the economics of exchange rates - wharton financefinance.wharton.upenn.edu/~bodnarg/courses/...eign...

35
Journal of Economic Literature, Vol. XXXIII (March 1995) Taylor: Exchange Rates Journal of Economic Literature Vol. XXXIII (March 1995), pp. 13–47 The Economics of Exchange Rates By MARK P. TAYLOR University of Liverpool and Centre for Economic Policy Research, London I am grateful to three anonymous referees for constructive com- ments on a previous draft. I am also indebted to the large number of people who provided helpful and often detailed comments on earlier versions of the paper, including Andrew Atkeson, Leonardo Bartolini, Tamim Bayoumi, Giuseppe Bertola, Stanley Black, William Branson, Guillermo Calvo, Michael Dooley, Hali Edison, Robert Flood, Jeffrey Frankel, Jacob Frenkel, Kenneth Froot, Peter Garber, Robert Hodrick, Peter Isard, Peter Kenen, Ronald MacDonald, Bennett McCallum, Marcus Miller, Maurice Obstfeld, Lawrence Officer, David Papell, Kenneth Rogoff, Nouriel Roubini, Alan Stockman, Lars Svensson, Myles Wallace, and John Williamson. Responsibility for any remaining errors of omission or interpretation remains with the author. This paper was written largely while the author was on the Staff of the Research Department of the International Monetary Fund, Washington D.C., although the views represented in the paper are solely those of the author and are not necessarily those of the International Monetary Fund or of its member authori- ties. I. Introduction This paper reviews the literature on exchange rate economics over the last two decades, with particular reference to recent developments. Exchange rate eco- nomics has been one of the most ac- tive—if challenging—areas of economic research over the last twenty years, and the amount of ground covered here is correspondingly vast. Thus, we can only hope to give a selective survey of the ter- rain and of its major promontories. In particular, we discuss the evidence on foreign exchange market efficiency (Sec- tion II), the theory and evidence relating to the determination of exchange rates (Sections III and IV respectively), recent work on the effectiveness of foreign ex- change intervention (Section V), and the recent literature on exchange rate behav- ior within target zones (Section VI). The emerging literature on foreign exchange market microstructure is also briefly dis- cussed (Section VII). In the concluding section of the paper we attempt to draw 13

Upload: others

Post on 19-Apr-2020

13 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

Journal of Economic Literature, Vol. XXXIII (March 1995)Taylor: Exchange RatesJournal of Economic LiteratureVol. XXXIII (March 1995), pp. 13–47

The Economics of Exchange Rates

By MARK P. TAYLOR

University of Liverpooland

Centre for Economic Policy Research, London

I am grateful to three anonymous referees for constructive com-ments on a previous draft. I am also indebted to the large numberof people who provided helpful and often detailed comments onearlier versions of the paper, including Andrew Atkeson,Leonardo Bartolini, Tamim Bayoumi, Giuseppe Bertola, StanleyBlack, William Branson, Guillermo Calvo, Michael Dooley, HaliEdison, Robert Flood, Jeffrey Frankel, Jacob Frenkel, KennethFroot, Peter Garber, Robert Hodrick, Peter Isard, Peter Kenen,Ronald MacDonald, Bennett McCallum, Marcus Miller, MauriceObstfeld, Lawrence Officer, David Papell, Kenneth Rogoff,Nouriel Roubini, Alan Stockman, Lars Svensson, Myles Wallace,and John Williamson. Responsibility for any remaining errors ofomission or interpretation remains with the author. This paperwas written largely while the author was on the Staff of theResearch Department of the International Monetary Fund,Washington D.C., although the views represented in the paperare solely those of the author and are not necessarily thoseof the International Monetary Fund or of its member authori-ties.

I. Introduction

This paper reviews the literature onexchange rate economics over the lasttwo decades, with particular reference torecent developments. Exchange rate eco-nomics has been one of the most ac-tive—if challenging—areas of economicresearch over the last twenty years, andthe amount of ground covered here iscorrespondingly vast. Thus, we can onlyhope to give a selective survey of the ter-rain and of its major promontories. In

particular, we discuss the evidence onforeign exchange market efficiency (Sec-tion II), the theory and evidence relatingto the determination of exchange rates(Sections III and IV respectively), recentwork on the effectiveness of foreign ex-change intervention (Section V), and therecent literature on exchange rate behav-ior within target zones (Section VI). Theemerging literature on foreign exchangemarket microstructure is also briefly dis-cussed (Section VII). In the concludingsection of the paper we attempt to draw

13

Page 2: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

out some broad themes in the researchprogram, and speculate on the likely ordesirable course of future research inthis area.1

II. Speculative Efficiency

In an efficient speculative market,prices should fully reflect informationavailable to market participants and itshould be impossible for a trader toearn excess returns to speculation. Aca-demic interest in foreign exchange mar-ket efficiency can be traced to argumentsconcerning the information content offinancial market prices and the im-plications for social efficiency. In itssimplest form, the efficient markets hy-pothesis can be reduced to a joint hy-pothesis that foreign exchange marketparticipants are, in an aggregate sense a)endowed with rational expectation andb) risk-neutral. The hypothesis can bemodified to adjust for risk, so that itthen becomes a joint hypothesis of amodel of equilibrium returns (which mayadmit risk premia) and rational expecta-tions.

If the risk-neutral efficient marketshypothesis holds, then the expected for-eign exchange gain from holding onecurrency rather than another (the ex-pected exchange rate change) must bejust offset by the opportunity cost ofholding funds in this currency ratherthan the other (the interest rate differen-tial). This is the cornerstone parity con-dition for testing foreign exchange mar-ket efficiency—the uncovered interestrate parity condition:

∆kset+k = it − i*t (1)

where st denotes the logarithm of thespot exchange rate (domestic price of

foreign currency) at time t, it, and ii∗ are

the nominal interest rates available onsimilar domestic and foreign securitiesrespectively (with k periods to maturity),∆kst+k ≡ st+k − st, and superscript e de-notes the market’s expectation based oninformation at time t.

1. Testing Foreign Exchange Market Efficiency

Early efficiency studies tested for therandomness of exchange rate changes(e.g., William Poole 1967). However,only if the nominal interest rate differen-tial is identically equal to a constant, andexpectations are rational, does (1) implya random walk in the exchange rate (withdrift if the constant is non-zero). Gener-ally, the random walk model is inconsis-tent with the uncovered interest rateparity condition. Robert Cumby andObstfeld’s (1981) analysis is a logical ex-tension of this literature because theytest for—and reject—the randomness ofdeviations from uncovered interest rateparity. Notwithstanding this, however, itremains true that time series for the ma-jor nominal exchange rates over the re-cent float are extremely hard to distin-guish empirically from random walks(Michael Mussa 1984).

Another method of testing market effi-ciency is to test for the profitability offilter rules. A simple j-percent filter ruleinvolves buying a currency whenever itrises j percent above its most recenttrough and selling the currency when-ever it falls j percent below its most re-cent peak. If the market is efficient anduncovered interest rate parity holds, theinterest rate costs of such a strategyshould on average eliminate any profit. Anumber of studies do indicate the profit-ability of simple filter rules (Dooley andJeffrey Shafer 1983; Levich and LeeThomas 1993), although it is usually not

1 This paper can be viewed as an extension andupdate of earlier surveys, notably Richard Levich(1985) and MacDonald and Taylor (1992). All ofthe topics discussed are dealt with more fully inTaylor (forthcoming).

14 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 3: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

clear that the optimal filter rule sizecould have been chosen ex ante, andthere are often also important elementsof riskiness in that substantial subperiodlosses are often generated. Further, indi-rect evidence on the profitability of trad-ing rules is provided by Charles Engeland James Hamilton (1990), who showthat the dollar, from the early 1970s tothe late 1980s, was susceptible to “longswings” (largely uninterrupted trends),which are susceptible to mechanical(“trend-following”) trading rules.

More often, researchers have testedfor efficiency by regression-based analy-sis of spot and forward exchange rates.The forward rate is the rate agreed uponnow for an exchange of currencies atsome agreed future point in time. Theforward premium at a certain maturity isthe percentage difference between thecurrent forward rate of that maturityand the current spot rate.2 Assumingcovered interest parity (see equation (7)discussed below), the interest rate dif-ferential should be just equal to the for-ward premium. Under rational expecta-tions, the expected change in the ex-change rate should differ from the actualchange only by a rational expectationsforecast error. Hence, the uncovered in-terest rate parity condition (1) can betested by estimating a regression equa-tion of the form

∆kst+k = α + β(ft(k) − st) + ηt+k (2)

where ft(k) is the logarithm of the forward

rate for maturity k periods ahead andηt+k is a disturbance term.3 If agents are

risk-neutral and have rational expecta-tions, we should expect the slope pa-rameter, β, to be equal to one and thedisturbance term ηt+k—the rational ex-pectations forecast error under the nullhypothesis—to be uncorrelated with in-formation available at time t.4 Empiricalstudies of (2), for a large variety of cur-rencies and time periods, for the recentfloating experience, generally report re-sults which are unfavorable to the effi-cient markets hypothesis under risk neu-trality (e.g., Eugene Fama 1984). Indeedit is stylized fact that estimates of β, us-ing exchange rates against the dollar, aregenerally closer to minus unity than plusunity (Froot and Richard Thaler 1990). Anumber of authors have interpreted thestylized fact of a negative coefficient inthis regression—the so-called “forwarddiscount bias”—as evidence that the for-ward premium mispredicts the directionof the subsequent change in the spotrate, although such statements may bemisleading because they ignore the con-stant term in the regression (Hodrick1992). What the negativity of the esti-mated slope coefficient does imply, how-ever, is that the more the foreign cur-rency is at a premium in the forwardmarket at a certain term k, the less thehome currency—usually the dollar—ispredicted to depreciate over the k peri-

2 Some authors term this the forward discountrather than the forward premium. The choice isessentially arbitrary, because a premium is just anegative discount.

3 Regression relationships involving exchangerates are normally expressed in logarithms in orderto circumvent the so-called “Siegel paradox”(Jeremy Siegel 1972): because of Jensen’s inequal-ity, one cannot have, simultaneously, an unbiasedexpectation of, say the mark-dollar exchange rate(marks per dollar) and of the dollar-mark ex-

change rate (dollars per mark) because 1/E(S) ≠E(1/S). Although the problem seems to be avoidedif exchange rates are expressed in logarithms be-cause E(−s) = −E(s), agents must still form expec-tations of final payoffs S and 1/S, so that it is notclear that taking logarithms does avoid the prob-lem. Engel (1984), using an argument based onreal as opposed to nominal returns to speculation,derives an efficiency condition which is indepen-dent of the choice of numéraire currency. J. Hus-ton McCulloch (1975), using 1920s data, demon-strates the operational importance of the Siegelparadox to be slight.

4 This follows from the formal property ofrational expectations forecast errors thatE[ηt+k|Ωt] = 0, where E[.|Ωt] denotes the mathe-matical expectation conditioned on the informa-tion set available at time t, Ωt.

Taylor: Exchange Rates 15

Page 4: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

ods to maturity.5 This may imply an ex-pected appreciation of the home cur-rency, but the constant terms are rela-tively large and often it does not.

Early regression-based tests of foreignexchange market efficiency regressed thelogarithm of the spot rate onto thelagged logarithm of the forward rate(e.g., Frenkel 1976), and usually foundan estimated slope coefficient close tounity. It was subsequently realized, how-ever, that standard regression analysis(or at least standard inferential statisticaltheory) was invalid with such a relation-ship, because of the nonstationarity ofthe series. Moreover, it should be notedthat the two relationships (2) and

st+k = α + β ft(k) + ηt

′+k (3)

are identical only under the null hypo-thesis β = 1. In particular, suppose (2)holds with β ≠ 1. Then (2) may bereparameterized as

st+k = α + β ft(k) + [(1 − β)st + ηt+k], (4)

so that the error term in (3), ηt′+k, is seen

to be [(1 − β)st + ηt+k]. Now, if st is non-stationary, then its sample variance willbe very high. But the ordinary leastsquares estimator works by minimizingthe residual variance in a regression re-lationship. Thus, ordinary least squaresapplied to (3) will tend to drive the esti-mated value of β toward unity, regard-less of the true value of β.

As noted above, a stylized fact con-cerning major exchange rates over therecent float is that they are not only non-stationary but are extremely hard to dis-tinguish from simple random walks. Ifthe exchange rate did literally follow a

random walk, then the estimated valueof β in (2) should be close to zero, re-gardless of whether the market is effi-cient. Moreover, because the best pre-dictor of future values of the spot rate is,under the assumption of a random walk,just the current spot rate, then the sim-ple efficiency hypothesis combined withthe random walk hypothesis would implyft(k) = st

e+k = st, so that the regressor in (2)should be close to zero, in which case βwould be unidentified. In practice, theobserved variation in (ft

(k)) − st) would al-most certainly be non-zero, even underthese assumptioms, if only because ofmeasurement errors.

Thus, regressions of the form (2) or(3) as tests of simple efficiency are seri-ously confounded by the near-random-walk behavior of spot exchange rates.Given these problems, perhaps a betterway of testing the simple efficiency hy-pothesis is to test the orthogonality ofthe forward rate forecast error (the errormade in forecasting the future spot rateusing the current forward rate6) with re-spect to a given information set by im-posing the restriction β = 1 in (2) andtesting the null hypothesis that Ψ = 0 inregressions of the form:

st+k − ft(k) = ΨIt + ηt+k (5)

where It is a vector of variables selectedfrom the information set available attime t. Orthogonality tests of this kind,using lagged forecast errors of the ex-change rate in question in It, generallyhave rejected the simple, risk-neutral ef-ficient markets hypothesis; even strongerrejections are usually obtained when ad-ditional information is included in It(Lars Hansen and Hodrick 1980).

A discernible trend in the efficiencyliterature since the 1970s has been to-

5 Equivalently, via the covered interest arbi-trage condition, these findings indicate that themore U.S. interest rates exceed foreign interestrates, the more the dollar tends on average to ap-preciate over the holding period, not to depreciateso as to offset on average the interest differentialin favor of the home currency.

6 This term, (st+k − ft(k)), alternatively may bethought of as the return to forward speculation.Some authors term it the “excess return” becauseno allowance is made for risk.

16 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 5: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

ward increasing econometric sophistica-tion. Thus, early tests of efficiency,which involved simple tests for a randomwalk in the spot rate, were supplanted bybasic linear regression analyses of uncov-ered interest parity, which were in turnsupplanted by application of the use ofsophisticated rational expectations esti-mators which allowed the use of datasampled more finely than the term of theforward contract involved (Hansen andHodrick 1980).7 By and large, this in-creasing sophistication has generated in-creasingly strong evidence against thesimple, no-risk-premium speculative effi-ciency hypothesis.

2. Rethinking Efficiency I: Risk Premia8

The rejection of the simple, risk-neu-tral efficient markets hypothesis may bedue to the risk-aversion of market par-ticipants or to a departure from the purerational expectations hypothesis, or both.If foreign exchange market participantsare risk-averse, the uncovered interestparity condition (1) may be distorted bya risk premium, ρt say, because agentsdemand a higher rate of return than theinterest differential in return for the riskof holding foreign currency. Thus, arbi-trage will ensure that the interest ratecost of holding foreign currency (i.e., theinterest rate differential) is just equal tothe expected gain from holding foreign

currency (the expected rate of deprecia-tion of the domestic currency) plus a riskpremium:9

it − it∗ = ∆kst+k

e + ρt. (6)

If the risk premium is time-varying andcorrelated with the forward premium orthe interest rate differential, this wouldconfound efficiency tests of the kind out-lined above (Fama 1984). This reasoninghas led to a search for stable empiricalmodels of the risk premium on the as-sumption of rational expectations. Be-cause of the theoretical relationship be-tween risk and the second moments ofasset price distributions, researchershave often tested for a risk premium as afunction of the variance of forecast er-rors or of exchange rate movements(Frankel 1982b; Ian Domowitz and Hak-kio 1985; Alberto Giovannini and Phil-lipe Jorion 1989). In common with otherempirical risk premium models, such aslatent variables formulations (Hansenand Hodrick 1983), such models havegenerally met with mixed and somewhatlimited success, and have not been foundto be robust when applied to differentdata sets and sample periods. As notedby Lewis (forthcoming), for credible de-grees of risk aversion, empirical risk pre-mium models have so far been unable toexplain to any significant degree the vari-ation in the excess return from forwardmarket speculation.

3. Rethinking Efficiency II: Expectations

An alternative explanation of the rejec-tion of the simple efficient markets hy-

7 An additional, econometrically sophisticatedmethod of testing the simple efficient markets hy-pothesis—which also generally has led to rejec-tions of the hypothesis—has involved testing thenonlinear cross-equation restrictions which the hy-pothesis imposes on a vector autoregression (VAR)in spot and forward rates. This was originally sug-gested in the context of foreign exchange rates byCraig Hakkio (1981) although, as the subsequentcointegration literature revealed, a VAR in firstdifferences alone is not appropriate for spot andforward rates.

8 See Karen Lewis (forthcoming) for a recentsurvey of the literature on the foreign exchangerisk premium and departures from the rational ex-pectations paradigm. Frankel (1988) surveys theempirical work on risk premia.

9 Our use of the term “premium” rather than“discount” is again arbitrary and follows standardusage in the literature; risk premia can, however,be negative. Note also that (6) is an arbitrage con-dition rather than a behavioral relationship. Inparticular, (6) could just as well be written with ρton the left-hand-side, in which case it would haveto redefined as -1 times its present implicit defini-tion, ρt ≡ it − it∗ − ∆kst+ke .

Taylor: Exchange Rates 17

Page 6: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

pothesis is that there is a failure, in somesense, of the expectations component ofthe joint hypothesis. Examples in thisgroup are: the “peso problem” originallysuggested by Rogoff (1979); rationalbubbles; learning about regime shifts(Lewis 1989); or inefficient informationprocessing, as suggested, for example, byJohn Bilson (1981). The peso problemrefers to the situation where agents at-tach a small probability to a large changein the economic fundamentals, whichdoes not occur in sample. This will tendto produce a skew in the distribution offorecast errors even if agents’ expecta-tions are rational, and thus may generateapparent evidence of non-zero excess re-turns from forward speculation. In com-mon with peso problems, the presence ofrational bubbles may also show up asnon-zero excess returns even whenagents are risk-neutral. Similarly, whenagents are learning about their environ-ment they may be unable fully to exploitarbitrage opportunities which are appar-ent in the data ex post. A problem withadmitting peso problems, bubbles orlearning into the class of explanations ofthe forward discount bias is that, asnoted above, a very large number ofeconometric studies—encompassing aneven larger range of exchange rates andsample periods—have found that the di-rection of bias is the same, i.e., that theestimated uncovered interest rate parityslope parameter, β in (2), is generallynegative and closer to minus unity thanplus unity. For example, Lewis (1989), ina study of the relationship of the early1980s dollar appreciation with learningabout the U.S. money supply process,notes a persistence in the forward rateerrors which, in itself, is prima facie evi-dence against the learning explanation:agents cannot forever be learning abouta once-for-all regime shift. Similarly, thepeso problem is essentially a small-sam-ple phenomenon; it cannot explain the

overall stylized fact that estimates of βare negative.

4. Rethinking Efficiency III: Survey Data Studies

A problem with much of the empiricalwork on the possible rationalizations ofthe rejection of the simple, risk-neutralefficient markets hypothesis, is that intesting one leg of the joint hypothesis,researchers typically have assumed thatthe other leg is true. For instance, thesearch for a stable empirical risk pre-mium model generally has been condi-tioned on the assumption of rational ex-pectations. Thus, some researchers haveemployed survey data on exchange rateexpectations to conduct tests of eachcomponent of the joint hypothesis (Frootand Frankel 1989; see Shinji Takagi 1991for a survey of survey data studies). Ingeneral, the overall conclusion thatemerges from survey data studies ap-pears to be that both risk aversion anddepartures from rational expectations areresponsible for rejection of the simpleefficient markets hypothesis.10,11

5. Other Parity Conditions

Although uncovered interest rate par-ity is the basic parity condition for as-

10 In an influential study, Froot and Frankel(1989) did not reject the hypothesis that the bias isdue entirely to systematic expectational errors. Inparticular, they found a slope coefficient insignifi-cantly different from one in the regression of themarket survey forecast onto the forward premium.Hodrick (1992) notes, however, that the R2 in thisregression is far from perfect, as it should be if theforward premium is the market’s expected rate ofdepreciation and risk factors are insignificant.

11 McCallum (1994) suggests that the negativityof the estimated uncovered interest rate parityslope coefficient is consistent with a simultaneityinduced by the existence of a government reactionfunction in which the interest rate differential isset in order to avoid large current exchange ratemovements as well as to smooth interest rate move-ments. This is a special case of the general pointmade by Fama (1984) that negativity of estimatedβ requires negative covariation between the riskpremium and the expected rate of depreciation.

18 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 7: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

sessing the efficiency of the foreign ex-change market, two other arbitrage con-ditions which receive considerable atten-tion in the literature are covered interestrate parity and purchasing power parity.12

(a) Covered interest rate parity. Ifthere are no barriers to arbitrage acrossinternational financial markets, then ar-bitrage should ensure that the interestdifferential on similar assets, adjustedfor covering in the forward foreign ex-change market the movement of curren-cies at the maturity of the underlying as-sets, be continuously equal to zero, sothat covered interest rate parity shouldhold:13

(it − it∗) − (ft

(k) − st) = 0. (7)

Frenkel and Levich (1975, 1977) testfor departures from (7) for a number ofmajor exchange rates during the 1970s,allowing for transactions costs, and findvery few departures when Euro-deposit

rates are used, but significantly more(some 20%) when Treasury bill discountsare used. Further evidence supportive ofcovered interest rate parity for severalmajor exchange rates during the recentfloat is provided by Kevin Clinton(1988). Taylor (1987, 1989) uses veryhigh quality, high frequency, contempo-raneously sampled data for spot and for-ward exchange rates and Euro-depositrates, and finds, inter alia, that there arefew profitable violations of covered in-terest rate parity, even in periods of mar-ket uncertainty and turbulence.14

(b) Purchasing power parity.15 Abso-lute purchasing power parity implies thatthe exchange rate is equal to the ratio ofthe two relevant national price levels.Relative purchasing power parity positsthat changes in the exchange rate areequal to changes in relative nationalprices. Thus, in estimates of equations ofthe form

st = α + βpt + β∗pt∗ + ut, (8)

a test of the restrictions β = 1, β* = −1would be interpreted as a test of abso-lute purchasing power parity, while atest of the same restrictions applied tothe equation with the variables in firstdifferences would be interpreted as atest of relative purchasing power parity.The real exchange rate, in logarithmicform

πt ≡ st − pt + pt∗, (9)

12 Tests of these two parity conditions have dif-ferent implications from tests for uncovered inter-est parity. Covered interest parity, for example,should hold independently of agents’ attitudes to-ward risk and their method of forming expecta-tions, so that tests of this parity condition are re-ally tests of barriers to arbitrage. While purchasingpower parity can be given an efficient markets in-terpretation, its major importance lies in the linkbetween economic fundamentals (the determi-nants of price movements) and exchange ratemovements. Nevertheless, these parity conditionsrecur in theoretical and empirical exchange ratework: covered interest parity was used above toderive equation (2) from equation (1), for exam-ple, and both parity conditions have been usedregularly in work on exchange rate determination,as we shall see below. Thus, a brief discussion ofthe empirical evidence relating to these parityconditions is warranted in the present context.

13 It is clearly important in this connection toconsider home and foreign assets which are com-parable in terms of maturity, and also in terms ofother characteristics such as default and politicalrisk; most often, researchers have used offshore,Euro-currency interest rates. A typical barrier toarbitrage would be capital controls; deviationsfrom covered interest parity using domestic secu-rity interest rates (or the spread between offshoreand onshore rates) have often been used as an in-direct indicator of the presence and effectivenessof these (Dooley and Isard 1980).

14 Another test of covered interest rate parity—where the forward premium is regressed onto theinterest differential—has also been strongly sup-portive of this parity condition (e.g., Branson1969). Regression-based tests of covered interestrate parity should, however, be interpreted withcaution: while a researcher may be unable to re-ject the hypothesis that the intercept and slopecoefficients are respectively zero and unity, the fit-ted residuals may themselves represent substantialarbitrage opportunities.

15 See Froot and Rogoff (forthcoming) for acomprehensive survey of the literature on pur-chasing power parity and long-run real exchangerates.

Taylor: Exchange Rates 19

Page 8: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

can be interpreted as a measure of thedeviation from absolute purchasingpower parity. Purchasing power parity(PPP) has variously been viewed as atheory of exchange rate determination,as a short- or long-run equilibrium con-dition, and as an efficient arbitrage con-dition in either goods or asset markets(Officer 1976; Frenkel 1976, 1978; Rudi-ger Dornbusch 1987a). The professionalconsensus on the validity of purchasingpower parity has shifted radically overthe past two decades or so. Prior to therecent float, the consensus appeared tosupport the existence of a fairly stablereal exchange rate (e.g., Milton Fried-man and Anna Schwartz 1963; HenryGaillot 1970). As we discuss below, how-ever, the prevailing orthodoxy of theearly 1970s, largely associated with themonetary approach to the exchange rate,assumed the much stronger propositionof continuous purchasing power parity(e.g., Frenkel 1976; and the studies inFrenkel and Harry Johnson 1978). In themid to late 1970s, in the light of the veryhigh variability of real exchange rates(the “collapse” of PPP; Frenkel 1981a)this extreme position was largely aban-doned. Subsequently, studies publishedmostly in the 1980s, which could not re-ject the hypothesis of random walk be-havior in real exchange rates (MichaelAdler and Bruce Lehmann 1983), re-duced further the confidence in purchas-ing power parity and led to the ratherwidespread belief that PPP was of littleuse empirically and that real exchangerate movements were highly persistent(Dornbusch 1988). More recently, re-searchers have tested for cointegrationbetween the nominal exchange rate andrelative prices—interpreted as testingfor long-run purchasing power parity—by testing for mean reversion or station-arity in the real exchange rate or in theresidual of an equation such as (8). Ear-lier cointegration studies generally re-

ported a failure of significant mean re-version of the exchange rate toward pur-chasing power parity for the recent float-ing experience (Taylor 1988; NelsonMark 1990), but were supportive of re-version toward purchasing power parityfor the interwar float (Taylor and PatrickMcMahon 1988), for the 1950s U.S.- Canadian float (Robert McNown andWallace 1989), and for the exchangerates of high-inflation countries (TaufiqChoudhry, McNown, and Wallace 1991).Very recent applied work on long-runpurchasing power parity among the ma-jor industrialized economies has, how-ever, been more favorable toward thelong-run purchasing power parity hy-pothesis for the recent float (e.g., Yin-Wong Cheung and Kon Lai 1993 andMacDonald 1993, who relax the con-straints β = −β* = 1 in (8)). A number ofauthors have argued that the data periodfor the recent float alone may simply betoo short to provide any reasonable de-gree of test power in the normal statisti-cal tests for stationarity of the real ex-change rate (Frankel 1990), andresearchers have sought to remedy thisby various means. Niso Abuaf and Jorion(1990) increase the power of their testsby using longer time series and by utiliz-ing systems estimation methods and areable to reject the unit root (randomwalk) hypothesis for the real exchangerate. Francis Diebold, Steven Husted,and Mark Rush (1991) apply fractionalintegration techniques to nineteenthcentury data and find evidence of long-run purchasing power parity. James Lo-thian and Taylor (forthcoming) utilizesterling-dollar and sterling-franc ex-change rate data spanning the two centu-ries ending in 1990, and find strong evi-dence in favor of mean reversion in thereal exchange rate. Robert Flood andTaylor (forthcoming) find strong supportfor mean reversion toward long-run pur-chasing power parity using panel data

20 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 9: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

on 21 industrialized countries over thefloating rate period and regressing five-,ten-, and twenty-year average exchangerate movements on average inflation dif-ferentials against the U.S.

III. Models of Exchange RateDetermination

Prior to the 1970s, the dominant inter-national macro model was, broadlyspeaking, an open Keynesian modelwhich had been developed in its essen-tials by James Meade (1951). The modelwas further developed in a series of pa-pers by Robert Mundell (e.g., Mundell1963) and J. Marcus Fleming (1962), andcame to be known as the Mundell-Fleming model. Although the integrationof asset markets and capital mobility intoopen-economy macroeconomics was animportant contribution of the Mundell-Fleming model, its treatment of assetmarket equilibrium is, however, inade-quate in that the stock-flow implicationsof interest rate differential changes arenot worked out.16 The distinguishingfeature of exchange rate models devel-oped during the 1970s is that they arebased on considerations of stock equilib-rium in international financial markets.

1. The Monetary Model I: Flexible Prices

The monetary approach to the ex-change rate, which emerged as the domi-nant exchange rate model at the start ofthe recent float in the early 1970s,17 starts from the definition of the ex-

change rate as the relative price of twomonies and attempts to model that rela-tive price in terms of the relative supplyof and demand for those monies. The de-mand for money, m, is assumed to de-pend on real income, y, the price level,p, and the level of the nominal interestrate, i (foreign variables are denoted byan asterisk). With all variables except in-terest rates expressed in logarithms,monetary equilibria in the domestic andforeign country respectively are given by

mt = pt + κyt − θit (10)

mt∗ = pt

∗ + κ∗yt∗ − θ∗it

∗. (11)

An important assumption in the flex-ible price monetary model is continuouspurchasing power parity. Setting β =− β* = 1 and normalizing the price indi-ces so that α = 0 in (8), the purchasingpower parity condition is:

st = pt − pt∗. (12)

The domestic money supply determinesthe domestic price level and hence theexchange rate is determined by relativemoney supplies. Solving (10), (11), and(12) for the exchange rate gives

st = mt − mt∗ − κyt + κ∗yt + θit − θ∗it

∗. (13)

Equation (13) is the fundamental flex-ible-price monetary equation. From (13),we can see that an increase in the do-mestic money supply, relative to the for-eign money stock, will lead to a rise inst—i.e., a depreciation of the domesticcurrency in terms of the foreign cur-rency. A rise in domestic real income,other things equal, creates an excess de-mand for the domestic money stock. Inan attempt to increase their real moneybalances, domestic residents reduce ex-penditure and prices fall until moneymarket equilibrium is achieved. Via pur-chasing power parity, falling domesticprices (with foreign prices constant) im-ply an appreciation of the domestic cur-

16 In his verbal exposition of his capital accounttheory, Meade had, in fact, worked through thestock equilibrium implications of a movement ininternational interest rate differentials, but did notfaithfully represent this feature in the mathemati-cal appendix to his volume. Mundell and Flemingfollowed Meade’s mathematical representationand thus abstracted from considerations of stockequilibrium.

17 See, for example, the studies collected to-gether in Frenkel and Johnson (1978).

Taylor: Exchange Rates 21

Page 10: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

rency in terms of the foreign currency.Similarly, a depreciation follows from anincrease in the domestic interest rate asthis reduces the domestic demand formoney.18

In practice, researchers often simplifythe model by imposing κ = κ* and θ = θ*

in (13). By invoking the uncovered inter-est parity condition we can then substi-tute ∆st+1

e for (it −it∗) in the resulting

equation to getst = mt − mt

∗ − κ(yt − yt∗) + θ∆st+1

e . (14)

The rational expectations solution to(14) is

st = (1 + θ) − 1∑ i=0

θ1 + θ

i

E[(m − m∗)t+i − κ(y − y∗)t+i|Ωt], (15)

where E[.|Ωt] denotes the mathematicalexpectation conditioned on the informa-tion set available at time t, Ωt. It is wellknown from the rational expectations lit-erature, however, that equation (15) isonly one solution to (14) from a poten-tially infinite set. If we denote the ex-change rate given by (15) by sct then (14)has multiple rational expectations solu-tions of the form

st = sct + ζt (16)

where the rational bubble term ζt satis-fies

E[ζt+1|Ωt] = θ − 1(1 + θ)ζt (17)

Rational bubbles represent significantdepartures from the fundamentals of themodel which would not be detected in aspecification such as (13). Thus, testingfor the presence of bubbles can be inter-

preted as an important specification testof the model (Richard Meese 1986).

Open economy macroeconomics is es-sentially about six aggregate markets:goods, labor, money, foreign exchange,domestic bonds (i.e., non-money assets)and foreign bonds. But the flexible-pricemonetary model concentrates directly onequilibrium conditions in only one ofthese markets—the money market. Thisis implicitly achieved in the followingfashion. By assuming perfect substitut-ability of domestic and foreign assets,the domestic and foreign bond marketsessentially become a single market. As-suming that the exchange rate adjustsfreely to equilibrate supply and demandin the foreign exchange market and alsoassuming equilibrium in the goods mar-ket (through perfectly flexible prices)and in the labor market (through flexiblewages) then implies equilibrium in threeof the five remaining markets. By Wal-ras’ law (the principle that equilibrium inn-1 markets of an n-market system im-plies equilibrium in the n-th market), theequilibrium of the full system is then de-termined by equilibrium conditions forthe money market. The flexible-pricemonetary model is thus, implicitly, amarket-clearing general equilibriummodel in which continuous purchasingpower parity among national price levelsis assumed.

The very high volatility of real ex-change rates during the 1970s float, con-spicuously refuting the assumption ofcontinuous purchasing power parity, ledto the development of two furtherclasses of models: sticky-price monetarymodels and equilibrium models.

2. The Monetary Model II: Sticky Prices and Overshooting

Sticky-price monetary models, origi-nally due to Dornbusch (1976), allowshort-term overshooting of the nominaland real exchange rates above their long-

18 Because the domestic interest rate is endo-genous in the flexible-price monetary model, how-ever, it is, in fact, not completely logical to con-sider increases in i which are independent ofmovements in i* or domestic or foreign monies orincomes.

22 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 11: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

run equilibrium levels. This results asthe “jump variables” in the system—ex-change rates and interest rates—com-pensate for stickiness in other vari-ables—notably goods prices. Considerthe effects of a cut in the nominal do-mestic money supply. Because goodsprices are sticky in the short run, this im-plies an initial fall in the real money sup-ply and a consequent rise in interestrates in order to clear the money market.The rise in domestic interest rates thenleads to a capital inflow and an apprecia-tion of the nominal exchange rate (a risein the value of domestic currency interms of foreign currency). Investors areaware that they are forcing the value ofthe domestic currency up artificially andthat they may therefore suffer a foreignexchange loss when the proceeds of theirinvestment are used to repay liabilities inforeign currency. However, so long asthe expected foreign exchange loss (ex-pected rate of depreciation) is less thanthe known capital market gain (the in-terest differential), risk-neutral investorswill continue to borrow abroad in orderto buy domestic assets. A short-run equi-librium is achieved when the expectedrate of depreciation is just equal to theinterest differential (uncovered interestrate parity holds). Because the expectedrate of depreciation must then be non-zero for a non-zero interest differential,the exchange rate must have overshot itslong-run equilibrium (purchasing powerparity) level. In the medium run, how-ever, domestic prices begin to fall in re-sponse to the fall in money supply. Thisalleviates pressure in the money market(the real money supply rises) and domes-tic interest rates begin to decline. Theexchange rate then depreciates slowly tolong-run purchasing power parity.19

The essential characteristics of thesticky-price monetary model can be seenin a three-equation structural model incontinuous time, holding foreign vari-ables and domestic income constant(these are simplifying rather than neces-sary assumptions):

s. = i − i∗ (18)

m = p + κy_

− θi (19)

p. = γ[α + µ(s − p) − ψi − y_

]. (20)

Equation (18) is the uncovered inter-est parity condition expressed in continu-ous time and utilizing certainty equiva-lence because of the linearity of themodel. Equation (19) is a domesticmoney-market equilibrium condition andequation (20) is a Phillips curve relation-ship, relating domestic price movementsto excess aggregate demand, where ag-gregate demand has an autonomous com-ponent, a component depending uponinternational competitiveness, and acomponent which is interest-rate sensi-tive. If we use a bar to denote a variablein long-run (noninflationary) equilib-rium, we can reduce this system to atwo-equation differential equation sys-tem:20

s.p.

=

0γ µ 1/θ

−γ (µ + ψ/θ)

s − s_

p − p__

. (21)

The coefficient matrix in (21) has anegative determinant and so the systemhas a unique convergent saddle path.The qualitative solution to (21) is shown

19 Frankel (1979) argues that the sticky-pricemonetary model should allow a role for seculardifferences between inflation rates. In his real in-terest differential variant of the sticky-price mone-

tary model the long-term interest rate differential isincluded as a proxy for the long-term inflation ratedifferential. Willem Buiter and Miller (1981) also de-velop a variant of the Dornbusch model with steady-state (“core”) inflation, and use it to analyze, inter alia,the dynamic effects of natural resource discoveries onoutput and exchange rates.

20 Note that the level of the money stock is ex-ogenous to the model, assumed under the controlof the authorities. Thus, for any given level of themoney stock, the perfect foresight equilibrium in-volves assuming mm = m.

Taylor: Exchange Rates 23

Page 12: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

in Figure 1, where the saddle path slopesdown from left to right in (s,p)-space.

Now consider again the effects of acut in the money supply. In the long run,the price level will be lower, at pm1 insteadof the initial level pm 0 in Figure 2, be-cause of the neutrality of money in thismodel. Because long-run purchasingpower parity holds in the model, andholding foreign prices constant, the long-run exchange rate will appreciate pro-portionately (i.e., s will be lower), mov-ing from sm0 to sm1 along the 45o ray. Thestable saddle path, which originally wentthrough point A must now go throughthe new long run equilibrium B. Becauseprices take time to adjust, however, theeconomy cannot jump directly from A toB. Instead, prices initially remain fixedand the exchange rate jumps to s2 inorder to get on to the new saddle path.Prices then adjust slowly and the econ-omy moves along the saddle path from Cto the new long run equilibrium B. Thenet effect of the money supply cut is along-run appreciation of sm0 − sm1, with aninitial overshoot of s2 − sm1.

3. Equilibrium Models and Liquidity Models

Equilibrium exchange rate models ofthe type developed originally by AlanStockman (1980) and Robert Lucas

(1982) analyze the general equilibriumof a two-country model by maximizingthe expected present value of a repre-sentative agent’s utility, subject tobudget constraints and cash-in-advanceconstraints (by convention, agents are re-quired to hold local currency, the ac-cepted medium of exchange, with whichto purchase goods).21 In an importantsense, equilibrium models are an exten-sion or generalization of the flexible-price monetary model to allow for multi-ple traded goods and real shocks acrosscountries. A simple equilibrium modelcan be sketched as follows. Consider atwo country, two good world in whichprices are flexible and markets are inequilibrium, as in the flexible-pricemonetary model, but in which, in con-trast to the monetary model, agents dis-tinguish between domestic and foreigngoods in terms of well-defined prefer-ences. Further, for simplicity, assumethat all agents, domestic or foreign, haveidentical homothetic preferences.22

p

Figure 1. The Qualitative Solution to the Sticky-Price Monetary Model

s- s

p-

p.=0

s.=0

p

Figure 2. Overshooting Following a Monetary Contraction

s-1 s

p-0

s-0s2

p-1

45º

AC

B

21 See Stockman (1987) for a more extensive,largely nontechnical exposition. This literature isan offshoot of the real business cycle literature.

22 For nonhomothetic preferences, many distur-bances create transfer-problem-like conditionswith unpredictable terms-of-trade effects.

24 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 13: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

Then, given domestic and foreign out-put of y and y* respectively, the equilib-rium relative price of foreign output,Π say, must be the slope of a repre-sentative agent’s indifference curve atthe point (y*/n,y/n) in foreign-domesticoutput per capita space (where n/2 is thenumber of individuals in each economy),as in Figure 3. But, the relative price offoreign output is the real exchange ratewhich is defined in logarithmic form (π)by (9). Now interpret expressions (10)and (11) as linearizations of expressionsderived from maximizing the repre-sentative agent’s utility function subjectto a cash-in-advance constraint, assum-ing that government policy (and other in-fluences on the constancy of parameters)remain constant. Combining (21) with(10) and (11) and letting, for ease ofexposition, κ = κ* and θ = θ* = 0, wehave

st = mt − mt∗ − κ(yt − yt

∗) + πt. (22)

Equation (22) is, in this very simpleformulation, the key equation determin-ing the nominal exchange rate in theequilibrium model, and illustrates thefact that the equilibrium model can beviewed as a generalization of the mone-

tary model.23 Indeed, relative monetaryexpansion leads to a depreciation of thedomestic currency (one-for-one in thissimple formulation) as in the simplemonetary model. However, some of theimplications of the equilibrium modelare either qualitatively or quantitativelydifferent from those of the flexible-pricemonetary model or else incapable ofanalysis within a purely monetary frame-work. As an example of the latter, con-sider an exogenous shift in preferencesaway from foreign goods toward domes-tic goods, represented as a flattening ofindifference curves as in Figure 4 (fromI1 to I2). With per capita outputs fixed,this implies a fall in the relative price offoreign output (or conversely a rise inthe relative price of domestic output)—Π falls (from Π1 to Π2 in Figure 4).24 Assuming unchanged monetary policies,this movement in the real exchange ratewill, however, be brought about entirely(and swiftly) by a movement in the nomi-nal exchange rate without any movementin national price levels. Thus, demand

domestic goods

Figure 3. Determination of the Real Exchange Rate in the Equilibrium Model

y/n

y*/n foreign goodstan-1∏

domestic goods

Figure 4. The Effect on the Real Exchange Rate of a Preference Shift Toward Foreign Goods

y/n

y*/n foreign goodstan-1∏1

I2I1

tan-1∏2

23 Because of the strict cash-in-advance con-straint, the assumed preferences, and the fixedoutput levels, intertemporal considerations are ab-sent from this very simple formulation.

24 Note that upper case pi denotes the real ex-change rate, lower case pi the logarithm of thereal exchange rate.

Taylor: Exchange Rates 25

Page 14: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

shifts are capable of explaining the ob-served volatility of nominal exchangerates in excess of volatility in relativeprices in equilibrium models. The fall ins, in this case matching the fall in π, willbe observed as a decline in domesticcompetitiveness. It would be a mistake,however, to infer that this rise in therelative price of domestic goods wascaused by the appreciation of the domes-tic currency: both are the result of theunderlying demand shift.

In this simple equilibrium model, anincrease in domestic productivity (outputper capita) has two analytically separateeffects. The first effect—the “relativeprice effect”—involves a reduction in therelative price of domestic output and soan increase in π which, through (22), willtend to raise s (depreciate the domesticcurrency). The second effect—the“money demand effect”—will tend to ap-preciate the domestic currency (i.e., de-press s) as the transactions demand formoney rises, exactly as in the monetarymodel. Whether the exchange rate risesor falls depends upon the relative size ofthese effects which, in turn, dependsupon the degree of substitutability be-tween domestic and foreign goods—thehigher the degree of substitutability, thesmaller the relative price effect. Thus,supply disturbances will generate volatil-ity of nominal exchange rates in excess ofthe volatility of relative prices only if thedegree of substitutability between do-mestic and foreign goods is relativelysmall (Obstfeld and Stockman 1985).25

In the very recent literature on liquid-ity models of the exchange rate, someauthors have extended the equilibriummodel framework by incorporating anextra cash-in-advance constraint onagents. In these models, agents arerequired to hold cash not only withwhich to purchase goods but also withwhich to purchase assets (as originallysuggested in a closed-economy contextby Lucas, 1990). In the two-countrymodel of Vittorio Grilli and Roubini(1992), for instance, the money supplyand bond issue of each country arelinked through the government budgetconstraint, and agents must decide onhow much domestic and foreign cur-rency to hold in order to purchase do-mestic and foreign goods and assets.Once this decision is made, subsequentshocks to the bond and money supplieswill affect nominal interest rates (in or-der to clear bond markets) and also, be-cause the expected rate of monetarygrowth (as opposed to the level of themoney supply) and hence expected infla-tion is unaffected, will affect real inter-est rates. This in turn affects the nominaland real exchange rates. It is interestingto contrast liquidity models with thesticky-price monetary model, becausethe latter assumes sticky goods pricesand instantaneous portfolio adjustment,while liquidity models essentially assumeslow portfolio adjustment and perfectlyflexible goods prices. Many of the impli-cations of the two types of models aresimilar, and they are, for example, obser-vationally equivalent with respect to theimpact of monetary shocks: A positiveshock to the money supply generates adecline in real and nominal interest ratesin both models, the domestic currencyappreciates against the foreign currencyin both real and nominal terms, and out-put rises (in response to lower real inter-est rates) until prices and portfolios areagain in equilibrium.

25 In the simple equilibrium model we havesketched here, we have implicitly made a host ofsimplifying assumptions. Chief among these is theassumption that individuals in either economyhold exactly the same fractions of their wealth inany firm, domestic or foreign. If this assumption isviolated, then supply and demand shifts will alterthe relative distribution of wealth between domes-tic and foreign residents as, for example, onecountry becomes relatively more productive. This,in turn, will affect the equilibrium level of the ex-change rate (Stockman 1987).

26 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 15: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

4. The Portfolio Balance Model

The key distinguishing feature of theportfolio balance model is the assumedimperfect substitutability between do-mestic and foreign assets.26 We do nothave space here to set out a fully speci-fied portfolio balance model, but we cangive a flavor of this class of models bysetting out a very basic model and trac-ing through a simple example. Considera simple model in which the net financialwealth of the private sector (W) is di-vided into three components: money(M), domestically issued bonds (B) andforeign bonds denominated in foreigncurrency and held by domestic residents(B*). B can be thought of as governmentdebt held by the domestic private sector;B* is the level of net claims on foreignersheld by the private sector. Because, un-der a free float, a current account sur-plus on the balance of payments must beexactly matched by a capital accountdeficit (i.e., capital outflow and hence anincrease in net foreign indebtedness tothe domestic economy), the currentaccount must give the rate of accumula-tion of B* over time. With foreign anddomestic interest rates given by i and i*

as before, we can write down our defini-tion of wealth and simple domestic de-mand functions for its components asfollows:27

W ≡ M + B + SB∗ (23)

M = M(i,i∗ + cS e)W M1 < 0, M2 < 0 (24)

B = B(i,i∗ + cS e)W B1 > 0, B2 < 0 (25)

SB∗ = B∗( i, i∗ + cS e)W B1

∗ < 0, B2∗ > 0 (26)

B.

∗ = T(S/P) + i∗B∗ T1 > 0 (27)

where Sce denotes the expected rate ofdepreciation of the domestic currency.Relation (23) is an identity definingwealth and (24), (25), and (26) are stan-dard asset demand functions.28 Equation(27) gives the rate of change of B*, thecapital account, as equal to the currentaccount which is in turn equal to thesum of the trade balance, T(.), and netdebt service receipts, i*B*. The tradebalance depends positively on the levelof the real exchange rate (a devaluationimproves the trade balance). For sim-plicity, assume static expectations, i.e.,Sc e = 0. Now consider an open marketpurchase of domestic bonds by theauthorities, paid for by printing money.In order to induce agents to hold moremoney and fewer bonds, the domestic in-terest rate falls (the price of domesticbonds rises) and, as agents attempt tocompensate for the reduction in theirportfolios of domestic interest-bearingassets by buying foreign bonds, the ex-change rate will depreciate, driving upthe domestic currency value of foreignbonds. The net impact effect is a lowerdomestic interest rate and a depreciatedcurrency, say from S0 to S1 (AC) in Fig-ure 5. Suppose that the economy was in-itially in equilibrium with a trade bal-ance of zero and net foreign assets ofzero (and hence a current account ofzero). This is depicted in Figure 5 at thepoint corresponding to time t0. Figure 5is drawn so that the initial (t0) values ofthe price level and the exchange rate arenormalized to unity. Assuming that the

26 A comprehensive treatment of the portfoliobalance model is given in Branson and Dale Hen-derson (1985).

27 Xk denotes the partial derivative of X(.) withrespect to the k-th argument, for X=M, B, B*, andT. The shift to upper case letters here indicatesthat variables are in levels rather than logarithms.As throughout, interest rates are in percentageterms.

28 Note that, as is standard in most expositionsof the portfolio balance model, the scale variableis the level of wealth, W, and the demand func-tions are homogeneous in wealth; this allows themto be written in nominal terms (assuming homoge-neity in prices and real wealth, prices cancel out).Note, however, that goods prices are indetermi-nate in this model. In what follows we assumelong-run neutrality.

Taylor: Exchange Rates 27

Page 16: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

Marshall-Lerner condition holds, the im-provement in competitiveness will im-prove the trade balance from zero to apositive amount (FG) (ignoring any “J-curve” effects). This means that the cur-rent account goes into surplus and do-mestic residents begin to acquire netforeign assets; equation (27). Residentswill then attempt to sell some of theirforeign assets in order to rebalance theirportfolios, and the exchange rate will be-gin to appreciate—in the diagram frompoint C along CD. This erosion of com-petitiveness will lead to a deteriorationof the trade balance along GH. Mean-while, the increase in the money supplywill have begun to increase prices alongthe path AB toward the new long runequilibrium price level P1; this adds tothe deterioration of competitiveness andhence the trade balance. At point E(time t1), the exchange rate and the pricelevel are equal in value and hence theirratio is unity—the same as the initial ra-tio at time t0. Because we have implicitlyheld foreign prices constant, this meansthat the real exchange rate is back to itsoriginal level and so the trade balance

must also be back to its original level,i.e., zero. This is no longer enough to re-store long-run equilibrium, however.Domestic residents have now acquired apositive level of net foreign assets andwill be receiving a stream of interest in-come i*B* from abroad. Thus, they arestill acquiring foreign assets—equation(27)—and so the exchange rate is still ap-preciating as agents attempt to rebalancetheir portfolios and sell these foreign as-sets. In order for the current accountbalance to be zero, the trade balancemust actually go into deficit. This re-quires a further appreciation of the ex-change rate to its long run equilibriumlevel S2, by which time the price levelhas reached its long run equilibriumlevel P1 and the current account just bal-ances (−T(S2/P1) = i*B*) so that there isno further net accumulation of foreignassets. The overall effect of the openmarket purchase on the exchange rate isa long-run depreciation from S0 to S2,with an initial overshoot of S1 − S2.29

IV. The Empirical Evidence onExchange Rate Models

1. Testing the Monetary Class of Models

In an influential paper, Frenkel (1976)finds evidence strongly supportive of theflexible-price monetary model for theGerman mark-U.S. dollar exchange rateduring the German hyperinflation of the

P0,S0= 1

Figure 5. The Dynamics of an Open-Market Purchase of Domestic Bonds in the Portfolio

Balance Model

Time

0

S2

P1

S1

S,P

H

B

D

E

C

K

A

F

G

t0 t1

Trade balance

29 Our exposition of the portfolio balance modelassumes that exchange rate expectations are static,so that the expected rate of depreciation is zero.In fact, most of the properties of the model re-main intact when rational expectations are intro-duced, except that impact effects become muchmore pronounced and a key distinction must nowbe drawn between anticipated disturbances (theeffect of which will already be discounted into thecurrent exchange rate level) and unanticipated dis-turbances (which require an initial jump in the ex-change rate and then a slow adjustment to the newequilibrium). A more completely specified modelwould also allow for interaction between produc-tion, consumption, saving, and the level of wealth.See Branson and Henderson (1985).

28 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 17: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

1920s. The subsequent accumulation ofdata during the 1970s allowed estimationof the model for the major exchangerates during the recent float, and theseearly studies were also broadly suppor-tive of the flexible-price monetary model(e.g., Bilson 1978; Dornbusch 1979). Be-yond the late 1970s, however, the flex-ible-price monetary model (or its real in-terest differential variant) ceases toprovide a good explanation of variationsin exchange rate data: the estimatedequations break down, providing poorfits, exhibiting incorrectly signed coeffi-cients and failing general equation diag-nostics (Frankel 1993b). In particular,estimates of equations for the dollar-mark rate often produce coefficientswhich imply that increases in Germany’smoney supply during this period causedits currency to appreciate.30 Someauthors sought to explain this breakdownas due to econometric misspecification,while others argued that large currentaccount deficits or surpluses during theperiod generated important wealth ef-fects which are not adequately capturedin the simple monetary model (e.g.,Frankel 1982a, 1993b).

The evidence for the sticky-pricemonetary model is also weak when thedata period is extended beyond the late1970s. For example, while RobertDriskell (1981), reports single-equationestimation results largely favorable tothe sticky-price monetary model for theSwiss franc-U.S. dollar, 1973–77, DavidBackus (1984) finds little support usingU.S.-Canadian data for the period 1971–80. Another implication of the sticky-price monetary model is proportionalvariation between the real exchange rateand the real interest rate differential.This follows from the basic assumptionsof the overshooting model: slowly adjust-

ing prices and uncovered interest rateparity, and can be derived by subtractingrelative inflation from either side ofequation (1). A number of studies havefailed to find strong evidence of this re-lationship, notably Meese and Rogoff(1988) who could not find cointegrationbetween real exchange rates and real in-terest rate differentials. (See also Edisonand Dianne Pauls 1993.) Recent worksuggests, however, that this may be dueto omitted variables which are determi-nants of the equilibrium real exchangerate or the risk premium (e.g., AdrianThroop 1993). Marianne Baxter (1994)uses band-spectral regression techniquesand finds that that there may be a signifi-cant positive correlation between real in-terest rate differentials and real ex-change rate changes at “business-cycle”(six to thirty-two quarters) and “trend”(more than thirty-two quarters) frequen-cies.

More recently, MacDonald and Taylor(1993, forthcoming) apply multivariatecointegration analysis and dynamic mod-eling techniques to a number of ex-change rates and find some evidence tosupport the monetary model as a long-run equilibrium toward which the ex-change rate converges, while allowingfor complicated short-run dynamics. Be-cause all of the monetary models col-lapse to an equilibrium condition of theform (13) in the long run, these testshave no power to discriminate betweenthe alternative varieties. The usefulnessof the cointegration approach suggestedby these studies should, moreover, betaken as at most tentative: their robust-ness across different data periods and ex-change rates has yet to be demonstrated.

Robert Flood and Andrew Rose(1993), observing the increased volatilityof exchange rates under floating as op-posed to fixed exchange rate regimes,argue that any tentatively adequate ex-change rate model should have funda-

30 Frankel (1982a) called this the “mystery ofthe multiplying marks.”

Taylor: Exchange Rates 29

Page 18: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

mentals which are also much more vola-tile during floating rate regimes. In fact,these authors find little shift in the vola-tility of economic fundamentals sug-gested by flexible-price or sticky-pricemonetary models across different nomi-nal exchange rate regimes for a numberof OECD exchange rates. Similar evi-dence is reported by Baxter and Stock-man (1989), who examine the time-seriesbehavior of a number of key macro-eco-nomic aggregates for 49 countries overthe postwar period. Although they detectevidence of increased real exchange ratevariability under flexible exchange ratesthan under pegged nominal exchangerate systems, Baxter and Stockman findno systematic differences in the behaviorof the macroeconomic aggregates underalternative exchange rate arrangements.Again, this suggests that there are specu-lative forces at work in the foreign ex-change market which are not reflected inthe usual menu of macroeconomic fun-damentals.31

2. Testing the Portfolio Balance Model

Much less empirical work has beencarried out on the portfolio balance ap-proach to the exchange rate than on themonetary class of models, presumablybecause of the problems which re-searchers have encountered in mappingtheoretical portfolio balance models intoreal-world financial data. This clearlyraises important methodological issuessuch as what non-money assets to in-clude in the empirical model as well asimportant practical issues such aswhether the data is actually available, es-

pecially on a bilateral basis. Log-linearversions of reduced-form portfolio bal-ance exchange rate equations, using cu-mulated current accounts for the stockof foreign assets, have, however, beenestimated for many of the major ex-change rates for the 1970s float, withpoor results: estimated coefficients areoften insignificant and there is a persist-ent problem of residual autocorrelation(e.g., Branson, Hannu Halttunen, andPaul Masson 1977).32

The imperfect substitutability of do-mestic and foreign assets which is as-sumed in the portfolio balance model isequivalent to assuming that there is arisk premium separating expected depre-ciation and the domestic-foreign interestdifferential, and in the portfolio balancemodel this risk premium will be a func-tion of relative domestic and foreigndebt outstanding. An alternative, indirectmethod of testing the portfolio balancemodel, therefore, is to test for empiri-cal relationships of this kind. Investiga-tions of this kind have usually reportedstatistically insignificant relationships(Frankel 1982b; Rogoff 1984). In a re-cent study of the effectiveness of ex-change rate intervention for dollar-markand dollar-Swiss franc during the 1980s,Kathryn Dominguez and Frankel (1993b)measure the risk premium using surveydata and show that the resulting measurecan in fact be explained by an empiricalmodel which is consistent with the port-

31 A number of studies have noted that, underthe recent float, nominal exchange rates haveshown much greater variability than importantmacroeconomic fundamentals such as price levelsand real incomes—see e.g., Dornbusch andFrankel (1988), Frankel and Froot (1990a), Rich-ard Marston (1989).

32 Dooley and Isard (1982) estimate the portfo-lio-balance model under rational expectations forthe dollar-mark exchange rate. While their esti-mated model provides predictions which are supe-rior to the forward rate as a spot-rate predictor,the coefficient of correlation between observedand predicted changes in the exchange rate is only0.4. The authors conclude that beating the for-ward rate is evidence in favor of the existence ofrisk premia, while the poor goodness of fit sug-gests that their simplifications of the portfolio bal-ance model may be too severe, or perhaps thatthere are substantial unexpected shifts in the un-derlying macroeconomic variables.

30 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 19: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

folio balance model with the additionalassumption of mean-variance optimiza-tion on the part of investors. In someways, the relative success of the Domin-guez and Frankel (1993b) study is consis-tent with the recent empirical literatureon foreign exchange market efficiency,discussed in Section II, which suggeststhe existence of significant foreign ex-change risk premia and non-rational ex-pectations.

3. Testing Equilibrium and Liquidity Models

In order to specify and solve an equi-librium model it is necessary to make aset of assumptions, such as uniform pref-erences or a specific utility function,which no one would seriously expect tohold exactly in the real world, eventhough the qualitative predictions of themodel may be valid. Thus, these modelsare not amenable to direct econometrictesting. Rather, researchers have soughtto test the broad rather than specific im-plications of this class of models for ex-change rate behavior.

Stylized facts or characteristics of therecent float include the high volatility ofreal exchange rates, the very high corre-lation of changes in real and nominalexchange rates and the absence ofstrongly mean-reverting properties ineither series. As we discussed above,equilibrium models are capable of ex-plaining the variability of nominal ex-change rates in excess of relative pricevariability (and hence the variability ofreal exchange rates), but so is the sticky-price monetary model. Some authorshave argued, however, that the difficultyresearchers have experienced in reject-ing the hypothesis of non-stationarity inthe real exchange rate is evidenceagainst the sticky-price model and infavor of equilibrium models. Explain-ing the persistence in both real and

nominal exchange rates over the recentfloat within the framework of the sticky-price model, it is argued, involves as-suming either implausibly sluggish priceadjustment or else that movements innominal exchange rates are due largelyto permanent real disturbances (Stock-man 1987). Equilibrium models, on theother hand, are not contradicted bypersistence in real and nominal ex-change rate movements. In the simpleequilibrium model sketched above, forexample, permanent shocks to technol-ogy could affect the slope of the “budgetline” in Figure 3 and so permanentlyaffect the equilibrium real exchangerate. Moreover, as we demonstratedwith the aid of Figure 4, a permanentshift in preferences could lead, in thismodel, to a permanent shift in nominaland real exchange rates with domesticprice levels unchanged—illustrating howequilibrium models are consistent withpersistent, correlated movements in realand nominal exchange rates. There aretwo basic responses to this line of argu-ment. First, as Frankel (1990) arguesforcibly, noncontradiction is not thesame as confirmation: simply being con-sistent with the facts is not enough todemonstrate the empirical validity of atheory. Secondly, as discussed above,there is now emerging evidence that realexchange rates may, in fact, be mean re-verting.33

One testable implication of the sim-plest equilibrium models is the neutral-ity of the exchange rate with respect tothe exchange rate regime: because thereal exchange rate is determined by realvariables such as tastes and technology,its behavior ought to be independent ofwhether the nominal exchange rate is

33 Note, however, that mean reversion in thereal exchange rate is not inconsistent with thebroad class of equilibrium models, because it ispossible that real shocks themselves are mean re-verting, e.g., because of technology transfer.

Taylor: Exchange Rates 31

Page 20: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

pegged or allowed to float freely. Al-though the major real exchange rateshave been demonstrably more volatileduring the recent floating-rate period,this may be due to the greater variabilityof underlying real shocks during this pe-riod. Stockman’s (1983) study of 38countries over a variety of time periods(including, for example, countries whosecurrencies remained pegged to the dollarafter 1973) does conclude, however, thatreal exchange rates are significantlymore volatile under floating nominal rateregimes; see also Mussa (1986) and Bax-ter and Stockman (1989). This evidencedoes indeed, constitute a rejection of thesimplest equilibrium models, although itis possible that the evidence is to someextent confounded by the endogeneity ofthe choice of exchange rate regime—i.e.,countries experiencing greater real dis-turbances are more likely to chooseflexible exchange rate systems. More-over, Stockman (1983) also shows thatthe assumptions necessary for regime-neutrality are in fact quite restrictive in afully specified equilibrium model, andinclude Ricardian equivalence, nowealth-distribution effects of nominalprice changes, no real effects of infla-tion, no real effects of changes in thelevel of the money supply, complete as-set markets, completely flexible prices,and identical sets of government policiesunder different exchange-rate systems.Because it is unlikely that all of theseconditions will be met in practice, Stock-man argues that only the simplest classof equilibrium models should be re-jected, and that equilibrium modelsshould be developed which relax some orall of these assumptions. In a more re-cent paper, Stockman (1988) notes thatcountries with fixed exchange rates aremore likely to introduce controls ontrade or capital flows in order to controllosses of international reserves. Thus, adisturbance that would tend to raise the

relative price of foreign goods (e.g., apreference shift toward foreign goods)will raise the probability that the domes-tic country will, at some future point, im-pose capital or trade restrictions that willraise the future relative world price ofdomestic goods. With intertemporal sub-stitution, this induces a higher world de-mand for domestic goods now, serving tooffset partly the direct effect of the dis-turbance, which was to raise the relativeprice of the foreign good, and hence toreduce the resulting movement in thereal exchange rate. Thus, countries withpegged exchange rates will experiencelower volatility in the real exchange ratethan countries with flexible exchangerates.

Empirical studies of the implicationsof liquidity models include MartinEichenbaum and Charles Evans (1993)(for the U.S.) and Grilli and Roubini(1993) (for the other G-7 countries).These researchers provide evidence thatunanticipated monetary contractionslead to increases in the level of domesticinterest rates and an appreciation of thedomestic currency in both real and nomi-nal terms, which is inconsistent withmost equilibrium models in which nomi-nal shocks should not affect real vari-ables, but is consistent with liquiditymodels with asset-market cash-in-ad-vance constraints.34 As discussed in Sec-tion III.3, however, sticky-price mone-tary models and liquidity models areobservationally equivalent in this re-spect.

Overall, although the empirical evi-dence rejects the very simplest equilib-rium models, it is not possible at thisstage to draw any firm conclusions con-cerning the empirical validity of thewhole class of equilibrium or liquiditymodels.

34 These papers are examples of “news” studiesdiscussed more generally in the next subsection.

32 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 21: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

4. News Effects and Exchange Rate Movements35

There is a strand of the literaturewhich has sought to test the relevance ofeconomic fundamentals for exchangerate movements through an examinationof the effect of “news” about fundamen-tals on unexpected exchange rate move-ments. If the market is efficient, the un-expected exchange rate change, (st+k −se

t+k), can only be due to news arrivingbetween the time the expectation wasformed and time t + k. Thus, if thesenews effects could be measured, theyshould be significantly correlated withthe unexpected change. Empirical imple-mentation of this approach involveschoosing a vector of variables, zt say,which are thought to affect the exchangerate, getting a measure of agents’ expec-tation of zt+k based on information attime t, ze

t+k, and measuring news aboutfundamentals as (zt+k − ze

t+k). A regres-sion of (st+k − se

t+k) onto the news termshould then yield a significant estimatedcoefficient. Typically, researchers haveused the interest rate differential or haveused exchange rate theory to choose zt,and have used either time-series meth-ods (Frenkel 1981b) or survey data(Dornbusch 1980) to form ze

t+k. The ex-pected value of st+k has most often beentaken as equal to the forward rate attime t (thus implicitly assuming a zerorisk premium). An advantage of the newsapproach is that it allows one to test theinfluence of the underlying fundamen-tals without having to specify a precisefunctional form for the exchange rateequation. Using this approach and a vari-ety of choices of elements of zt, a num-ber of researchers have reported signifi-cant news effects, thus indicating theimportance of fundamentals in explain-

ing exchange rate movements (e.g.,Dornbusch 1980; Sebastian Edwards1982). Some researchers have found,however, that the full effects of news onthe exchange rate are often observedwith a lag (e.g., Eichenbaum and Evans1993).36

5. The Out-of-Sample Forecasting Performance of Exchange Rate Models

Another way of examining the empiri-cal content of exchange rate theories isto examine their out-of-sample forecast-ing performance. In a landmark paper,Meese and Rogoff (1983a) compare theout-of-sample forecasts produced byvarious exchange rate models with fore-casts produced by a random walk model,by the forward exchange rate, by a uni-variate regression of the spot rate, andby a vector autoregression. They use roll-ing regressions to generate a successionof out-of-sample forecasts for eachmodel and for various time horizons. Theconclusion which emerges from thisstudy is that, on a comparison of rootmean square errors (RMSEs), none ofthe asset-market exchange rate modelsoutperforms the simple random walk,even though actual future values of theright-hand-side variables are allowed inthe dynamic forecasts (thereby giving themodels a very large informational advan-tage).37 Further work by the sameauthors (Meese and Rogoff 1983b) sug-

35 See Frankel and Rose (1994) for a recentmore comprehensive survey of work on “news”and foreign exchange markets.

36 Frankel and Rose (1994) point out that slowadjustment of the exchange rate to monetary sur-prises in a sticky-price monetary model might ex-plain the negative correlation of forward rate fore-cast errors and forward premia, discussed inSection II.

37 Meese and Rogoff compare random-walkforecasts with those produced by the flexible-pricemonetary model, Frankel’s (1979) real interestrate differential variant of the monetary model,and a synthesis of the monetary and portfolio bal-ance models suggested by Peter Hooper and JohnMorton (1982).

Taylor: Exchange Rates 33

Page 22: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

gested that the estimated models mayhave been affected by simultaneity bias.Imposing coefficient constraints takenfrom the empirical literature on moneydemand, Meese and Rogoff find that al-though the coefficient-constrained assetreduced forms still fail to outperform therandom walk model for most horizons upto a year, combinations of parameterconstraints can be found such that themodels do outperform the random walkmodel for horizons beyond twelvemonths. Even at these longer horizons,however, the models are unstable in thesense that the minimum RMSE modelshave different coefficient values at dif-ferent horizons.

Although beating the random walk stillremains the standard metric by which tojudge empirical exchange rate models,researchers have found that one key toimproving forecast performance basedon economic fundamentals lies in the in-troduction of equation dynamics. Thishas been done in various ways: by usingdynamic forecasting equations for theforcing variables in the forward-looking,rational expectations version of the flex-ible-price monetary model, by incorpo-rating dynamic partial adjustment termsinto the estimating equation, by usingtime-varying parameter estimation tech-niques, and—most recently—by usingdynamic error correction forms.38 A par-ticularly interesting example of the latteris Mark’s (1993) study of long-horizonpredictability. Mark estimates equationsof the form:

∆kst + k = α + β(zt − st) + νt + k (28)

where zt is the exchange rate fundamen-tal suggested by the monetary class ofmodels, zt ≡ [(mt − mt

∗) − κ(yt − yt∗)] , and

νt+k is a disturbance term.39 In a seriesof forecasting tests over very long hori-zons for a number of quarterly dollar ex-change rates, Mark finds that (28) is agood predictor of the nominal exchangerate, particularly at the four-year hori-zon. Moreover, both the goodness of in-sample fit and the estimated value of βkrise as the horizon k rises. Mark inter-prets this as evidence that, while quar-ter-to-quarter exchange rate movementsmay be noisy, systematic movements re-lated to the fundamentals become appar-ent in long-horizon changes.

V. Official Intervention

Official intervention in foreign ex-change markets occurs when the authori-ties buy or sell foreign exchange, nor-mally against their own currency and inorder to affect the exchange rate. Steril-ized intervention occurs when theauthorities—simultaneously or with avery short lag—take action to offset or“sterilize” the effects of the resultingchange in official foreign asset holdingson the domestic monetary base. The ex-change rate effects of intervention—inparticular sterilized intervention—havebeen an issue of some debate in the lit-erature. At the outset of the currentfloat, in the early 1970s, a “clean” ratherthan a “dirty” or managed float was fa-vored. Then, in the late 1970s, the U.S.was frequently criticized for its reluc-tance to intervene in support of the dol-lar. By the early 1980s, however, the pre-vailing consensus among economists,policy makers, and foreign exchangemarket practitioners appeared to be thatintervention, particularly sterilized inter-vention, could have at most very small

38 E.g., Throop (1993), MacDonald and Taylor(1993, forthcoming). Kees Koedijk and PeterSchotman (1990) estimate an error-correction realexchange rate equation and show that this is supe-rior, in-sample, to a random-walk model.

39 Mark chooses a priori values of λ of 1.0 and0.3. Equation (28) can in fact be derived from theDornbusch (1976) model, and Mark (1993) showsthat β tends in that case to unity as the horizon, k,gets longer.

34 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 23: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

and transitory effects on the exchangerate. This view was enshrined in the con-clusions of the Jurgensen Report (1983),commissioned by the 1982 G-7 Eco-nomic Summit of Heads of Governmentat Versailles. After the meeting of theG-5 economic leaders at the Plaza Hotelin New York in September 1985, how-ever, official views on the usefulness ofintervention appeared to switch again.Since that time, intervention in the mar-kets for the major exchange rates hasbeen regular and, at times, heavy.40

1. The Channels of Influence

Sterilized intervention may influencethe exchange rate through either of twochannels: by changing the relative sup-plies of assets, and by signaling policy in-tentions. Its effects through the firstchannel can be analyzed within theframework of the portfolio balancemodel of exchange rate determinationwhich was outlined above. Suppose, forexample, that the authorities purchaseforeign exchange and carry out an openmarket sale of domestic bonds in orderto sterilize the effect of the rise in offi-cial reserves on the money supply. If do-mestic and foreign bonds are perfectsubstitutes in private agents’ portfolios(so that the portfolio balance model es-sentially collapses to a monetary model),and assuming that agents’ portfolioswere initially in stock equilibrium, thenthey will sell foreign bonds one for onewith the increase in domestic bonds.Thus, the private sector will sell thesame amount of foreign currency thatthe authorities bought, and there will beno net effect on the level of the ex-change rate.41 If domestic and foreign

bonds are less than perfectly substitut-able, however, private sales of foreignbonds will be less than the increase inthe stock of bonds privately held, and sothe intervention will have a net effect onthe level of the exchange rate.

The second channel of influence—thesignaling or expectations channel (Mussa1981)—allows for the intervention to af-fect exchange rates by providing themarket with relevant information thatwas previously not known or incorpo-rated into the current exchange rate.This presumes that the authorities havesuperior information to other marketparticipants and that they are willing toreveal this information through their ac-tions in the foreign exchange market.Even in a simple flexible-price monetarymodel, sterilized intervention could af-fect the exchange rate through the sig-naling channel by altering agents’ ex-change rate expectations about futuremovements in relative money or income,which then feeds back into the currentexchange rate; see equation (15).

2. The Empirical Evidence on the Effectiveness of Intervention42

Empirical work on the effectiveness ofintervention via the portfolio balancechannel has generally taken one of thefollowing forms. First, researchers havedirectly estimated structural asset de-mand equations of portfolio balancemodels (Obstfeld 1983; Lewis 1988).Secondly, researchers have estimated in-verted asset demand equations wherethe ex post difference in the rate of re-turn between domestic and foreign as-sets is regressed onto a range of vari-ables: under the joint null hypothesis ofperfect substitutability and rational ex-pectations, the estimated regression co-efficients should be insignificantly differ-

40 Dominguez and Frankel (1993a) provide anup to date account of the history of official inter-vention over the recent float.

41 See Kenen (1982) for a thorough analysis ofthe long- and short-run effects of intervention inthe framework of the portfolio balance model.

42 Recent surveys of the empirical literature onforeign exchange market intervention includeEdison (1993) and Lewis (forthcoming).

Taylor: Exchange Rates 35

Page 24: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

ent from zero (Rogoff 1984; DeborahDanker et al. 1987). A third approach in-volves estimating asset demand equa-tions derived in a specific optimizationframework, such as mean-variance analy-sis (Frankel 1982b). Much of this litera-ture has found difficulty in rejecting thehypothesis that the exchange rate effectsof intervention through the portfolio bal-ance channel are very small and at bestvery short lived. A recent treatment byObstfeld (1990), for example, argues thatsterilized intervention in itself has notplayed an important role in promotingexchange rate realignment since the1985 Plaza Accord, but rather that re-alignments have occurred as the result ofappropriate macro-policy coordination.This conventional wisdom has been chal-lenged in several recent papers. Domin-guez and Frankel (1993b), for instance,using survey data on dollar-mark anddollar-Swiss franc exchange rate expecta-tions to construct measures of the riskpremium as the deviation from uncov-ered interest rate parity [(it − it*) −∆se

t+k] (see equation (6)) find that inter-vention variables have significant ex-planatory power for the risk premium, inaddition to terms in the second momentsof exchange rate changes.

Tests of the influence of sterilized in-tervention through the signaling channelinvolve testing for the significance of in-tervention variables in equations relatingto exchange rate expectations. Studies ofthe signaling effect have generally pro-duced very mixed results (Owen Hum-page 1989; Dominguez 1986). In a re-cent study, Dominguez and Frankel(1993c), again using survey data on dol-lar-mark exchange rate expectations, findthat official announcements of exchangerate policy and reported intervention sig-nificantly affect exchange rate expecta-tions and that, overall, the effectivenessof intervention, in terms of significantlyaffecting both weekly and daily exchange

rate changes, is very much enhanced if itis publicly announced.

Although Dominguez and Frankel(1993b, 1993c) offer no explanation as towhy their results concerning the effec-tiveness of intervention, through eitherchannel, stand in contrast to much of theearlier literature, Edison (1993) pointsout that the more obvious distinguishingfactors include the use of survey datarather than reliance on an assumption ofrational expectations, the use of a bilat-eral rather than a multi-currency model,and concentration on data for the 1980s.

Each of the Dominguez and Frankelstudies cited above employs daily inter-vention data obtained from the U.S. andGerman authorities. Pietro Catte, Giam-paolo Galli, and Salvatore Rebecchini(1992) additionally employ daily data onintervention by the Japanese authoritiesand carry out a statistical analysis of co-ordinated G-3 intervention over the1985–1991 period. Of seventeen epi-sodes of coordinated intervention identi-fied over this period, the authors claimthat all were successful in the sense ofreversing the trend in the dollar and, inthe case of the Plaza episode (late 1985),making it resume its fall. Out of ten ma-jor turning points in the dollar-mark ex-change rate over the period, the authorsidentify nine of these as coinciding ex-actly with periods of concerted interven-tion. Note that Catte et al. make no at-tempt to disentangle sterilized fromunsterilized intervention; neither do theyattempt to disentangle the portfolio bal-ance and signalling effects. In interpret-ing their results, however, Catte et al. doseem to favor the signaling channel ex-planation. Given the magnitude of un-certainty about the link between ex-change rates and fundamentals, theysuggest that the signals provided by in-tervention may help coordinate agents’expectations, inducing them to convergeon a particular model of the economy

36 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 25: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

and pick a value of the exchange ratethat is similar to that targeted by theauthorities.

Overall, therefore, the evidence on theeffectiveness of official intervention isunclear, although some recent studies dosuggest a significant link.

VI. Exchange Rates and Target Zones

A “target zone” is a range within whichthe authorities are committed to keepingthe nominal exchange rate. The Ex-change Rate Mechanism of the Euro-pean Monetary System is an example of amultilateral target zone, the theory ofwhich has not yet been fully worked out.The best examples of unilateral targetzones, to which much of existing targetzone theory applies, are those pursuedby the three Nordic countries outside ofthe ERM: Finland, Norway, and Swe-den. Following an original paper by PaulKrugman, circulated since 1987 and pub-lished in 1991 (Krugman 1991), a sub-stantial literature on this topic has ap-peared with remarkable speed.43

1. The Basic Target Zone Model

Consider a flexible-price monetarymodel of the exchange rate expressed incontinuous time:

s = m + v + θE[dsΩ(t)]/dt. (29)

Equation (29) is the continuous-timeanalogue of equation (14) under the as-sumption of rational expectations, wherev lumps together all of the right-hand-side fundamentals in (14) beside domes-tic money and the expected exchangerate change.44 The money supply is as-sumed to be a policy variable under the

control of the authorities. The shift vari-able, v, is assumed to follow a Brownianmotion, which is the continuous-timeanalogue of a random walk. Under a freefloat, the authorities are assumed not toalter m to offset movements in v, so that,from (29), s itself follows a Brownianmotion process and the expected changein the exchange rate is zero,E[ds|Ω(t)]/dt=0. Thus, in a plot of the ex-change rate against the fundamentals,m + v, s would lie on a 45o ray (the “free-float” line FF in Figure 6). Under a fixedexchange rate regime, the authoritieswould alter m in order to offset move-ments in v and the expected exchangerate change would again be zero but aplot of s against m + v would be concen-trated at a single point. In the basic tar-get zone model, the authorities standready to intervene at the upper (smax)and lower (smin) edges of the band,where they will alter the level of the fun-damentals (i.e., change the level of m)just enough to keep the exchange ratewithin the band.45 Technically, the solu-tion to the basic target zone model im-plies that the relationship between theexchange rate and the fundamentals is

43 Svensson (1992) provides a comprehensivesurvey of the target zone literature. A usefulsource reference is Krugman and Miller (1992).

44 Ω(t) again denotes the information set avail-able at time t; the shift in notation to Ω(t) from Ωtreflects the shift to continuous time.

Figure 6. The Basic Target Zone Model

smax F Z

T Fsmin

m + v

45º

45 Clearly, intervention is assumed to be unster-ilized because it affects the exchange rate throughchanges in the money supply.

Taylor: Exchange Rates 37

Page 26: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

described by a nonlinear, S-shaped curvepasted smoothly onto the upper andlower edges of the target zone (TZ inFigure 6). This characteristic S-shape isthe graphical representation of the twomain features of the Krugman model:the “honeymoon effect” and the “smooth-pasting conditions.”

The honeymoon effect entails that, ina plot of s against m + v under a fullycredible target zone arrangement, s willlie on a curve which is less steep thanthe 45o line (if m + v is on the horizontalaxis). Intuitively, if s is near the top ofthe band (above the central parity) theprobability is higher that the exchangerate will touch the edge of the band andtrigger intervention by the authorities.Thus, the probability that the exchangerate will fall is higher than the prob-ability that it will rise. Thus, marketagents will bid the exchange rate belowthe level it would be at if there were noprobability of intervention—i.e., s mustlie below the 45o “free float line.” By asymmetric argument, when s is nearer tothe lower edge of the band, it must beabove the free float line. The honey-moon effect thus implies that a credibletarget zone is stabilizing in the sense thatthe range of variation in the exchangerate will be smaller, for any given rangeof variation in the fundamentals, thanunder a free float.

The smooth-pasting conditions areboundary conditions for the solution ofthe basic target zone model which entailthat in (m + v, s)-space, the permissi-ble exchange rate path must “paste”smoothly onto the upper and loweredges of the band. This result is againquite intuitive: If the exchange rate weresimply to bump into the edge of theband at an angle, traders would be of-fered a one-way bet, because they knowthat the authorities would intervene tobring the rate back into the band. Be-cause traders would start taking positions

in anticipation of the one-way bet beforeit occurred, this will tend to work againstthe influence of the fundamentals as theband is approached—e.g., a currency de-preciating because of weak fundamentalswill be bought near the edge of the bandin anticipation of official support. Thus,the exchange rate becomes increasinglyless responsive to movements in the fun-damentals as the edges of the band areapproached and, in the limit, the slope ofTZ, which measures the responsivenessof the exchange rate to the fundamen-tals, tends to zero.46

The formal solution to the basic (sym-metric) target zone model is an S-shapedfunction of the form

s = m + v + A[eα(m + v) − e−α(m + v)] (30)

where α = (2/θσ2)1/2, σ is the variance ofthe innovation in the fundamentals, andA is uniquely determined by the smooth-pasting conditions.47

2. Testing the Basic Target Zone Model

The basic target zone model has beentested empirically on data from theEuropean Monetary System, the Nordiccountries, the Bretton Woods system,and the Gold Standard. Invariably, thesetests have rejected the basic model (see,for example, R. Flood et al. 1991). Forexample, plots of the exchange rateagainst the fundamentals (variously de-fined), as well as more sophisticatedtests, do not reveal anything resemblingthe predicted S-shape (Meese and Rose1990).

3. Modifications to the Basic Target Zone Model

Given the empirical rejection of thebasic target zone model, a number ofauthors have sought to rehabilitate the

46 An alternative derivation of the smooth-past-ing result is given by R. Flood and Garber (1991).

47 This solution assumes symmetry of the targetzone and zero drift in the fundamentals.

38 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 27: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

model by modifying its underlying as-sumptions to allow for imperfect credi-bility of the arrangement, intramarginalintervention, sticky prices, and so on.For example, incorporating intramargi-nal intervention substantially reduces theimpact of the smooth-pasting conditions.This is because, as s approaches theedges of the band, the authorities are al-ready known to be intervening. The per-ceived probability of hitting the edge ofthe band is therefore lower than undermarginal intervention. The probability ofa riskless arbitrage opportunity occurringwill therefore be lower and the slope ofthe curve relating the exchange rate tothe fundamentals will be closer to astraight line, with smooth pasting occur-ring only when the exchange rate is veryclose to the band. Thus, the presence ofintramarginal intervention may explainwhy researchers have found little evi-dence of nonlinearities or the charac-teristic S-shaped curve (Hans Lindbergand Paul Söderlind 1992).

In assuming an underlying flexible-price monetary exchange rate model, thetarget zone literature may also have beenunduly naive, because the empirical evi-dence on the monetary models, albeitunder regimes closer to a free float thana target zone, is so overwhelmingly nega-tive. The implicit assumption of marketefficiency in the target zone model, inspite of mounting evidence to the con-trary, has also been criticized. Krugmanand Miller (1993), for example, arguethat policy makers have in the past beenwilling to enter into target zone arrange-ments precisely because of their skepti-cism with respect to the efficiency offoreign exchange markets and the ratio-nality of foreign exchange market partici-pants. If the motivation behind targetzones is in fact largely due to the fear ofirrational runs on a currency, then it isironic that the standard target zonemodel rules this out by assumption.

VII. Market Microstructure

As seen from the news literature, stud-ies of long-term exchange rate move-ments, and studies of economies experi-encing pathologically large movementsin the fundamentals such as a hyperinfla-tion, macroeconomic fundamentals areimportant influences on exchange rates.Nevertheless, the empirical literature asa whole demonstrates that there areoften large and persistent movements inexchange rates which are apparently un-explained by the macro fundamentals.One motivation for the emerging litera-ture on market microstructure has beenthe desire to understand the mechanismsgenerating these deviations from thefundamentals. In this literature, re-searchers focus on the behavior of mar-ket agents and market characteristicsrather than on the influence of macrofundamentals.48

One way in which a speculative move-ment away from the level consistent withthe macro fundamentals could begin, forexample, is if some agents have destabi-lizing expectations—e.g., if a five per-cent rise leads them to expect a ten per-cent rise, which causes them to buy theappreciating currency, causing it to risefaster in a self-fulfilling fashion. “Band-wagon” effects of this kind have been ex-amined by Frankel and Froot (1987),Helen Allen and Taylor (1990), Taka-toshi Ito (1993), and others, using surveydata on expectations, and this work ten-tatively suggests that expectations may

48 A comprehensive survey on the literature onthe theory and evidence relating to foreign ex-change market microstructure would warrent afull-length paper in itself, and we have space hereonly to flag some of the salient issues and to pro-vide a starting point for further reading. MarkFlood (1991) provides a good, nontechnical surveyof market microstructure theory applied to theforeign exchange market. The volume edited byFrankel, Galli, and Giovannini (forthcoming) con-tains a selection of recent scholarship in this area.

Taylor: Exchange Rates 39

Page 28: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

be more destabilizing over shorter hori-zons.

Another suggested explanation ofanomalous exchange rate movements isthe widespread influence of foreign ex-change analysts who do not base theirpredictions on economic theory—thefundamentals—but on the identificationof supposedly recurring patterns ingraphs of exchange rate movements—i.e.,“technical” or “chart” analysts (CharlesGoodhart 1988; Frankel and Froot1990a, 1990b; Allen and Taylor 1990).Questionnaire surveys conducted by theGroup of Thirty (1985) and Taylor andAllen (1992) reveal that extremely highproportions of traders employ technicalor chartist analysis, especially when fore-casting over shorter horizons. Goodhart(1988) presents a discussion of how ex-change rate misalignments might occurby considering the possibility that therate is determined by the balance ofchartist and fundamentalist predic-tions.49 A similar approach is developedmore formally by Frankel and Froot(1990a), who explain the sharp rise inthe demand for the U.S. dollar over the1981–85 period as a shift in the weightof market opinion away from fundamen-talists and toward chartists. Bilson (1990)emphasizes that technical traders em-ploying “overbought or oversold” indica-tors (“oscillators”), will tend to impartnonlinearity into exchange rate move-ments because small exchange ratechanges which do not trigger the oscilla-tor will tend to be positively correlatedbecause of the effect of trend-following

trading programs, while larger exchangerate movements, which trigger an oscilla-tor, indicating that a currency has been“oversold” or “overbought,” will be nega-tively correlated. Bilson (1990) estimatessimple nonlinear exchange rate equa-tions which are consistent with this pat-tern of serial correlation and which aremoderately successful in capturing ex-change rate changes.50

In some respects, however, the ques-tions posed by microstructural analysesare often quite different from those ap-plicable to macroeconomic studies.Thus, researchers have begun to analyzemicrostructural topics such as the deter-minants of the bid-ask spread (e.g., Hen-drik Bessembinder 1994) and the volumeof trade in foreign exchange markets,and why, for example, volume is verymuch higher on a gross basis (among for-eign exchange dealers and brokers) thanon a net basis (involving nonfinancialcompanies).51 Frankel and Froot(1990b), for example, show that the vol-ume of trade and market volatility is re-lated to the heterogeneity of exchangerate expectations, as reflected in disper-sion in survey expectations.52

It should be noted, moreover, that thevery notion of a dispersion of expecta-tions across market participants runscounter to more traditional analyses ofexchange rates which assume rational ex-pectations and hence homogeneous ex-pectations across market participants. Al-though it might be argued that the

49 Analyzing the accuracy of a number of indi-vidual technical analysts’ one-week and four-weekahead forecasts of three major exchange rates, Al-len and Taylor (1990) find that some chartistswere able to outperform a range of alternativeforecasting procedures. Earlier studies by Levich(1980) and Stephen Goodman (1979) found that,in terms of profitability, chartist foreign exchangeforecast services’ forecasts outperformed the for-ward rate in qualitative tests, while the forwardrate outperformed fundamentals based services.,

50 David Hsieh (1988) also detects evidence ofnonlinearities in exchange rate movements. Seealso Paul De Grauwe, Hans Dewachter, and MarkEmbrechts (1993).

51 Frankel and Froot (1990b), for example, sug-gest that over 95 percent of market transactions inmajor exchange rates are typically between foreignexchange market dealers or brokers, while RichardLyons (1993) suggests that over 80 percent is be-tween market makers alone.

52 Ito (1990) and MacDonald (1992) also reportevidence of heterogeneity of expectations, usingdisaggregated survey data.

40 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 29: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

rational expactations equilibrium may bepoliced by only a few well informed trad-ers, homogeneity of expectations is aworking assumption in most traditionalstudies. In contrast, many microstructu-ral studies often place the heterogeneityof expectations at center stage: the de-termination of the equilibrium exchangerate becomes endogenous to the wholeprocess of information transmission andprice discovery (Hal Varian 1989).

Other studies in the emerging marketmicrostructure literature have looked atthe way information is processed andtransmitted through the market, and therelationship of information processing tomarket volatility and volume. Lyons(1993), for example, reports evidencethat volume affects the bid-ask spreadthrough the information signalled bymarket volume as well as the desire ofmarket makers to control their inventoryof currencies. Tim Bollerslev and Mi-chael Melvin (1994) also provide empiri-cal support for a model in which the bid-ask spread is determined by underlyinguncertainty concerning exchange ratemovements. In related work on marketvolatility, several authors have docu-mented regularities in market volatility(Hsieh 1988; Goodhart and Marcello Gi-ugale 1988), and the contagion of ex-change rate volatility across foreign ex-change markets (the so-called “meteorshower”), which may be interpreted asevidence of information processing(Robert Engle, Ito, and Wen-Ling Lin1990).

VIII. Conclusion

In this final section we identify someof the broad trends in the literature dur-ing the 1980s and early 1990s, andspeculate as to the likely future direc-

tions of the research program.53 We re-marked in Section II on the trend towardincreasing econometric sophistication intests of foreign exchange market effi-ciency. This observation also holds moregenerally of empirical work on exchangerates. Thus, applied researchers on ex-change rates have begun to apply re-cently developed, sophisticated time se-ries techniques. Attention has alsotended to shift from examination ofmacroeconomic models of exchangerates toward work related to the foreignexchange market as a financial marketper se (e.g., risk premium models, mod-els of deviations from speculative effi-ciency such as peso problems and learn-ing models, microstructural analyses).The two areas of the literature whichhave conspicuously bucked this trend arethe work on target zones and that onequilibrium exchange rate models. Theformer of these, at least in its initial in-carnation, has met with spectacular em-pirical failure. Advocates of the latterhave tended to eschew formal economet-ric analysis as a means of judging its rele-vance, preferring to adduce support byappeal to their consistency with broadstylized facts of exchange rate andmacrovariable behavior.

The emerging finding that the dataprovide support for some of the long-runrelationships suggested by economic the-ory suggests that progress might bemade by concentrating on the long-rundeterminants of exchange rates. But itseems that further attempts to provideexplanations of short-term exchange ratemovements based solely on macroeco-nomic fundamentals may not prove suc-cessful54 —and this perhaps accounts for

53 The reader is cautioned that the speculationsin this concluding section are, by necessity, sub-jective. They are, however, by no means esoteric.

54 Not only has the search for macroeconomicfundamentals been extensive, but the results ofBaxter and Stockman (1989), R. Flood and Rose(1993), and others, as noted above, suggest thatthe usual set of macroeconomic fundamentals isunlikely to be capable of explaining exchange rate

Taylor: Exchange Rates 41

Page 30: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

the shift toward more purely financialmodels of exchange rate movements andheightened interest in market micro-structure. The macroeconomic funda-mentals are clearly important in settingthe parameters within which the ex-change rate moves in the short term, butthey do not appear to tell the wholestory. It is in this context that the emerg-ing literature on foreign exchange mar-ket microstructure seems especiallypromising. In the light of our discussionof the recent literature on foreign ex-change market intervention, furtherwork on the microeconomics of the in-teraction of the central bank and markettraders may also be warranted.55

But economics has an important nor-mative as well as positive element, and itis clear that in terms of assessing the ap-propriateness of exchange rate behaviorand policy, the macro fundamentals areof supreme importance. Viewed fromthis perspective, the macroeconomicfundamentals thus provide an indispen-sable framework for policy debate andanalysis.

REFERENCES

ABUAF, NISO AND JORION, PHILIPPE. “PurchasingPower Parity in the Long Run,” J. Finance,Mar. 1990, 45(4),, pp. 157–74.

ADLER, MICHAEL AND LEHMANN, BRUCE. “De-viations from Purchasing Power Parity in theLong Run,” J. Finance, Dec. 1983, 38(4), pp.1471–87.

ALLEN, HELEN L. AND TAYLOR, MARK P.“Charts, Noise and Fundamentals in the For-eign Exchange Market,” Econ. J., Mar. 1990,100(400), pp. 49–59.

BACKUS, DAVID K. “Empirical Models of the Ex-change Rate: Separating the Wheat from theChaff,” Can. J. Econ., Nov. 1984, 17(4), pp.824–46.

BAXTER, MARIANNE. “Real Exchange Rates and

Real Interest Differentials: Have We Missedthe Business-Cycle Relationship?” J. Monet.Econ., Feb. 1994, 33(1), pp. 5–37.

BAXTER, MARIANNE AND STOCKMAN, ALAN C.“Business Cycles and the Exchange Rate Re-gime: Some International Evidence,” J. Monet.Econ., May 1989, 23(3), pp. 377–400.

BESSEMBINDER, HENDRIK, “Bid-Ask Spreads inthe Interbank Foreign Exchange Market,” J. Fi-nan. Econ., June 1994, 35(3), pp. 317–48.

BILSON, JOHN F. O. “Rational Expectations andthe Exchange Rate,” in The economics of ex-change rates. Eds.: JACOB A. FRENKEL ANDHARRY G. JOHNSON. Reading, MA: Addison-Wesley, 1978, pp. 75–96.

———. “The ‘Speculative Efficiency’ Hypothesis,”J. Bus., July 1981, 54(3), pp. 435–51.

———. “ ‘Technical’ Currency Trading,” in Thecurrency-hedging debate. Ed.: LEE R.THOMAS. London: IFR Publishing, 1990, pp.257–75.

BOLLERSLEV, TIM AND MELVIN, MICHAEL. “Bid-ask Spreads and Volatility in the Foreign Ex-change Market—An Empirical Analysis,” J. Int.Econ., May 1994, 36(3/4), pp. 355–72.

BOSSAERTS, PETER AND HILLION, PIERRE. “Mar-ket Microstructure Effects of Government In-tervention in the Foreign Exchange Market,”Rev. Finan. Stud., 1991, 4(3), pp. 513–41.

BRANSON, WILLIAM H. “The Minimum CoveredInterest Differential Needed for InternationalArbitrage Activity,” J. Polit. Econ., Dec. 1969,77(6), pp. 1028–35.

BRANSON, WILLIAM H.; HALTTUNEN, HANNUAND MASSON, PAUL. “Exchange Rates in theShort Run: The Dollar-Deutschemark Rate,”Europ. Econ. Rev., Dec. 1977, 10(4), pp. 303–24.

BRANSON, WILLIAM H. AND HENDERSON, DALEW. “The Specification and Influence of AssetMarkets,” in RONALD W. JONES AND PETERB. KENEN, eds. 1985, Vol. 2, pp. 749–805.

BUITER, WILLEM H. AND MILLER, MARCUS H.“Monetary Policy and International Competi-tiveness: The Problems of Adjustment,” OxfordEcon. Pap., July 1981, 33(Supplement), pp.143–75.

CATTE, PIETRO; GIAMPAOLO GALLI AND REBEC-CHINI, SALVATORE, “Concerted Interventionsand the Dollar: An Analysis of Daily Data.” Pa-per presented at the Rinaldo Ossola MemorialConference, Banca d’Italia, July 1992.

CHEUNG, YIN-WONG AND LAI, KON S. “Long-runPurchasing Power Parity During the RecentFloat,” J. Int. Econ., Feb. 1993, 34(2), pp. 181–92.

CHOUDHRY, TAUFIQ; MCNOWN, ROBERT ANDWALLACE, MYLES. “Purchasing Power Parityand the Canadian Float in the 1950s,” Rev.Econ. Statist., Aug. 1991, 73(3), pp. 558–63.

CLINTON, KEVIN. “Transactions Costs and Cov-ered Interest Arbitrage: Theory and Evidence,”J. Polit. Econ., Apr. 1988, 96(2), pp. 358–70.

CUMBY, ROBERT E. AND OBSTFELD, MAURICE.

movements on its own. As Dornbusch (1987b) notes:“Research on exchange rate economics has grown tiredsearching for risk premia determinants or for newmacroeconomic models.”

55 This is suggested by Lewis (forthcoming); Pe-ter Bossaerts and Pierre Hillion (1991) and Garberand Michael Spencer (forthcoming) are examplesof this work.

42 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 31: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

“A Note on Exchange-Rate Expectations andNominal Interest Differentials: A Test of theFisher Hypothesis,” J. Finance, June 1981,36(3), pp. 697–703.

DANKER, DEBORAH J. ET AL. “Small EmpiricalModels of Exchange Market Intervention: Ap-plication to Germany, Japan, and Canada,” J.Policy Modeling, Spring 1987, 9(1), pp. 143–73.

DE GRAUWE, PAUL; DEWACHTER, HANS ANDEMBRECHTS, MARK. Exchange rate theory:Chaotic models of foreign exchange markets.Oxford: Blackwell, 1993.

DIEBOLD, FRANCIS X.; HUSTED, STEVEN ANDRUSH, MARK. “Real Exchange Rates Under theGold Standard,” J. Polit. Econ., Dec. 1991,99(6), pp. 1252–71.

DOMINGUEZ, KATHRYN M. “Does Sterilized In-tervention Influence Exchange Rates? A Test ofthe Signalling Hypothesis.” Mimeo, Harvard U.,1986.

DOMINGUEZ, KATHRYN M. AND FRANKEL, JEF-FREY A. Does foreign exchange interventionwork? Consequences for the dollar. Washing-ton, DC: Institute for International Economics,1993a.

———. “Does Foreign Exchange InterventionMatter? The Portfolio Effect,” Amer. Econ.Rev., Dec. 1993b, 83(4), pp. 1356–69.

———. “Foreign Exchange Intervention: An Em-pirical Assessment,” 1993c in JEFFREY A.FRANKEL 1993a, pp. 327–45.

DOMOWITZ, IAN AND HAKKIO, CRAIG S. “Condi-tional Variance and the Risk Premium in theForeign Exchange Market,” J. Int. Econ., Aug.1985, 19(1/2), pp. 47–66.

DOOLEY, MICHAEL P. AND ISARD, PETER. “Capi-tal Controls, Political Risk, and Deviations fromInterest-Rate Parity,” J. Polit. Econ., Apr. 1980,88(2), pp. 370–84.

———. “A Portfolio-Balance Rational-Expecta-tions Model of the Dollar-Mark ExchangeRate,” J. Int. Econ., May 1982, 12(3/4), pp.257–76.

DOOLEY, MICHAEL P. AND SHAFER, JEFFREY R.“Analysis of Short-run Exchange Rate Behavior:Mar. 1973 to 1981,” in Exchange rate and tradeinstability. Eds.: David Bigman and Teizo Taya.Cambridge, MA: Ballinger, 1983, pp. 43–69.

DORNBUSCH, RUDIGER. “Expectations and Ex-change Rate Dynamics,” J. Polit. Econ., Dec.1976, 84(6), pp. 1161–76.

———. “Monetary Policy under Exchange-RateFlexibility,” in Managed exchange-rate flexibil-ity: The recent experience. Federal ReserveBank of Boston, Conference series, No. 20.Boston: Author, 1979, pp. 90–122.

———. “Exchange Rate Economics: Where DoWe Stand?” Brookings Pap. Econ. Act., 1980, 1,pp. 143–85.

———. “Purchasing Power Parity,” in The newPalgrave dictionary of economics. Eds.: JOHNEATWELL, MURRAY MILGATE, AND PETERNEWMAN. London: MacMillan; New York:Stockton Press, 1987a.

———. “Exchange Rate Economics: 1986,” Econ.J., Mar. 1987b, 97(385), pp. 1–18.

———. “Real Exchange Rates and Macro-economics: A Selective Survey.” National Bu-reau of Economic Research Working Paper No.2775, 1988.

DORNBUSCH, RUDIGER AND FRANKEL, JEFFREYA. “The Flexible Exchange Rate System: Expe-rience and Alternatives,’ in International fi-nance and trade. Ed.: SILVIO BORNER, London:MacMillan, 1988, pp. 151–97; reprinted in JEF-FREY FRANKEL 1993a, pp. 5–40.

DRISKILL, ROBERT A. “Exchange-Rate Dynamics:An Empirical Investigation,” J. Polit. Econ.,Apr. 1981, 89(2), pp. 357–71.

EDISON, HALI J. “The Effectiveness of CentralBank Intervention: A Survey of the LiteratureAfter 1982.” Special Papers in InternationalEconomics No. 18, Princeton U. InternationalFinance Section, July 1993.

EDISON, HALI J. AND PAULS, B. DIANNE. “A Re-Assessment of the Relationship Between RealExchange Rates and Real Interest Rates: 1974–1990,” J. Monet. Econ., Apr. 1993, 31(2), pp.165–87.

EDWARDS, SEBASTIAN. “Exchange Rates andNews: A Multi-Currency Approach,” J. Int.Money Finance, Dec. 1982, 1(3), pp. 211–24.

EICHENBAUM, MARTIN AND EVANS, CHARLES.“Some Empirical Evidence on the Effects ofMonetary Policy Shocks on Exchange Rates.”National Bureau of Economic Research Work-ing Paper No. 4271, 1993.

ENGEL, CHARLES. “Testing for the Absence ofExpected Real Profits from Forward MarketSpeculation,” J. Int. Econ., Nov. 1984, 17(3/4),pp. 299–308.

ENGEL, CHARLES AND HAMILTON, JAMES D.“Long Swings in the Dollar: Are They in theData and Do Markets Know It?” Amer. Econ.Rev., Sept. 1990, 80(4), pp. 689–713.

ENGLE, ROBERT F.; ITO, TAKATOSHI AND LIN,WEN-LING, “Meteor Showers or Heat Waves?Heteroskedastic Intra-Daily Volatility in theForeign Exchange Market,” Econometrica, May1990, 58(3), pp. 525–42.

FAMA, EUGENE F. “Forward and Spot ExchangeRates,” J. Monet. Econ., Nov. 1984, 14(3), pp.319–38.

FLEMING, J. MARCUS, “Domestic Financial Poli-cies under Fixed and Floating Exchange Rates,”Int. Monet. Fund Staff Papers, Nov. 1962, 9(4),pp. 369–80.

FLOOD, MARK D. “Microstructure Theory andthe Foreign Exchange Market,” Fed. Res. BankSt. Louis Rev., Nov./Dec. 1991, 73(6), pp. 52–70.

FLOOD, ROBERT P. AND GARBER, PETER M.“The Linkage Between Speculative Attack andTarget Zone Models of Exchange Rates,”Quart. J. Econ., Nov. 1991, 106(4), pp. 1367–72.

FLOOD, ROBERT P. AND ROSE, ANDREW K. “Fix-ing Exchange Rates: A Virtual Quest for Funda-

Taylor: Exchange Rates 43

Page 32: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

mentals.” Centre for Economic Policy ResearchDiscussion Paper No. 838, Sept. 1993.

FLOOD, ROBERT P. ET AL. “An Empirical Explora-tion of Exchange Rate Target Zones,” Carnegie-Rochester Conf. Ser. Public Pol., 1991, 35, pp.7–77.

FLOOD, ROBERT P. AND TAYLOR, MARK P. “Ex-change Rate Economics: What’s Wrong withthe Conventional Macro Approach?” in JEF-FREY FRANKEL, GIAMPAOLO GALLI, AND AL-BERTO GIOVANNINI, forthcoming

FRANKEL, JEFFREY A. “On the Mark: A Theory ofFloating Exchange Rates Based on Real Inter-est Differentials,” Amer. Econ. Rev., Sept.1979, 69(4), pp. 601–22; reprinted in JEFFREYFRANKEL 1993a, pp. 77–93.

———. “The Mystery of the Multiplying Marks: AModification of the Monetary Model,” Rev.Econ. Statist., Aug. 1982a, 64(4), pp. 515–19.

———. “In Search of the Exchange Risk Pre-mium: A Six-Currency Test Assuming Mean-Variance Optimization,” J. Int. Money Finance,Dec. 1982b, 1(3), pp. 255–74; reprinted in JEF-FREY FRANKEL 1993a, pp. 219–34.

———. “Recent Estimates of Time-Variation inthe Conditional Variance and in the ExchangeRisk Premium,” J. Int. Money Finance, Mar.1988, 7(1), pp. 115–25; reprinted in JEFFREYFRANKEL 1993a pp. 235–44.

———. “Zen and the Art of Modern Macro-economics: The Search for Perfect Nothing-ness: A Commentary,” in Monetary policy for avolatile global economy. Eds: WILLIAM S.HARAF AND THOMAS D. WILLETT. Washing-ton, DC: American Enterprise Institute forPublic Policy Research, 1990, pp. 117–23; re-printed in JEFFREY FRANKEL 1993a, pp. 173–84.

———. On exchange rates. Cambridge, MA: MITPress, 1993a.

———. “Monetary and Portfolio Balance Modelsof the Determination of Exchange Rates,”1993b in JEFFREY FRANKEL 1993a, pp. 95–116.

FRANKEL, JEFFREY A. AND FROOT, KENNETH A.“Short-term and Long-term Expectations of theYen/Dollar Exchange Rate: Evidence from Sur-vey Data,” J. Japanese Int. Economies, Sept.1987, 1(3), pp. 249–74.

———. “Chartists, Fundamentalists and the De-mand for Dollars,” in Private behaviour andgovernment policy in interdependent econo-mies. Eds.: ANTHONY S. COURAKIS AND MARKP. TAYLOR. Oxford U. Press, 1990a, pp. 73–126.

———. “Chartists, Fundamentalists, and Tradingin the Foreign Exchange Market,” Amer. Econ.Rev., May 1990b, 80(2), pp. 24–38; reprinted inJEFFREY FRANKEL 1993a, pp. 317–26.

FRANKEL, JEFFREY A.; GALLI, GIAMPAOLO ANDGIOVANNINI, ALBERTO, The microstructure offoreign exchange markets. Chicago: U. of Chi-cago Press for the National Bureau of Eco-nomic Research, forthcoming.

FRANKEL, JEFFREY A. AND ROSE, ANDREW K.“An Empirical Characterization of Nominal Ex-

change Rates.” Mimeo, Department of Eco-nomics, U. of California, Berkeley, 1994; inGENE GROSSMAN AND KENNETH ROGOFF,forthcoming.

FRENKEL, JACOB A. “A Monetary Approach to theExchange Rate: Doctrinal Aspects and Empiri-cal Evidence,” Scand. J. Econ., Mar. 1976,78(2), pp. 200–24.

———. “Purchasing Power Parity: Doctrinal Per-spective and Evidence from the 1920s,” J. Int.Econ., May 1978, 8(2), pp. 169–91.

———. “The Collapse of Purchasing Power Pari-ties During the 1970s,” Europ. Econ. Rev., May1981a, 16(1), pp. 145–65

———. “Flexible Exchange Rates, Prices and theRole of ‘News’: Lessons from the 1970s,” J.Polit. Econ., Aug. 1981b, 89(4), pp. 665–705.

———, ed. Exchange rates and internationalmacroeconomics. Chicago: U. of Chicago Pressfor the National Bureau of Economic Research,1983.

FRENKEL, JACOB A. AND LEVICH, RICHARD M.“Covered Interest Arbitrage: Unexploited Prof-its?” J. Polit. Econ., Apr. 1975, 83(2), pp. 325–38.

———. “Transaction Costs and Interest Arbitrage:Tranquil versus Turbulent Periods,” J. Polit.Econ., Dec. 1977, 85(6), pp. 1207–24.

FRIEDMAN, MILTON AND SCHWARTZ, ANNA J. Amonetary history of the United States: 1867–1960. Princeton, NJ: Princeton U. Press for theNational Bureau of Economic Research, 1963.

FROOT, KENNETH A. AND FRANKEL, JEFFREY A.“Forward Discount Bias: Is It an Exchange RiskPremium?” Quart. J. Econ., Feb. 1989, 104(1),pp. 139–61; reprinted in JEFFREY FRANKEL1993a, pp. 245–60.

FROOT, KENNETH A. AND ROGOFF, KENNETH.“Perspectives on PPP and Long-Run Real Ex-change Rates.” Mimeo, Department of Eco-nomics, Princeton U. 1994; in GENE GROSS-MAN AND KENNETH ROGOFF, forthcoming.

FROOT, KENNETH A. AND THALER, RICHARD H.“Anomalies: Foreign Exchange,” J. Econ. Per-spectives, Summer 1990, 4(2), pp. 179–92.

GAILLOT, HENRY J. “Purchasing Power Parity asan Explanation of Long-Term Changes in Ex-change Rates,” J. Money Credit Banking, Aug.1970, 2(3), pp. 348–57.

GARBER, PETER M. AND SPENCER, MICHAEL G.“Dynamic Hedging and the Interest Rate De-fense,” in JEFFREY FRANKEL, GIAMPAOLOGALLI, AND ALBERTO GIOVANNINI, forthcom-ing.

GIOVANNINI, ALBERTO AND JORION, PHILIPPE.“The Time Variation of Risk and Return in theForeign Exchange and Stock Markets,” J. Fi-nance, June 1989, 44(2), pp. 307–25.

GOODHART, CHARLES A. E. “’The Foreign Ex-change Market: A Random Walk with a Drag-ging Anchor,” Economica, Nov. 1988, 55(220),pp. 437–60.

GOODHART, CHARLES A. E. AND GIUGALE, MAR-CELLO. “From Hour to Hour in the Foreign Ex-

44 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 33: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

change Market.” London School of Economics,Financial Markets Group Discussion Paper No.33, 1988.

GOODMAN, STEPHEN H. “Foreign Exchange RateForecasting Techniques: Implications for Busi-ness and Policy,” J. Finance, May 1979, 34(1),pp. 415–27.

GRILLI, VITTORIO AND ROUBINI, NOURIEL. “’Li-quidity and Exchange Rates,” J. Int. Econ.,May 1992, 32(3/4), pp. 339–52.

———. “Liquidity and Exchange Rates: PuzzlingEvidence from the G-7 Countries.” Mimeo, De-partment of Economics, Yale U., 1993.

GROSSMAN, GENE AND ROGOFF, KENNETH.Handbook of international economics. 2nd. ed.,Amsterdam: North Holland, forthcoming.

GROUP OF THIRTY. The foreign exchange mar-ket in the 1980s. New York: Group of Thirty,1985.

HAKKIO, CRAIG S. “Expectations and the ForwardExchange Rate,” Int. Econ. Rev., Oct. 1981,22(3), pp. 663–78.

HANSEN, LARS P. AND HODRICK, ROBERT J.“Forward Exchange Rates as Optimal Predic-tors of Future Spot Rates: An EconometricAnalysis,” J. Polit. Econ., Oct. 1980, 88(5), pp.829–53.

———. “Risk-Averse Speculation in the ForwardForeign Exchange Market: An EconometricAnalysis of Linear Models,” in JACOBFRENKEL, ed. 1983, pp. 113–42.

HODRICK, ROBERT J. The empirical evidence onthe efficiency of forward and futures foreign ex-change markets, London: Harwood, 1987.

———. “Risk, Uncertainty, and Exchange Rates,”J. Monet. Econ., May 1989, 23(3), pp. 433–59.

———. “An Interpretation of Foreign ExchangeMarket Efficiency Tests.” Mimeo, Northwest-ern U., Kellog Graduate School of Manage-ment, 1992.

HOOPER, PETER AND MORTON, JOHN. “Fluctua-tions in the Dollar: A Model of Nominal andReal Exchange Rate Determination,” J. Int.Money Finance, Apr. 1982, 1(1), pp. 39–56.

HSIEH, DAVID A. “The Statistical Properties ofDaily Foreign Exchange Rates: 1974–1983,” J.Int. Econ., Feb. 1988, 24(1–2), pp. 129–45.

HUMPAGE, OWEN F. “On the Effectiveness of Ex-change Market Intervention.” Mimeo, Fed.Res. Bank Cleveland, June 1989.

ITO, TAKATOSHI. “Foreign Exchange Rate Expec-tations: Micro Survey Data,” Amer. Econ. Rev.,June 1990, 80(3), pp. 434–49.

———. “Short-Run and Long-Run Expectationsof the Yen-Dollar Exchange Rate.” National Bu-reau of Economic Research Working paper No.4545, 1993.

HUMPAGE, OWEN F. “On the Effectiveness of Ex-change Market Intervention.” Mimeo, Fed.Res. Bank f Cleveland, June 1989.

JONES, RONALD W. AND KENEN, PETER B., eds.Handbook of international economics. Two vol-umes. Amsterdam: North-Holland, 1985.

JURGENSEN, PHILIPPE (CHAIRMAN). Report of

the working group on exchange market inter-vention. Washington, DC: U.S. Treasury De-partment, 1983.

KENEN, PETER B. “Effects of Intervention andSterilization in the Short Run and the LongRun,” in The international monetary system un-der flexible exchange rates. Eds.: Richard N.Cooper et al. Cambridge, MA.: Ballinger, 1982,pp. 51–68.

KOEDIJK, KEES G. AND SCHOTMAN, PETER.“How to Beat the Random Walk: An EmpiricalModel of Real Exchange Rates,” J. Int. Econ.,Nov. 1990, 29(3/4), pp. 311–32.

KRUGMAN, PAUL R. “Target Zones and ExchangeRate Dynamics,” Quart. J. Econ., Aug. 1991,106(3), pp. 669–82.

KRUGMAN, PAUL R. AND MILLER, MARCUS H.,eds. Exchange rate targets and currency bands.Cambridge: Cambridge U. Press, 1992.

———. “Why Have a Target Zone? A Comment,”Carnegie-Rochester Conf. Ser. Public Pol.,1993, 38 pp. 279–318.

LEVICH, RICHARD M. “Analyzing the Accuracy ofForeign Exchange Advisory Services: Theoryand Evidence,” in Exchange risk and exposure.Eds: RICHARD M. LEVICH AND CLAUSWIHLBORG. Lexington, MA: DC Heath, 1980,pp. 99–127.

———. “Empirical Studies of Exchange Rates:Price Behavior, Rate Determination, and Mar-ket Efficiency,” in RONALD W. JONES AND PE-TER B. KENEN, Vol. II, 1985.

LEVICH, RICHARD M. AND THOMAS, LEE R.“The Significance of Technical Trading-RuleProfits in the Foreign Exchange Market: ABootstrap Approach,” J. Int. Money Finance,Oct. 1993, 12(5), pp. 451–74.

LEWIS, KAREN K. “Testing the Portfolio BalanceModel: A Multilateral Approach,” J. Int. Econ.,Feb. 1988, 24(1/2), pp. 109–27.

———. “Changing Beliefs and Systematic Ra-tional Forecast Errors with Evidence from For-eign Exchange,” Amer. Econ. Rev., Sept. 1989,79(4), pp. 621–36.

———. “International Financial Markets.”Mimeo, Wharton School, U. of Pennsylvania,1994; in GENE GROSSMAN AND KENNETH RO-GOFF, forthcoming.

LINDBERG, HANS AND SÖDERLIND, PAUL. “Tar-get Zone Models and Intervention Policy: TheSwedish Case.” Stockholm: Institute for Inter-national Economic Studies Seminar Paper No.496, 1992.

LOTHIAN, JAMES R. AND TAYLOR, MARK P. “RealExchange Rate Behavior: The Recent Floatfrom the Perspective of the Past Two Centu-ries.” International Monetary Fund, 1995; in J.Polit. Econ., forthcoming.

LUCAS, ROBERT E., JR. “Interest Rates and Cur-rency Prices in a Two-Country World,” J.Monet. Econ., Nov. 1982, 10(3), pp. 335–59.

———. “Liquidity and Interest Rates,” J. Econ.Theory, Apr. 1990, 50(2), pp. 237–64.

LYONS, RICHARD K. “Tests of Microstructural Hy-

Taylor: Exchange Rates 45

Page 34: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

potheses in the Foreign Exchange Market.” Na-tional Bureau of Economic Research WorkingPaper No. 4471, Sept. 1993.

MACDONALD, RONALD. “Exchange Rate SurveyData: A Disaggregated G-7 Perspective,” Man-chester Sch. Econ. Soc. Stud., June 199260(Supplement), pp. 47–62.

———. “Long-run Purchasing Power Parity: Is itfor Real?,” Rev. Econ. Statist., Nov. 1993, 75(4)pp. 690–95.

MACDONALD, RONALD. AND TAYLOR, MARK P.Exchange Rate Economics: A Survey,” Interna-tional Monetary Fund Staff Papers, Mar. 1992,39(1) pp. 1–57.

———. “The Monetary Approach to the ExchangeRate: Rational Expectations, Long-Run Equilib-rium, and Forecasting,” Int. Monet. Fund StaffPapers, Mar. 1993, 40(1), pp. 89–107.

———. “The Monetary Model of the ExchangeRate: Long-Run Relationships, Short-Run Dy-namics, and How to Beat a Random Walk,” J.Int. Money Finance, forthcoming.

MARK, NELSON C. “Real and Nominal ExchangeRates in the Long Run,” J. Int. Econ., Feb.1990, 28(1/2), pp. 115–36.

———. “Exchange Rates and Fundamentals: Evi-dence on Long-Horizon Predictability andOvershooting.” Working Paper, Ohio State U.,July 1993.

MARSTON, RICHARD C. “Real and Nominal Ex-change Rate Variability,” Empirica, 1989, 16(2),pp. 147–60.

MCCALLUM, BENNETT T. “A Reconsideration ofthe Uncovered Interest Rate Parity Relation-ship,” J. Monet. Econ., Feb. 1994, 33(1), pp.105–32.

MCCULLOCH, J. HUSTON. “Operational Aspects ofthe Siegel Paradox: Comment,” Quart. J. Econ.,Feb. 1975, 89(1), pp. 170–72.

MCNOWN, ROBERT F. AND WALLACE, MYLES S.“National Price Levels, Purchasing Power Par-ity and Cointegration: A Test of Four High In-flation Economies,” J. Int. Money Finance,Dec. 1989, 8(4), pp. 533–45.

MEADE, JAMES E. The balance of payments. Withmathematical appendix. Oxford: Oxford U.Press, 1951.

MEESE, RICHARD A. “Testing for Bubbles in Ex-change Markets: A Case of Sparkling Rates?” J.Polit. Econ., Apr. 1986, 94(2), pp. 345–73.

MEESE, RICHARD A. AND ROGOFF, KENNETH.“Empirical Exchange Rate Models of the Sev-enties: Do They Fit Out of Sample,” J. Int.Econ., Feb. 1983a, 14(2), pp. 3–24.

———. “The Out-of-Sample Failure of EmpiricalExchange Rate Models: Sampling Error or Mis-specification?” 1983b in JACOB A. FRENKEL,ed. 1983, pp. 67–105.

———. “Was It Real? The Exchange Rate-Inter-est Differential Relationship Over the ModernFloating-Rate Period,” J. Finance, Sept. 1988,43(4), pp. 933–48.

MEESE, RICHARD A. AND ROSE, ANDREW K.“Nonlinear, Nonparametric, Nonessential Ex-

change Rate Estimation,” Amer. Econ. Rev.,May 1990, 80(2), pp. 192–96.

MUNDELL, ROBERT A. “Capital Mobility and Sta-bilization Policy under Fixed and Flexible Ex-change Rates,” Canadian J. Econ. Polit. Sci.,Nov. 1963, 29(4), pp. 475–85.

MUSSA, MICHAEL. “The Role of Intervention.”Group of Thirty Occasional Paper No. 6, NewYork, 1981.

———. “The Theory of Exchange Rate Determi-nation,” in Exchange rate theory and practice.Eds: JOHN F. O. BILSON AND RICHARD C.MARSTON. Chicago: U. of Chicago Press, 1984,pp. 13–58.

———. “Nominal Exchange Rate Regimes and theBehavior of Real Exchange Rates: Evidence andImplications,” Carnegie-Rochester Conf. Ser.Public Pol., Autumn 1986, 25, pp. 117–213.

OBSTFELD, MAURICE. “Exchange Rates, Inflation,and the Sterilization Problem: Germany 1975–1981,” Europ. Econ. Rev., Mar./Apr. 1983,21(1/2), pp. 161–89.

———. “Floating Exchange Rates: Experienceand Prospects,” Brookings Pap. Econ. Act.,1985, 2, pp. 369–450.

———. “The Effectiveness of Foreign ExchangeIntervention: Recent Experience, 1985–1988,”in International policy coordination and ex-change rate fluctuations. Eds: WILLIAM H.BRANSON, JACOB A. FRENKEL, AND MORRISGOLDSTEIN. Chicago: U. of Chicago Press,1990, pp. 146–52.

OBSTFELD, MAURICE AND STOCKMAN, ALAN C.“Exchange Rate Dynamics,” in RONALD W.JONES AND PETER B. KENEN, eds. 1985, Vol. 2,pp. 917–77.

OFFICER, LAWRENCE H. “The Purchasing PowerParity Theory of Exchange Rates: A Review Ar-ticle,” Int. Monet. Fund Staff Papers, Mar.1976, 23(1), pp. 1–60.

POOLE, WILLIAM. “Speculative Prices as RandomWalks: An Analysis of Ten Time Series of Flex-ible Exchange Rates,” Southern Econ. J., Apr.1967, 33(2), pp. 468–78.

ROGOFF, KENNETH. “Expectations and ExchangeRate Volatility.” Unpublished Ph.D. thesis,Massachusetts Institute of Technology, 1979.

———. “On the Effects of Sterilized Intervention:An Analysis of Weekly Data,” J. Monet. Econ.,Sept. 1984, 14(2), pp. 133–50.

SIEGEL, JEREMY J. “Risk, Interest, and ForwardExchange,” Quart. J. Econ., May 1972, 86(2),pp. 303–09.

STOCKMAN, ALAN C. “A Theory of Exchange RateDetermination,” J. Polit. Econ., Aug. 1980,88(4), pp. 673–98.

———. “Real Exchange Rates under AlternativeNominal Exchange Rate Systems,” J. Int.Money Finance, Aug. 1983, 2(2), pp. 147–66.

———. “The Equilibrium Approach to ExchangeRates,” Fed. Res. Bank Richmond Econ. Rev.,Mar./Apr. 1987, 73(2), pp. 12–30.

———. “Real Exchange Rate Variability UnderPegged and Nominal Floating Exchange Rate

46 Journal of Economic Literature, Vol. XXXIII (March 1995)

Page 35: The Economics of Exchange Rates - Wharton Financefinance.wharton.upenn.edu/~bodnarg/courses/...eign exchange gain from holding one currency rather than another (the ex-pected exchange

Systems: An Equilibrium Theory.” National Bu-reau of Economic Research Working Paper No.2565, 1988.

SVENSSON, LARS E. O. “An Interpretation of Re-cent Research on Exchange Rate TargetZones,” J. Econ. Perspectives, Fall 1992, 6(4),pp. 119–44.

TAKAGI, SHINJI. “Exchange Rate Expectations: ASurvey of Survey Studies,” Int. Monet. FundStaff Papers, Mar. 1991, 38(1), pp. 156–83.

TAYLOR, MARK P. “Covered Interest Parity: AHigh-Frequency, High-Quality Data Study,”Economica, Nov. 1987, 54(216), pp. 429–38.

———. “An Empirical Examination of Long-RunPurchasing Power Parity Using CointegrationTechniques,” Applied Econ., Oct. 1988, 20(10),pp. 1369–82.

———. “Covered Interest Arbitrage and MarketTurbulence,” Econ. J., June 1989, 99(396), pp.376–91.

———. The economics of exchange rates. Cam-bridge and New York: Cambridge U. Press,forthcoming.

TAYLOR, MARK P. AND MCMAHON, PATRICK C.“Long-run Purchasing Power Parity in the1920s,” Europ. Econ. Rev., Jan. 1988, 32(1), pp.179–97.

TAYLOR, MARK P. AND ALLEN, HELEN L. “TheUse of Technical Analysis in the Foreign Ex-change Market,” J. Int. Money Finance, June1992, 11(3), pp. 304–14.

THROOP, ADRIAN W. “A Generalized UncoveredInterest Parity Model of Exchange Rates,”Econ. Rev., Federal Reserve Bank of San Fran-cisco, 1993, No. 2, pp. 3–16.

VARIAN, HAL R. “Differences of Opinion in Fi-nancial Markets,” in Financial risk: Theory, evi-dence, and implications. Ed.: COURTENAY C.STONE. Boston: Kluwer Academic, 1989, pp. 3–37.

Taylor: Exchange Rates 47