that when the debt ratio is fixed, and notadjusted optimally, the real discount rate remains unchanged, andsaving and investment hence remain unaffected. In contrast, whenthe debt ratio is adjusted optimally, equation (21) implies that thereal discount rate falls, thereby reducing saving and investment. Amatching reduction in investment obtains because of the fall indebt ratio that leads to an increase in the real cost of capital. Thisapparently counterintuitive result—that the real cost of capitalincreases in the presence of optimal debt behavior designed tominimize the cost of capital, but not so when the debt ratio isfixed—is due mainly to the effect of inflation on the risk premiumrequired by investors, an effect that was ignored in analyzing thecase of the fixed debt ratio. Indeed, when the risk premium is notaffected by inflation—so that p^ = 0—equation (21) implies thatas long as the real administered interest rate is kept unchanged theoptimal debt ratio does not change, and the real discount rate alsoremains unaffected—just as in the case of the fixed debt ratio.

If the controlled interest rate is raised by more than the increasein inflation, then the issue of the effect on saving under optimaldebt policy can be analyzed by examining the sign of

Substituting from equation (21), it can be verified that saving willimprove if and only if

8 Gordon (1982) has analyzed the effect of inflation on the debt ratio in theU.S. economy. Preliminary empirical tests suggest that pLX) is a validassumption for the Republic of Korea. The sign will probably depend on thelevel of inflation, and no a priori judgments are possible.


458 V. SUNDARARAJAN

From the second-order condition for optimal debt ratio it isknown that the second term on the right-hand side of the aboveinequality is positive. Thus, when debt is chosen optimally, thecritical upper limit on the debt-equity ratio is smaller than in thecase of the fixed debt ratio. This tighter upper limit on the debt-equity ratio continues to apply as the condition for an increase inthe controlled interest rate to improve saving, even when there isno change in inflation. In other words, the likelihood that anactive interest rate policy will have adverse effects on saving in-creases when firms choose the debt ratio optimally.

Predetermined Amortization Schedule

If the rate of amortization on loans cannot be adjusted wheninflation accelerates, then the real value of amortization paymentswill be reduced, and the average and marginal debt ratios willbegin to diverge. To understand the implications of these devel-opments, it is convenient to abstract from the existence of foreign-currency debt and to assume that all debt is denominated in do-mestic currency and offered at the rate R , which is lower than thediscount rate p. Firms strive to reach an average debt ratio in thelong run, however, and therefore adjust the marginal debt ratio inline with inflation and growth prospects. For simplicity it will beassumed that the average debt ratio a is a predetermined targetand is not chosen optimally. Under these assumptions, the cost ofcapital can be expressed as

where

The differentiation of the saving-investment equilibrium condi-tion with respect to inflation yields

where



An analysis of the expression above reveals that the size of theFisher effect depends not only on interest rate policy (dRIdv), butalso on whether the real discount rate initially exceeds or fallsshort of the expected real growth of capital stock. This result canbe summarized as follows:

The implication of this result is best illustrated when dR/dir = 1.In this case,

Thus, when the administered rate is maintained in real terms,the real discount rate would rise, and so would saving and in-vestment, if initially the real discount rate is greater than theexpected growth of capital stock. Otherwise the real discount ratewould fall, and with it saving and investment. These results implythat, with a fixed amortization rate and a real discount rate thatis large to begin with, inflation will make the real discount rateeven larger, and the slope of the IS curve will become steeper asthe discount rate rises, possibly intersecting the LM curve twice.The intersection corresponding to the higher discount rate will beunstable because of the steeper slope of the IS curve when thediscount rate is large (see equations (22) and (22a)).

The dependence of the effects of macroeconomic policies onthe size of the real discount rate is related to two conflicting forcesthat act on the real cost of capital when the rate of amortizationis fixed. If both the discount rate and the controlled interest rateare assumed to increase temporarily as inflation rises, by the sameamount as inflation, it follows that the present value of debtservice costs — R +a/p + a — also increases, thereby raising thecost of capital. But the increase in inflation induces a larger use ofdebt at the margin, and the marginal debt ratio — a(g + TT + a)/(g + 8) — increases, thereby lowering the cost of capital. The neteffect on the cost of capital depends on the initial size of p, whichaffects the present value of debt service in relation to the expected

if and only if

where


460 V. SUNDARARAJAN

magnitude of the growth in assets (g + TT), which in turn affectsthe marginal debt ratio.

The conditions under which saving improves in response tointerest rate policy depends, as before, on the size of the debt-equity ratio, but the critical limit on this ratio is influenced notonly by the initial size of the discount rate, and of the interest rate,but also by interest rate policy. In more formal terms, if it isassumed that dR/dn > 1, the condition for improvement in savingand investment is

Note that the limit on the debt-equity ratio has been derived byevaluating the derivative at a = 8 - IT. Without this simplifyingassumption, the expressions become more complex, but the con-clusions remain unaffected.

If p - TT >g, then B is positive and can be extremely large forsmall increments in the real controlled interest rate. In this casethe condition for improved saving will always hold because theleft-hand side of inequality (23) is negative. Moreover, for largechanges in the real controlled interest rate, the improvement insaving depends on a much less stringent condition on the debtratio than in the case of flexible amortization. If p - TT <g, thenB is negative, and the condition on the size of the debt ratio ismuch more stringent than in the case of flexible amortization. Theearlier condition p — TT > g, however, is most likely to hold undernormal circumstances: the larger is the real discount rate, themore efficient is the use of capital stock, implying that, for anygiven target of output growth, the required growth in real capitalstock is likely to be less. This association of a larger real discountrate p - TT with a smaller real asset expansion g would lead to thecondition p - TT >g. In any case, the results clearly demonstratethat, depending on the initial conditions in the financial marketsand of interest rate policy, the behavior of the rate of amortizationcan make an important difference to the impact of interest ratepolicy.

As in previous cases, the effect of interest rate policy, unac-

where

if and only if



companied by a change in inflation, is also governed by the size ofthe debt-equity ratio. An increase in the administered interestrate will improve saving if and only if

where, as before, TJ is the interest-sensitivity ratio. The upper limitT] now applies to the marginal debt-equity ratio. Because the rateof amortization is assumed to be fixed, the average and marginaldebt ratios can diverge. Therefore, although the average debtratio is not large (in relation to T]), it is conceivable that a highlevel of inflation and the associated credit policies induce a largemarginal use of debt. If so, an increase in the controlled interestrate could lead to adverse effects on the cost of capital and saving.The analysis suggests that such adverse effects can be mitigated byallowing for a more flexible amortization schedule and, at thesame time, restraining the share of debt in project finance. If theaverage debt ratio is already quite large, however, the adverseeffects cannot be avoided.

It is legitimate to ask why firms would try to achieve the targetfor the average debt ratio by raising the marginal ratio in line withan increase in inflation, when the target can be readily reached bysimply amortizing the existing loans at a slower rate (the im-portance of such funding operations for the validity of the formu-las for the weighted-average cost of capital is noted in Linke andKim (1974) and Beranek (1975)). From the firm's point of view,the latter option may be preferable—if it is available—but theoption chosen will depend on particular institutional circum-stances and historical practices. For example, many countries im-pose norms on debt-equity ratios for project finance, a practicethat would restrict the freedom at the margin. If firms have exten-sive overdraft facilities with banks, then the rate of amortizationcan be easily adjusted by varying the use of overdraft limits. Oftenit may be easier to raise a loan that is a larger than normal fractionof the project's current value than to obtain rollover credits for theprincipal amounts falling due. Depending on which option is thenorm—or on how the marginal debt ratio is determined—theimpact of interest rate policy will be changed.9 Such differences in

9 In this subsection the marginal debt ratio has been assumed to depend on thetarget average debt ratio, the rate of inflation, the rate of growth, the rate ofamortization, and the rate of depreciation. More interesting formulations arepossible. For example, the rate of amortization or the average maturity of loanscan be made a function of growth and interest rates. For a discussion of thedeterminants of the rate of amortization of corporate debt in the United States,see Morris (1976).


462 V. SUNDARARAJAN

the effects of policies can be significant, as has been demon-strated.

Foreign-Currency Debt with PredeterminedTarget Debt Ratio and Fixed Amortization Rate

To highlight the effect of foreign-currency debt on the cost ofcapital, considerations of optimal choice of debt are ignored. Theaverage domestic- and foreign-currency debt ratios are given astargets that remain fixed. Domestic and foreign interest rates areallowed to diverge by assuming that capital mobility is subject torestrictions. In view of the limited, and often uncertain, access tointernational capital markets in many developing countries, therate of amortization of foreign-currency loans is assumed to befixed. Amortization of domestic-currency loans, however, is as-sumed to be flexible, so that the marginal and average domesticdebt ratios are identical. Under these assumptions, the cost ofcapital is given by:

where

The expression above reveals that the cost of capital dependson, among other things, the expected rates of increase in thenominal and real exchange rates, which influence the presentvalue of the debt service on foreign-currency loans as well as theshare of such loans in investment finance. This relation suggeststhat the type of exchange rate regime influences the effect ofinterest rate policies.

First, suppose that the expected change in the nominal ex-change rate is fixed a priori. This would be the case when thenominal exchange rate is pegged to a currency basket (x = 0) or,for instance, the path of the depreciation of the nominal exchangerate is preannounced (x > 0) independently of other relevant vari-ables. Under such exchange rate regimes, the expected change inthe real exchange rate varies with inflation and thereby influencesthe marginal debt ratio (that is, C2 varies with 0). Differentiating



the saving-investment equilibrium condition allows the Fisher ef-fect to be obtained as

With the terms of equation (25) rearranged, it can be verifiedthat

The sign of the term D2 is in general negative. Only when theforeign interest rate substantially exceeds the domestic discountrate would D2 be positive. The likelihood of this happening in adeveloping economy is quite remote. Therefore, equation (26)implies that the real discount rate will increase with inflation aslong as the controlled interest rate is not increased by significantlymore than the increase in inflation. In other words, when the realcontrolled interest rate is raised slightly, but within limits set byequation (26), saving will improve because the induced increase inthe real discount rate will reinforce the positive impact on savingof a rise in the controlled interest rate. Despite the increases inreal interest rates, a matching increase in investment occurs be-cause the real cost of capital actually falls—owing both to anincrease in the share of foreign-currency loans, which is cheaperthan domestic equity, and to the rise in the nominal discount rate,which reduces the opportunity cost of external debt service pay-ments.

The result is illustrated in Figure 3 for the special case ofdR/dv = 1. When inflation rises from ir0 to TTI, the saving scheduleeither remains fixed or shifts to the right, to the extent that foreignsaving (current account deficit) increases because of the apprecia-tion of the real exchange rate. The investment schedule also shiftsto the right, thereby raising saving and investment. The upward

if and only if

where


464 V. SUNDARARAJAN

Figure 3. Effect of Increase in Inflation GivenForeign-Currency Debt and Fixed Nominal Exchange Rates

shift in investment can be deduced from equation (24), in whichthe present value of the projected debt service payments on exter-nal loans is given by the expression [(/?/+ 0/)/(p + af-x)]. Thisterm is clearly reduced (thereby reducing the cost of capital) whenp increases because of higher inflation, but x remains fixed. More-over, the share of foreign-currency debt in financing investmentrises because of the appreciation of the real exchange rate (0 falls,raising C2; the increased share of foreign-currency debt in fi-nancing investment may be facilitated by the increase in foreignsavings, but this need not be the case). This development alsoreduces the cost of capital and contributes to the upward shift ininvestment.

Conditions under which saving and investment improve can besummarized as follows. With dR/d^>l and foreign saving as-sumed to be fixed,10 saving will improve (dSId^ > 0) if and only if

10The assumption of fixed foreign saving is for expository convenience only.Making foreign saving a function of the real exchange rate does not alter thequalitative conclusions about the effects of interest rates and exchange ratepolicy on investment.


The left-hand side of inequality (27) is approximately the ratio ofdomestic-currency debt to equity. The right-hand side is the sumof the interest-sensitivity ratio and a positive term. This positiveterm is quite large when dRIdts is close to unity, and inequality(27) always holds; therefore, saving improves regardless of thesize of the debt ratios. When the controlled rate is raised substan-tially in real terms, however, the size of the debt ratio becomesbinding. But the upper limit on the debt ratio now applies only todomestic debt; moreover, the limit is larger than in the case of noforeign-currency debt.

When the controlled interest rate is allowed to decline in realterms as inflation rises (or is allowed to rise as inflation falls),inequality (27) is necessary and sufficient to ensure a reduction insaving and an increase in the real cost of capital. This underscoresthe need for an active interest rate policy in the presence ofsubstantial use of foreign-currency debt. If foreign-currency debtis substantial and the domestic debt ratio is therefore smallenough to satisfy inequality (27), then the emergence of a negativeinterest rate (owing to increased inflation) will reduce investment,thus producing precisely the opposite of the effect intended.

When there is no change in inflation, interest rate policy willimprove saving if and only if the domestic debt ratio is sufficientlysmall:

It is important to note that, even if the average domestic debtratio a(3 is small, the above inequality may be violated if, at themargin, firms use substantial amounts of foreign currency loans tofinance investment in the hope that the debt ratio will revert totarget levels in the long run. In this case the left-hand side ofinequality (28) could become large (that is, C2 is quite largebecause of high growth and expectations of inflation), therebyviolating the necessary (and sufficient) condition for ensuring pos-itive saving response. This type of adverse outcome for interestrate policy can be avoided if foreign-currency loans can be amor-tized more flexibly, thereby permitting a reduction in marginaldebt ratios.


where


466 V. SUNDARARAJAN

So far the analysis has been based on the assumption that thepath of the nominal exchange rate is fixed. Similar analysis can bereadily completed under the assumption that the real exchangerate (or its rate of change) is fixed a priori, so that the nominalexchange rate varies with changes in inflation. Therefore, thepresent value of external debt service payments varies because ofchanges in the nominal exchange rate, but the marginal share offoreign-currency debt remains unaffected (C2 does not change,since ir^ and 9 remain fixed). As a result, the effects of interestrate policy turn out to be different from the case of a fixed nominalexchange rate. But the safe limit on the domestic debt ratio re-mains as that shown in inequality (28), which now applies to boththe pure interest rate action as well as to interest rate responsesto changes in inflation. Thus, irrespective of the exchange rateregime, the availability of foreign-currency loans serves to easethe constraint governing the improvement of saving and in-vestment through interest rate policy.

The discussion above suggests that the impact of interest ratepolicy can be sensitive to the exchange rate regime. Proper anal-ysis of this aspect of interest policy, however, would require theexplicit incorporation of the effect of exchange rates on inflation,output, and capital flows. Such analysis is beyond the scope of thepresent paper.

III. Conclusions and Policy Implications

In many developing countries, enterprises rely largely on debtfinance. Equity capital remains scarce, in part because the bank-ing system and (in some cases) the unregulated segments of thefinancial system such as the curb markets have together providedsubstitutes for stock issue in the form of long- and short-termloans, whereas the flow of foreign saving has been mainly in theform of debt rather than equity. For example, the average debt-equity ratio of firms in the industrial sector in the Republic ofKorea has grown from about 100 percent in the early 1960s toabout 500 percent in recent years, in part because of the rapidgrowth of the banking system after interest rate reform in 1965.The resultant overleveraged financial structure of firms is oftenperceived to restrict the economic policy options open to theauthorities. The purpose of this paper has been to analyze themacroeconomic consequences that flow from enterprises fi-



nancing their investment with a large share of debt in relation toequity.

For this purpose, the paper developed a model of saving, in-vestment, portfolio adjustments, and the debt ratio in developingcountries that are characterized by segmented financial markets,controls on the banking system, and substantial reliance on debt,including external debt. The financial structure of firms, and plau-sible behavioral assumptions regarding how firms adjust their fi-nancing patterns, were explicitly built into the model by appropri-ately defining the cost of capital in developing economies, therebyemphasizing the linkage between debt behavior and incentives forsaving and investment. The model was used to analyze the impactof interest rate policy on stability and growth.

The major conclusions of the analysis are as follows. The debt-equity ratios of firms make a sizable difference for the impact ofstabilization policies, particularly of interest rate policies. Whendebt ratios used by firms are large, pursuit by the authorities of apassive interest rate policy—that is, maintenance of the controlledinterest rate unchanged when inflation changes—can lead to mac-roeconomic instability that is characterized by perverse effects ofmonetary policy and accelerating inflation or deflation. There-fore, in economies in which firms tend to have a large debt-equityratio, appropriate adjustments in the real administered interestrate become necessary to achieve macroeconomic stability. Theimpact of such action on saving and investment, however, is con-ditioned by the relative shares of domestic- and foreign-currencydebt and by the ability of firms to adjust these relative shares andto change the debt ratio in general.

There usually exists a safe upper limit on the debt-equity ratioof firms in the aggregate, defined as the limit that, if exceeded,leads to perverse effects on saving and investment when the realinterest rate is raised. This limit depends mainly on the interestsensitivity of saving, but it is also influenced by a host of otherconsiderations, including the initial conditions in domestic fi-nancial markets, the ability of firms to adjust the rate of amortiza-tion and the target debt ratio, the terms and availability of foreign-currency loans, and the size of the increase in the controlledinterest rate. For example, when the debt ratio is governed pri-marily by the availability of financial saving—as would be the casein a financially repressed environment—the safe limit on the debt-equity ratio becomes less stringent than in the case of a fixed debtratio. In contrast, more stringent limits apply when the debt ratio


468 V. SUNDARARAJAN

is determined by demand, so that firms are able to adjust thedebt-equity ratio optimally to balance the benefits of additionalsubsidized credit from banks with the associated costs arising fromthe increased riskiness of investment. A more stringent limit alsoapplies when the rate of amortization is fixed and the discount rateis low. The availability of foreign capital, however, serves to easethe constraint on the debt-equity ratio.

When the debt ratio exceeds the safe limit, appropriate in-creases in the real interest rate to ensure stability would alsoinvolve considerable cost in growth forgone. Because of this highcost, maintenance of low and stable inflation is the optimal policyin economies with firms relying on high leverage.

In view of the significant implications of the debt-equity mixused by firms, an evaluation of the financial structure of firms andthe institutional framework of the financial system that underliessuch structure is important for a proper assessment of the impactof stabilization policies. Often the effectiveness of stabilizationpolicies, particularly of interest rate policies, can be enhanced byimplementing appropriate financial reform measures that includesteps to reduce the debt-equity ratios of firms. To the extent thatthe financing patterns used by firms are conditioned by the insti-tutional framework of the financial system, substantial changes inthe debt-equity mix can be brought about only in the long runthrough institutional reforms (for example, through promotingcorporate saving, developing capital markets, and establishingdebt-equity norms). An assessment of the safe limit on the debt-equity ratio that is based on the macroeconomic frameworksuggested in the paper can help to devise debt-equity norms forproject finance and to decide the extent to which the reforms ofthe financial system should emphasize a restructuring of companyfinance. In any event, the financial reform package should striveto reduce segmentation of the financial markets and to reduceinterest subsidy because, as demonstrated in the paper, such ac-tions can also contribute to macroeconomic stability, improve theeffectiveness of interest rate policy, and eventually reduce thecost, in growth forgone, of stabilization policies. In addition, ap-propriate adjustments in lending practices relating to the rolloverof credits, both domestic and foreign, and to the provision ofadequate access to foreign capital can complement stabilizationpolicies in a significant way.



APPENDIX I

Derivation of the Expressionfor Real Cost of Capital

Two approaches to the derivation of the equation for real capital cost areconsidered. The first is based on an optimal control technique, the second on thewell-known Modigliani-Miller (1963) theorem (their proposition I).

Optimal Control Approach

The problem is to minimize total cost,

with respect to the control variable /, subject to the following differential equa-tions on the state variables K, G, and F:

The notation is explained below (for convenience, the time subscript as wellas the superscript m to denote the marginal ratio have been omitted):

p Discount rate, or the opportunity cost of funds to the ownersof the firm

C Variable cost of productionQ* Target outputK Real capital stock1 - a Share of investment financed by equity (including curb mar-

ket loans)a Marginal debt ratioP Share of domestic-currency debt in total debt/ Real gross investmentP Price level (Pt - e™, where TT is the rate of inflation)Rd Domestic interest rate (the controlled interest rate)Rf Foreign interest ratead Rate of amortization of domestic debtaf Rate of amortization of foreign debtG Domestic-currency debt outstandingF Foreign-currency debt outstandingE Exchange rate (domestic-currency units per unit of foreign

currency; Et = E0ext, where x is the rate of change in £)

(Rd + ad)G Debt service payments on domestic-currency loans(Rf + df) FE Debt service payments on foreign-currency loans in domestic

currency.


470 V. SUNDARARAJAN

The Hamiltonian for the control problem is

The first-order conditions are given by

Solving the differential equations (33) and (34) yields the present value of oneunit of domestic-currency loan,

and the present value in foreign currency of one unit of foreign-currency loan,

It is assumed that the exchange rate at time t is given by Et = E0 exp (xt),where x is the expected rate of change in the nominal exchange rate.

Substituting the values of X2 and X3 given in equations (36) and (37) intoequation (35) and regrouping terms allows X! to be expressed as

where the price level P is entered as exp (irt), with TT denoting the rate ofinflation. The initial price level has been normalized to unity.

Differentiating both sides of equation (38) with respect to time allows analternative expression for Xi to be given by

On substituting equations (38) and (39) into equation (32) and rewriting, it isseen that along the optimal path the following relation should hold:

where rfc is the real cost of capital.In the special case when there is no foreign-currency loan (P = 1), the cost of

capital can be written as

Modigliani-Miller Approach

An alternative approach to deriving the cost of capital formula (41) is basedon proposition I of Modigliani and Miller (1963). For simplicity the role offoreign-currency debt will be ignored.



Consider a project, with an earnings stream given by X exp [(TT - 8)r], that isfinanced by an initial debt D0 that is amortized at the rate a. Thus, the streamof debt outstanding over time is given by Do exp (-at). As before, TT is the rateof inflation (perfectly anticipated), but 8 now stands for the rate of output decay(deriving from the real economic depreciation of the underlying equipment). LetS denote the value of equity and V the value of the project. By definition,

where p is the required return to equity from the point of view of the equityinvestors in the project, and R is the rate of interest on debt. From Modiglianiand Miller's proposition I, the overall cost of capital c0 is fixed given the riskcharacteristics of the project. Therefore, the value of the project is given by

which should equal, at the margin, the total initial cost of the project, /. Assumethat a proportion a of the initial cost of the project is financed by debt; for themarginal project,

Because by definition V = S + Do, substituting from equations (42), (43), and(44) and solving for c0 yields

This is exactly the expression for cost of capital shown in equation (41) ob-tained from the optimization exercise.

An interesting implication of equation (45) is that the required return to equityp is a nonlinear function of a for any given c0, TT, 8, and with a =£ 8 - TT. Whena = 8 - TT, then the familiar linear function derived in Modigliani and Miller'sproposition II is obtained. With the assumption that a =/= 8 - IT, the requiredreturn to equity p is the positive root of the following quadratic equation:

APPENDIX II

Relation Between Marginaland Average Debt Ratios

First the relation between the marginal and average debt ratios will be derived.Solving the differential equation (30) of Appendix I allows the level of domesticdebt of a firm to be expressed as

where G0 is the initial level of debt, assumed to be zero, and amd is the marginal


This is equation (10) of the text, in which aad is the limit of af as t tends toinfinity.

A similar analysis for external debt (by solving equation (31) of Appendix I)yields

where E0 is the initial exchange rate, and x denotes the expected rate of increasein the nominal exchange rate. Therefore E0 exp (xi) denotes the exchange rateexpected at time t. From equation (48) it is clear that, when the rate of amorti-zation on foreign loans af equals 8-ir + jc, the average external debt ratioFtEJKt • exp (irt) equals the marginal ratio oimf. The condition af = 8 - TT + xreduces to af - 8 - TTW + 0 if the real exchange rate is expected to change at therate 0, so that * = ir - IT*- + 6, where TTW is the foreign rate of inflation. It will beassumed that the rate of foreign inflation and the rate of amortization of foreignloans are fixed. With the procedure as before, it can be verified that the marginaland the long-run average external debt ratios are related as follows:

This is the same as equation (11) of the text.

472 V. SUNDARARAJAN

debt ratio for domestic loans. Integrating by parts allows equation (46) to berewritten as

If it is assumed for simplicity that real capital stock is expected to grow at the rateg, so that Ks = KQ exp (gs), then

By using equation (47a), the average debt ratio oiatd at time t can be expressed as

In the long run, the last term of the above equation approaches zero, and thefollowing relation between the average and marginal debt ratios emerges:



REFERENCES

Ang, James S., "The Intertemperal Behavior of Corporate Debt Policy," Jour-nal of Financial and Quantitative Analysis (Seattle, Washington), Vol. 11(November 1976), pp. 555-60.

Beranek, William, "A Little More on the Weighted Average Cost of Capital,"Journal of Financial and Quantitative Analysis (Seattle, Washington), Vol.10 (December 1975), pp. 892-96.

, "Research Directions in Finance," Quarterly Review of Economics andBusiness (Urbana, Illinois), Vol. 21 (Spring 1981), pp. 6-24.

Ekman, Elon V., "A Dynamic Financial Model of a Managerial Firm," inOptimal Control Theory and Economic Analysis, First Viennese Workshopon Applications of Control Theory, ed. by Gustav Feichtinger (Amsterdam:North-Holland, 1982; New York: Elsevier, 1982), pp. 79-105.

Eriksson, Goran, "The Effects of Taxation on the Firm's Investment and Fi-nancial Behavior," Scandinavian Journal of Economics (Stockholm), Vol.82 (No. 3, 1980), pp. 362-77.

Feldstein, Martin S., Jerry R. Green, and Eytan Sheshinski, "Inflation andTaxes in a Growing Economy with Debt and Equity Finance," Journal ofPolitical Economy (Chicago), Vol. 86 (April 1978), pp. S53-S70.

Feldstein, Martin S., and Jerry R. Green, "Corporate Financial Policy andTaxation in a Growing Economy," Quarterly Journal of Economics (Cam-bridge, Massachusetts), Vol. 93 (August 1979), pp. 411-32.

Fry, Maxwell J., "Models of Financially Repressed Developing Economies,"World Development (Oxford, England), Vol. 10 (September 1982),pp. 731-50.

Galbis, Vincente, "Financial Intermediation and Economic Growth in Less-Developed Countries: A Theoretical Approach," Journal of DevelopmentStudies (London), Vol. 13 (January 1977), pp. 58-72.

Gordon, Roger H., "Interest Rates, Inflation, and Corporate Financial Policy,"Brookings Papers on Economic Activity: 2 (1982), The Brookings Institu-tion (Washington), pp. 461-91.

Kite, Gailen L., "Leverage, Output Effects, and the M-M Theorems," Journalof Financial Economics (Amsterdam), Vol. 4 (March 1977), pp. 177-202.

Kim, E. Han, "A Mean-Variance Theory of Optimal Capital Structure andCorporate Debt Capacity," Journal of Finance (New York), Vol. 33 (March1978), pp. 45-63.

Linke, Charles M., and Moon K. Kim, "More on the Weighted Average Cost ofCapital: A Comment and Analysis," Journal of Financial and QuantitativeAnalysis (Seattle, Washington), Vol. 9 (December 1974), pp. 1069-80.

McKinnon, Ronald L, Money and Capital in Economic Development (Washing-ton: The Brookings Institution, 1973).

Madan, B.K., Report on a Study of the Debt-Equity Ratio Norms (New Delhi:Management Development Institute, 1978).

Marsh, Paul, "The Choice Between Equity and Debt: An Empirical Study,"Journal of Finance (New York), Vol. 37 (March 1982), pp. 121-44.


474 V. SUNDARARAJAN

Modigliani, Franco F., "Debt, Dividend Policy, Taxes, Inflation, and MarketValuation," Journal of Finance (New York), Vol. 37 (May 1982),pp. 255-73.

, and M.H. Miller, "Corporation Income Taxes and the Cost of Capital:A Correction," American Economic Review (Nashville, Tennessee), Vol. 53(June 1963), pp. 433-43.

Morris, James R., "A Model for Corporate Debt Maturity Decisions," Journalof Financial and Quantitative Analysis (Seattle, Washington), Vol. 11 (Sep-tember 1976), pp. 339-57.

Myers, Stewart C., "Determinants of Corporate Borrowing," Journal of Fi-nancial Economics (Amsterdam), Vol. 5 (November 1977), pp. 147-75.

Nickell, Stephen J., The Investment Decisions of Firms (Welwyn, England:James Nisbet and Co., 1978).

Patrick, Hugh T., "Finance, Capital Markets and Economic Growth in Japan,"in Financial Development and Economic Growth: The Economic Con-sequences of Underdeveloped Capital Markets, ed. by Arnold W. Sametz(New York: New York University Press, 1972).

Rubinstein, Mark E., "A Mean-Variance Synthesis of Corporate Financial The-ory," in Issues in Management Finance, ed. by Stewart C. Myers (New York:Praeger, 1976), pp. 46-60.

Sakong II, "An Overview of Corporate Finance and the Long-Term SecuritiesMarket" in Essays on the Korean Economy, Vol. 1: Planning Model andMacroeconomic Policy, ed. by Kim Chuk Kyo (Seoul: Korean DevelopmentInstitute, 1977), pp. 228-62.

Shaw, Edward S., Financial Deepening in Economic Development (New York:Oxford University Press, 1973).

Sundararajan, V., and Subhash Thakur, "Public Investment, Crowding Out,and Growth: A Dynamic Model Applied to India and Korea," Staff Papers,International Monetary Fund (Washington), Vol. 27 (December 1980),pp. 814-55.

Taylor, Lance, Structuralist Macroeconomics: Applicable Models for the ThirdWorld (New York: Basic Books, 1983).

van Wijnbergen, Sweder, "Credit Policy, Inflation, and Growth in a FinanciallyRepressed Economy," Journal of Development Economics (Amsterdam),Vol. 13 (August-October 1983), pp. 45-65.


Effects of Exchange Rate

Volatility on Trade

Some Further Evidence

PADMA GOTUR*

rate movements witnessed since the beginning of generalizedfloating have led to widespread interest in the nature and extentof the impact of such movements on trade. A principal concern isthat exchange rate volatility appears to increase the risk and un-certainty in international transactions and may therefore ad-versely affect trade and investment flows. This concern hasstrengthened in recent years in response to increasing protec-tionist trends and slowing growth of world trade, and numerousempirical studies have been written on the subject. The generallyinconclusive findings of these studies, however, have failed toprovide any empirical basis for the view that exchange rate vol-atility has discouraged international trade. Indeed, a recent sur-vey of the empirical studies examining the effects of increasedexchange rate volatility on international trade concluded that "thelarge majority of empirical studies . . . are unable to establish asystematically significant link between measured exchange ratevariability and the volume of international trade, whether on anaggregated or on a bilateral basis" (International Monetary Fund(1984, p. 36)). A recent paper by Akhtar and Hilton (1984a)examines afresh the issue of whether exchange rate uncertainty,

*Ms. Gotur, an economist in the Research Department, is a graduate ofGeorge Washington University, where she was an assistant professor of eco-nomics.

The author is grateful to M. A. Akhtar and R. Spence Hilton for kindly provid-ing the data used in their study, and she acknowledges the beneficial conversa-tions held with her colleagues in the Research Department and with BonnieLoopesko and P.A.V.B. Swamy.

475

TTHE VOLATILITY, frequency, and erratic pattern of exchange


476 PADMA GOTUR

proxied by observed exchange rate volatility, has had statisticallysignificant adverse effects on international trade.1

The results of Akhtar and Hilton's study differ from the find-ings of other researchers. They find that exchange rate volatility,as measured by the standard deviation of indices of nominal effec-tive exchange rates, has had significant adverse effects on theaggregate trade in manufactured goods of the United States andthe Federal Republic of Germany. On the basis of regressionresults for export and import price and volume equations, theauthors report a marginally significant adverse effect of exchangerate volatility on U.S. export volumes and U.S. import prices andsignificant adverse effects on German export and import volumes.Therefore, the authors conclude that nominal exchange rate un-certainty has had a significant negative effect on trade. AlthoughAkhtar and Hilton's results from a similar exercise based on ameasure of real exchange rate volatility are less conclusive, theyfind the weight of the evidence sufficient to conclude that "fromthe perspective of international trade, it is desirable to reduceexchange rate uncertainty or variability" (1984a, p. 73). They goon to suggest that this objective may be accomplished throughchanges in macroeconomic policies, official intervention, or sub-stantial changes in the exchange rate system. The authors do notethat, notwithstanding the possible adverse effect of exchange rateuncertainty on trade, other considerations may still support pres-ent floating exchange rate arrangements.

The purpose of the present study is to test the robustness ofAkhtar and Hilton's empirical results, with their basic theoreticalframework taken as given. The analysis has two parts. The firstsimply extends Akhtar and Hilton's analysis, which was limited tothe United States and Germany, to include France, Japan, and theUnited Kingdom. The second examines the robustness of theirresults with respect to changes in the choice of sample period,volatility measure, and estimation techniques.

The main conclusion of the analysis is that the Akhtar-Hiltonmethodology fails "to establish a systematically significant linkbetween measured exchange rate variability and the volume ofinternational trade" (International Monetary Fund (1984, p. 36)).The results obtained are not sufficiently robust to indicate the

Unless otherwise noted, all subsequent references to Akhtar and Hilton'swork are to their more comprehensive paper (1984a); their second paper (1984b)is a condensation of this first paper and reports only selective results.


EXCHANGE RATE VOLATILITY AND TRADE 477

presence of such a link. This is not to say that significant adverseeffects cannot be detected in individual cases, but rather that,viewed in the large, the results tend to be insignificant or unstable.Specifically, the results suggest that straightforward applicationof the Akhtar-Hilton methodology to three additional countries(France, Japan, and the United Kingdom) yields mixed results;that the Akhtar-Hilton methodology seems to be flawed in severalrespects, and that correction for such flaws has the effect of weak-ening their conclusions; that the estimates seem to be quite sensi-tive to fairly minor variations in methodology; and that "revised"estimates for the five countries do not, for the most part, supportthe hypothesis that exchange rate volatility has had a systemati-cally adverse effect on trade. In sum, the empirical results do not,in the author's judgment, provide strong grounds for modifyingthe agnostic conclusions of the Fund survey cited above. Needlessto say, and as already noted in that survey, "the failure to establisha statistically significant link . . . does not, of course, prove thata causal link does not exist" (International Monetary Fund (1984,p. 36)).

The remainder of the paper is organized as follows. Section Ioutlines the model used by Akhtar and Hilton and discusses itsempirical implementation. Section II presents empirical resultsfor the United States, Germany, Japan, France, and the UnitedKingdom based on the Akhtar-Hilton methodology and the con-clusions they suggest. Section III discusses a number of technicalproblems with Akhtar and Hilton's estimations and illustrates theempirical significance of these shortcomings by reference toAkhtar and Hilton's results for the United States and Germany.Section IV outlines preferred methodological procedures and ap-plies them to data for the five countries. Section V presents theconclusions and outlines some avenues for further research. Datadefinitions and sources, and a full set of regression results for thefive countries, are given in the two appendices.

I. The Akhtar-Hilton Model

One of the principal arguments against floating exchange rateshas been that they lead to heightened risk and uncertainty ininternational transactions and thus discourage trade and in-vestment flows. If market participants are risk averse, exchangerate uncertainty and the need to provide against unfavorable


478 PADMA GOTUR

changes will lead to supply and demand decisions that will yieldhigher prices or reduced levels of transactions at any given price.In addition, exchange rate uncertainty and the resultant uncer-tainty about the price to be paid or received in international trademay lead, other things being equal, to a preference for domesticover foreign markets. This preference in turn may lead to a grad-ual reduction in the volume of trade through a backward shift insupply and demand schedules. The size of the shift will depend ontraders' perceptions of the risks involved, on the extent of ex-change rate uncertainty, and on the elasticities of supply anddemand.2

International traders can, of course, avoid or minimize foreigncurrency uncertainty in a short-term trading transaction by hedg-ing in the forward market. But forward markets for maturitiesbeyond one year are not well developed, and thus most tradingactivity, which requires decisions made with respect to a medium-to long-term time horizon, is unprotected by forward cover. Evenif forward cover were available for longer maturities, such marketscould not eliminate exchange rate uncertainty as long as tradersare unable to predict the magnitude and timing of all their foreignexchange payments or earnings (Lanyi (1969)).

Akhtar and Hilton, therefore, disregard the possibilities forforward cover and postulate a standard set of demand and priceequations, with each equation augmented to include the exchangerate volatility variable. In the volume equations this variable isexpected to reflect the effect on demand of the price uncertaintyassociated with exchange rate uncertainty when invoices are de-nominated in a currency other than that of the demander. Simi-larly, in the price equations the volatility variable is expected toreflect the increase in supply prices induced by increased ex-change rate uncertainty when invoices are denominated in a cur-rency other than that of the supplier.

Specifically, Akhtar and Hilton postulate the following fourequations. The export demand equation is

2 Besides increasing costs through uncertainties, exchange rate fluctuationsmay result in costly snifts of resources between economic activities in responseto changing price incentives or to greater riskiness perceived for the traded-goods sector. Large and persistent changes in real exchange rates can involveserious adjustment costs that affect investment decisions and trade patterns.Such resource shifts and the related economic costs are not, however, directlyassociated with the more short-term volatility in nominal exchange rates beingexamined in the present analysis and, therefore, are not of particular relevanceto it. (For a review of the mechanisms by which exchange rate volatility couldaffect trade flows, see International Monetary Fund (1984).)



where XV is the quantity index of total manufacturing exportsdelivered; YF is the real foreign activity level; PX is the price ofmanufacturing exports, in domestic currency; r is the foreign cur-rency price of domestic currency; PFf is the price of foreign-produced substitutes of exports, in foreign currency; and SXf isthe exchange rate risk facing demanders of exported goods.

The import demand equation is

where MV is the quantity index of total manufacturing importsdelivered; YD is the real domestic activity level; PD is the priceof domestically produced substitutes for imported manufactures,in domestic currency; PM is the price of foreign-produced manu-facturing goods faced by domestic consumers, in domestic cur-rency; and SMd is the exchange rate risk facing demanders ofimported goods.

The export supply equation is

where UCD represents input costs of domestic manufactured out-put, in domestic currency, and SXd is the exchange rate risk facingdomestic producers of the exported commodity.

Finally, the import supply equation is

where UCFf represents input costs of foreign manufactured out-put, in foreign currency; r is the foreign currency price of domesticcurrency; and SMf is the exchange rate risk facing foreign pro-ducers of the imported commodity.

It is assumed that prices are set on the date a contract is maderather than on the delivery date. The price term in equations (1)and (2) is the relative price that exporters and importers expect toreceive or to pay on delivery, which is when payment is assumedto be made. If the contract price is quoted in domestic currency —say, for the importer — the importer faces no price uncertainty.When payment must be made in foreign currency, however, the(domestic) price of imports cannot be known if uncertainty existsabout future exchange rates or if the importer does not hedge.This exchange rate risk is denoted by SMd. A similar reasoningapplies to the export demand equation, in which the exchangerate risk is denoted by SXd.


480 PADMA GOTUR

Equations (3) and (4) imply that the foreign supply of importedmanufactured goods and the domestic supply of manufacturedexports are both assumed to be perfectly elastic with respect to thevolume of trade. Although not very realistic, this assumptionpermits one to use ordinary least-squares regression proceduresfor estimating the structural equations, since it implies that thesupply price of traded goods is unaffected by the volume of trade.Akhtar and Hilton note that this assumption permits a distinctionbetween the different effects of volatility on demanders and sup-pliers in terms of the price and quantity of traded commodities.They point out that the perfectly competitive market structureimposed by this assumption implies that, whereas the uncertaintyfaced by suppliers cannot directly affect the volume of trade de-manded, such uncertainty can indirectly affect trade volumes byraising the price of traded goods.

An important methodological issue concerns specification ofthe exchange rate volatility measure. A first question is whetherto base the measure on the nominal or the real exchange rate, aquestion that hinges on which rate better captures the risk oruncertainty faced by traders, particularly over the medium-termplanning horizon adopted by them. It is frequently argued that,over this time horizon, the real exchange rate is the more relevantmeasure because the effects of uncertainty on a firm's revenuesand costs that arise from fluctuations in the nominal exchange rateare likely to be offset in large part by movements in costs andprices. Akhtar and Hilton, however, opt for the nominal exchangerate: first, because of the highly unpredictable nature of exchangerate changes and, second, because of the lack of empirical supportfor purchasing power parity over the medium term. Given theshort time horizon over which exchange rate variations are exam-ined in Akhtar and Hilton's analysis, it is probably correct to sup-pose that most of the variability in the real exchange rate comesfrom the variability in the nominal rate.

Given the choice of nominal over real exchange rates, a secondquestion concerns the volatility measure that is appropriate foruse in empirical work. The various measures that have been usedinclude the more conventional ones, such as the standard devi-ation of the levels of exchange rates or of the changes in theserates, and others, such as absolute percentage first differences ofexchange rates, nonparametric measures such as Gini's mean dif-ference, and measures based on the estimated ex ante (rather thanex post) exchange rate. Each of these has advantages and draw-



backs. (For a discussion of the merits and limitations, see Brodsky(1984), Kenen (1979), Lanyi and Suss (1982), and Rana (1981).)The measure chosen by Akhtar and Hilton is the standard devi-ation of the level of the daily effective exchange rate during eachquarter. They experimented with other volatility measures, butthese results are not reported in their paper. These alternativemeasures were the natural log of the volatility measure, the stan-dard deviation of the natural log measure, the standard deviationof the daily percentage changes of the multilateral exchange rateindices, the trade-weighted averages of the standard deviations ofthe daily observations of a country's bilateral exchange rates, andGini's mean difference coefficient. Akhtar and Hilton note thatuse of alternative volatility measures in general yielded similarresults, although all alternative measures were not used in eachsupply and demand equation for the two countries they studied(the United States and the Federal Republic of Germany). Therationale underlying their use of the standard deviation of thedaily effective exchange rate in each quarter is that the averageexchange rate for the quarter is the best predictor of the expectedrate for each day of the quarter. (For further discussion of thechoice of volatility measure, see the ninth paragraph of Sec-tion III.)

However volatility is measured, the relation between tradeflows and exchange rate uncertainty may not be independent of,and cannot easily be separated from, other uncertainties faced bytraders (such as those relating to other features of the economicenvironment: for instance, the increase in uncertainty in the eco-nomic environment in the last decade may also be attributed tothe oil price shocks, which overlapped with the advent of floatingexchange rates). There is, therefore, need for considerable cau-tion in interpreting empirical results. To the extent that othersources of risk faced by traders are partially offset by exchangerate fluctuations (or vice versa), the empirical results will over-state (or understate) the effects of exchange rate uncertainty ontrade.

Akhtar and Hilton modified the basic structural equations,equations (1) through (4) above, for empirical purposes as fol-lows: variables for domestic and foreign capacity utilization wereadded to each of the equations; domestic and foreign unit costvariables were proxied by series for domestic and foreign pricesfor manufactures; price equations were extended to include com-petitor price variables; seasonal dummy variables were added to


482 PADMA GOTUR

all equations; and dock-strike dummy variables were included inthe volume equations for the United States.

Independent variables other than the price and volatility vari-ables were assumed to enter the equations either contemporane-ously or with a one-quarter lag. The relative price terms in thevolume equations were permitted a lag of up to eight quarters toaccount for order-delivery lags. A similar lag structure was im-posed on the volatility measure to take into account not onlyorder-delivery lags but also the gradual adjustment of expectedvolatility to actual volatility. Lags of up to eight quarters wereimposed on the exchange rate volatility and relative price vari-ables in the volume equations and on the volatility variable in theprice equations. In all these cases, a second-degree polynomial lagstructure was imposed with a zero (far) end-point constraint. Aone-iteration Cochrane-Orcutt (CO) procedure was employed tocorrect for first-order serial correlation in all equations. Exceptfor the volatility measure, the natural log of all variables was used,and hence the estimated coefficients represent elasticities. Theresultant equations were estimated in Akhtar and Hilton's study(1984a) using U.S. and German quarterly data for various esti-mation periods between 1974 and 1982. Akhtar and Hilton fo-cused on the results for the period from the first quarter of 1974through the fourth quarter of 1981, although they also presentedresults for the extended period through the last quarter of 1982.

II. Empirical ResultsBased on the Akhtar-Hilton Methology

Table 1 presents, for five countries, regression estimates for thevolatility coefficient that result from the model and estimationprocedures just described. The results for the United States andGermany are, of course, those reported in Akhtar and Hilton'spaper. As noted previously, these estimates suggest that nominalexchange rate volatility has had a marginally significant adverseeffect on U.S. export volume, and significant negative effects onGerman export and import volumes. For U.S. imports, volatilitywas found to reduce import volumes indirectly through a mar-ginally significant positive effect on import prices. In quantitativeterms, the results indicate that a 10 percent increase in exchangerate volatility can lead to a 2 percent reduction in German exportsand a reduction in U.S. exports of about l/2 percent. On the import



Table 1. Regression Results for Exchange Rate Volatility VariableBased on Akhtar-Hilton (A-H) Methodology

DependentVariable

Export volume

Import volume

Export price

Import price

UnitedStates

-0.04(1.82)0.005

(0.28)-0.002(0.31)0.02

(1.94)

Germany,Fed. Rep. of

-0.22**(3.24)

-0.12*(2.51)0.001

(0.10)0.01

(0.31)

France

-0.02(0.81)

-0.002(0.03)0.03**1

(5.24)0.03**

(3.18)

Japan

0.04*(2.59)0.02

(0.48)0.05**

(3.90)0.06*

(2.79)

UnitedKingdom

0.04(0.81)0.05

(0.88)-0.03*(2.02)

-0.005(0.32)

MemorandumImplied total (direct and indirect) elasticity

of trade flows with respect to volatility variable2

Exports -0.05 -0.20 — -0.18 —Imports -0.06 -0.12 -0.04 -0.16 —

Note: Figures in parentheses are t -statistics; * denotes statistical significanceat the 5 percent level; ** denotes statistical significance at the 1 percent level.

Sources: For the United States and Germany, A-H (1984a); for France, Japan,and the United Kingdom, own calculations (see the text and Appendices Iand II).

1 The relative price variable in the export volume equation is not statisticallysignificant.

2 In the calculations all coefficients that were not significant at the 5 percentlevel were assumed to be equal to zero. For the United States, the coefficientsin the export volume and import price equations are significant only for a one-tailtest.

side, a 10 percent increase in volatility can cause a 1 percentdecline in German imports and a l/2 percent decline in U.S. im-ports. It is these results that are the basis for the authors' conclu-sion that exchange rate uncertainty has had a significant adverseeffect on the volume of international trade.

Table 1 also presents results of the Akhtar-Hilton methodologyfor three additional countries — France, Japan, and the UnitedKingdom. These estimates were obtained by replicating theAkhtar-Hilton methodology as closely as possible. The main dif-ference is that the volatility measures for these three countrieswere derived as the standard deviations of daily observationswithin each quarter of the effective exchange rate index for eachcountry, weighted according to the Fund's multilateral exchangerate model, MERM (see Artus and McGuirk (1981); the com-putation and sources of this and other data series for these coun-tries are presented in Appendix I).


484 PADMA GOTUR

The results for France, Japan, and the United Kingdom arerather different from those obtained by Akhtar and Hilton for theUnited States and Germany. In the first place, all the coefficientsfor the United Kingdom are not only insignificant but are also ofthe "wrong" sign. Second, whereas the effects in Akhtar andHilton's results for the United States and Germany came throughprimarily on the volume side, the results for France and Japansuggest instead that any adverse effect of volatility on trade isindirect rather than direct. Thus, the volatility variables in thevolume equations are either of the "wrong" sign or quite insignif-icant for France and Japan but are of the "right" sign and signifi-cant in the price equations. Given Akhtar and Hilton's inter-pretation of the export and import price equations as domesticand foreign supply equations, respectively, the results have therather paradoxical implication that, whereas French and Japaneseexporters bear the exchange rate risk (a not unrealistic result ifmost export contracts are denominated in foreign currency),French and Japanese importers do not. Rather, it is the foreignsuppliers to French and Japanese markets who ostensibly bear theexchange rate risk and who accordingly raise their prices to coverthemselves for that risk. This result is not very plausible, es-pecially in light of the magnitudes of the corresponding coeffi-cients in the import price equations for the United States andGermany, which suggest that suppliers to those markets bear asubstantially smaller exchange risk. (See Section IV, footnote 7.)

The interpretation of the coefficients in Table 1 is somewhatobscured by the fact—stemming from the particular specificationchosen by Akhtar and Hilton—that these coefficients are notelasticities. Nevertheless, because the average value of the vol-atility variable often tends to be around unity, the coefficients inTable 1 can be roughly interpreted as elasticities. Indeed, re-gression equations identical to those used to produce the resultsof Table 1, except for a logarithmic transformation of the volatilityvariable, yield regression coefficients very similar to those shownin Table 1. The only significant change in the results is that threeof the eight significant and "right sign" coefficients lose theirstatistical significance (the coefficients for the U.S. and for theJapanese import price equations, and the coefficient for the U.S.export volume equation; therefore, the results with the loga-rithmic transformation of the volatility variable show no effect—direct or indirect—of volatility on U.S. import and export vol-umes). Be that as it may, the statistically significant parameters in



Table 1, together with the price elasticity parameters from thevolume equations, may be used to calculate the overall (that is,direct and indirect) elasticity of trade flows with respect to vol-atility. These implied total elasticities are shown, as a memoran-dum, in the lower tier of Table 1. Taken at face value, they suggestthat German trade and Japanese trade are relatively sensitive toexchange rate volatility, French trade and U.S. trade are rela-tively insensitive to exchange rate volatility, and British trade isquite insensitive to such volatility.

III. Shortcomings of the Akhtar-Hilton Methodology

The reliability of the results reported in Table 1 is underminedby several technical problems associated with the methodologyemployed by Akhtar and Hilton. These problems are discussed inturn below, and their quantitative significance is illustrated byreference to modified results for the United States and Germany.The section ends with the presentation of a set of improved pro-cedures, which are then applied in the following section to thedata for each of the five countries shown in Table 1. It should beunderstood that the critique that follows pertains only to theempirical procedures adopted by Akhtar and Hilton. Their basicanalytical framework is taken as given.

Although Akhtar and Hilton state that they have tested eachequation for first-order serial correlation, it appears that theyhave instead applied the CO correction for serial correlation to allequations as a routine procedure without a preliminary check forthe presence of serial correlation in the ordinary least-squaresestimation. This seems likely because all equations reported bythem incorporate a CO correction, and when their equations areestimated without correction there is no evidence of serial cor-relation in several of the equations. The use of the CO procedurein equations where there was no serial correlation implies anincorrect assumption about the structure of the error term in thoseequations. Moreover, as shown below, the ordinary least-squaresand CO results were significantly different for many equations.

A second unusual feature of the Akhtar-Hilton methodology,also related to their procedure for correcting for serial correlation,is that the equations were estimated using a one-iteration COprocedure, which belongs to the class of "two-stage" generalizedleast-squares correction procedures for serial correlation, rather


486 PADMA GOTUR

than the more customary iterative CO procedure. In the absenceof serial correlation, and with it assumed that the other conditionsof the classical regression model hold, ordinary least-squares esti-mators have all the desirable small-sample and asymptotic proper-ties. Some of these properties are lost, however, when serial cor-relation is present. In this case, generalized least squares, iterativeCO, and maximum likelihood estimators are all consistent andasymptotically equivalent to best-linear unbiased estimators. Butthe small-sample properties of these estimators are difficult toderive analytically, and any choice between them that is based onsampling properties must be made on the strength of Monte Carloevidence (see Judge and others (1980, p. 187), and Kmenta (1971,p. 292)). Such evidence has not led to a clear-cut choice. Inapplied work, however, researchers usually choose the CO iter-ative procedure over the two-stage procedures.

For reasons noted above, a preferable set of procedures forcorrecting for autocorrelation would be first to estimate the equa-tions by ordinary least-squares, check the Durbin-Watson (DW)statistic to test for serial correlation, and, for those equations inwhich serial correlation is present, estimate the equations usingthe iterative CO technique. The resultant estimates for the UnitedStates and Germany are presented in Table 2. The table showsthat the results for Germany turn out to be relatively robust (atleast with respect to the present change in procedures), but thatresults for the United States are not. For that country, the ordi-nary least-squares estimates of the export volume equation do notindicate the presence of serial correlation. Moreover, the ordinaryleast-squares estimate of the volatility coefficient is not statisti-cally significant, thus invalidating the evidence for the direct (ad-verse) effect of exchange rate volatility on U.S. export volumesfound by Akhtar and Hilton in using the CO correlation. The newresults do, however, continue to indicate an indirect effect onimport volumes via import prices, as reported by Akhtar andHilton. Nevertheless, the adjustment in the procedure to correctfor serial correlation has the overall effect of undermining afourth of the evidence adduced by these authors in favor of theirconclusion.

A second problem with Akhtar and Hilton's procedures is theirchoice of sample period. The issue is twofold. First, why do theyfeature results for the 1974-81 period, even though they alsoreport results for the 1974-82 period? Second, why do they in-clude 1974, when that date implies, given the eight-quarter lags



Table 2. Effect on A-H Volatility Results of Changes in ProceduresUsed to Correct for Serial Correlation

DependentVariable

Export volume

Import volume

Export price

Import price

United

A-H result

-0.04(1.82)0.005

(0.28)-0.002(0.31)0.02

(1.94)

States

Modifiedresult

-0.04(1.36)0.01

(0.49)-0.001(0.24)0.02*

(2.01)

Germany, Fed.

A-H result

-0.22**(3.24)

-0.12*(2.51)0.001

(0.10)0.01

(0.31)

Rep. of

Modifiedresult

-0.22**(3.23)

-0.12*(2.34)0.001

(0.10)-0.01(0.24)

Note: Figures in parentheses are f-statistics; * denotes statistical significanceat the 5 percent level; ** denotes statistical significance at the 1 percent level.

Sources: A-H (1984a) and own calculations.

used in their analysis, use of data from the period before floatingexchange rates? These questions are not trivial. On the first point,it is particularly important that the marginally significant negativeeffect of volatility on U.S. export volume for the sample period1974-1981 (which is barely significant at the 95 percent level ofconfidence by a one-tail test) disappears altogether when the esti-mation period is extended to 1982 (see Table 3). Although thecontradictory results for the extended sample period are reportedby Akhtar and Hilton in their longer research paper (1984a), theirmain article on this research (1984b) focuses only on the resultsfor the shorter sample period and presents only those results.Because it is desirable to examine the effects of volatility on tradeflows over the longest available estimation period of the floatingexchange rate regime, it is more appropriate to look at the extend-ed estimation period, through 1982, rather than to stop with 1981.In sum, although the results for Germany do not differ signifi-cantly over the two sample periods (1974-81 versus 1974-82), in the case of the United States the effect of volatility on tradeis reduced, under the extended period, to an indirect effect onimport volumes via import prices, without any effect on exports.

The significance of the choice of sample period becomes morecrucial when the analysis turns to the choice of a starting point.Akhtar and Hilton place considerable emphasis on their resultsbeing superior to those of most previous researchers in part be-cause they have used observations only from the period of flexible


488 PADMA GOTUR

exchange rates. Yet, although their estimation period does indeedbegin in 1974 (the initial observation for the dependent variable isfor the first quarter of 1974), their analysis necessarily includesobservations from the period of fixed rates, since they use aneight-quarter lag structure for the exchange rate volatility variableand for the relative price variables. In effect, therefore, their esti-mates in part reflect developments reaching back to early 1972.Thus it is possible that bias in specification may be introducedowing to the change in the exchange rate regime. To avoid thispossibility and to exclude observations from the period of fixedrates, the sample period should begin with the first quarter of1975.

But such a truncation of the sample period has a major impacton the estimates. As can be seen from Table 3, the equationsfailed to reproduce all four pieces of significant evidence reportedby Akhtar and Hilton for the 1974-81 period. For the UnitedStates, the results do not support either a direct or indirect effectof exchange rate volatility on export or import volumes. ForGermany, the re-estimation fails to corroborate the direct effectof volatility on export and import volumes and, indeed, suggestsa "perverse" negative effect on import prices not detected inAkhtar and Hilton's estimation for the period. Overall, Akhtarand Hilton's results are evidently quite sensitive to the inclusionof observations for the transition years preceding adoption of

Table 3. Effect on A-H Volatility Results of Changes in Sample Period

United States Germany, Fed. Rep. of

A-H A-H Modified A-HDependent result result result result

Variable (1974-81 ) (1974-82 ) (1975-81 ) (1974-81 )

Export

Import

Export

Import

volume

volume

price

price

-0(10,

(0-0.(00

(1

.04

.82)

.005

.28),002.31).02.94)

0.01(0.61)

-0.003(0.14)0.001

(0.20)0.02*

(2.61)

0.(0.

-0.(0.0.

(1.0.

(0.

,001,05),02,80),02,55),002,13)

-0(3-0(20,

(00

(0

.22**

.24)

.12*

.51)

.001

.10)

.01

.31)

A-Hresult

(1974-82)

0.19*(2.54)

-0.12**(3.03)0.01

(0.73)-0.002(0.07)

Modifiedresult

(1975-81)

-0,(0.

-0.(0,

-0.(0,

-0,(2.

,05,39),03.59),02,79).07*,16)

Note: Figures in parentheses are t-statistics; * denotes statistical significanceat the 5 percent level;** denotes statistical significance at the 1 percent level.

Source: A-H (1984a) and own calculations.



floating rates. Not only does inclusion of data from the period offixed rates introduce specification bias into the estimation results,it also complicates derivation of the standard deviations of ex-change rates over this period. One may have to rely on monthlyrather than daily observations of the exchange rate to derive thequarterly standard deviations and to compare end-period valueswith period averages, given the discrete rate changes that oc-curred during this period. Moreover, the appropriate measure forexchange rate volatility during the period of fixed exchange ratesmay be different from that for the period of floating rates. Theseproblems are all solved by restricting the estimation period strictlyto the period of floating rates.

Another potential problem with the Akhtar and Hilton's pro-cedures is the arbitrary basis for their specification of the second-degree polynomial lag structure, which they defended on thegrounds of simplicity. This specification is questionable chieflybecause of the extreme sensitivity, amply documented by otherresearchers, of regression results based on this technique tochanges in the specification of the lag structure.3 It was thusdeemed important to test the robustness of Akhtar and Hilton'sresults under alternative specifications. Such testing sometimesyielded significantly different regression results for the UnitedStates. For instance, with a third-degree polynomial, the adverseeffect of volatility on U.S. export volumes disappears once again,and there is no indication of a direct effect on import volumes(Table 4). Moreover, the effect on import prices found with asecond-degree polynomial (by Akhtar and Hilton) loses its signifi-cance. The effect on export volumes also disappears when alterna-tive lag structures and end-point constraints are used in the esti-mation procedure. The results for Germany are more robust whensubjected to alternative dynamic specifications of the basic equa-tions. As shown in Table 4, the results for Germany obtained witha third-degree polynomial lag continue to show a direct effect ofvolatility on trade volumes (given, of course, the 1974-81 sampleperiod featured by Akhtar and Hilton).

A further aspect of Akhtar and Hilton's procedures that meritsattention is their specification of the effective exchange rates usedto compute the volatility variables for each country. These vari-

3 Misspecification of the lag renders the coefficient estimates biased and incon-sistent, and the standard tests of hypotheses unreliable. See Judge and others(1980).


490 PADMA GOTUR

Table 4. Effect on A-H Volatility Results of Change in PolynomialSpecification of Volatility Variable

Polynomial Lag Structure

United States Germany, Fed. Rep. of

A-H Modified A-H ModifiedDependent result result result result

Variable (Second degree) (Third degree) (Second degree) (Third degree)

Export volume

Import volume

Export price

Import price

-0.04(1.82)0.005

(0.28)-0.002(0.31)0.02

(1.94)

-0.04(1.62)

-0.003(0.20)

-0.01(1.70)0.02

(1.74)

-0.22**(3.24)

-0.12*(2.51)0.001

(0.10)0.01

(0.31)

-0.23**(3.23)

-0.14*(2.86)

-0.001(0.07)0.01

(0.27)

Note: Figures in parentheses are t-statistics; * denotes statistical significanceat the 5 percent level;** denotes statistical significance at the 1 percent level.

Sources: A-H (1984a) and own calculations (see the text).

ables were derived on the basis of trade-weighted exchange rateindices constructed with respect to major trading partners (9countries for the United States and 13 for Germany). These mea-sures for the effective exchange rate may be somewhat narrow.The broadest possible measure for the effective exchange rate ismore likely to capture accurately the way in which exchange ratefactors influence trading uncertainty. Furthermore, the Akhtar-Hilton method of weighting by trading partners allows each coun-try's effective exchange rate to vary depending on which countrieswere omitted in that computation. It may, instead, be moreappropriate to treat all countries being examined symmetricallyby standardizing cross-country comparisons through use of thesame set of countries (or bilateral exchange rates) for computingthe effective exchange rates. (This point is made by Kenen andRodrik (1984), although not with regard to Akhtar and Hilton'sresearch.) Therefore, in computing the effective exchange rate, itis more appropriate to use a range of trading partners that is widerthan that used by Akhtar and Hilton.

In light of these difficulties, and with a view to extending theresults to France, Japan, and the United Kingdom, an alterna-



tive measure for exchange rate volatility was computed. Thismeasure—again the standard deviation of the daily observationswithin each quarter—was calculated by using a daily version of theFund's MERM-weighted effective exchange rate index (see Artusand McGuirk (1981) for a description of the MERM weights).This index uses bilateral exchange rates for 18 industrial countriesand, therefore, satisfies both the conditions noted in the previousparagraph. Charts 1 and 2, for the United States and Germany,respectively, compare movements in the volatility variables usedby Akhtar and Hilton and in the alternative measures describedabove. Although the two measures of volatility move together,that used by Akhtar and Hilton in general exhibits greater vari-ance, which might account, in part, for the differences in theestimation results described below.

For the United States, use of the alternative volatility measureyielded an insignificant coefficient estimate for this variable in theexport volume equation (Table 5). The coefficient of the volatilityvariable continued to have the perverse, positive sign in the U.S.import volume equation obtained by using Akhtar and Hilton'svolatility measure. The results also continued to show a weaklysignificant effect of volatility on U.S. import prices over the1974-81 period. The substitution of the MERM-based volatilitymeasure had a somewhat greater effect on the estimates for Ger-many. On the one hand, the volatility coefficient in the exportvolume equation became smaller, and the adverse effect on im-port volume lost its statistical significance. On the other hand, apossible positive effect of exchange rate volatility on import pricescame to the fore.

A final point that requires comment is Akhtar and Hilton's useof dummy variables to correct for seasonality in the regressionequations. Although both seasonally adjusted and unadjustedvariables appear in the regression equations, the authors do notindicate which variables were adjusted and which were not. Incorrecting for seasonality, it is preferable to start with all non-seasonally adjusted series and correct for seasonality usingdummy variables or other techniques or, alternatively, to use allseries that are seasonally adjusted. In practice, time series areoften published only in the adjusted form, thus giving rise to thelikelihood of both adjusted and unadjusted series appearing in theestimation equation. Upon examination of Akhtar and Hilton'sdata sources, it appears that the U.S. export volume series is the


Chart 1. Exchange Rate Volatility, United States, 1973-83

Sources: A-H (1984a) and own calculations using Fund effective exchange rate indices weighted according to the multipleexchange rate model (MERM; see the text and Artus and McGuirk (1981)).


Chart 2. Exchange Rate Volatility, Federal Republic of Germany, 1973-83

Sources: See Chart 1.


494 PADMA GOTUR

Table 5. Effect on A-H Volatility Results of Change in Volatility Measure

DependentVariable

Export volume

Import volume

Export price

Import price

United

A-H measure

-0.04(1.82)0.005

(0.28)-0.002(0.31)0.02

(1.94)

States

Alternativemeasure

-0.04(1.65)0.006

(0.29)-0.002(0.31)0.02

(1.88)

Germany, Fed

A-H measure

-0.22**(3.24)

-0.12*(2.51)0.001

(0.10)0.01

(0.31)

. Rep. of

Alternativemeasure

-0.17*(2.60)

-0.10(1.64)0.01

(1.03)0.04

(1.81)

Note: Figures in parentheses are f-statistics; * denotes statistical significanceat the 5 percent level; ** denotes statistical significance at the 1 percent level.

Sources: A-H (1984a) and own calculations (see the text).

only dependent variable that was seasonally adjusted. Moreover,the seasonal dummy variables were statistically insignificant inthis equation. The export volume equation was, therefore, re-estimated without such variables. This re-estimation continued toshow a weakly significant adverse effect of volatility on U.S. ex-port volumes. It is noteworthy that the volatility coefficient lost itssignificance when the dock-strike dummy variable was droppedfrom the estimation equation. As a separate issue, it would be in-teresting to examine any possible seasonality patterns in the ex-change rate volatility series.

The foregoing findings on the implications for Akhtar andHilton's results of various changes in empirical methods castdoubts on the robustness of these results and thus raise questionsregarding the main policy conclusion that the authors have de-rived from them. The direct, adverse effect of volatility on U.S.export volumes appears to be highly tentative and unstable andtherefore may be disregarded. In addition, the replication resultsfail to provide a strong, consistent basis for any indirect effect ofexchange rate volatility on trade volumes through its effect onprices of traded goods. The results for Germany are somewhatmore robust than those for the United States but appear to bequite sensitive to the choice of sample period. Taken together, theapparent lack of stability and uniformity of the empirical resultsfor both the United States and Germany significantly underminesthe validity of findings reported in Akhtar and Hilton's paper.



IV. Revised Results

Given the shortcomings associated with certain aspects of theAkhtar-Hilton methodology, the equations for the United States,Germany, France, Japan, and the United Kingdom reported inTable 1 were re-estimated using revised estimation procedures.The following changes were made to the Akhtar-Hilton meth-odology.

• An iterative CO correction procedure for serial correlationwas used, but only when the ordinary least-squares estimationindicated the presence of serial correlation (determined on thebasis of the DW statistic and the pattern of residuals).

• 1975-83 was chosen as the sample period, a period that, afterallowance for lags, excludes observations from the period of fixedexchange rates but includes the experience of the entire period offloating rates thus far.4

• Seasonal dummy variables were included in only those equa-tions in which the dependent variable was not seasonally adjusted(for the United States, Germany, Japan, and the United King-dom) and were excluded when it was seasonally adjusted (forFrance); the U.S. export volume series was seasonally adjusted,and the export volume equation for the United States was esti-mated excluding the dummy variables.

• The volatility measures were derived on the basis of standarddeviations of daily observations within each quarter of the nomi-nal MERM-weighted, effective exchange rate index for eachcountry, as described in Section III.

• With some misgivings, a polynomial lag specification was im-posed on the volatility and relative price variables. It was consid-ered important to use some empirical criterion to determine theappropriate dynamic specification for the equations; that is, thelength of the lag, the degree of the polynomial, and the choice ofend-point constraints. The Akhtar-Hilton procedure of arbitrarilypicking one specification for estimation was not used because it is

4 The data used in the estimations reported for the United States and Germanyfor the period 1974-81 were those provided to the author by Akhtar and Hilton.Because it was difficult to extend these series through 1983, new data series werecomputed by the author, for the entire 1973-83 period, using the same pro-cedures adopted for computing the data for France, Japan, and the UnitedKingdom. Although the computation methods are broadly similar to those usedby Akhtar and Hilton, the new series for the United States and Germany do notappear to be strictly comparable to those provided by Akhtar and Hilton.


496 PADMA GOTUR

unlikely that the identical specification would necessarily be ap-propriate for all four equations for all the countries. Furthermore,given the sensitivity of the results to the dynamic specification ofthe equations, it was deemed important to try alternative poly-nomial specifications.5 Predictive testing was used to chooseamong the various specifications. In this procedure, one makesalternative assumptions about the correct lag length and the cor-rect degree of the polynomial and chooses among them on thebasis of their predictive ability. For this purpose, the equationswere estimated for the 1975-81 and 1975-82 periods and extrapo-lated through 1983. The polynomial lag specification that yieldedthe smallest root mean square error was chosen as the preferredspecification, to be used in the re-estimation over the full sampleperiod, 1975-83.

The new results are shown in Table 6 (for a complete set ofregression results, see Appendix II). It is difficult to interpretthem as supportive of the hypothesis that exchange rate volatilityhas systematically undercut world trade.6 On examining the new

5 The problems and difficulties that arise when using the Almon lag specifica-tion are well known. (See Judge and others (1980) and Schmidt and Waud (1973)for a review of the issues involved.) Thus it is sometimes preferable to opt foralternative lag specifications. In the present analysis, this option could not beused because the objective was to determine whether the results of Akhtar andHilton's analysis could be extended to other industrial countries, and thereforetheir broad empirical specification had to be adopted. Some researchers havehandled the problem of determining the correct dynamic specification of anAlmon lag estimation by using a procedure that searches over several possiblevalues for the degree of the polynomial and the length of the lag and then bychoosing the combination that minimizesjhe residual variances (maximizes thecorrected coefficient of determination, R2). The changes in R2 for alternativeestimations, however, are frequently too small to permit meaningful selection ofone estimation over another.

6 Other studies do not help resolve the difficulty. Two recent studies in this areaby Justice (1983) and Kenen and Rodrik (1984), published after the Fund survey(International Monetary Fund (1984)), have reported evidence that may beregarded as being more suggestive than conclusive. Despite testing with alterna-tive measures for exchange rate volatility, Justice did not find a significant effectof such volatility on the volume of U.K. exports of manufactures in the periodof floating rates. He did, however, find some tentative evidence suggesting thatvolatility had influenced export pricing behavior, but this result was heavilydependent on the particular volatility measure used in the estimation. Kenen andRodrik have presented new data on the short-term volatility of real exchangerates for the members of the Group of Ten plus Switzerland and have analyzedthe effect of such volatility on trade volumes. Their results are also mixed. Foronly three of the seven countries examined did they find a significant negativeeffect of volatility on export volumes; for three others, volatility was actuallyfound to stimulate exports, whereas for the remaining five countries the coeffi-cient of the volatility variable was not significantly different from zero. With



Table 6. Revised Regression Results for Exchange RateVolatility Variable, 1975-83

DependentVariable

Export volume

Import volume

Export price

Import price

UnitedStates

0.14*1

(2.70)-0.02(1.36)0.10**

(4.16)-0.005(0.23)

Germany,Fed. Rep. of

-0.12*(2.55)

-0.05(1.07)0.01

(0.58)-0.02(0.70)

France1

-0.01(0.40)0.05

(0.61)0.04*3

(2.73)0.03*3

(2.67)

Japan

0.03(1.50)0.09

(2.01)0.03

(1.33)0.06*

(2.43)

UnitedKingdom2

0.04(0.98)0.07

(1.31)-0.01(0.68)

-0.01(0.82)

MemorandumImplied total (direct and indirect) elasticity

of trade flows with respect to volatility variable4

Exports 0.10 0.10 — — —Imports — — — 0.14 —

Note: Figures in parentheses are £-statistics; * denotes statistical significanceat the 5 percent level; ** denotes statistical significance at the 1 percent level. Forthe complete estimation results, see Appendix II.

Source: Own calculations.Estimation equation excludes seasonal dummies. Estimation results did not

change substantively upon inclusion of seasonal dummies.2 Because of unavailability of data, equations were estimated through the

second quarter of 1983.3 The (relative) price variable was not statistically significant in the correspond-

ing volume equation.4 In the calculations all coefficients that were not significant at the 5 percent

level were assumed to be equal to zero.

results for the United States and Germany, one notes that, of thefour statistically significant results that formed the basis of Akhtarand Hilton's policy conclusions, three no longer are either of the"correct" sign or statistically significant. The only robust coeffi-cient in this respect is that in the equation for German exportvolumes. On the one hand, the adverse effect on German importvolumes reported by Akhtar and Hilton is now quite insignificant,as are the effects on U.S. export volumes and import prices, whichare now of the opposite sign. On the other hand, the new resultspoint to a positive effect on U.S. export prices that was not de-

respect to import volumes, they found a significant negative effect of volatilityfor four countries, a significant positive effect for two others, and a coefficientfor the volatility variable not significantly different from zero for the remainingfive countries.


498 PADMA GOTUR

tected by Akhtar and Hilton, an effect that would indirectly havean adverse effect on U.S. export volumes. The evident instabilityof these parameters points to the more general conclusion that thetesting procedures used here do not appear to lend themselves toany very confident conclusions.

If one broadens the analysis of the results in Table 6 to includeFrance, Japan, and the United Kingdom, a first point to note isthe paucity of any direct adverse effect of exchange rate volatilityon trade volumes. Of the ten such coefficients in Table 6, only oneis statistically significant and of the correct sign—the one forGermany. Moreover, half of the coefficients have the wrong sign.The same is true of the price equations, where again close to halfof the coefficients are negative. Of the positive coefficients in theprice equations, however, four meet the standard test of statisticalsignificance, suggesting that exchange rate volatility may ad-versely affect trade volumes by significantly raising internationaltrade prices.

This conclusion, however, must be qualified in several impor-tant respects. s First, the significance of two of these four coeffi-cients (those for France) is undermined by the nonsignificance ofthe relative price terms of the volume equation. Second, the pat-tern of the coefficients is at odds with commonsense notions aboutthe nature of the links between the various countries and theworld market. Taken at face value, the export price equationssuggest that U.S. and French exporters bear the exchange raterisk on their exports, whereas exporters from Germany, Japan,and the United Kingdom do not. In light of conventional beliefsabout the relative market positions of countries in world trade andof information on the currency composition of countries' trade, itis difficult to provide a clear rationale for this pattern.7 Rather,one would expect that Japanese and French exporters—and, per-haps to a lesser extent, British exporters—would bear relatively

7 Although information on currency composition has not been fully compiled,findings by Magee (1974) and Page (1981) clarify the issue. Citing estimates for1979 and 1980, Page shows that over 95 percent of U.S. exports and 85 percentof U.S. imports are invoiced in U.S. dollars; 80 percent of German exports and40 percent of German imports are invoiced in deutsche mark; 75 percent of U.K.exports and 40 percent of U.K. imports are invoiced in pound sterling; 60percent of Japanese exports and 90 percent of Japanese imports are invoiced inU.S. dollars (with about 30 percent of exports invoiced in yen); 60 percent ofFrench exports and 35 percent of French imports are invoiced in francs; and,finally, 30 percent of German, U.K., and French imports, respectively, areinvoiced in U.S. dollars.



more exchange rate risk. The results on the import side are in thisrespect scarcely more reasonable. From Akhtar and Hilton's per-spective, the import price equations represent the world's supplyprice to the country in question. Hence the positive volatilitycoefficients for France and Japan suggest that the world bears theexchange risk in exporting to these countries, a risk that it doesnot incur when exporting to the United States, Germany, or theUnited Kingdom. Again, these results are not very plausible.Rather, one would expect that Japanese importers—and to asmaller extent French, British, and German importers—wouldbear the exchange risk, whereas the world would bear the risk inexporting to the United States.

With respect to the results by country, the main points to benoted, subject to the qualifications made above, are as follows.For the United Kingdom, the revised estimates do not show eitherdirect or indirect effects of exchange rate volatility on trade, aresult that corroborates the findings of another re-estimation ofAkhtar and Hilton's results for the United Kingdom.8 For France,the results indicate positive effects on trade prices, but theseeffects are undermined by the nonsignificance of the relative priceterms in the volume equations. For Japan, the results suggest thatvolatility might have reduced import volumes through its effect onprices of imported goods. Even here, however, there is no indi-cation of a direct impact on trade volumes. In quantitative terms,the results for Japan suggest that a 10 percent increase in volatilitycould indirectly reduce import volumes by close to \Vi percent.For the United States, the results indicate that volatility mighthave reduced export volumes through an effect on export prices,with a 1 percent reduction in export volumes likely to result froma 10 percent increase in volatility. Once again, however, there isno evidence of a direct effect on export or import volumes. Onlyfor Germany is there evidence of a direct adverse effect of vol-atility on export volumes. In quantitative terms, the results sug-gest that a 10 percent increase in volatility could reduce Germanexport volumes by about 1 percent.

8 Applied to the equations for U.K. manufacturing trade volumes and for unitvalues in the Bank of England's short-term model, the Akhtar-Hilton method-ology yielded no significant effect of short-term volatility on trade for the UnitedKingdom (Bank of England (1984)). No significant direct effect was found ontrade volumes; the relevant coefficients in the import price equation tended tobe negative rather than positive, and those in the export price equation tendedto be of the expected sign but statistically insignificant.


500 PADMA GOTUR

V. Conclusions and Avenues for Further Research

In sum, the empirical results for the five countries do not pro-vide conclusive evidence that exchange rate volatility has had astatistically significant effect on trade flows. The results suggestthat, even if there is some residual effect of exchange rate uncer-tainty on trade, this effect has not operated in a stable and consis-tent manner. Although the present analysis incorporated someimprovements in the statistical methods used by Akhtar andHilton, it would be advisable to adopt further refinements andstability tests in future work, given that the results are sensitive toalternative specifications of equations and sample periods. (Forinstance, to deal with the problem of choosing the optimal poly-nomial specification, a new estimator derived by Kashyap andothers (1984) might be used. This is a Bayesian distributed-lagestimator that is consistent with Shiller's (1973) "smoothnessprior" and uses sample data to improve the operating character-istics of both Almon's and Shiller's estimators; it has been shownto be superior in forecasting when compared with either Almon'sor Shiller's estimator.)

More fundamentally, the equations should be re-estimated us-ing a volatility measure that reflects only the unpredictable ele-ment of exchange rate movements and, thereby, the short-termvolatility or deviations of exchange rates around a long-termtrend. The observed variability in flexible exchange rates typicallyreflects both systematic rate movements, which are largely pre-dictable, and uncertain rate movements, which are largely un-predictable. To the extent that risk from predictable rate changescan be diversified away, it may be argued that it is the unantici-pated component of exchange rate movements that is the appro-priate proxy for the uncertainty in exchange rate transactions. Incalculating exchange rate volatility, it may be important to elimi-nate the movements along some predictable long-term trend,since these movements are unlikely to reflect the risk associatedwith exchange rate transactions. But the standard deviation (usedas a proxy for exchange rate risk) of a nonstationary process, suchas the observed exchange rate series, reflects the short-term vol-atility in the time series as well as the movement along a long-termtrend. The principal volatility measure used by Akhtar and Hiltonis derived as the standard deviations of (the levels of) the observedexchange rate series and, therefore, includes both the trend move-ment and the short-term volatility in the exchange rate. With a



view to focusing on the unpredictable and short-term volatility inexchange rates, an alternative volatility measure, derived as thestandard deviation of percentage changes in the exchange rate,was calculated in the present analysis. Preliminary testing with thismeasure led to only one significant change in the results: the directadverse effect of volatility on German export volumes was nolonger evident, thus invalidating the only direct effect of volatilityon trade volumes, as reported in Table 6. Akhtar and Hilton alsoundertook some re-estimation with this measure, but these resultswere not reported in their paper (see Section II).

It is also not clear that the variation in daily exchange rates isthe most appropriate unit of observation. As Solomon (1984,p. 16) has pointed out, daily fluctuations around a "steady or, atleast, predictable" trend in the exchange rate may not dissuadetraders from making transactions because traders can choose topostpone or speed up foreign exchange conversions over briefperiods. It is also noteworthy that earlier empirical findings sug-gest that it is the longer-term (rather than weekly or daily) move-ments in exchange rates that might influence the decisions oftraders (see a review of the literature by Farrell, with De Rosa andCrown (1983)). Finally, because Akhtar and Hilton's theoreticalspecification excludes the more fundamental economic deter-minants of the exchange rate (and, thus, the basic cause of itsvariability) and of trade volumes and prices, the identification ofa robust empirical relationship between volatility and trade flowscould merely reflect the common effects of such omitted variableson these two factors rather than any causal link between them.Conclusive evidence on the effect of exchange rate volatility ontrade may, therefore, hinge on the specification of a markedlymore comprehensive model.

APPENDIX I

Data Definitions and Sources

This Appendix presents a description of the construction of, and the datasources for, the variables used in estimations. The data used in the estimationsreported for the United States and the Federal Republic of Germany for theperiod 1974-81 were those provided to the author by Akhtar and Hilton. Be-cause it was difficult to extend these series through 1983, new data series werecomputed by the author, for the entire 1973-83 period, using the same pro-cedures adopted for computing the data for France, Japan, and the UnitedKingdom (see Sections II and IV of the text). Although the computation


502 PADMA GOTUR

methods are broadly similar to those used by Akhtar and Hilton, the new seriesfor the United States and Germany do not appear to be strictly comparable tothose provided by those authors.

In general, all variables were constructed on a basis conceptually similar tothat in Akhtar and Hilton (1984a).

Import and Export Volume and Average Value Indices

Import and export volume and average value indices for manufactured goodswere used to denote the trade volume and price variables, respectively. Non-seasonally adjusted quantity and average value indices from the Organization forEconomic Cooperation and Development, Trade Series A (Paris: OECD, vari-ous issues) were used for Germany, Japan, and the United Kingdom. Trade vol-ume and price data for the United States were obtained from U.S. Departmentof Commerce, International Economic Indicators (Washington: GovernmentPrinting Office, various issues) and other Department of Commerce data bases.Of the four series for the United States, only the export volume data are pub-lished on a seasonally adjusted basis. For France, seasonally adjusted series ofexport and import value in constant prices were used because quantity indicescould not be obtained. These variables and the export and import price indiceswere obtained from Institut National de la Statistique et des Etudes Eco-nomiques, Les comptes nationaux trimestriels: Series tongues, 1963-83 (Paris:INSEE, 1984).

Exchange Rate Volatility

The exchange rate volatility variable was defined as the standard deviation ofthe daily observations of the Fund's MERM-weighted effective exchange rateindex within each quarter over the period from the second quarter of 1973through the fourth quarter of 1983. Observations for the exchange rate index forthe period from the first quarter of 1972 through the first quarter of 1973 wereobtained from the Treasurer's Department of the Fund. For this period, the stan-dard deviations were based on monthly observations of the effective exchangerate index within each quarter. The MERM weights used in computing the effec-tive exchange rate indices are described and listed in Artus and McGuirk (1981).

Domestic Real Income

The domestic real income variable for each country was denoted by an indexof each country's seasonally adjusted real gross national product (GNP), or grossdomestic product (GDP). These indices were obtained from a data base, main-tained in the Fund's Research Department, that comprises data from nationalsources provided by desk economists throughout the Fund. These data series areeither identical or quite similar to those maintained in International FinancialStatistics (IPS), International Monetary Fund (Washington).

Domestic Capacity Utilization _

A measurement of capacity utilization was obtained for each country by takingthe ratio of an index of seasonally adjusted industrial production to an indexreflecting trend growth in industrial production. The data for industrial pro-duction were taken from the IPS. Measures of trend growth were computed fordifferent sample periods.



Domestic Prices

The domestic price variable was represented by the wholesale price index formanufactured goods as reported in the IFS.

Foreign Real IncomeThe foreign real income variable was computed as a trade-weighted average

of the real GNP (GDP) indices for all industrial country and developing countrytrading partners. These data were derived from the GEE (Global EconomicEnvironment) data base, maintained in the Fund's Research Department. Theindustrial countries included in this data base are the United States, the UnitedKingdom, Germany, France, Japan, Canada, Italy, Austria, Belgium, Denmark,the Netherlands, Norway, Sweden, Switzerland, Spain, Ireland, Finland, Aus-tralia, and New Zealand. Both oil exporting and non-oil developing countries areincluded in the data base, with individual weights assigned to the larger countriesin these two groups.

Foreign Capacity Utilization and Prices

The foreign capacity utilization and foreign price variables were computed astrade-weighted averages of the corresponding indices for all industrial countrytrading partners. The data were drawn from the GEE data base (for the indus-trial production data underlying the capacity utilization variables) and from theIFS (for the wholesale price indices).

Base-year (1975) total export values were used in computing the trade weightsfor all of the three "foreign" variables described above.

Exchange Rates

Period average, bilateral exchange rates drawn from the IFS were used asconversion factors in the computation of the foreign real income and relativeprice variables.

APPENDIX II

Regression Results by Country

This Appendix provides complete regression results for each country thatcorrespond to those reported in Table 6 of the text. The f-statistics are given inparentheses under the coefficient estimates. All variables, except for the dummyvariables and the volatility variable, enter the equation in natural log form. Forsome equations, more than one polynomial lag specification yielded similar rootmean square errors. Only one specification is presented in these instances,because the coefficient estimates did not differ significantly for these alternativeestimations.

As can be seen from the estimates, the coefficient of the real income variableis of the right sign and is significant in most equations. The results for thecapacity utilization variable and the price variables are somewhat mixed. In a fewinstances, when a low Durbin-Watson (DW) test statistic was found, re-estimation using the iterative Cochrane-Orcutt (CO) procedure yielded an in-significant estimate of the autocorrelation coefficient. This finding may indicate


504 PADMA GOTUR

the possible misspecification of the structural equation rather than the presenceof serial correlation.

Variables were defined as follows:

XV Volume of manufacturing exportsMV Volume of manufacturing importsXP Price of manufacturing exports, in domestic currencyMP Price of manufacturing imports faced by domestic consumers, in

domestic currencyDl, Z)2, Z)3 Quarterly seasonal dummy variables

RDY Real domestic activity levelRFY Real foreign activity levelCUF Measure of foreign capacity utilization

CU Measure of domestic capacity utilizationS Nominal exchange rate volatility

RPl Ratio of price of manufacturing exports in foreign currency (do-mestic currency price times exchange rate) to price of foreignsubstitutes of these exports in foreign currency

RP2 Ratio of price of imports of manufacturing goods in domesticcurrency to price of domestically produced substitutes for im-ported manufacturing goods in domestic currency

RP3 Price of foreign substitutes for imports of manufacturing goods,in domestic currency (foreign currency price converted at theexchange rate)

PD Price of domestic substitutes for imported manufacturing goods,in domestic currency.

The regression results are arrayed by country in the following subsections. R2

is the adjusted coefficient of determination; e is the error term.

United States

Export volume equation, 1975-83:XVt = 7.9 + 1.57 O7F,-! + 0.10 RFY^ + 0.14 S - 0.82 RPl + 0.25 et-i

(4.72) (2.56) (0.22) (2.70) (4.05) (1.94)

DW = 1.90; R2 = 0.879

Lag Length (Quarters)

Variable 0 1

S

RPl -0.43(2.04)

0.03(2.54)

-0.26(3.16)

0.03(3.00)

-0.13(2.74)

0.03(2.54)

-0.04(0.46)

0.03(2.15)

0.01(0.13)

0.02(1.90)

0.03(0.37)

0.01(1.73)

Export price equation, 1975-83:

XPt = 1.30 + 0.002 Dl- 0.004 D2- 0.006 D3 + 0.01 O/,-i + 0.55 PA-i(2.97) (0.75) (1.00) (1.96) (0.14) (4.22)

+ 0.10 S + 0.17 RP3t-l + 0.85 et-i(4.16) (2.01) (36.1)

DW = 1.91; ̂ 2 - 0.993

2 3 4 5 6



Variable

5

0 1

— 0.01(2.14)

2

0.01(4.27)

Lag

3

0.02(4.72)

Length (Quarters)

4

0.02(4.75)

5

0.02(4.17)

6

0.01(3.17)

7

0.01(2.22)

8

0.004(1.50)

Import volume equation, 1975-83:MVt = 2.46 - 0.01 Dl + 0.04 D2 - 0.002 D3 + 2.33 RDYt-{

(1.74) (0.50) (2.14) (0.12) (15.4)

-0.68(0/F/ClOr-i - 0.02 5- 1.87 RP2(2.26) (1.36) (8.56)

DW = 1.72; R2 = 0.973


Variable

S

RP2

0

—

-0.31(2.08)

1

-0.004(0.54)

-0.37(5.66)

2

-0.01(1.13)

-0.39(8.03)

3

-0.01(1.30)

-0.36(5.21)

4

-0.004(0.97)

-0.28(3.93)

5

-0.003(0.77)

-0.17(3.30)

6

-0.002(0.66)

—

Import price equation, 1975-83:

MP = 0.10 + 0.01 Dl + 0.004 D2-0.002 D3-0.01 CUFt-i(0.32) (0.91) (0.62) (0.42) (0.07)

+ 0.41 PD,_(4.87)

DW = 2.04; R2 = 0

Variable 0 1

S — 0.003(0.80)

i - 0.005 S + 0.57 R(0.23)

.994

2

0.0004(0.11)

(6.86)

Lag Length

3

-0.001(0.35)

!P3,-! + 0.49 e,_,(3.70)

(Quarters)

4 5

-0.003 -0.003(0.57) (0.68)

6

-0.002(0.75)

Federal Republic of Germany

Export volume equation, 1975-83:

XVt = 4.11 - 0.05 Dl- 0.04 D2- 0.09 D3 + 0.52 CUF^ + 1.30 RFY^(2.17) (4.05) (3.17) (6.80) (1.88) (7.72)

- 0.12 S-1.14 KP1(2.55) (4.58)

DW = 1.60; R2 = 0.957

Re-estimation using the CO correction procedure yielded an insignificant auto-correlation coefficient (whose value was 0.10); therefore, the ordinary least-squares procedure was used for estimation.


506 PADMA GOTUR


Variable

5

RPl

0

—

-0.20(2.56)

1

-0.01(0.97)

-0.19(3.54)

2

-0.02(1.72)

-0.17(4.51)

3

-0.02(2.40)

-0.15(4.39)

4

-0.02(2.74)

-0.13(3.54)

5

-0.02(2.79)

-0.11(2.80)

6

-0.02(2.72)

-0.09(2.29)

7

-0.01(2.63)

-0.06(1.94)

8

-0.01(2.53)

-0.03(1.69)


XPt = 0.89 + 0.01 Dl - 0.001 D2 - 0.0001 D3 + 0.02 O/,-i + 0.88 P£>,-i(9.89) (1.82) (0.21) (0.05) (0.47) (10.32)

+ 0.01 5 - 0.07 RP3t-! + 0.38 et-i(0.58) (0.92) (2.39)

DW = 1.85; R2 = 0.994


Variable 0 1 2 3 4 5 6 7 8

S — 0.002 0.002 0.002 0.002 0.002 0.001 0.001 0.001(0.55) (0.62) (0.60) (0.55) (0.49) (0.45) (0.41) (0.38)

Import volume equation, 1975-83:

MVt = 12.1 - 0.03 Dl- 0.01 £>2- 0.09 Z>3 + 1.70 /M>y,-i(2.10) (1.95) (0.71) (6.08) (4.41)

+ 0.22 (CUFICU)<-i - 0.05 S - 3.29 RP2(0.29) (1.07) (3.76)

DW = 1.60; R2 = 0.975

Re-estimation using the CO correction procedure yielded an insignificant auto-correlation coefficient (whose value was 0.10); therefore, the ordinary least-squares procedure was used for estimation.


Variable

S — -0.006 -0.008 -0.008 -0.008 -0.008 -0.007 -0.005 -0.003(0.46) (0.75) (1.01) (1.12) (1.11) (1.05) (0.99) (0.94)

RP2 -0.24 -0.35 -0.43 -0.47 -0.48 -0.45 -0.39 -0.30 -0.17(0.88) (2.17) (3.72) (3.56) (2.94) (2.53) (2.27) (2.10) (1.98)


MPt = 0.89 + 0.02 Dl + 0.01 D2 + 0.01 D3 + 0.23 CUF,-i + 0.43 PA-i(7.50) (4.41) (2.53) (3.62) (2.72) (3.77)

-0.02 S + 0.37 RP3t-! + 0.35 et-i(0.70) (3.73) (2.84)

DW = 2.21', R2 = 0.991

0 1 2 3 4 5 6 7 8




Variable

5 _ -0.0002 -0.0003 -0.001 -0.002(0.04) (0.08) (0.28) (0.57)

FranceExport volume equation, 1975-83:

XV< = -0.34 + 0.34 CUFt-i + 1.52 RFYt-l(0.11) (1.23) (5.45)

- 0.44 RPI + 0.39^-i(1.03) (2.51)

DW = 1.19; R2 = 0.977

-0.004 -0.004 -0.004(0.88) (1.06) (1.09)

- 0.01 S(0.40)

-0.003(1.06)


Variable

S

RP1

0

—

-0.28(2.97)

1

0.005(0.83)

-0.18(2.61)

2

0.001(0.33)

-0.11(1.73)

3

-0.001(0.26)

-0.04(0.71)

4

-0.003(0.59)

0.003(0.05)

5

-0.004(0.75)

0.03(0.54)

6

-0.004(0.84)

0.05(0.86)

7

-0.004(0.88)

0.05(1.08)

8

-0.002(0.92)

0.03(1.24)


XP = 0.50 + 0.22 CUt-i + 0.57 PA-i + 0.04 S+0.32/^3,-1 + 0.19e,-i(11.33) (2.52) (8.57) (2.73) (5.05) (1.02)

DW = 1.85-,R2 = 0.997


Variable 0

5 —

1

-0.006(1.57)

2

0.0001(0.02)

3

0.004(2.15)

4

0.007(3.50)

5

0.009(3.98)

6

0.009(4.13)

7

0.007(4.17)

8

0.004(4.17)


MVt = -2.90 + 2.21 RDYt-t + 0.67 (CUFICU)^ + 0.05 S(0.48) (2.95) (1.34) (0.61)

- 0.59 RP2 + 0.64e,_!(1.07) (5.38)

DW = 1.82;R2 = 0.975


Variable

S

RP2

0

—

-0.20(1.56)

1

-0.01(1.29)

-0.15(1.73)

2

-0.003(0.26)

-0.11(1.46)

3

0.01(0.47)

-0.07(0.89)

4

0.01(0.80)

-0.04(0.47)

5

0.01(0.96)

-0.02(0.22)

6

0.01(1.06)

-0.01(0.06)

7

0.01(1.12)

0.003(0.04)

8

0.01(1.16)

0.005(0.12)

0 1 2 3 4 5 6 7 8


508 PADMA GOTUR


MP = 1.31 + 0.41 CUFt-! + 0.19 PD,-i + 0.03 S + 0.52 RPl^(28.7) (4.36) (2.49) (2.67) (7.26)

DW = 1.84; R2 = 0.995


Variable 0

5 —

1

-0.001(0.18)

2

0.002(0.86)

3

0.005(2.32)

4

0.006(3.01)

5

0.007(2.98)

6

0.006(2.83)

7

0.005(2.70)

8

0.003(2.60)

JapanExport volume equation, 1975-83:

XVt = 8.12 - 0.09 Dl - 0.04 D2 - 0.03 D3 + 0.01 C£/F,-i + 1.22 RFY,^(3.71) (8.45) (3.72) (3.37) (0.11) (5.85)

+ 0.03 S-1.96 ftPl(1.50) (7.08)

DW = 1.84; R2 = 0.991


Variable

S

RPl

0

—

-0.22(2.35)

1

-0.005(1.01)

-0.28(7.27)

2

-0.003(0.86)

-0.31(12.5)

3

0.0001(0.01)

-0.31(11.5)

4

0.004(1.24)

-0.28(10.7)

5

0.01(2.70)

-0.23(7.95)

6

0.01(3.77)

-0.17(4.67)

7

0.01(4.05)

-0.11(2.69)

8

0.01(3.91)

-0.05(1.60)

Export price equation, 1975-83:XP = 2.08-0.005 Dl + 0.01 D2-0,002 D3 + 0.22 CC/,-i + 0.21 PD,-i

(2.44) (0.52) (0.85) (0.25) (1.02) (0.79)

+ 0.03 S + 0.32 JRP3,-i + 0.65 et-i(1.33) (2.71) (5.13)

DW = 1.70; R2 = 0.942


Variable

5

0 1

— -0.005(1.14)

2

0.00003(0.01)

3

0.004(1.08)

4

0.006(1.69)

5

0.007(2.00)

6

0.007(2.17)

7

0.006(2.26)

8

0.004(2.32)

Import volume equation, 1975-83:MVt = 5.57 - 0.04 D1 +0.03 D2-0.02 D3 +1.11 KDY,-i

(1.97) (2.03) (1.60) (1.41) (3.47)

- 0.80 (CUF/CU)t-, + 0.09 S - 1.33 RP2 + 0.51 et-i(1.14) (2.01) (2.73) (3.96)

DW = 1.91;K2 = 0.967




Variable

5

RP2

0

—

-0.50(4.11)

1

-0.01(0.98)

-0.37(4.04)

2

-0.01(0.81)

-0.25(3.44)

3

0.002(0.23)

-0.16(2.34)

4

0.02(1.46)

-0.08(1.25)

5

0.02(2.61)

-0.03(0.43)

6

0.03(3.37)

0.01(0.13)

7

0.03(3.63)

0.02(0.52)

8

0.02(3.59)

0.02(0.80)


MPt = 2.70 + 0.01 D 1 + 0.003 D2 + 0.001 £>3 + 0.27 CUFt-i - 0.48FA_!(3.43) (0.48) (0.16) (0.06) (1.09) (1.59)

+ 0.06 5 + 0.89 RP3t-i + 0.30 et-i(2.43) (5.06) (1.74)

DW = 1.66; R2 = 0.906


Variable 0

S —

1

-0.004(0.61)

2

0.003(0.64)

3

0.008(2.10)

4

0.01(2.79)

5

0.01(2.95)

6

0.01(2.94)

7

0.01(2.91)

8

0.01(2.87)

United Kingdom

Export volume equation, 1975-83:

XVt = 1.74 - 0.05 £>! + 0.02 D2- 0.06 D3 + 0.22 CW-i + 1.44 RFYt^(1.86) (2.36) (1.06) (2.86) (0.54) (3.52)

+ 0.04 S-0.82 #P1(0.98) (3.26)

DW = 2.32; R2 = 0.654


Variable

5

RPl

0

—

-0.12(1.16)

1

0.002(0.23)

-0.12(2.10)

2

0.01(0.77)

-0.12(3.53)

3

0.01(1.04)

-0.11(2.86)

4

0.01(1.14)

-0.10(1.96)

5

0.01(1.18)

-0.09(1.49)

6

0.01(1.19)

-0.07(1.23)

7

—

-0.05(1.07)

8

—

-0.03(0.96)


XP< = 0.36 + 0.003 Dl - 0.003 D2 - 0.005 D3 + 0.18 CUt-i + 0.76 PA-i(2.34) (0.84) (0.94) (1.51) (2.04) (24.7)

- 0.01 5 + 0.17 flP3,_! + 0.57 e(0.68) (3.60) (5.17)

DW = 1.39: R2 = 0.997


510 PADMA GOTUR


Variable

5

0 1

— 0.005(1.91)

2

0.001(0.48)

3

-0.001(0.47)

4

-0.003(0.99)

5

-0.004(1.29)

6

-0.004(1.48)

7

-0.004(1.61)

8

-0.002(1.71)


MVt = -4.86 - 0.01 D1 + 0.06 D2- 0.03 D3 + 2.91 RDYt^(2.04) (0.03) (2.57) (1.26) (8.25)

- 0.54 (CUF/CU) ,-i + 0.07 S - 0.87 RP2(1.25) (1.31) (3.87)

DW = 1.89; R2 = 0.938


Variable

5

RP2

0

—

-0.02(1.48)

1

0.02(1.54)

-0.10(1.50)

2

0.02(1.53)

-0.15(4.46)

3

0.01(1.31)

-0.18(3.19)

4

0.01(1.05)

-0.18(2.45)

5

0.01(0.84)

-0.14(2.11)

6

0.005(0.68)

-0.09(1.92)

7

0.002(0.57)

—

Import price equation, 1975-83:MPt = 0.49 - 0.01 D 1-0.01 D2-0.01 D3

(4.24) (0.73) (0.81) (1.29)

+ 0.36 CUF^ + 0.34 PD - 0.01 5 + 0.57 RP3t-l(3.03) (10.37) (0.82) (11.85)

DW = 1.85;£2 = 0.993


Variable 0

5 —

1

0.01(2.17)

2

0.003(0.97)

3

-0.001(0.47)

4

-0.004(1.32)

5

-0.01(1.75)

6

-0.01(1.97)

7

-0.005(2.11)

8

-0.003(2.20)

REFERENCES

Akhtar, M.A., and R. Spence Hilton (1984a), "Exchange Rate Uncertainty andInternational Trade: Some Conceptual Issues and New Estimates forGermany and the United States," Research Paper No. 8403 (New York:Federal Reserve Bank of New York, May).

(1984b), "Effects of Exchange Rate Uncertainty on German and U.S.Trade," Federal Reserve Bank of New York Quarterly Review (New York),Vol. 9 (Spring), pp. 7-16.

Artus, Jacques R., and Anne Kenny McGuirk, "A Revised Version of the Multi-lateral Exchange Rate Model," Staff Papers, International Monetary Fund(Washington), Vol. 28 (June 1981), pp. 275-309.



Bank of England, "The Variability of Exchange Rates: Measurement and Ef-fects," Bank of England Quarterly Bulletin (London), Vol. 24 (September1984), pp. 346-49.

Brodsky, David A., "Fixed Versus Flexible Exchange Rates and the Mea-surement of Exchange Rate Instability," Journal of International Eco-nomics (Amsterdam), Vol. 16 (May 1984), pp. 295-310.

Farrell, Victoria S., with Dean A. DeRosa and T. Ashby McCown, Effects ofExchange Rate Variability on International Trade and Other Economic Vari-ables: A Review of the Literature, Staff Studies No. 130 (Washington: Boardof Governors of the Federal Reserve System, December 1983).

International Monetary Fund, Exchange Rate Volatility and World Trade: AStudy by the Research Department of the International Monetary Fund,Occasional Paper No. 28 (Washington, July 1984).

Judge, George G., William E. Griffiths, R. Carter Hill, and Tsoung-Chao Lee,The Theory and Practice of Econometrics (New York: Wiley, 1980).

Justice, G., The Impact of Exchange Rate Variability on International TradeFlows, Discussion Paper (Technical Series) No. 4 (London: Bank ofEngland, December 1983).

Kashyap, A.K., P.A.V.B. Swamy, J.S. Mehta, and R.D. Porter, EstimatingDistributed Lag Relationships Using Near-Minimax Procedures, SpecialStudies Paper No. 187 (Washington: Board of Governors of the FederalReserve System, Division of Research and Statistics, September 1984).

Kenen, Peter B., "Exchange Rate Instability: Measurement and Implications,"International Finance Section Research Memorandum (Princeton, New Jer-sey: Princeton University, June 1979).

, and Dani Rodrik, "Measuring and Analyzing the Effects of Short-TermVolatility in Real Exchange Rates," International Finance Section WorkingPaper in International Economics No. G84-01 (Princeton, New Jersey:Princeton University, March 1984).

Kmenta, Jan, Elements of Econometrics (New York: Macmillan, 1971).Lanyi, Anthony, The Case for Floating Exchange Rates Reconsidered, Essays in

International Finance No. 72 (Princeton, New Jersey: Princeton University,February 1969).

, and Esther C. Suss, "Exchange Rate Variability: Alternative Measuresand Interpretation," Staff Papers, International Monetary Fund (Washing-ton), Vol. 29 (December 1982), pp. 527-60.

Magee, Stephen P., "U.S. Import Prices in the Currency-Contract Period,"Brookings Papers on Economic Activity: 7(1974), The Brookings Institu-tion (Washington), pp. 117-64.

Page, S.A.B., "The Choice of Invoicing Currency in Merchandise Trade," Na-tional Institute Economic Review, National Institute of Economic and SocialResearch (London), No. 98 (November 1981), pp. 60-72.

Rana, Pradumna B., "Exchange Rate Risk Under Generalized Floating: EightAsian Countries," Journal of International Economics (Amsterdam),Vol. 11 (November 1981), pp. 459-66.


512 PADMA GOTUR

Schmidt, Peter, and Roger N. Waud, "The Almon Lag Technique and theMonetary Versus Fiscal Policy Debate," American Statistical AssociationJournal, Vol. 68 (March 1973), pp. 11-19.

Shiller, Robert J., "A Distributed Lag Estimator Derived from SmoothnessPriors," Econometrica (Evanston, Illinois), Vol. 41 (July 1973), pp. 775-88.

Solomon, Robert, The Consequences of Exchange-Rate Variability, BrookingsDiscussion Papers in International Economics No. 24 (Washington: TheBrookings Institution, December 1984).


Real Exchange Rates, Import

Penetration, and Protectionism

in Industrial Countries

ERIC V. CLIFTON*

protectionism have been widely discussed. The focus of muchof the attention has been on the role of real exchange rate move-ments in prompting protectionist pressures from management andunions in adversely affected industries. Nevertheless, despite thepopularity of the view that appreciating real exchange rates fosterprotectionism, little research attention has actually been paid tothe question, and virtually no empirical evidence to support orrefute this view has been forthcoming (see Bergsten and William-son (1983) and Corden (1984)). This paper is an attempt to fillpart of that gap in the literature.

The paper addresses the interaction between exchange rateappreciation and increased protection (specifically, nontarifftrade barriers) in terms of a two-stage process. In the first stage,an appreciating real exchange rate leads to increases in importpenetration beyond what would typically accompany the normalgrowth of international trade. In the second, rising import pene-tration leads to increased protection. Recent studies by Cline(1984) and others have established empirically the second stage ofthis process. The present paper tests for the occurrence of the firststage by estimating industry-level import penetration ratios asfunctions of industry-level real exchange rates and the level of acountry's aggregate international trade.

The plan of the paper is as follows. Section I describes themodel underlying the hypothesized relation between the import

*Mr. Clifton, an economist in the Western Hemisphere Department, holdsdegrees from the University of Kentucky and Indiana University. This paper wasprepared while he was with the Research Department.

513

THE CAUSES and effects of the recent trend toward increasedT


514 ERIC V. CLIFTON

penetration ratio and the real exchange rate and outlines theempirical tests used. Section II discusses the countries and indus-tries that are studied. Section III reviews the empirical results.Section IV discusses some implications of the results and showshow they can be used to assess some of the economic effects oftrade restrictions. A quantitative illustration is developed to showhow a policy to restrict import penetration into an industry willlead to measurable increases in economic rents for domestic pro-ducers in terms of a depreciation of the "shadow" real exchangerate for that industry.

I. The Model

Cline (1984) examined the determinants of nontariff quan-titative restrictions at the industry level in five major industrialcountries. He estimated protection functions for the UnitedStates, Canada, the United Kingdom, the Federal Republic ofGermany, and France and concluded that, on the basis of thesimilarity of the coefficients in the individual country equations,the protection process is broadly similar in all five countries. Healso found that, in general, the import penetration ratio is the keyvariable triggering protection, with its influence moderated to theextent that the home country is also an exporter of goods inthe same industrial category. Cline's "best estimates" (defined asthe set of explanatory variables for each equation that producedthe highest-percentage explanation of protection) of the protec-tion functions for Germany, the United Kingdom, and the UnitedStates included as major explanatory variables the import pene-tration ratio and an industry's share of the total manufacturinglabor force.

Whereas Cline's and other studies have shown that increasingimport penetration is an important determinant of protection, fewhave directly explored the variables that might lead to increases inthe import penetration ratio itself (although this issue is addressedindirectly, of course, by empirical studies of import demand func-tions). In addressing this question, this paper examines the role ofthe industry-specific real exchange rate in determining the importpenetration ratio. Throughout the analysis, the real exchange rateis defined as an index of the labor cost of a unit of gross output inthe relevant domestic industry divided by an index of the unit


REAL EXCHANGE RATES AND PROTECTIONISM 515

value of imports of those goods, with all values converted intodomestic currency.

In the underlying model it is assumed that in each countryconsumers—including both final consumers and sectors that usethe goods as intermediate inputs—purchase one product fromeach industry y. Apparent consumption, Ch consists of goodsproduced by foreign and domestic manufacturers as the goods M,and Dh respectively, where D, is domestic production, Ph minusexports, Eh or

where D; = P} - £,.The partial equilibrium proportion of the total purchases of

product/ that domestic residents will satisfy through imports, IP*,is assumed to be a function of the industry-specific real exchangerate indicator and the real level of the country's internationaltrade; that is,

where Rj is the real exchange rate for product/, and Tis the tradevariable, equal to the sum of real total imports plus total exports(rdoes not have a subscript because it applies to aggregate trade,not only to the industry's trade).1 Other factors that influence thetotal domestic demand for the product are assumed not to bedifferentiated between foreign and domestic goods.

Equation (2) represents a demand-determined relation be-tween the import penetration ratio, the real exchange rate, and

lrThis measure is similar to the measure of "real openness" devised byBeenstock and Warburton (1983). Their measure was real imports plus realexports, divided by real gross national product (GNP). Beenstock and Warbur-ton pointed out that this type of measure has the advantage of abstracting fromterms of trade movements and responding only to changes in trade volumes. Inthe real trade measure used in this study, real imports and exports are notdivided by real GNP because real GNP contains real net exports, which wouldbe correlated with the real exchange rate variable.

Over a very long sample period it would be appropriate to scale T because IPis theoretically bounded from above by unity. As IP is measured in practice (asin the Market Penetration System Data Base maintained by the World Bank'sEconomic Analysis and Projections Department), however, it is not boundedfrom above because of the possibility that imports are added to inventory or arere-exported. Over the sample period examined in this paper, the relation be-tween IP and T is approximately log-linear. Estimation of the model using ascaled version of T did not produce significantly different results.


516 ERIC V. CLIFTON

the level of trade. The real exchange rate variable reflects therelative prices of domestic and imported goods, which are in turnfunctions of their relative costs to the domestic market.2 Thisspecification also makes the simplifying assumption that move-ments in each industry's real effective exchange rate are ex-ogenous to the industry. Because the exchange rate is endogenousto the macroeconomy, this assumption is obviously a rather strongone, particularly in the long run. Although consumers are as-sumed to be indifferent to the source of Ch this indifference doesnot necessarily imply that M/ and D, must sell for the same price.As Goldstein and Khan (1984) have noted, many empirical studieshave shown that the "law of one price" does not appear to holdcontinuously within countries, even at disaggregated commoditylevels. Aside from having differences in quality, delivery dates,after-sales service, and other factors, a given product need not besold by two different suppliers for the same price in the short tomedium term because selling prices of a product reflect the costconditions of its respective producers, with the exact cost-pricerelations depending on the competitive structure of the industry.In a competitive market, of course, the low-cost producer wouldeventually drive out all competitors. But the low-cost producermay not drive out the competition in a flexible exchange ratesystem in which importers perceive that a foreign supplier's costadvantage (or disadvantage) is the temporary result of an over-valuation (or undervaluation) of the home country's currency inreal terms. Equation (2) is thus a composite function representinga relation between the share of imports in total domestic con-sumption of product Cj and the relative price of domesticallyproduced good; with respect to imported goody, which is in turna function of the costs of domestic producers relative to the costof the imported substitute.3

2 Such an approach makes the implicit assumption that the selling price of thedomestic good is a given proportional markup over the labor cost of production.Of course, restricting the indicator to labor cost alone misses other aspects of thecost of production. Labor cost is, however, a major element in determining theoverall cost structure and international competitiveness of the domestic industry.See International Monetary Fund (1984) for a discussion of real exchange ratemeasures based on unit labor cost.

3Equivalently, equation (2) can be viewed as the reduced-form relation thatsummarizes both the supply and the demand functions for the share of importsin the domestic consumption of product C, in the case where the elasticity of thesupply of imports is assumed to be infinite. As noted by Goldstein and Khan(1984), it is more plausible to argue that the elasticity of supply is infinite in thecase of a country's imports than in the case of its exports.



It is assumed that domestic purchasers do not respond immedi-ately to changes in their ex ante demand for imports relative tototal consumption because of adjustment costs associated withaltering their sources of supply. Letting IPt equal the actual im-port penetration ratio at time t (the j subscript is deleted below fornotational simplicity), the change in the import penetration ratiois assumed to follow4

where X is the coefficient of adjustment and IP* is the desiredvalue of the import penetration ratio. Corresponding to IP* inequation (2), IP? is a function of the real exchange rate and thelevel of real trade,

Substituting equation (4) into equation (3) and rearranging termsgives

Whereas several specifications of the relation between the importpenetration ratio and the real exchange rate were estimated, onlythe results of estimating equation (5) are reported in this paper.5

4 The issues associated with using this type of lag structure have been exten-sively covered in the literature, recently by Goldstein and Khan (1984). As notedin footnote 5, alternative specifications of the lag structure did not producesignificant variations in the estimates; the choice of a simple partial adjustmentmodel, therefore, appears to have been justified.

5 For example, a version of equation (5) with a dummy variable to account forthe effects of trade restrictions was estimated for each case, but the coefficientson the dummy variables were not significant except in the case of the Germanclothing industry (see Section III). A more complex form of the model, withequation (3) as

where A is the difference operator, was also estimated. This examination did notproduce significantly different estimates of Pj and p2. The specification of theadjustment process more simply as text equation (3), which has a geometricallydeclining lag, does not appear to have biased the estimates of Pi and p2 signifi-cantly. Similarly, forms of the model with equation (4) as

and

were also estimated and did not produce significantly different results.


518 ERIC V. CLIFTON

The hypotheses are that the signs of pi and (32 are positive, andthat X, the adjustment coefficient, is between zero and unity. Thecoefficient (J2 is the elasticity of the import penetration ratio withrespect to the real exchange rate. Given the definition of theimport penetration ratio—imports divided by domestic consump-tion—p2 is equal to the elasticity of imports with respect to thereal exchange rate, minus the elasticity of domestic consumptionwith respect to the real exchange rate, or

Thus the hypothesis that f}2 is greater than zero is equivalent to thehypothesis that em is greater than ec, with em assumed to be posi-tive. Alternatively, equation (6) can be expressed as

where eP and eE are the elasticities of domestic production andexports with respect to the real exchange rate. Thus the hypothe-sis that p2 is positive is equivalent to the hypothesis that

The left-hand term, em, is usually assumed to be positive. Esti-mates of the price elasticity of import demand contained in Stern,Francis, and Schumacher (1976) suggest that in the long run thedemand for imports in the countries and industries examined inthis paper is fairly responsive to changes in relative prices. Suchestimates do not necessarily imply, however, that import demandis responsive to movements in real exchange rates. In Inter-national Monetary Fund (1984), Jacques Artus and MalcolmKnight estimated the elasticity of the import volume of manu-factures with respect to a real exchange rate measure—relativenormalized unit labor costs—for a sample of industrial countriesthat included the three examined in this paper. They found that,in both the short run (less than six months) and the long run, theresponsiveness of imports to movements in relative normalizedunit labor costs was estimated to be positive, although not verylarge.

On the right-hand side of equation (8), e^ is usually assumed tobe negative; eP is also assumed to be negative.6 The rationale for

6 The findings of Artus and Knight in International Monetary Fund (1984,Table 2, p. 15) support the assumption that €E, as defined here, is negative.Results reported by Deardorff, Stern, and Greene (1979, Table 5.2, p. 136)support the assumption that eP, as defined here, is negative in the industriesexamined.



the assumption about eP is as follows. As their costs increase,domestic producers can either raise their prices or cut their profitmargins. Because of price competition from imports, at least partof the impact of increased production costs falls on profit margins,and in the long term increases in domestic costs of production leadto declines in domestic production. Regardless of the assumptionabout the sign of ep, however, equation (7) makes clear that thesign of p2 is an empirical question. Traditional assumptions fromtrade theory about the response of imports and exports to move-ments in real exchange rates do not automatically imply that theimport penetration ratio will respond in a certain direction toexchange rate changes.

II. Countries and Manufacturing Industries Examined

To examine empirically the determinants of the import pene-tration ratio, quarterly data from four manufacturing industries inthe United States, the United Kingdom, and the Federal Republicof Germany over the period 1963 through 1980 were used.7 Thesecountries were chosen for their importance in international tradeand because they experienced relatively large movements intheir real exchange rates over this period, as measured by rela-tive normalized unit labor costs in manufacturing, adjusted forexchange rate changes (see International Monetary Fund (1984,pp. 10-12)). In addition, in all of these countries there has beenan upsurge in protectionist sentiment over the past several years(see Cline (1983) for a discussion of recent developments in com-mercial policy). The manufacturing industries that were consid-ered are: textiles, clothing, iron and steel, and transport equip-ment.8 These industrial categories accounted for about 40 percentof world trade in manufactures in 1980 and have all been subjectto protectionist pressures over the past decade (see Anj aria andothers (1982)).

Charts 1-3 present annual data on import penetration ratiosand real exchange rates by industry. The data sources and thederivation of the variables used in these charts are described in

7 Slightly shorter time periods were used for the iron and steel sector (1963-79)and the transport equipment sector (1963-75) in the United Kingdom. Fordetails, see the Appendix.

8 These sectors are as defined at the two-digit Standard International TradeClassification (SITC) level and the three-digit International Standard IndustrialClassification (ISIC) level.


520 ERIC V. CLIFTON

Chart 1. Import Penetration Ratios and Real Exchange Rates by Industry,United States, 1963-80

Note: The real exchange rates are defined as an index of the labor cost ofa unit of domestic gross output in the industry divided by an index of the unitvalue of imports of those goods, with all values converted to domestic currency.An increase thus represents a real appreciation. Import penetration ratios arereal imports divided by apparent real consumption.

Source: Fund staff estimates. See the Appendix for notes on the data in thisgraph.

Real exchange rate(Left scale: Average value for 1963-80 = 100)

Import penetration ratio(Right scale: Percent)



Chart 2. Import Penetration Ratios and Real Exchange Rates by Industry,Federal Republic of Germany, 1963-80

Note: The real exchange rates are defined as an index of the labor cost of aunit of domestic gross output in the industry divided by an index of the unit valueof imports of those goods, with all values converted to domestic currency. Anincrease thus represents a real appreciation. Import penetration ratios are realimports divided by apparent real consumption.


Real exchange rate(Left scale: Average value for 1963-80 =100)

Import penetration ratio(Right scale: Percent)


522 ERIC V. CLIFTON

Chart 3. Import Penetration Ratios and Real Exchange Rates by Industry,United Kingdom, 1963-80

Real exchange rate(Left scale: Average value for 1963-80 =100)

- — — Import penetration ratio(Right scale: Percent)

Note: The real exchange rates are defined as an index of the labor cost of aunit of domestic gross output in the industry divided by an index of the unit valueof imports of those goods, with all values converted to domestic currency. Anincrease thus represents a real appreciation. Import penetration ratios are realimports divided by apparent real consumption.




detail in the Appendix. In brief, industry-level real exchange rateindicators are defined as an index of the labor cost of a unit ofgross output in the relevant domestic industry divided by an indexof the unit value of imports of those goods, with all values con-verted into the domestic currency. Thus a decline in the solid linein Charts 1-3 indicates a depreciation of the home country's realexchange rate for the relevant industry. The import penetrationratios used are equal to real imports in each sector divided byapparent real consumption, which is defined as gross productionplus imports minus exports. The import penetration ratios are inreal terms, so that changes in the ratios represent changes inthe volume of consumption accounted for by imports, not justchanges in the price of imports.

In general, over the 1963-80 period import penetration ratioswere increasing for all the countries considered, with the U.S.textile industry the only exception. As regards the measures ofindustry-level real exchange rates, the indicators for the UnitedStates were in general depreciating over the period, whereas thosefor Germany were in general appreciating. The real exchange ratemeasures for the United Kingdom fluctuated widely during1963-80 but on balance appear to have shown relatively lesschange than in the other two countries.

III. Estimation Results

Equation (5) was estimated for each of the four industries andthree countries in the study on the basis of a sample of 72 quarterlyobservations extending over the period 1963 through 1980 (asnoted earlier, slightly shorter periods were used for two Britishindustries). Estimates of equation (5), which are presented inTable 1, provide support for the hypothesis that increases in im-port penetration are positively related to appreciating real ex-change rates. All variables, except dummy variables, are in theform of natural logarithms. These 12 equations were indepen-dently estimated, using a nonlinear estimation technique for iter-ative minimum distance estimation as implemented by Berndt andothers (1974), and quadratic interpolation was used as the searchmethod for determining the parameter step size at each iteration.Three seasonal dummy variables were used in the estimations,although these coefficients have been omitted from Table 1. Therewas no evidence of significant serial correlation, and thus no cor-


524 ERIC V. CLIFTON

Table 1. Estimated Import Penetration Equations, 1963-80

(In IPt = \PO + \pi In Tt + \& In Rt + (1 - X) In 7P,_0

Estimated Coefficient

Country andIndustry

United StatesTextiles

Clothing

Iron and steel

Transport equipment

\

0.320(3.94)0.395

(4.96)0.279

(3.78)0.105

(2.90)

Po

-5.458(2.75)

-5.479(4.56)

-9.810(3.16)

-15.806(2.89)

Pi

0.585(2.68)1.475

(10.87)1.170

(4.82)2.110

(4.36)

Dummy(32 variable

0.845 —(4.06)0.300 —

(2.34)1.410 —

(3.17)1.856 —

(2.52)

R2

Level

0.787

0.971

0.882

0.983

Change

0.342

0.323

0.282

0.105

Germany, Fed. Rep. ofTextiles

Clothing

Iron and steel

Transport equipment

0.257(4.29)0.386

(5.67)0.192

(2.83)0.434

(5.02)

-3.779(2.76)

-3.627(6.10)

-3.490(3-55)

-4.437(3.98)

0.768(5.40)1.455

(11.34)0.409

(1.69)0.789

(3.11)

0.792(1.89)0.187

(0.81)0.907

(2.26)0.925

(1.92)

-0.0711

(3.13)

-0.1381

(3.63)

0.986 0.602

0.996

0.977

0.669

0.381

0.955 0.621

United KingdomTextiles

Clothing

Iron and steel2

Transport equipment2

0.267(3.88)0.570

(5.45)0.470

(4.71)0.321

(3.10)

-39.065(5.73)

-31.222(3.44)

-6.815(1.76)

-24.121(3.81)

4.894(15.28)

7.036(10.62)

4.250(5.85)4.479

(3.71)

4.243 —(3.31)0.568 —

(0.38)-2.262 —(1.81)1.602 —

(0.72)

0.983

0.899

0.762

0.930

0.294

0.316

0.353

0.290

Note: The numbers in parentheses are the ratio of the parameter estimate tothe standard error for that parameter and are asymptotically normally distrib-uted. All variables are defined as natural logarithms. The equations were j^sti-mated with seasonal dummy variables, which are omitted from this table. R2 isthe adjusted coefficient of determination.

1 These dummy variables are to take account of restrictive trade actions thatoccurred during the sample period and other special factors. See the Appendixfor details.

2The time period for the U.K. iron and steel industry is 1963-79. The timeperiod for the U.K. transport industry is the first quarter of 1963 through the firstquarter of 1976.

rection was made for it; the tests for autocorrelation were basedon the /i-statistic and the Durbin-Watson bounds test.

In 8 of 12 individual cases the coefficient p2 has the expectedpositive sign and is significant at least at the 10 percent level. Inanother three cases (32 has its expected positive sign but is notstatistically significant. The evidence of a link between the real



exchange rate and the import penetration ratio is most persuasivein the cases of the United States and Germany. (The tests in thispaper do not establish "causality" between the variables; how-ever, it seems unlikely that increasing import penetration wouldcause the real exchange rate to appreciate.) For the United King-dom, the link is in general less evident. In addition, the estimatesof equation (5) provide strong evidence that increases in importpenetration are related to the level of the country's internationaltrade. In all 12 cases, the sign of fr is positive at statisticallysignificant levels.

In the last two columns of Table 1 a£e presented values of theadjusted coefficient of determination, /?2, for each equation. Thefirst of these columns, entitled "Level," shows the portion of thetotal variance of the levels of the import penetration ratio that theestimated equations are able to explain. Based on F-statistics,these equations are all significant at the 1 percent level. Thesecond of the columns, entitled "Change," contains 7?2s for aversion of each equation in which the dependent variable is thechange in In IP at time t. The R2s based on changes in the de-pendent variable provide an indication of the portion of the vari-ance of changes in the import penetration ratio that the estimatedequations are able to explain. (See Boughton (1984) for a similaruse of R2s for levels and changes.) On the basis of /^-statistics, theequations are all significant at the 1 percent level except the equa-tion for the U.S. transport equipment industry, which is signifi-cant at the 5 percent level.

Table 2 presents information on the mean time lag of the re-sponse of the import penetration ratio (that is, the time requiredfor just over 60 percent of the adjustment to be completed), alongwith estimates of the long- and short-run elasticities with respectto the real trade variable and the real exchange rate. Concerningthe mean lags, Goldstein and Khan (1984) reported that studies ofimport demand equations that have used Koyck-type lag struc-tures have found that most of the adjustment to price changesoccurs within a year. The median lag of the equations in Table 1is about six months, so that the results are basically consistent withearlier empirical work.

Artus and Knight in International Monetary Fund (1984) esti-mated impact and long-run elasticities of import volume withrespect to relative normalized unit labor costs for the UnitedStates, Germany, and the United Kingdom using semiannualdata. Although p2 is not the same as the elasticity of demand for


526 ERIC V. CLIFTON

Table 2. Mean Lags and Long- and Short-Run Elasticities

Mean Lag1 Real Trade Variable2 Real Exchange Rate2

Country and Industry

United StatesTextilesClothingIron and steelTransport equipment

Germany, Fed. Rep. ofTextilesClothingIron and steelTransport equipment

United KingdomTextilesClothingIron and steelTransport equipment

(In quarters)

2.11.52.68.5

2.91.64.21.3

2.70.81.12.1

Long run

0.5851.4751.1702.110

0.7681.4550.4090.789

4.8947.0364.2504.479

Short run

0.1880.5830.3270.222

0.1970.5620.0790.342

1.3064.0071.9971.436

Long run

0.8450.3001.4101.856

0.7920.1870.9070.925

4.2430.568

-2.2621.602

Short run

0.2710.1190.3940.195

0.2030.0720.1740.401

1.1320.324

-1.0630.514

Source: Table 1.TThe formula for the mean time lag is (1 -2 The short-run elasticities are \p/ and the long-run elasticities are p/. The

short-run elasticity shows the percentage change in the import penetration ratioduring the first quarter for a given percentage change in an independent variable.The long-run elasticity gives the total percentage change in the import pene-tration ratio after all adjustment has taken place.

imports with respect to the real exchange rate (see equation (7)),the difference between these two measures will be small if totaldomestic consumption is not highly responsive to changes in thereal exchange rate. The results reported in Table 2 for the UnitedStates and Germany are similar to those reported in InternationalMonetary Fund (1984), but the results for the United Kingdom inTable 2 are in general larger in absolute value.

For the United States, the results in Table 1 strongly support thehypothesis that the real value of trade and the real exchange rateare significant determinants of the import penetration ratio, atleast for these industries. Based on two-sided tests for the signifi-cance of "r-statistics," the coefficients for pi are all significant atthe 1 percent level, as are the (32 coefficients for the textile andiron and steel industries.9 The (32 coefficients for the clothing andtransport equipment industries are significant at the 5 percentlevel.

As discussed earlier, Cline (1984) found a significant relation

9 The statistic that is referred to as the "f-statistic" is actually asymptoticallynormally distributed.



between the import penetration ratio and protection in the UnitedStates. Thus the evidence in Table 1, taken in conjunction with theCline study, tends to support the hypothesis that appreciation inindustry-level real exchange rates may indeed lead to increasedprotection in the United States. The results also suggest, however,that the growth and increasing openness of U.S. trade can lead toincreases in import penetration even if the real exchange rate isdepreciating. Section IV uses the empirical results for the U.S.transport equipment industry from Table 1 to illustrate the effectsof a hypothetical trade restriction.

For Germany, the results in Table 1 also support, although notas strongly, the hypothesis that the real level of international tradeand the real exchange rate are significant determinants of theimport penetration ratio. The "^-statistics" for & are significant atthe 1 percent level except in the iron and steel industry, where thesignificance level is 10 percent; for (32, the coefficient is signifi-cantly different from zero at the 5 percent level in the iron andsteel industry and at the 10 percent level for the textile and trans-port equipment industries. In two of the German industries,dummy variables were used to account for special factors. (See theAppendix for further details on the dummy variables.) In the caseof the clothing industry, the special factor was the Multi-FiberArrangement (MFA) that went into effect on January 1, 1974.Experiments using a similar dummy variable in the cases of theUnited Kingdom and the United States did not yield significantcoefficients for the dummy variables. It may be that the effects ofthe MFA were, in terms of the estimated equations, relativelymore severe for Germany, which is a much less important pro-ducer of clothing and textiles than the United Kingdom and theUnited States. In the case of the transport equipment industry,the special factor was rising oil prices; the significance of thepositive sign of (32 for the industry is dependent on the inclusionof this dummy variable. Cline (1984) found a statistically signifi-cant relation between the import penetration ratio and protec-tionist actions in Germany. Thus, as in the case of the UnitedStates, the results in Table 1 tend to support the hypothesis of apositive relation between the real exchange rate and protection.

For the United Kingdom, the results in Table 1 do not providemuch support for the hypothesis that the real exchange rate is adeterminant of the import penetration ratio. Only in the textileindustry is the sign of p2 significantly positive (at the 1 percentlevel). In contrast, the level of British trade does appear to be an


528 ERIC V. CLIFTON

important determinant of the import penetration ratios for thevarious industries. The Pi coefficients are all significant at the1 percent level. The failure to uncover a significant link betweenmovements in industry-level real exchange rates and import pene-tration at the industry level may be due to data problems thatwere encountered in this case. Alternatively, as can be seen fromChart 3, the British industry-level real exchange rates showedrelatively less change over the sample period compared with thosein the United States and Germany, and this fact may have compli-cated the estimation procedure. Of interest is that Cline (1984)did not uncover a significant link between British nontariff tradebarriers and import penetration, although the estimated relationwas positive.

IV. Implications of the Analysis

This section illustrates how estimates such as those presented inTable 1 might be used as a tool for examining the effects of aquantitiative limitation on imports. Specifically, the analysisshows that it is possible to estimate the "shadow" real exchangerate associated with any given level of quantitative trade re-strictions by solving the estimated equation (5) for the level of thereal exchange rate when given a fixed import penetration ratio.Comparison of the actual exchange rate with the (depreciated)shadow exchange rate then gives a quantitative impression of theeconomic effect of the quota.10 For example, suppose that in thefourth quarter of 1980 a decision had been taken to limit thepenetration of imports into the U.S. transport equipment industryto 10 percent of the domestic market for the next five years (theactual penetration of imports in the fourth quarter of 1980 wasabout 14.5 percent in this U.S. industry). In this example, the year1980 was chosen because it coincides with the end of the esti-mation period for the results in Table 1, and it enables one toexamine what the effects of the trade restriction would have been

10 The difference between the shadow real exchange rate and the actual realexchange rate can provide a general impression of the relative severity of protec-tion in various industries. For example, import penetration in two industriesmight be limited to the same level, but, because of differences between thesupply and demand characteristics of the two industries, one receives substan-tially greater protection. That more protected industry would have a shadow realexchange rate that was relatively more depreciated from its actual real exchangerate.



over 1981-84. Although this restriction is hypothetical, it isroughly analogous to the proposed U.S. Fair Trade in Steel Act of1984. In the debate in the U.S. Congress, this Act was proposedto limit U.S. imports of carbon steel to 15 percent of the domesticmarket for five years, compared with an actual ratio of 20.5 per-cent in 1984 (The New York Times, August 23, 1984).

By using the estimated equation (5) for the U.S. transportequipment industry, an implied value of the real exchange rateover the period from 1981 through mid-1984 can be derived on thebasis of the assumption that the import penetration ratio is heldfixed at 10 percent.11 In Chart 4 the shadow exchange rate isextrapolated from end-1980 through mid-1984 against an esti-mated measure of the actual movement of the real exchange ratefor the industry over this period. This experiment suggests that thetrade restriction would have had little effect on the real exchangerate at first, since the recession in the United States at the time ledto a decline in the demand for imports. After early 1983, however,the continuing real appreciation of the U.S. dollar and thestrengthening of the U.S. economy would have caused the de-mand for imports to increase, and the import restriction wouldhave corresponded to a depreciating real exchange rate for thetransport equipment industry. By mid-1984 the shadow real ex-change rate implied by the trade restriction is estimated to beabout 25 percent below the level it actually held without therestriction.

If it is assumed that the trade restriction is imposed in a sectorfor which home-country consumption is small relative to the sizeof the total market, then the restriction would cause a correspond-ing increase in the domestic price of the good. As is well known,the higher domestic selling price would cause economic rents to begenerated for domestic producers and for those foreign producersallowed to fill the quota. Given that the actual value of the real ex-change rate appreciated from 1981 through mid-1984 in Chart 4,the rent generated for domestic producers would have graduallyrisen to the equivalent of as much as 25 percent of the labor costof gross output in mid-1984.

The short-run effect of the limitation on import penetration in

11 Because of the variable T in the estimated equation, the severity of theimport restriction can change over time even if the restriction remains fixed. Forexample, changes in the overall propensity to import would change the strin-gency of the import restriction even if factors specific to the industry remainunchanged.


530 ERIC V. CLIFTON

Chart 4. Industry-Level Estimated Real Exchange Rates,U.S. Transport Equipment, 1981-mid-1984

Note: The real exchange rates are defined as an estimate of unit labor cost inthe industry divided by the unit value of imports of those goods, with all valuesconverted to domestic currency.


1 Estimated on the assumption of no changes in trade restrictions.2 Estimated on the assumption of an import penetration ratio fixed at

10 percent.

Index (first quarter of 1981 = 100)



the U.S. transport equipment industry can be simply illustrated.Figure 1 presents a partial equilibrium picture of this industry inthe short run. The initial price of the good is P0 (the initial pricesof the imported and domestic goods are assumed to be the same).

Figure 1. Illustration of a Restriction on Import Penetration,U. S. Transport Equipment Industry

At this price, domestic consumption would be C0, and domesticproduction (excluding exports) would be D0. The import pene-tration ratio would be IP0 = (C0 - D0)/C0. The trade restrictionis to set IPi equal to a constant less than IP0. Given the supply and


532 ERIC V. CLIFTON

demand conditions in Figure 1, this implies Cx and DI, and theinternal price rises to Pl for the given external price P0. (Ofcourse, the demand and supply conditions could be such that animport penetration ratio would imply several different possiblecombinations of C, D, and P.) Thus, in the short run the quan-titative restriction bestows economic rents on domestic producerswhenever IP1 is less than IP0. That is, as the domestic selling pricerises, so do the sizes of the rents that are the result of the quan-titative restriction on imports. The estimate of the economic rentaccruing to domestic producers in the short run is equivalent tothe area PÂBP^ in Figure 1.

The quantitative restriction causes the domestic selling price ofimports to rise (it is assumed to remain fixed externally), allowingdomestic producers to raise their prices (although by less than therise in the price of imports) and to capture a larger share of thedomestic market. The quantitative restriction allows domesticproducers to earn economic rents, although the rents may declinein the long run depending on the cost structure and conditions ofentry into the domestic industry.

The exercise above is meant only as an illustration and not as ananalysis of any actual or proposed U.S. trade policy.12 None-theless, it is interesting to compare the example with the historyof the voluntary restraints on Japanese automobile exports to theUnited States that took effect in 1981.13 This program limitedexports of Japanese passenger cars to the United States to a rateabout 8 percent below the 1980 level. After the imposition of thisrestraint, U.S. automobile producers were able to increase theirprofits to record levels in 1983. The sales volume for the domesticU.S. auto industry in 1983 was similar to that in 1980, a year whenthe industry had incurred large losses.

The estimates of the elasticities of the import penetration ratiowith respect to the real exchange rate in Table 2 provide someindication of where the greatest economic rents would be gen-erated by quantitative restrictions. The more inelastic is the re-

12The U.S. transport equipment industry was chosen for the example becausefor it the effects of the posited import restrictions would be particularly severe.In some other cases, a restriction on import penetration might have had littleeffect over the period from 1981 through mid-1984.

13The illustration is based on the entire U.S. transport equipment industry,not just on automobiles; the decline in imports in the example is much largerthan the decline that actually occurred in the U.S. automobile industry.



lation between the real exchange rate and the import penetrationratio, the more a given quantitative restriction on import pene-tration will cause a divergence between external and internalprices. The estimates for the United States and Germany suggestthat the economic rents generated over the long run by a quantita-tive restriction would likely be the greatest in the textile andclothing industries (with it assumed that cost structures and condi-tions of entry are the same across countries and industries).

It is also interesting that the empirical estimates reported in theprevious section are at least broadly consistent with the popularview that exchange rate volatility can lead to a steady increase inprotectionist pressure, even if the average level of a country's realexchange rate does not appreciate over time. Such heightenedprotection can happen because, although an appreciation of anexchange rate may lead to protectionist actions, a depreciation isunlikely to generate moves to dismantle existing impediments totrade. If this is the case, real exchange rate fluctuations may leadto the absolute level of protection "ratcheting" continuallyhigher, even if the level of the real exchange rate shows littlechange in the long run.14 The effects of exchange rate volatility onprotection have been discussed elsewhere, but the relation usuallyhas been described somewhat differently. For example, Anjariaand others (1982; see also Bergsten and Cline (1983)) discussedthe possibility that increasing exchange rate volatility might leadfirms to experience difficulties in competing internationally be-cause of rapid shifts in exchange rates. In the struggle to survive,these firms might seek protection from imports.

Finally, although the estimates in Table 1 suggest that exchangerate developments may explain some of the pressure for pro-tection, these developments do not justify either generalized orsectoral protection. Moreover, given the recent drift toward pro-tectionism, any extrapolation based on past trends in the associ-ation between import penetration and protectionism may not holdin the future. Protectionist measures in the industrial countries,being the outcome of policy deliberations, are not influenced in amechanical way by the import penetration ratio or by any othervariable.

14 The "ratchet" effect discussed here is also described in Bergsten andWilliamson (1983) and in Corden (1984). A formal test of this hypothesis is notattempted here (or in the two works cited).


534 ERIC V. CLIFTON

APPENDIX

Data Sources and Definitions

This Appendix provides further details on the sources of the data and on themethods used to calculate the real exchange rates and import penetration ratiosfor the various industries examined. The real exchange rate indicators for eachindustry were constructed by using data on labor costs in the following Inter-national Standard Industrial Classification (ISIC) categories:

321 Textiles322 Wearing apparel (except footwear)371 Iron and steel basic industries384 Transport equipment.

Index numbers for the labor cost of a unit of gross output (LCU) in localcurrency were derived from data on labor cost and production in United Nations,Yearbook of Industrial Statistics, Vol. 1—General Industrial Statistics (New York,various issues). Industry-specific data were available only on an annual basis.The yearly data were benchmarked by using the procedure of Denton (1971) toform quarterly data. The benchmark series comprised data on normalized unitlabor costs for each country's entire manufacturing sector.

Each LCU index series was divided by an index-series estimate of the unitvalue of imports in that industry. For the United States, the unit value of importswas an aggregate measure taken from the International Monetary Fund's Inter-national Financial Statistics (IPS) (Washington, various issues). For the FederalRepublic of Germany, data on import unit values in the Standard InternationalTrade Classification (SITC) categories 6, 7, and 8 were used. These data weretaken from Federal Statistical Office, Foreign Trade Series 5, Special TradeAccording to the Classification for Statistics and Tariffs (Wiesbaden, Germany,various issues), and Foreign Trade According to the Standard International TradeClassification (SITC—Rev. II)—Special Trade (Wiesbaden, various issues). Forthe United Kingdom, data on import values in disaggregated SITC sectors weretaken from Central Statistical Office, Monthly Digest of Statistics (London), foryears after 1971. Before 1971, data on the aggregate unit value of imports weretaken from the IFS.

The import penetration ratios were calculated as real imports divided by realdomestic consumption, which was defined as gross output minus exports plusimports. The trade data covered the following SITC Revision I categories:

65 Textile, yarn, fabrics, and so forth84 Clothing67 Iron and steel73 Transport equipment.

Except for minor differences, these categories are equivalent to the previouslymentioned ISIC categories. Annual trade data from the Organization for Eco-nomic Cooperation and Development, Trade Series C (Paris: OECD, variousissues) were used in forming the import penetration ratios. These data werebenchmarked to form quarterly data using the procedure of Denton (1971). Thebenchmark data comprised quarterly trade data from the previously mentionedGerman and British statistical publications and the OECD's Trade Series A



(Paris, various issues). In the case of the United Kingdom, the quarterly importpenetration ratios for the iron and steel industry in the year 1980 and for thetransport equipment industry in the years 1976-78 proved unreliable; hence theywere omitted from the estimations in Table 1 of the text. Estimations for thesetwo industries using annual data over the period 1963-80 did not produce signifi-cantly different results from those reported in Table 1. In Chart 3, the graph ofthe import penetration ratio for the transport equipment industry is interpolatedbetween 1975 and 1979. The data on production for all three countries weretaken from the previously mentioned United Nations publication and theOECD's Indicators of Industrial Activity (Paris, various issues).

The trade variable used is an index series of the volume of merchandiseimports plus exports. The data were taken from the IPS.

Dummy variables were used in two of the estimated equations for Germany.For the clothing industry, a dummy variable was used to account for the Multi-Fiber Arrangement (MFA) that went into effect on January 1,1974. Before thatdate the dummy variable is equal to unity; afterward it is equal to zero. For theGerman transport equipment industry, a dummy variable was used to accountfor rising oil prices. The dummy variable is equal to zero through the fourthquarter of 1973. Afterward it equals 1 except in the third quarter of 1978, thirdquarter of 1979, and from the second quarter of 1980 through the fourth quarterof 1980, when it equals 2.

REFERENCES

Anjaria, Shailendra J., Zubair Iqbal, Naheed Kirmani, and Lorenzo L. Perez,Developments in International Trade Policy, Occasional Paper No. 16(Washington: International Monetary Fund, November 1982).

Beenstock, Michael, and Peter Warburton, "Long-Term Trends in EconomicOpenness in the United Kingdom and the United States," Oxford Eco-nomic Papers (London), Vol. 35 (March 1983), pp. 130-42.

Bergsten, C. Fred, and William R. Cline, "Trade Policy in the 1980s: An Over-view," in Trade Policy in the 1980s, ed. by William R. Cline (Washington:Institute for International Economics, 1983), pp. 59-98.

Bergsten, C. Fred, and John Williamson, "Exchange Rates and Trade Policy,"in Trade Policy in the 1980s, ed. by William R. Cline (Washington: Institutefor International Economics, 1983), pp. 99-120.

Berndt, E.K., B.H. Hall, R.E. Hall, and J.A. Hausman, "Estimation and In-ference in Nonlinear Structural Models," Annals of Economic and SocialMeasurement (New York), Vol. 3 (October 1974), pp. 653-65.

Boughton, James M., "Exchange Rate Movements and Adjustment in FinancialMarkets: Quarterly Estimates for Major Currencies," Staff Papers,International Monetary Fund (Washington), Vol. 31 (September 1984),pp. 445-68.

Cline, William R., ed., Trade Policy in the 1980s (Washington: Institute forInternational Economics, 1983).

, Exports of Manufactures from Developing Countries: Performance andProspects for Market Access (Washington: The Brookings Institution,1984).


536 ERIC V. CLIFTON

Corden, W.M., The Revival of Protectionism, Occasional Paper No. 14 (NewYork: Group of Thirty, 1984).

Deardorff, Alan V., Robert M. Stern, and Mark N. Greene, "The Sensitivity ofIndustrial Output and Employment to Exchange Rate Changes," in Tradeand Payments Adjustment Under Flexible Exchange Rates, ed. by John P.Martin and Alasdair Smith (London: Macmillan for the Trade Policy Re-search Centre, 1979).

Denton, Frank T., "Adjustment of Monthly or Quarterly Series to AnnualTotal: An Approach Based on Quadratic Minimization," American Statisti-cal Association Journal (Washington), Vol. 66 (March 1971), pp. 99-102.

Goldstein, Morris, and Mohsin S. Khan, "Income and Price Effects in ForeignTrade," in Handbook of International Economics, Vol. 2, ed. by Ronald W.Jones and Peter B. Kenen (Amsterdam: North-Holland, 1984; New York:Elsevier, 1984), pp. 1041-1105.

International Monetary Fund, Issues in the Assessment of the Exchange Rates ofIndustrial Countries: A Study by the Research Department of the Inter-national Monetary Fund, Occasional Paper No. 29 (Washington, July 1984).

Stern, Robert M., Jonathan Francis, and Bruce Schumacher, Price Elasticities inInternational Trade (London: Macmillan for the Trade Policy ResearchCentre, 1976).


SUMMARIES

Interest Rate Determination in Developing Countries: A Conceptual Frame-work—SEBASTIAN EDWARDS and MOHSIN S. KHAN (pages 377-403)

As some developing countries move toward more liberalized financial systems,policymakers in these countries face the question of how interest rates respondto foreign influences and domestic policies. Most existing studies of interest ratestypically treat only the extreme cases, either of a fully open economy, in whichsome form of interest rate arbitrage holds, or of a completely closed economy,in which interest rates are determined solely by domestic monetary factors.Developing economies, however, in general fall somewhere between these twoextremes, so that the standard models of interest rate determination would notseem to be relevant to them.

The purpose of this paper is to outline a theoretical framework that can serveas a starting point for analyzing interest rate behavior in those developing coun-tries that are in the process of removing controls on the financial sector andrestrictions on capital flows. The approach suggested here combines elements ofmodels developed for both closed and open economies; thus it is able to incor-porate the influences on domestic interest rates of foreign interest rates, ex-pected changes in exchange rates, and domestic monetary developments. Aninteresting feature of the model presented is that the approximate degree offinancial openness, defined as the extent to which domestic interest rates arelinked to foreign interest rates, can be determined from the data of the countryanalyzed.

To illustrate the empirical validity of the proposed model, it was applied to twocountries—Colombia and Singapore. These two countries are quite different intheir degrees of financial development and openness, and thus they provide auseful first test of the general nature of the model. The model is able to representboth these cases quite adequately. The estimates indicate that in Colombia bothforeign and domestic factors are important, whereas domestic interest rates inSingapore are fully determined by foreign interest rates and by variations in theexchange rate. These results are precisely those expected, given the character-istics of the respective financial systems in the two countries.

Exchange Rate Changes and Exports of Selected Japanese Industries—DANIELCITRIN (pages 404-29)

This paper presents the results of an empirical investigation of the impact overtime of exchange rate variation on export prices and volumes of three majorJapanese industries—motor vehicles, consumer electronics, and iron and steel.Statistical analysis is based on a theoretical model that specifies the followingdynamic elements in the adjustment of export prices and volumes: the extent towhich, over time, suppliers pass through an exchange rate change to the price ofexports quoted in foreign currencies; the impact of relative export price ex-

537


538 INTERNATIONAL MONETARY FUND STAFF PAPERS

pectations on the response of export orders (demand) to relative export pricechanges; lags in the adjustment of export orders (demand) to relative exportprice changes; and delivery lags between orders and shipments.

Since doubt on the efficiency of the exchange rate as a tool for achievingadjustment in Japan's trade account may be a motive for the adoption of variousprotectionist measures by some of Japan's major trading partners, determiningthe extent to which the exchange rate influences trade flows is especially relevantin the Japanese context. Consequently, the study looks at five specific exportproducts: subcompact passenger cars, color television sets, and, in the steelindustry, heavy steel plate, tin plate, and galvanized steel sheet. Analysis on adisaggregate basis reflects theoretical considerations and a recognition that ten-sions between Japan and its trading partners have tended to focus on specificindustries in recent years.

The statistical analysis suggests several conclusions. For all five Japaneseexport items, the impact of exchange rate changes on exports is subject to lagson both the supply and the demand side, with the adjustment of demand beingparticularly slow. Furthermore, prices of foreign competitors appear to have asignificant influence on the pricing decisions of suppliers. As a result, in the longrun as well as in the short run, exchange rate changes are not fully passed throughto export prices quoted in foreign currency. The analysis indicates that exchangerate effects differ by product. The exchange rate does seem to be an effective toolfor adjusting exports of Japanese subcompact passenger cars and color tele-visions, even though adjustment may be particularly slow in the case of auto-mobiles. Nevertheless, it does not seem to be as effective with respect to Japan'siron and steel industry. In fact, for the three products analyzed in this industry,export values in foreign currency are estimated to increase subsequent to appre-ciation of the yen exchange rate.

Debt-Equity Ratios of Firms and Interest Rate Policy: Macroeconomic Effects ofHigh Leverage in Developing Countries—v. SUNDARARAJAN (pages 430-74)

The paper analyzes the macroeconomic consequences that flow from enter-prises' financing their investment with a large share of debt in relation to equity.For developing countries characterized by segmented financial markets, controlson the banking system, and substantial reliance on debt, including external debt,the paper develops a model of saving, investment, portfolio adjustments bysavers, and adjustments in the financial structure of firms. The model is used toanalyze the impact of interest rate policy on stability and investment incentives.

The major conclusions are as follows. The debt-equity ratios of firms make asizable difference in the impact of national stabilization policies, particularlyinterest rate policies. When debt ratios used by firms are large, pursuit by theauthorities of a passive interest rate policy—that is, maintaining the controlledinterest rate unchanged when inflation changes—can lead to macroeconomicinstability characterized by perverse effects of monetary policy and acceleratinginflation or deflation. Therefore, in economies in which firms are highly lever-aged, appropriate adjustments in the real administered interest rate becomenecessary to achieve macroeconomic stability. An increase in the real interestrate to ensure stability, however, reduces saving and investment whenever thedebt-equity ratio of firms in the aggregate exceeds a safe limit. This limit depends


SUMMARIES 539

mainly on the interest sensitivity of saving, but it is also influenced by a varietyof other considerations, such as the initial conditions in domestic financial mar-kets, induced adjustments in the debt ratio, the maturity structure of domesticand foreign currency loans, and exchange rate policy.

In view of these implications of the high leverage of nonfinancial firms, anevaluation of the financial structure of firms and the underlying institutionalframework of the financial system is important for a proper assessment of theimpact of stabilization policies. Often the effectiveness of stabilization policies,particularly of interest rate policies, can be enhanced by implementing appropri-ate financial reform measures that include steps to lower the debt-equity ratiosof firms (for example, promoting corporate saving or establishing debt-equitynorms), to reduce segmentation in the financial markets, and to minimize inter-est subsidies. Such policies can contribute not only to macroeconomic stability,but also eventually to reducing the cost, in growth forgone, of stabilizationpolicies.

Effects of Exchange Rate Volatility on Trade: Some Further Evidence—PADMAGOTUR (pages 475-512)

A recent survey of the empirical studies examining the effects of exchange ratevolatility on international trade concluded that "the large majority of empiricalstudies... are unable to establish a systematically significant link between mea-sured exchange rate variability and the volume of international trade, whetheron an aggregated or on a bilateral basis" (International Monetary Fund, Ex-change Rate Volatility and World Trade, Washington, July 1984, p. 36). A recentpaper by M.A. Akhtar and R.S. Hilton ("Exchange Rate Uncertainty andInternational Trade," Federal Reserve Bank of New York, May 1984), in con-trast, suggests that exchange rate volatility, as measured by the standard devi-ation of indices of nominal effective exchange rates, has had significant adverseeffects on the trade in manufactures of the United States and the Federal Repub-lic of Germany.

The purpose of the present study is to test the robustness of Akhtar andHilton's empirical results, with their basic theoretical framework taken as given.The study extends their analysis to include France, Japan, and the United King-dom; it then examines the robustness of the results with respect to changes in thechoice of sample period, volatility measure, and estimation techniques.

The main conclusion of the analysis is that the methodology of Akhtar andHilton fails to establish a systematically significant link between exchange ratevolatility and the volume of international trade. This is not to say that significantadverse effects cannot be detected in individual cases, but rather that, viewed inthe large, the results tend to be insignificant or unstable. Specifically, the resultssuggest that straightforward application of Akhtar and Hilton's methodology tothree additional countries (France, Japan, and the United Kingdom) yieldsmixed results; that their methodology seems to be flawed in several respects, andthat correction for such flaws has the effect of weakening their conclusions; thatthe estimates are quite sensitive to fairly minor variations in methodology; andthat "revised" estimates for the five countries do not, for the most part, supportthe hypothesis that exchange rate volatility has had a systematically adverseeffect on trade.



Real Exchange Rates, Import Penetration, and Protectionism in Industrial Coun-tries—ERIC v. CLIFTON (pages 513-36)

The paper addresses the interaction between exchange rate appreciation andincreased protection, specifically nontariff trade barriers, in terms of a two-stageprocess. In the first stage, an appreciating real exchange rate leads to increasesin import penetration beyond what would typically accompany the normalgrowth of international trade. In the second stage, rising import penetrationleads to increased protection, a result that has been fairly well established byrecent empirical studies.

The paper tests for the occurrence of the first stage by estimating industry-level import penetration ratios as functions of industry-level real exchange ratesand the level of a country's aggregate international trade. For the statistical tests,indices of real exchange rates at the industry level for textiles, clothing, iron andsteel, and transport equipment are used that were derived for the United States,the Federal Republic of Germany, and the United Kingdom. The empirical worksuggests that, as the real exchange rate of a given domestic industry begins toappreciate, import penetration—as distinct from the level of imports—in thatindustry is likely to increase. Over time, of course, the level of import pene-tration can be expected to rise in sectors in which domestic producers arerelatively less efficient than foreign producers as international trade expands inresponse to the forces of comparative advantage. The estimates in the papersuggest, however, that an appreciating real exchange rate is associated withincreases in import penetration beyond what can be accounted for by the seculargrowth of international trade.

The analysis also shows how the empirical results can be used to estimate someof the economic effects of trade restrictions. A quantitative illustration is devel-oped to demonstrate how a policy to restrict import penetration into marketsserved by a given domestic industry will lead to measurable increases in eco-nomic rents for domestic producers in the form of a depreciation of the"shadow" real exchange rate for that industry.


RESUMES

Determination du taux d'interet dans les pays en developpement: cadretheorique—SEBASTIAN EDWARDS et MOHSIN s. KHAN (pages 377-403)

Au moment ou certains pays en developpement se rapprochent de systemesfinanciers plus liberaux, leurs dirigeants se preoccupent de la fagon dont les tauxd'interet reagissent aux influences exterieures et aux politiques interieures. Dansla plupart des etudes consacrees aux taux d'interet, les auteurs ne traitenthabituellement que les cas extremes: soit une economic pleinement ouverte, oula determination des taux en question fait 1'objet d'un arbitrage sous une formeou sous une autre, soit, au contraire, une economic completement fermee, ouseuls les facteurs monetaires interieurs les determinent. Les economies endeveloppement se situent cependant, en general, entre ces deux poles, de sorteque les modeles types de determination ne semblent pas leur etre applicables.

Les auteurs de la presente etude se sont donne pour tache d'esquisser un cadretheorique susceptible de servir de base a 1'analyse du comportement des tauxd'interet dans les pays en developpement ou les autorites sont en train d'abolirles controles qu'elles exergaient sur le secteur financier, ainsi que les restrictionsen matiere de mouvements de capitaux. La m£thode qu'ils proposent associe leselements de modeles elabores aussi bien pour les economies fermees que pourles economies ouvertes; elle peut done tenir compte des influences exercees surles taux d'interet interieurs par leurs contreparties exterieures, les variationsprevues des taux de change et 1'evolution monetaire interieure. Le modelepresente se distingue par une caracteristique interessante: en se basant sur lesdonnees du pays analyse, on peut determiner son degre d'ouverture financiereapproximative, defini comme la mesure dans laquelle les taux d'interet interieurssont lies a ceux du dehors.

Pour montrer la validite empirique du modele propose, les auteurs 1'qnt appli-que a deux pays: la Colombie et Singapour. Comme ces deux pays se trouventa des degres de developpement financier et d'ouverture tout a fait differents, ilsfournissent un premier test utile de la nature generate du modele, qui est capablede representer ces deux cas de fagon tres satisfaisante. D'apres les estimations,en Colombie les facteurs exterieurs et interieurs sont, les uns comme les autres,importants, tandis qu'a Singapour les taux d'interet interieurs sont entierementdetermines par les taux de 1'etranger et les variations du taux de change. Etantdonne les particularites des systemes financiers respectifs dans les deux pays, cesresultats correspondent exactement a ceux que Ton attendait.

Variations du taux de change et exportations de diverses industries japo-naises—DANIEL CITRIN (pages 404-29)

L'auteur presente les resultats d'une recherche empirique relative aux effetsque produisent, a la longue, les fluctuations du taux de change sur les prix et les

541



volumes a 1'exportation de trois grandes industries japonaises: les automobiles,1'electronique grand public, le fer et 1'acier. II fonde son analyse statistique surun modele theorique specifiant les elements dynamiques suivants dans1'ajustement des prix et des volumes a 1'exportation: la mesure dans laquelle, ala longue, les fournisseurs repercutent une modification du taux de change surles prix d'exportation libelles en devises; les effets des anticipations, en matierede prix relatif a Importation, sur la reaction des commandes de 1'etranger(demande) aux variations du prix relatif a 1'exportation; les retards de1'ajustement des commandes de 1'etranger (demande) aux variations du prixrelatif a 1'exportation; et les decalages de livraison entre commandes et expedi-tions.

Les doutes nourris au sujet de I'efficacite" du taux de change comme instru-ment d'ajustement de la balance commerciale du Japon peuvent inciter certainsde ses principaux partenaires commerciaux a adopter diverses mesures protec-tionnistes: dans le contexte de ce pays, il importe done tout particulierement dedeterminer jusqu'a quel point le taux de change influence les courantsd'echange. L'auteur, dans son etude, examine cinq produits d'exportation: lespetites automobiles, les recepteurs de television en couleurs et, dans 1'industriede 1'acier, la grosse tole, la ferblanterie et la tole d'acier galvanise. L'analyse surbase desagregee s'inspire de considerations theoriques et reconnait, implici-tement, que les tensions entre le Japon et ses partenaires commerciaux ont eupour origine, ces dernieres annees, certaines industries bien precises.

Diverses conclusions se degagent de 1'analyse statistique. Pour les cinq pro-duits d'exportation sur lesquels porte I'etude, 1'incidence des variations du tauxde change est sujette a des decalages, tant du cote de 1'offre que de la demande;1'ajustement de celle-ci est particulierement lent. Qui plus est, semble-t-il, lesprix exiges par les concurrents etrangers influencent beaucoup les decisions desfournisseurs en matiere de calcul des prix. A longue echeance comme a courtterme, done, les variations de taux de change ne sont pas integralement reper-cutees par le biais des prix a 1'exportation exprimes en devises. L'analyse montreque les effets du taux de change varient selon le produit. II semble vraiment quece taux constitue un instrument efficace lorsqu'il s'agit d'ajuster les exportationsde petites voitures et de recepteurs de television japonais, meme si 1'ajustementest des plus lents dans le cas des automobiles. II ne semble pas aussi efficace, enrevanche, pour ce qui est de la siderurgie nipponne. En fait, pour les troisproduits analyses dans cette industrie, on estime que les valeurs a 1'exportationen devises n'augmentent qu'apres une appreciation du taux de change du yen.

Ratio d'endettement des entreprises et politique de taux d'interet: effets macro-economiques du niveau eleve de ce ratio dans les pays en developpement—v. SUNDARARAJAN (pages 430-74)

Cette etude est consacree a une analyse des consequences qu'entrame, sur leplan macroeconomique, le comportement des entreprises qui financent leursinvestissements au moyen d'un pourcentage eleve de capitaux d'emprunt parrapport aux capitaux propres. Appliquee aux pays en developpement, lesquelssont caracterises par des marches de capitaux compartimentes, des systemesbancaires r^glementes et un recours important a Pendettement, notamment1'endettement exterieur, cette etude propose un modele qui met en parallele les


RESUMES 543

ajustements operes par les epargnants sur leurs economies et leurs investisse-ments, et les ajustements correspondants de la structure financiere des en-treprises. Ce modele permet d'analyser 1'incidence d'une politique de tauxd'interet sur la stabilite economique et les incitations a investir.

Les principales conclusions auxquelles aboutit 1'etude sont les suivantes:1'efficacite des politiques nationales de stabilisation, notamment des politiquesen matiere de taux d'interet, depend pour une large part des ratiosd'endettement des entreprises. Quand ces ratios sont eleves, une politique pas-sive en matiere de taux d'interet—c'est-a-dire le maintien d'un taux d'interetreglemente alors que le taux d'inflation varie — peut entrainer, sur le planmacroeconomique, une certaine instabilite caracterisee par des effets pervers surle plan monetaire et une acceleration de 1'inflation ou de la deflation. Parconsequent, dans les pays ou les entreprises ont un fort degre d'endettement, ildevient necessaire, pour assurer la stabilite sur le plan macroeconomique, deproceder a des ajustements adequats du taux d'interet reglemente reel. Tout-efois, un relevement du taux d'interet reel en vue d'assurer la stabilite entraineune diminution de 1'epargne et de 1'investissement chaque fois que le ratiod'endettement global des entreprises depasse une limite raisonnable. Cettelimite depend essentiellement du degre d'elasticite de 1'epargne par rapport auxtaux d'interet, mais elle est egalement fonction d'une serie d'autres facteurs,notamment la situation initiale des marches financiers nationaux, les ajuste-ments induits du ratio d'endettement, 1'etalement des echeances des empruntsen monnaie locale et en devises, enfin la politique de taux de change.

Etant donne qu'un ratio d'endettement eleve entraine de telles consequences,il est important, pour determiner avec precision 1'incidence des politiques destabilisation, d'evaluer la structure financiere des entreprises et le cadre institu-tionnel du systeme financier dans lequel elles s'inserent. L'efficacite des poli-tiques de stabilisation, en particulier des politiques de taux d'interet, peut sou-vent etre renforcee par la mise en place de reformes financieres appropriees quicomportent notamment des mesures visant a faire baisser le ratio d'endettementdes entreprises (par exemple, en encourageant 1'epargne des entreprises ou enetablissant des normes d'endettement), a reduire la compartimentation desmarches de capitaux et a limiter au maximum les bonifications d'interet. Desmesures de ce type peuvent contribuer non seulement a la stabilite sur le planmacroeconomique, mais aussi, finalement, a une reduction du cout en evitant lesacrifice de gains de croissance des politiques de stabilisation.

Effets de 1'instabilite des taux de change sur le commerce mondial: nouvellesconstatations—PADMA GOTUR (pages 475-512)

Un recent apergu des etudes empiriques consacrees aux effets de 1'instabilitedes taux de change sur le commerce international conclut que «dans leur grandemajorite, les etudes empiriques... ne reussissent pas a etablir un lien significatifet systematique entre la variabilite mesuree des taux de change et le volumedu commerce international, que celui-ci soit exprime sous forme globale oubilaterale» (Fonds monetaire international, Exchange Rate Volatility and WorldTrade, Washington, juillet 1984, page 36). Par contre, un article public recem-ment par M.A. Akhtar et R.S. Hilton («Exchange Rate Uncertainty and Inter-national Trade», Federal Reserve Bank of New York, mai 1984) soutient que



rinstabilite' des taux de change, mesuree par 1'ecart type des indices des taux dechange effectifs nominaux, a eu un effet defavorable significatif sur le com-merce de produits manufactures des Etats-Unis et de la Republique federated'Allemagne.

La presente etude a pour objet d'evaluer la solidite des resultats empiriquespr6sent6s par Akhtar et Hilton, en prenant comme donne leur cadre theoriquede base. L'auteur etend 1'analyse au cas de la France, du Japon et du Royaume-Uni; elle cherche ensuite dans quelle mesure ces resultats restent valables si Tonmodifie la periode de reference, la mesure de 1'instabilite et les techniquesd'estimation.

La principale conclusion de cette etude est que la methode utilisee par Akhtaret Hilton n'etablit pas de lien significatif et systematique entre Tinstabilite destaux de change et le volume du commerce international. Ceci ne veut pas direque Ton ne puisse pas constater dans certains cas particuliers des effets de-favorables significatifs, mais plutot que, pris dans leur ensemble, les resultatssont peu significatif s ou peu stables. Plus precisement, cette etude laisse en-tendre qu'une application systematique de la methode d'Akhtar et Hilton a troispays supple*mentaires (France, Japon et Royaume-Uni) donne des resultats miti-ges; que leur me*thode semble presenter plusieurs defauts et que la correction deces defauts a pour effet d'affaiblir la portee de leurs conclusions; que leursestimations sont tres sensibles a des variations relativement mineures de lamethode utilisee et que la plupart des estimations «revisees» pour les cinq paysne confirment pas I'hypothese selon laquelle 1'instabilite des taux de changeaurait eu un effet systematiquement negatif sur le commerce international.

Taux de change reels, penetration des importations et protectionnisme dans lespays industrialises—ERIC v. CLIFTON (pages 513-36)

Cette etude traite de 1'interaction entre la hausse du taux de change et lamontee du protectionnisme, principalement les barrieres commerciales non tari-faires, selon un processus en deux etapes. Dans un premier temps, 1'augmenta-tion du taux de change conduit a un accroissement de la penetration des importa-tions superieur a celui qui accompagnerait en principe la croissance normale ducommerce international. Pendant la seconde etape, 1'accroissement de la pene-tration des importations conduit a une montee du protectionnisme, resultat quia etc en grande partie confirme par des travaux empiriques recents.

L'auteur teste le deroulement de la premiere etape en estimant les ratios dela penetration des importations au niveau des industries en fonction des taux dechange reels dans 1'industrie et du volume du commerce exterieur global d'unpays. Pour les tests statistiques, Pauteur utilise des indices de taux de changereels pour les industries du textile, de 1'habillement, du fer et de 1'acier et desmateriels de transport; ces indices ont ete calcules pour les Etats-Unis, la Repu-blique federate d'Allemagne et le Royaume-Uni. II ressort des donnees etudieesque, lorsque le taux de change reel pour une industrie nationale donnee com-mence a s'apprecier, la penetration des importations — qui est distincte du vo-lume des importations — dans cette industrie a tendance a s'accroitre. II est,evidemment, normal qu'au fil des annees le degre de penetration des importa-tions augmente dans les secteurs ou les producteurs nationaux sont relativementmoins efficients que les producteurs etrangers, puisque le commerce interna-


RESUMES 545

tional se developpe du fait du jeu des avantages comparatifs. Toutefois, d'apresles estimations presentees dans le document, il semble bien qu'une hausse dutaux de change reel s'accompagne d'un accroissement de la penetration desimportations superieur a celui qui peut s'expliquer par la croissance tendancielledu commerce international.

Dans son analyse, Tauteur indique aussi comment les resultats empiriquespeuvent servir a estimer certains effets economiques des restrictions commer-ciales. A 1'aide d'un exemple chiffre, il montre comment une politique visant alimiter la penetration des importations sur les marches approvisionnes par uneindustrie interieure donnee conduira a des augmentations mesurables des renteseconomiques des producteurs nationaux sous forme d'une baisse du taux dechange reel «virtuel» pour cette industrie.


RESUMENES

Determinacion de los tipos de interes en los paises en desarrollo: Un marcoconceptual—SEBASTIAN EDWARDS y MOHSIN s. KHAN (paginas 377-403).

A medida que algunos paises en desarrollo van adoptando sistemas financierosmas liberates, quienes elaboran su politica economica tienen que determinar laforma en que los tipos de interes reaccionan ante las influencias exteriores y dela politica interna. Lo tradicional en la mayoria de los estudios existentes sobretipos de interes es considerar s61o los casos extremos: una economia totalmenteabierta en la que se da alguna forma de arbitraje de los tipos de interes, o unaeconomia totalmente cerrada en la cual los tipos de interes resultan exclu-sivamente de la acci6n de factores monetarios internes. Pero las economias endesarrollo, por lo general, se encuentran a mitad de camino entre esos dosextremos, por lo cual no les ser£n aplicables los modelos estandar de deter-minaci6n de los tipos de interns.

La finalidad del presente trabajo consiste en esbozar un marco teorico quesirva de punto de partida para analizar la evolution de los tipos de interes en lospaises en desarrollo que ban comenzado a eliminar los controles al sector finan-ciero y las restricciones a las corrientes de capital. En el enfoque aqui sugeridose combinan elementos extraidos de modelos elaborados para economias ce-rradas y para economias abiertas, lo cual permite incorporar los efectos de lostipos de interns externos, de la variaci6n prevista de los tipos de cambio y de laevoluci6n monetaria interna sobre los tipos de interes internos. Una caracte-ristica interesante del modelo presentado es que permite determinar el gradoaproximado de apertura financiera—que se define como la medida en que lostipos de interes internos estan vinculados a los externos—mediante el examende los datos del pais analizado.

A modo de ejemplo de la validez empfrica del modelo propuesto, se le aplicoa dos paises: Colombia y Singapur. Estos dos paises difieren considerablementeen cuanto al grado de apertura y desarrollo financieros, por lo cual resultanid6neos para una primera comprobaci6n de la aplicabilidad general del modelo.Este logra representar ambos casos muy adecuadamente. Las estimaciones indi-caron que en Colombia tanto los factores externos como los internos revistenimportancia, en tanto que en Singapur los tipos de interes internos estan deter-minados exclusivamente por los tipos de interes externos y por las variaciones deltipo de cambio. Esto es precisamente lo que cabia esperar de acuerdo con lascaracteristicas del sistema financier© de estos dos paises.

Las variaciones de los tipos de cambio y las exportaciones de industrias japonesasseleccionadas—DANIEL CITRIN (paginas 404-29)

El presente trabajo expone los resultados de una investigation empiricaacerca de la influencia a lo largo del tiempo de la variation de los tipos de cambio

546


RESUMENES 547

sobre los precios y los volumenes de exportation de tres de las principalesindustrias japonesas: la del automovil, la de aparatos electronicos para el con-sumidor y la siderurgia. El analisis estadistico se basa en un modelo teorico queespecifica los siguientes elementos dinamicos del ajuste de los precios y vo-lumenes de exportation: el grado en que, con el transcurso del tiempo, losproveedores trasladan las variaciones del tipo de cambio a las exportaciones conprecio marcado en moneda extranjera; el impacto de las expectativas relativasde los precios de exportation sobre la reaction de los pedidos (la demanda) deexportation ante la variation relativa de los precios de exportation; el desfasedel ajuste de los pedidos (la demanda) de exportation ante la variation relativade los precios de exportation, y el desfase entre el pedido y la entrega.

Como las dudas acerca de la eficacia del tipo de cambio como instrumentopara lograr el ajuste de la cuenta comercial de Japon pueden hacer que algunosde los principales paises con los cuales este comercia adopten diversas medidasproteccionistas, es especialmente importante en el contexto japones averiguar enque medida el tipo de cambio influye sobre las corrientes comerciales. Para ellose examinan en el documento cinco productos de exportation especificos: auto-moviles de pasajeros subcompactos, televisores a color y, en la siderurgia, chapapesada, chapa de estano y lamina de acero galvanizado. Se ha optado por elanalisis desagregado en virtud de consideraciones teoricas y en reconocimientode que en los ultimos anos las tensiones reinantes entre Japon y los paises conlos cuales este comercia se han concentrado por lo general en industriasespecificas.

Del analisis estadistico se desprenden varias conclusiones. El impacto de lavariation del tipo de cambio sobre la exportation de los cinco productos japo-neses mencionados presenta desfases tanto en el lado de la oferta como en el dela demanda, siendo particularmente lento el ajuste en esta ultima. Ademas, losprecios de demanda de los competidores extranjeros parecen ejercer una in-fluencia significativa sobre las decisiones de los proveedores en materia deprecios. Como consecuencia, el efecto de la variation de los tipos de cambio nose traslada enteramente, ni a largo ni a corto plazo, a los precios de las expor-taciones citados en divisas. Del analisis se deduce que los efectos de los tipos decambio varian segun los productos de que se trate. Efectivamente, el tipo decambio parece ser un instrumento eficaz para ajustar la exportation de auto-moviles de pasajeros subcompactos y de televisores en color japoneses, si bienel ajuste puede ser especialmente lento en el caso de los automoviles. En cam-bio, parece ser menos eficaz en lo que se refiere a la industria siderurgica. Dehecho, se estima que los valores de exportation en divisas de los tres productosanalizados en esta industria solo aumentan cuando se aprecia el tipo de cambiodel yen.

Las razones deuda/capital de las empresas y la politica de tipo de interes: Efectosmacroeconomicos de los coeficientes de endeudamiento elevados en los paises endesarrollo—v. SUNDARARAJAN (paginas 430-74)

En este trabajo se analizan las consecuencias macroeconbmicas derivadas deque las empresas financien sus inversiones con una proporci6n elevada de deudaen relaci6n con los recursos propios. Se construye un modelo de ahorro, inver-sion, ajuste de la cartera del ahorrista y ajuste de la estructura financiera de lasempresas para paises en desarrollo que se caracterizan por mercados financieros



segmentados, controles del sistema bancario y planificaci6n economica basadasustancialmente en la deuda, incluida la deuda externa. El modelo se usa paraanalizar el impacto de la politica de tipo de interes en la estabilidad y losincentives a la inversion.

Las principals conclusiones son las siguientes. Las razones deuda/capital delas empresas hacen variar considerablemente el impacto de las politicas deestabilizaci6n nacional, en especial, las de tipo de interes. Cuando las empresasutilizan elevados coeficientes de endeudamiento, la aplicacion por las auto-ridades de una politica pasiva de tipo de interes—o sea, el mantenimiento de untipo de interes controlado invariable cuando la inflacion varia— puede conducira la inestabilidad macroeconomica, caracterizada por efectos perversos de lapolitica monetaria y aceleracidn de la inflacion o la deflacion. De modo que enlas economias cuyas empresas tienen un elevado coeficiente de endeudamientoes necesario ajustar en forma apropiada el tipo de interes real administrado paraque haya estabilidad macroeconomica. Un aumento del tipo de interes real conese fin, sin embargo, reduce el ahorro y la inversion cuando la razon deuda/capital de las empresas en conjunto supera un limite de seguridad. Este limitedepende principalmente de la sensibilidad del ahorro respecto al tipo de interes,pero tambien de varias otras consideraciones, tales como las condiciones ini-ciales de los mercados financieros nacionales, los ajustes inducidos del co-eficiente de deuda, la estructura de vencimientos de los prestamos en monedanacional y extranjera y la politica de tipo de cambio.

Dadas estas repercusiones de los coeficientes de endeudamiento elevados delas empresas no financieras, es importante evaluar la estructura financiera de lasempresas y el marco institucional basico del sistema financiero para determinaradecuadamente el impacto de las politicas de estabilizacion. Muchas veces esposible acrecentar la eficacia de las politicas de estabilizacion, especialmente lasde tipo de interes, aplicando medidas adecuadas de reforma financiera entre lascuales figuren medidas para reducir las razones deuda/capital de las empresas(por ejemplo, la promoci6n del ahorro empresarial o la adoption de normassobre las razones deuda/capital), para disminuir la segmentation de los mercadosfinancieros y para minimizar las subvenciones de intereses. Esas politicas ademasde coadyuvar a la estabilidad macroecon6mica, tambien pueden en definitivareducir el costo en perdida de crecimiento de las politicas de estabilizacion.

Efectos de la inestabilidad de los tipos de cambio en el comercio internacional:Alguna evidencia adicional—PADMA GOTUR (paginas 475-512)

En un examen reciente de los estudios empiricos sobre los efectos de lainestabilidad de los tipos de cambio en el comercio internacional se llega a laconclusion de que "la gran mayoria de estos analisis empiricos no consiguendemostrar sistematicamente un vmculo significativo entre los diferentes gradosde variabilidad cambiaria y el volumen del comercio internacional, tanto sea enterminos agregados como bilaterales". (Fondo Monetario Internacional, Ex-change Rate Volatility and World Trade, Washington, julio de 1984, pag. 36). Unestudio reciente de M.A. Akhtar y R.S. Hilton ("Exchange Rate Uncertaintyand International Trade," Banco de la Reserva Federal de Nueva York, mayo de1984) indica, por el contrario, que la inestabilidad de los tipos de cambio,expresada segun la desviacion estandar de los indices de los tipos de cambio


RESUMENES 549

efectivos nominales, ha tenido efectos negatives considerables en el comercio deproductos manufacturados de Estados Unidos y de la Republica Federal deAlemania.

El presente estudio tiene por objeto comprobar la solidez de los resultadosempiricos de Akhtar y Hilton, tomando como base de partida su marco teoricobasico. El estudio amplia su analisis incluyendo a Francia, Japon y el ReinoUnido, pasando luego a examinar la solidez de los resultados con respecto avariaciones en la selecci6n del periodo de la muestra, medida de la inestabilidady tecnicas de estimation.

La conclusibn principal del analisis es que la metodologia de Akhtar y Hiltonno logra establecer un vinculo significative sistematico entre la inestabilidad delos tipos de cambio y el volumen del comercio internacional. Esto no quiere decirque no puedan obsevarse en casos especificos efectos negatives importantes, sinomas bien que, en terminos generales, los resultados no suelen ser ni conside-rables ni estables. En concrete, de los resultados se desprende que la aplicaciondirecta de la metodologia de Akhtar y Hilton a tres nuevos paises (Francia,Japon y el Reino Unido) arroja resultados dispares; que esta metodologia pareceser defectuosa en varies aspectos y que la correction de tales deficiencias tienecomo efecto el debilitamiento de sus conclusiones; que las estimaciones son muysensibles a modificaciones poco importantes de la metodologia, y que las esti-maciones "revisadas" para los cinco paises no confirman, en su mayor parte, lahipotesis de que la inestabilidad de los tipos de cambio ha ejercido un efectonegative sistematico en el comercio exterior.

Tipos de cambio reales, penetration de las importaciones y proteccionismo en lospaises industriales—ERIC v. CLIFTON (paginas 513-36)

En este trabajo se examina la interacci6n entre la apreciacion del tipo decambio y el aumento del proteccionismo, especificamente las barreras no aran-celarias al comercio, como un proceso en dos etapas. En la primer a etapa, untipo real de cambio que se aprecie da lugar a aumentos de la penetracion de lasimportaciones superiores a los que normalmente acompanarian al crecimientonormal del comercio internacional. En la segunda etapa, el aumento de lapenetracion de las importaciones origina un mayor proteccionismo, resultadobastante respaldado por estudios empiricos recientes.

Se efectiian pruebas de que ha ocurrido la primera etapa estimando los co-eficientes de penetracion de las importaciones a nivel de las industrias comofunciones de los tipos reales de cambio al nivel de las industrias y el nivel delcomercio internacional agregado de un pais. En las pruebas estadisticas se utili-zan indices de los tipos de cambio reales a nivel de las industrias textil, deprendas de vestir, sideriirgica y de equipo de transporte, que se calcularon paraEstados Unidos, la Republica Federal de Alemania y el Reino Unido. El trabajoempirico parece indicar que al empezar a apreciarse el tipo de cambio real de unadeterminada industria nacional, la penetracion de las importaciones—que sediferencia del nivel de importaciones—de esa industria posiblemente aumente.Con el tiempo, desde luego, es de prever que aumente el nivel de la penetracionde las importaciones en los sectores en que los productores nacionales sonrelativamente menos eficientes que los extranjeros, a medida que el comerciointernacional se acrecienta ante el impulse de la ventaja comparativa. Sin em-bargo, las estimaciones de este trabajo indican que un tipo de cambio real que



se aprecia esta relacionado con aumentos de la penetration de las importacionessuperiores a los explicables por el crecimiento secular del comercio inter-national.

El analisis indica tambien como pueden utilizarse los resultados empiricospara estimar algunos de los efectos econ6micos de las restricciones al comercio.Se elabora una ilustracion cuantitativa para demostrar la forma en que unapolitica destinada a restringir la penetration de las importaciones en los mer-cados abastecidos por una determinada industria nacional dara origen a unincremento mensurable de la renta economica de los productores nacionales enforma de depreciaci6n del tipo de cambio real de "sombra" correspondiente aesa industria.


This page intentionally left blank


In statistical matter (except in the resumes and resumenes} throughoutthis issue,

Dots ( . . . ) indicate that data are not available;

A dash (—) indicates that the figure is zero or less than half the finaldigit shown, or that the item does not exist;

A single dot (.) indicates decimals;

A comma (,) separates thousands and millions;

"Billion" means a thousand million, and "trillion" means a thousandbillion;

A short dash (-) is used between years or months (for example,1981-83 or January-October) to indicate a total of the years ormonths inclusive of the beginning and ending years or months;

A stroke (/) is used between years (for example, 1981/82) to indicate afiscal year or a crop year;

Components of tables may not add to totals shown because of rounding.


Documents

INTERNATIONAL MONETARY FUND...economists, and by others concerned with monetary and financial problems. Much of what is now presented is quite provisional. On some international monetary