Fler 10yrs

JEFFREY R.
SHAFER
Federal Reserve Bank of New York
BONNIE E. LOOPESKO
Federal Reserve Bank of New York
Floating Exchange Rates

after Ten Years
TEN YEARS AGO, in March 1973, the United States and other nations
abandoned efforts to maintain the Bretton Woods system of fixed
exchangeratesamongthe majorcurrencies.Theperiodof largelymarket-
determined(floating)exchangeratesbetween the dollarandothermajor
currenciesbegan.I This fundamentalchange in the internationalmone-
tary system was ratifiedby the RambouilletSummit in 1975 and the
Second Amendmentof the InternationalMonetaryFund's Articles of
Agreementin 1978.
The debateaboutwhetherexchangeratesamongthe majorcurrencies
oughtto be fixed,freelyfloating,or somewherein betweenhas continued
with considerableintensity. Stronglyheld but sharplycontrastingviews
on how exchangeratesbehaveandinteractwithothereconomicvariables
have been central to the debate, but differing interests and values
regardingthe properrole of governmentsin financialmarketshave also
We gratefullyacknowledgethe excellent assistance of Peter Borish and the help of

CatherineCrosbywiththedatabase. Wearealso indebtedto the membersof the Brookings
Panelandto MauriceObstfeldandKennethS. Rogofffor insightfulandhelpfulcomments
andsuggestions.The views expressedherearethose of the authorsandare not the official
views of the FederalReserveBankof New Yorkor the FederalReserveSystem.
1. The Canadiandollar had been floatingsince May 1970;the pound sterling,since
June 1972;andthe yen, since February1973.Transitionalfloatswere permittedon several
occasions in the lateryears of the BrettonWoodsperiod,but the changein U.S. policy in
March1973markedthe end of thatperiod.
1
2 Brookings Papers on Economic Activity, 1:1983
been important.2The strongest advocates of fixed exchange rates,
proponentsof the reestablishmentof a gold standard,and the strongest
advocates of a free float, the monetarists, tend to share a desire to
eliminateany discretionaryelement in economic policy. Those who see
a role for discretiongenerallytake intermediatepositions, but also vary
widely in how much flexibilitythey would permitin exchangerates.
In this paper we review what the ten years of experience with a
floatingexchange ratecan reveal abouthow exchangeratesbehave. We
conclude that no simple existing theory explains the behaviorof rates
well enough to provide a strong case for any mechanicalapproachto
exchange rate management. Uncertainty about how exchange rates
behave over periods as long as several years is itself one of the most
importantfacts to consider in choosing amongalternativestrategiesfor
managingexchange rates and, more generally, in formulatingmacro-
economic policies in an interdependentworld.
The Transition from Fixed to Floating Exchange Rates
Duringthe late 1960sand early 1970s, the system of fixed exchange

rates established at the Bretton Woods Conference in 1944 showed
2. The strongpresumptionunderlyingthe BrettonWoodssystem,thatfreemultilateral
traderequiresfixed exchangerates, was based in largeparton the dislocationsaccompa-
nyingthe interwarexperienceswith flexiblerates.The classic indictmentof the flexibility
of exchange rates duringthat period is by RagnarNurkse in InternationalCurrency
Experience: Lessons of the Inter-war Period (Geneva: League of Nations, 1944). Milton
Friedman'sinfluentialarticleis a responseto that indictment.See Friedman,"The Case
for Flexible Exchange Rates," Essays in Positive Economics (University of Chicago
Press, 1953),pp. 157-203. Notable later contributionsto the debate are Egon Sohmen,
FlexibleExchangeRates (Universityof ChicagoPress, 1969);J. E. Meade,"TheCasefor
VariableExchangeRates," ThreeBanksReview,vol.27 (September1955),pp.3-27; Harry
G. Johnson,"The Case for FlexibleExchangeRates, 1969,"FederalReserveBankof St.
Louis Review, vol. 51 (June 1969),pp. 12-24; CharlesP. Kindleberger,"The Case for
Fixed Exchange Rates, 1969," in Federal Reserve Bank of Boston, The International
AdjustmentMechanism,ConferenceSeries 2 (FRBB, 1970),pp. 93-108; and RobertA.
Mundell,"UncommonArgumentsfor CommonCurrencies,"in HarryG. Johnsonand
Alexander K. Swoboda, eds., The Economics of Common Cuirrencies(London: Allen and
Unwin Ltd., 1973),pp. 114-32. Recent assessmentsincludeJacquesR. Artus and John
H. Young, "Fixed and Flexible ExchangeRates:A Renewalof the Debate," IMF Staff
Papers, vol. 26 (December 1979), pp. 654-98; and Morris Goldstein, Have Flexible
Exchange Rates Handicapped Macroeconomic Policy? Special Papers in International
Economics, 14(PrincetonUniversity,InternationalFinanceSection, June 1980).
Jeffrey R. Shafer and Bonnie E. Loopesko 3
increasingsigns of stress. Despite widespreadcapital controls, capital
flows andconsequentinterventionrequirementsexpandedto levels that
alarmedpolicymakers.Imbalancesin the currentaccount seemed to be
giving off signals of fundamental disequilibriumwith growing fre-
quency. Monetaryauthoritiesincreasinglyobserved situationsin which
the monetary policy they believed was appropriatefrom a domestic
standpointand the policy that would bringaboutexternalbalancewere
in conflict. Par value changes and periods during which countries
permittedtheircurrenciesto floatbecamemorecommon.Two currency
realignmentsin December 1971 and February 1973, involving a net
depreciationof the dollar,failedto stemmountingpressureon the dollar.
It was againstthis backgroundthatthe decision to floatthe exchange
rate was made in March 1973. For nearly a year afterwardthe United
States continued to advocate the establishmentof a modifiedBretton
Woods system of fixed but adjustablepar values in the Committeeof
Twenty on the Reform of the InternationalMonetary System. The
turbulencein internationalmoney marketsfollowing the first oil shock
then persuadedofficials in the United States and other countries that
HumptyDumpty could not be put back togetheragain-at least for the
time being. U.S. negotiatingefforts shiftedto legitimizingandestablish-
ingrulesfor conductinga float.These effortsculminatedin the agreement
at the RambouilletSummitin November 1975.
A growingnumberof economists, particularlyin the United States,
had been advocatinga shift to flexible rates for some time. Expediency
in the face of an increasingly uncontrollable situation, rather than
economic theory, was apparentlythe strongest consideration in the
initialdecision to float; but the theoreticalargumentsdid influencethe
decisions to continue supportof floatingrates.
The advocates of floatingexchangerates madefour principalclaims.
First, price-adjustedor real exchange rates would be maintainedat
relatively constant values by stabilizingspeculationand would change
mainlyin response to shifts or trends in the equilibriumterms of trade
between economies. Second, external balance would be better main-
tained than under fixed rates. Imbalances in the overall balance of
payments(the officialsettlementsbalance)wouldnot ariseby definition;
andcurrentaccounts wouldbe kept roughlyin line with the fundamental
positions of countries as net suppliersor demandersof capital, deter-
mined by wealth accumulationand investment potential. Third, econ-
omies wouldbe relativelyinsulatedagainstmacroeconomicshocks from

abroad,and authoritieswould enjoy greaterindependenceof monetary
policies to pursue domestic stabilizationgoals. And fourth, national
policymakerswould findit unnecessaryto impose restrictionson trade
and capital flows for macroeconomicreasons, and existing restrictions
could be discarded,thus enhancingthe efficiencyof resourceallocation
in the world economy.
Policymakersin the United States and abroadeach saw an additional
advantagefrom floatingrates. For both, a floatingdollarwas seen as a
way of escapingthe asymmetryin the role of the dollarunderthe Bretton
Woods system-the so-called "n-plus first" currency problem. This
asymmetrywas evident in two aspects of the way the system worked.
Under the Bretton Woods system, other countries bore the principal
obligationto maintaintheircurrencieswithinfixed marginsvis a vis the
dollar through the purchase and sale of dollar assets. But they also
enjoyed greaterfreedom to choose par values againstthe dollarand to
change them from time to time; and the effective or weighted-average
parvalue of the dollarresultedfromthe parvalues chosen by others.
Foreign authorities had become increasingly concerned about the
consequences of the first aspect of the asymmetry:the privilege that
they felt the system bestowed on the United States to follow expansion-
ary policies unconstrainedby reserve losses, particularlyas U.S. fiscal
deficitsand monetaryexpansionbeganto be associated with the unpop-
ularwar in Vietnam.These foreignauthoritiessaw themselves as facing
anunpleasantchoice between acceptingindefinitelylargeaccumulations
of dollar reserves or getting in step with U.S. policy.3 Foreign official
holders of dollars were paid interest and sometimes were given an
exchange rate guaranteeon their dollar investments, but they did not
see this as balancingthe scales.
At about the same time there was growing distress among U.S.
authoritiesabout the consequences of the second aspect of the asym-
metry-the difficultyof achievingan appropriateeffective value of the
dollar. Changingthe value of the dollarentaileda systemic negotiation,
3. One statement of this view appears in Otmar Emminger, The D-Mark in the Conflict
between Internal and External Equilibrium, 1948-75, Essays in International Finance,
122 (Princeton University, International Finance Section, June 1977).
which had proven difficultand tedious when undertakenat the Smith-
sonianin December1971andagainin February1973.Moreover,because
of its special role as a reserve currency, devaluationof the dollar was
thoughtto threatenthe stabilityof the system.
Not all economists were enthusiastic about the move to floating
exchange rates, and many, if not most, policymakershad reservations.
Skeptics were most numerousoutside the United States, where there
was less confidencein the self-equilibrationof markets.They hadseveral
areas of concern: (1) Speculative capitalmovementswould be destabi-
lizing and would lead to wide swings in exchange rates. (2) National
policies, once freed of the disciplineof fixed rates, would be less stable
and, on the whole, more inclined to accommodateinflation.Policies,
togetherwith frequentspeculativedisturbancesto rates, would tend to
result in greaterdivergenceand higheraveragelevels of inflationrates.
(3)Exchangeratesthatwere not associatedwithunderlyingfundamental
developments would induce undesirable fluctuations in current ac-
counts, domestic output, employment, and inflation.(4) False signals
fromexchangeratesandexchangerateuncertainty,per se, were thought
to distort and inhibittrade and investmentwith adverse consequences
for the allocationof resources in the worldeconomy. (5) Tradebarriers
and restrictionson capitalflows would proliferateas officials soughtto
protect their economies from the destabilizingeffects of exchange rate
changes.
Problems in Evaluating the Period of Floating Rates
The period of floating exchange rates has been marked by great

instabilityin the world economy and relatively poor economic perfor-
mance when comparedwith the 1960s, and even when comparedwith
the 1970-72 period when the Bretton Woods system was becoming
unworkable.Table 1 highlightsthe generaldecline in economic perfor-
mance. But one cannot decide the argumentin favor of the opponents
of floatingon the basis of such casual observations. On the one hand,
one must keep in mind the problems with the Bretton Woods system
that led to its abandonment. On the other, the past ten years are
distinguishedfrom the 1960s in a numberof respects other than the
Table 1. Economic Performanceof the OECD Economies,
Selected Periods, 1960-82
Percent
Measure 1960-69 1970-72 1973-82
Average real GNP growthrate 4.95 4.12 2.45
Average inflationrate 2.84 5.20 9.88
Average unemploymentratea 2.73 3.40 5.34
Source: Organization for Economic Cooperation and Development, OECD EcononmicOutlook, various issues.
a. Data on total OECD unemployment commences in 1964.
exchange rate regime. These differencesinclude the following:(1) It is

unlikelythat growthin the economies of continentalEurope and Japan
could have been sustained at the rapidpace of the 1950s and 1960s, a
period that showed in part a catchingup after the Second WorldWar.
(2) The periodof floatingrates inheritedthe internationaldisequilibrium
and inflationarytendencies that had builtup graduallyduringthe 1960s.
(3) Internationaltrade grew in relative importanceand Eurocurrency
marketswere developing that were relatively free of direct control by
authorities. Consequentlythe exposure of all economies to economic
and financialdisturbancesfrom abroadwas potentially greater in the
1970s. (4) The majoroil marketshocks of 1973-74 and 1979disturbed
the world economy much more than any events that occurred in the
1960s.
Still another complication arises because the domestic monetary
policies andexchangeratepolicies thathave been pursuedhave changed
fromtime to time andhave often departedfromthe strictprescriptionof
the advocates of floating. The unsatisfactoryperformanceof national
economies and the adjustmentto the needs and possibilitiesof floating
exchangeratesled to considerableexperimentationwithmonetarypolicy
strategiesin the 1970s,especially attemptsto use monetarytargetsmore
or less stringently.
Moreover, exchange rates have not been permittedto float cleanly
since March 1973, nor have exchange rate policies in any country
remainedsettled for any extended period. Interventionstrategieshave
differedamong countries and over time, rangingfrom free floating,to
short-run smoothing, to rather heavy intervention directed toward
achievinga specifiedexchange rate. Capitalcontrolshave been a factor
influencingexchangerateson at least some occasions, althoughcontrols
have been a lesser influence as time has passed.4Domestic monetary
policies have rangedfrom benign neglect of exchange rates to giving
exchangerates considerableweight in settinginterestrates or monetary
growthrates, and Europeancountrieshave gone throughseveral stages
of experimentationwith maintainingnarrowmarginsamong their cur-
rencies.
Thus, sorting out how harsh a judgment the floating rate regime
deserves and whether there is a better way is difficult,even after ten
years. What can be done, however, is to take stock of what has been
learnedabout how floatingexchange rates behave. Differingviews on
whetherfloatingexchange rates are beneficial, what domestic policies
are appropriatewith a floating exchange rate, and whether exchange
ratescan and shouldbe managedmore tightly,independentof domestic
policies, follow fromdifferentmodels of exchange ratedetermination.
The theory of exchange rate determinationhas undergoneconsider-
able developmentover the past ten years as simple models were found
to explain current experience poorly and the lengtheningtime series
introduced new phenomena for theorists to explain. One important
question is how the predictions and policy prescriptionsof the early
advocatesof floatingshouldbe modified,given the evolutionof theoret-
ical understandingof exchange ratedetermination.A second important
question is how complacent economists should be about the current
understandingof exchangeratesandaboutany set of policy prescriptions
thatdependson a particularmodel.
The following section of this paper reviews the behavior of several
importantexchange rates (the dollar-deutschemark,the dollar-yen,the
dollar-poundsterling,and a weighted-averageratefor the dollaragainst
six major currencies), investigates the challenges that this behavior
posed for theory, and describes the evolutionof mainstreamtheoretical
views duringthe ten years of floatingexchange rates from 1973to 1983.
A subsequentsection takes a statisticallook at how muchof the variance
of these exchange rates can be relatedsystematicallyto other variables
4. Capitalcontrols were pervasive at the beginningof the floatingrate period. The

UnitedStates,Germany,andthe UnitedKingdomhaveliftedtheircontrols,andJapanhas
substantiallyliberalizedcapitalflows; these actions supportat least one predictionmade
by the advocates of floatingrates. Only France and Italy, among the large industrial
countries, now use capital controls as an importantshort-runpolicy tool, and these
countrieshave acceptedfixedparityobligationswithinthe EuropeanMonetarySystem.
that appearin the models. The central restrictionsof the models from
the standpointof exchangeratepolicy are then examined.A concluding
section presents an assessment of what economists do and do not know
aboutexchange rates and what policies are defensibleor indefensiblein
lightof this knowledge.
Exchange Rate Behavior and the Development

of Exchange Rate Theory
The theory of exchange rate determinationhas evolved largelyfrom

an asset view of exchange markets; that is, models have been built
around the determinantsof net outside supplies of stocks of assets
denominatedin differentcurrenciesand the demandsfor them. Prices
in asset markets are viewed as adjusting freely, and hence stock
equilibriumis assumed to be achieved very quickly, if not instanta-
neously, followinga disturbance.Expectationsare seen to play a central
role in the determinationof equilibriumand, reflectingthe fashion in
macroeconomicsmore generally, the rationalexpectations hypothesis
is the norm. Goods marketsare importantin varyingdegrees, although
they are generallykept in the background.At a minimum,they tie down
long-runexpectations and influencethe demandsfor assets.
THE FLEXIBLE-PRICE MONETARY MODEL
The model that provides the strongestcase for the advantagesof the
system of floatingexchange rates, the flexible-pricemonetarymodel, is
also the simplestmodel.SIt presentsthe exchangerate, S, as the relative
5. For some early contributionsto the monetaryapproachsee RudigerDornbusch,
"The Theoryof Flexible ExchangeRate Regimesand MacroeconomicPolicy," Scandi-
navian Journal of Economics, vol. 78, no. 2 (1976), pp. 255-75; Jacob A. Frenkel, "A
MonetaryApproachto the ExchangeRate:DoctrinalAspects andEmpiricalEvidence,"
Scandinavian Journal of Economics, vol. 78, no. 2 (1976), pp. 200-24; and Michael Mussa,
"The ExchangeRate, The Balanceof Paymentsand Monetaryand Fiscal Policy undera
Regimeof ControlledFloating," ScandinavianJournalofEconomics,vol. 78, no.2 (1976),
pp. 229-48. Earlyempiricalevidencewas providedby Lance Girtonand Don Roper, "A
MonetaryModelof ExchangeMarketPressureAppliedto the PostwarCanadianExperi-
ence," American Economic Review, vol. 67 (September 1977), pp. 537-48; John F. 0.
Bilson, "The MonetaryApproachto the Exchange Rate: Some EmpiricalEvidence,"
IMF StaffPapers, vol. 25 (March1978),pp. 48-75; andRobertJ. Hodrick,"An Empirical
Analysisof the MonetaryApproachto the Determinationof the ExchangeRate," in Jacob
A. Frenkel and Harry G. Johnson, eds., The Economics of Exchange Rates: Selected
Studies(Addison-Wesley,1978),pp. 97-116.
priceof two monies, Mand M* (anasteriskindicatinga foreigncountry),
each supplied as the liability of a central bank and demandedby the
residents of one country. The condition for zero excess demands in
money marketsdeterminesprice levels in the two respective countries,
P and P*. The assumptionof one tradablegood in the world implies a
constant real exchange rate reflecting purchasingpower parity, and
hence a path for the nominalexchange rate determinedby the paths of
money suppliesandoutputs, Yand Y*,in the two economies. The model
has an equationfor domestic monetaryequilibrium,
Mt= m(rt)PtYt;
for purchasingpower parity,

Pt = StPt*;t
andfor uncoveredinterestrateparity,
St = rt - rt*
where r is the nominalinterest rate, e denotes an expected value, and a

lowercase renderingof a level variabledenotes its percentagechange,
computed as the difference in naturallogarithms.In a small-country
model the foreign price level and nominal interest rate, P* and r*,
respectively, are taken as exogenously given.
The uncovered interest-rate parity condition implies that assets
denominatedindifferentcurrenciesareviewedby internationalinvestors
as perfect substitutes in portfolios. With perfect asset substitutability
and rationalexpectations, the fundamentalbehaviorof exchange rates
is governed by purchasingpower parity. If one country has a higher
expected monetary growth rate and consequently a higher expected
inflation rate, assets denominatedin its currency will carry a higher
interest rate that is exactly offset by an expected depreciationof its
exchange rate. So uncovered interest rate parityimplies that expected
rates of return on interest-bearingassets denominated in different
currencies are eciualized.6Since in this model the rate of exchange
6. One related propositionthat was settled ratherearly and has stood the test of
time is thatcoveredinterestarbitrage(the equalityof the annualizedforwardpremiumor
discount and the interest differential)holds extremelywell for assets having the same
characteristicsexceptforcurrencyof denomination(forexample,Eurocurrencydeposits).
This impliesthatforwardmarketsneed not be modeledexplicitly.It also impliesthatthe
uncovered interest arbitragecondition is equivalentto the condition that the forward
depreciation is just the difference in inflation rates between the two

countries, this propositionholds whetherreturnsare measuredin either
country'scurrencyor in realterms.The realinterestratein the domestic
economy is tied to the exogenous real rate in the foreigneconomy.
Althoughthe currentaccountdoes not appearexplicitlyin this model,
the model is not inconsistent with surplusesand deficits in the current
account financedby net accumulationof cross-borderclaims andliabil-
ities of a nonmonetarycharacter.However, imbalancesin the current
accountwould reflectdifferencesin propensitiesto save andinvestment
opportunities-considerations that are kept in the backgroundand are
assumed not to be influencedby exchangemarketdevelopments.They
would influenceexchange rates only to the extent that they alteredthe
expected futurepathof output.
The diagramopposite illustratesthe characteristicresponses of the
nominal and real exchange rates to a one-time increase in the money
supply and to a permanentincrease in the growth rate of the money
supply for the flexible-pricemonetarymodel. Prices and interest rates
in the rest of the world are taken as given. The time paths for variables
in the diagramassumeperfectforesightby economic agents, except that
the two policy changes are unanticipated.A one-time increase in the
domestic money stock occurs at time tl, resultingin a jump in the price
level and a rise in the price of foreign exchange (a depreciationof the
home currency) to new equilibriumvalues. The real exchange rate is
unaffected.An increasein the rateof growthof the money supplyoccurs
at time t2. The rate of price increase rises immediatelyby the same
amountas the increase in the growthrate of the money supply, and the
exchange rate begins risingat the same rate, that is, the home currency
exchangerateis equalto the expectedfuturespot rate, whichprovidesan alternativeand

essentiallyequivalentway of testingthe uncoveredinterestparitycondition.These tests
are discussed below. Interest rates in differentnationalmoney marketshave at times
deviatedfromcorrespondingEurocurrencymarketinterestratesandhencefromcovered
interestparitywhen capitalcontrolshave been tight.Tax requirements,reserverequire-
ments, and risk considerationsalso influencecovered interest paritydeviationsjust as
they lead to interest rate differentialswithin nationalmarkets,but for a wide rangeof
instrumentsthese differencesare relativelysmall. For evidence on the covered interest
arbitragecondition, see Jacob A. Frenkel and RichardM. Levich, "Covered Interest
Arbitrage: Unexploited Profits?" Journal of Political Economy, vol. 83 (April 1975), pp.
325-38; FrankMcCormick,"Comment,"Journal of Political Economy, vol. 87 (April
1979), pp. 411-17; and RichardC. Marston, "Interest Arbitragein the Euro-currency
Markets,"EuropeanEconomicReview,vol. 7 (January1976),pp. 1-13.
Nominalexchange
rate
Real exchangerate
Price level
- - Moneysupply
_~~~~~~._
_______'
tl t2
Time
depreciates. The exchange rate and price level also jump at time t2
because the demandfor money depends on the nominalinterest rate.
But even so, the realexchangerate is unaffected.
EXCHANGE RATE BEHAVIOR: MARCH 1973 TO LATE 1975
By late 1975it was becomingclear that purchasingpower paritywas

not well maintainedand that changes over time in real exchange rates
were much larger than those under fixed rates.7 Figure 1 shows the
behavior of the nominal and real exchange rates over the fixed and
7. Econometricstudiesof purchasingpowerparityincludeIsard'sstudyusingdisag-
gregatedpricedata. See PeterIsard,"How FarCanWe Push 'The Law of One Price?' "
American Economic Review, vol. 67 (December 1977), pp. 942-48. See also Paul R.
Krugman,"PurchasingPower Parity and Exchange Rates: Another Look at the Evi-
dence," Journal of International Economics, vol. 8 (August 1978), pp. 397-407; Hans
Genberg,"PurchasingPowerParityUnderFixedandFlexibleExchangeRates,"Journal
of InternationalEconomics, vol. 8 (May 1978),pp. 247-76; and Irving B. Kravis and
RobertE. Lipsey, "PriceBehaviorin the Lightof Balanceof PaymentsTheories,"Journal
of International Economics, vol. 8 (May 1978), pp. 193-246.
Figure 1. Nominal and Real ExchangeRatesa
Markper dollar(index,1973:1= 100)
150 Germany
100
100 _ \ ~~~~~~~~~~~~Real/
65 - Nominal/
.I
a
, , , , I .' | , l I I 'i
1960 1965 1970 1975 1980
Yenper dollar(index,1973:1= 100)
Japan
150 - ~ _ Real
a,-b- - -^ -_A
Nominal
100
65-
I ,,,, I , ,,I , ,, I , , I,
1960 1965 1970 1975 1980
Source: Federal Reserve Bank of New York, international macro data base.
a. Quarterly data. Price adjustments use consumer price indexes. The effective U.S. dollar is the weighted average
of six currencies-Canadian dollar, French franc, mark, lira, yen, and pound sterling (expressed in units of foreign
Figure 1 (continued)
Poundper dollar(index,1973:1= 100)
150 UnitedKingdom
100
- Nominal
65
L, I , I I I I I IIII I I I I I I I I I I I
1960 1965 1970 1975 1980
EffectiveU.S. dollar(index,1973:1= 100)
150 W UnitedStates
--' ominal ">

100 V---- %
65
1960 1965 1970 1975 1980

currency per dollar). Weights are from the International Monetary Fund's multilateral exchange rate model and are
normalized for the six countries.
floatingrate periods for the mark, yen, and pound as measuredagainst
the dollarand a weighted-averagevalue of the dollaragainst six major
currencies-'
Wheninflationrates divergedwidely, as they did between the United
States and the United Kingdomfrom mid-1972to 1977, the exchange
rate showed some tendency to change over time so as to contain
deviations from purchasingpower parity.9Nevertheless, most of the
nominal exchange rate fluctuationsin the period from March 1973 to
September 1975 were reflected in movements of real exchange rates.
These relativelyshort-termexchangerate swings were of greatconcern
to many at the time, but they shriveled almost to insignificancewhen
comparedwith the largesustainedswings that came later.
Exchangeratemovementshave often been analyzedandtheircauses
debated in terms of episodes that appearto be associated with specific
events. Indeed, until a ratherlengthy data record could be compiled,
moreformalstatisticaltechniqueswere of little value. Afterthe decision
to float exchange rates, the dollar continued to depreciatefor several
monthsin both nominalandrealterms,especiallyagainstthe mark.This
depreciationwas widely explainedas a continuationof the adjustment
to the misalignmentof the dollarand the markthat had built up during
the Bretton Woods period, but some marketobservers expressed con-
cern that speculation against the dollar was feeding on itself and was
carryingthe dollar to unrealisticallylow levels. The dollar ultimately
stabilizedand recovered somewhatafterJuly 1973,but not before U.S.
authoritieshad temperedtheir policy of benign neglect and intervened
to resist the slide.
The dollar strengthenedagainst most currenciesfollowing the Arab
oil embargoof October 1973and the quadruplingof oil prices at the end
of the year. The poundwas also remarkablystrongin the wake of the oil
shock, considering the sharp increase in wages and prices that had
occurredin the United Kingdom.The yen weakened sharply.
8. The six currenciesarethe Canadiandollar,Frenchfranc,deutschemark,lira, yen,
andpoundsterling.
9. Whilethe tendencyfornominalexchangeratechangesto offset differentialchanges
in price levels has been weak in the currentperiod of floatingrates, it has been more
markedwhen nominalinflationdifferentialshavebeen very large.See the evidenceon the
Germanhyperinflation inJacobA. Frenkel,"ExchangeRates,Prices,andMoney:Lessons
fromthe 1920s," American Economic Review, vol. 70(May1980,Papers andProceedings,
1979), pp. 235-42.
Three reasons were advancedto explainthese adjustmentsto the oil

shock. First, the United States appearedto be followinga relativelyless
accommodatingcourse than other countriesafterthe inflationaryimpe-
tus of the increases in oil prices. An explosion of the money supply,
wages, and prices in Japanstood out by contrast. Second, at that time
the United States was relatively less dependent on importedoil than
other countries. And third, the currentaccount surplusesaccumulated
by OPECmemberswere expected to be investedpredominantlyin dollar
andpoundassets.
The firstreason was roughlyconsistent with the flexible-pricemone-
tary model. The second could be reconciled with this model by recog-
nizingthatlargerealdisturbancescouldpermanentlyalterthe purchasing
power parity relation, even if monetarydisturbancescould not. Thus
considerableeffort was devoted to assessing how much the oil shock
couldbe expected to alterpermanentlycountries'competitivepositions.
The final consideration-portfolio preferences of OPEC investors
amongnonmoneyassets denominatedin differentcurrencies-was not
allowed for in the monetarymodel. But the shock was unprecedented
andthereforedid not necessarilydiscreditthe claimof the flexible-price
monetarymodel to explain a majorityof the persistent movements in
nominalexchange rates. Broadcapitalcontrolshadbeen inheritedfrom
the BrettonWoods periodand could also be arguedto have affected the
rangeandrelativeattractivenessof variouscurrenciesfor the investment
of OPECfinancialwealth.U.S. authoritiestookadvantageof the strength
of the dollarto lift all controls on capitalflows out of the United States
in January1974.This representeda furtherstep toward allowing rates
to be determinedby marketforces. But Germanycontinuedcontrols on
inflows, the United Kingdomon outflows, and Japanon both.
In mid-1974one could look back on the firstfifteenmonthsof floating
andbe somewhatcomplacentthat the system of floatingexchange rates
would function roughly as predicted by the flexible-price monetary
model despite the movementsof real exchangerates that had occurred.
The majordevelopments seemed explainableby continuedadjustment
to the disequilibriumbuilt up duringthe Bretton Woods period and by
the unprecedented oil shock. Moreover, it was not surprisingthat
marketswere taking some time to adjustto the new environment.The
continued prevalence of capital controls meant that the conditions
prescribedby the advocates of floatinghad not been fully established.
But the skeptics could note that the slide of the dollarin 1973had only
been reversed after concerted interventionby the centralbank. In any
event, the next episode posed a greaterchallengeto the theory.
The United States maintaineda relatively restrictive policy stance
throughmost of 1974as the U.S. economy experienceda less marked
drop in output than most other countries in the wake of the oil shock.
Thenin the fall of that year a sharprecession beganin the UnitedStates.
Short-terminterest rates on dollar assets fell relative to average rates
abroad and particularly relative to rates for the mark. The dollar
weakened, decliningby 10 percent againstthe markand 5 percent on a
weighted-averagebasis. Adjusting for prices, the weighted-average
depreciationwas even sharper,and against the mark it was nearly as
large.Thedropinthe dollarultimatelydroveU. S. authoritiesto intervene
once again.Finallythe slidecameto anendanda sharpreboundoccurred
with the strong recovery in the United States in 1975:2. Short-term
interest rates on dollar instrumentsrose relative to rates on the mark
and yen counterpartswhile U.S. inflationwas declining more rapidly
than the lower inflationrate in Germany.By Septemberthe dollarhad
recoveredfromits earlierdecline.
Figure2 shows the three real bilateralexchange rates, togetherwith
the correspondingthree-monthinterest differentialand smoothed-out
inflationratedifferential.The associationbetweenchangesinthe interest
rate and in the exchange rate in late 1974and early 1975is apparentfor
the mark-dollarrate. Some parallelmovementscan also be seen at other
times and in other currencies, especially when interest rates move
counter to inflationrates. But the relation is not close. Nevertheless,
changesin the interestrateplayeda prominentrole in day-to-daymarket
commentary. And the tendency, evident in late 1974 and 1975, for a
rising nominal interest rate to be associated with a strong currency,
contraryto the predictionof the flexible-pricemonetarymodel, stimu-
lated interest in findinga theoretical connection. The lack of a close
correspondencebetween money growthand exchange rate movements
in the early period of floating rates also suggested the need for a
modificationof the theory.
THE STICKY-PRICE MONETARY MODEL
The principaltheoreticalresponseto these observationswas renewed

recognition of the importance of price stickiness for understanding
JeffreyR. Shafer and Bonnie E. Loopesko 17
changesin the exchangeratesamongmajorcurrenciesin the 1970s.This
"Keynesian" assumption had fallen into disfavor in the inflationary
environmentof the early part of the decade, but it was revived as a
central assumption of the dynamic sticky-price monetary model of
exchange rate determination.As developed by Dornbusch,this model
predicted that, following an unanticipatedmonetary disturbance,ex-
change rate expectations would deviate from purchasingpower parity
for as long as it took goods prices to adjustfully to the new monetary
conditions and goods market clearing to be restored.10Moreover, the
exchange rate could (dependingon the parametervalues of the model)
overshoot its long-runpath.
The monetary equilibriumcondition and perfect substitutabilityof
nonmoneyassets were retainedfromthe flexible-pricemonetarymodel.
Althoughgoods marketsmight adjust slowly, asset marketswere still
assumedto clear continuously.Purchasingpower paritybecamea long-
runtendency since the equilibriumrealexchangeratewas unaffectedby
purelymonetaryfactors. The new elementswere an equationdetermin-
ing the inflationrate-a stylizedexpectations-augmented Phillipscurve-
and an IS locus incorporatingthe relative price of foreign and home
goods andthe realinterestrate. Workwiththismodelhas mainlyfocused
on the case of perfectlyforeseen changes in exchange rates.
A sticky-pricemonetarymodel,then, hasanequationforthe condition
of monetaryequilibrium,
Mt = m(rt) PtYt;
for the IS locus,
Yt = y(StP* /Pt, rt - pe);
for the expectations-augmentedPhillipscurve,

P = g(Yt - Y) + v;
andfor uncoveredinterestparity,
Se = rt - r*
10. TheDornbuschmodelpresentedin "ExpectationsandExchangeRateDynamics,"
Journal of Political Economy, vol. 84 (December 1976),pp. 1161-76, extended earlier
work by RobertA. Mundell,InternationalEconomics(Macmillan,1968)and J. Marcus
Fleming,"DomesticFinancialPoliciesunderFixed and FloatingExchangeRates," IMF
Staff Papers, vol. 9 (November 1962),pp. 369-80. Otherimportantwork in the sticky-
pricetraditionincludesMichaelMussa,"A Modelof ExchangeRateDynamics,"Journal
of PoliticalEconomy,vol. 90 (February1982),pp. 74-104.
Figure 2. Real Exchange Rates, Short-TermInterest Rates,
and Inflationa
Markper dollar Interestand inflationdifferential
(index,March1973 = 100) (percent)
150 15
Germany
1000 0 S
,, . Interestd~~infferetiondifreta
65 _'^ ,""'' _ -5
1975 1980
Yenper dollar Interestand inflationdifferential

150 15
Japan Interestdifferential r
10
100 Realexchangerate
65 -5
ation differential
I , , | I I |
, -~~~~~~~~~~~~~~15
1975 1980
Poundper dollar Interestand inflationdifferential

150 5
UnitedKingdom
Interestdifferential
d
100
65 - Real exchangerate -- 15
1975 1980
Source: Monthly data. Federal Reserve Bank of New York, international macro date base.
a. The interest rates are based on the three-month interbank rate for Germany, the sixty-day Treasury bill for
Japan, the ninety-one-day Treasury bill for the United Kingdom, and ninety-day bankers' acceptances for the United
States. The inflation rate is measured as a twelve-month centered moving average of consumer price changes.
Differentials are calculated as U.S. rates minus foreign rates.
where the rest of the world is taken as exogenous. The actual and
expected rates of change in the exchange rate are the same, except at
theinstantwhenanunanticipateddisturbanceoccurs. Thenthe exchange
ratejumps to a new path. Note that the Phillipscurve equationrequires
that, for a stable long-run equilibrium, Y, converge to Y. Long-run
purchasingpower parityis thenimpliedby the IS locus. The v parameter
is the expected equilibriuminflationrate, which equals the expected
growthrate of the money supply(fromthe monetaryequilibriumcondi-
tion).
The diagrambelow illustrates for the sticky-pricemonetary model
the response of nominaland real exchange rates and of prices to a one-
time increasein the money supplyat time tI andto a permanentincrease
in the growthrate of the money supplyat time t2. As in the firstdiagram
above, perfect foresightis assumed, except for the disturbances,which
are unanticipated. Following a monetary disturbance, the variables
Nominalexchange
|..
'. ',Real exchangerate
Price level
_/ - - , Moneysupply
/~~~~~~~~~~~~~~ ~~~~00,/
/ ,
_ __
tl t2
Time
ultimatelyconverge to the same paths as in the flexible-pricemonetary

model.11 But for some time the real exchange rate is deflectedfrom its
equilibriumvalue. For example, an unanticipatedmonetaryexpansion
at time tl leads to a transitoryrise in the real exchange rate (a real
depreciationof the home currency). The nominal exchange rate may
overshoot its long-runlevel or path, as illustratedin the diagram,but
this result dependson the parametervalues of the model.
The sticky-pricemonetarymodelwas consistentwith severalstylized
facts that did not fit with the flexible-pricemodel. Not only did it admit
deviationsfrompurchasingpowerparity,but it providedan explanation
for periods when a rising nominal interest rate was associated with a
11. Monetarydisturbancesmay includeunanticipatedchangesin monetarypolicy or
changesin moneydemandthatare expectedto persistwhile monetarypolicy is expected
to pursuean unchangedmonetarytarget.Changesin the IS locus, includingmodifications
in fiscalpolicy, can give rise to the same kindof exchangerateeffects if they do not alter
the long-runtermsof trade.
strong currency. In the flexible-pricemonetary model, a rise in the
interestratewas alwaysassociatedwithanincreasein the rateof inflation
and morerapiddepreciation(orless rapidappreciation)of the currency.
In the sticky-price monetary model a persistently higher level of the
interest rate would reflect higher inflation and be associated with a
weaker currency. But an increase in the interest rate and declining
inflationaryexpectationsmay be producedby a shiftto tightermonetary
policy, and then the currencywill strengthen.
The model suggeststhat changesin the exchangerate may not have a
simple relation to monetaryaggregateseven when money supply dis-
turbancesare the underlyingcause. Expectationsconcerningthe future
behavior of money supplies may be much more importantthan their
currentbehaviorin determiningthe exchangerate.Thuspoorcorrelation
between contemporaneouschanges in monetary aggregates and ex-
changerates can be explained.(Thisis also trueof the rationalexpecta-
tions versionof the flexible-pricemonetarymodel.) The overshootingof
the exchange rate associated with sticky prices also seems to account
for a widespread perception that rates tend to move too far in one
directionand then reverse themselves.
Withstickyprices,the assumptionthatassets denominatedindifferent
currenciesmust providethe same expected rateof returnto a particular
investor (uncovered interest parity) does not mean that real interest
rates must be the same for residents of differentcountries, as it does
with flexible prices. Indeed, differences in real interest rates across
residents of differentcountries must match expectationsof real appre-
ciation or depreciationof currencies. This point is discussed further
below.
Looking back over the first three years of floating, it appears that
sustainedexchange rate swings occurredover periods of three months
to a year. Thus the convergence to equilibriumwas assumed to be
relatively rapid. In empiricalwork real interest rate differentialswere
sought in short-termrates, while changes in long-term interest rates
were viewed as indicative of changinginflationexpectations. Frankel
achieved some success in interpretingthe modelthis way for explaining
the behavior of the mark-dollarrate.12 Althoughallowances had to be
12. JeffreyA. Frankel,"On the Mark:A Theoryof FloatingExchangeRates Based
on Real InterestDifferentials,"AmericanEconomicReview, vol. 69 (September1979),
pp. 610-22.
made for the initial adjustmentto floatingrates and the oil shock, the
sticky-pricemonetarymodel showed promise.
EXCHANGE MARKET DEVELOPMENTS:

LATE 1975 TO EARLY 1979
Developments in 1976 did not seriously challenge the sticky-price

monetarymodel. It was the most tranquilyear for the weighted-average
exchange rate of the dollarof the entire ten years of floating.But there
were majoradjustmentsamongforeign currencies.Authoritiesin Italy
and France were forced to abandon intra-Europeannarrow-margin
commitments,and theircurrenciesdepreciatedsharply.The downward
pressure then shifted to the pound. Some saw the weakness of this
currencyas reflectinga structuralshift away from the poundby OPEC
andotherinvestors. Thepoundstabilizedonly aftera standbyagreement
callingfor more restrictivepolicies was concludedbetween the United
Kingdomand the InternationalMonetaryFundat the end of the year.
The weakness of the lira, franc, and poundand subsequentaccelera-
tion of inflationin Italy, France, and the United Kingdomrevived and
sharpened the debate about whether floating exchange rates tend to
destabilizeinflation.Therewas considerablediscussionof "vicious and
virtuous cycles" in which high inflationand currency depreciationor
low inflationand currencyappreciationreinforceone another.13At least
some who saw such a process at workseemedto have in minda perpetual
cycle that was touched off by an exogenous disturbanceto exchange
rates. At the other pole were those who believed the source of such
apparentcycles was the differentmonetarypolicy choices of national
13. For more detail on variousaspects of the vicious circle hypothesis, see Rudiger
Dombusch and Paul Krugman,"Flexible Exchange Rates in the Short Run," BPEA,
3:1976,pp. 537-75;GiorgioBasevi and PaulDe Grauwe,"Vicious and VirtuousCircles:
A TheoreticalAnalysisand a Policy Proposalfor ManagingExchangeRates," European
EconomicReview, vol. 10 (December 1977),pp. 277-301; MauriceObstfeld, "Relative
Prices, Employment,andthe ExchangeRatein an Economywith Foresight,"Econome-
trica,vol. 50 (September1982),pp. 1219-42;andJohnF. 0. Bilson, "The 'ViciousCircle'
Hypothesis," IMF Staff Papers, vol. 26 (March1979),pp. 1-37. For a sortingout of the
theoreticalissues, see HenryWallichandJo AnnaGray,"StabilizationPolicyandVicious
and VirtuousCircles," in John S. Chipmanand CharlesP. Kindleberger,eds., Flexible
Exchange Rates and the Balance of Payments: Essays in Memory of Egon Sohmen
(Amsterdam:NorthHolland,1980),pp. 49-65.
authorities.They used the sticky-pricemonetarymodel to demonstrate
that an actual or expected shift towarda more expansionarymonetary
policy would be reflected more rapidlyin currency weakness than in
higherinflation.Thusto infercausalityfromthe observationthatchanges
inexchangeratessometimesprecededchangesininflationwas to commit
the fallacy of post hoc ergo propterhoc. In the middlewere some who
believed that disturbancescould originatein foreignexchange markets
as well as in domestic monetary policy, but that these disturbances
wouldbe self-limitingunless accommodatedby monetarypolicy.
After 1976, evidence less favorable to the sticky-price monetary
modelbeganto emerge. Oneanomalywas thatthe modeldidnot provide
a role for shifts in investors' preferences among currencies. Market
commentaryincreasingly associated exchange rate movements with
such portfolioshifts; and evidence began to accumulateas early as late
1976that the uncovered interestparityconditionmightnot hold or that
the expectations for which it held were not rational.Put anotherway,
there appeared to be systematic differences in the rates of return
achievable from holding open positions in different currencies if no
allowancewas madefor risk.
Early research supportingthis result was not conclusive and used
daily data, hence allowingfor the possibilitythat the anomalymightbe
a very short-runphenomenon.14 The apparentprofitopportunitiescould
also be dismissedas a temporaryphenomenonthatwoulddissipateonce
marketparticipantslearnedto operatein the new regime.Moreover,the
results had to be viewed as tentative until out-of-sampletests could
be conducted. But evidence has continued to accumulate that casts
strong doubt on the joint hypotheses embeddedin tests of uncovered
interest rate parity-that is, the hypotheses that otherwise identical
nonmoney assets denominatedin different currencies are viewed as
perfect substitutes (and hence bear the same expected rates of return)
andthat expectationsare rational.15Earliertests have been redonewith
14. See MichaelP. Dooley and JeffreyR. Shafer,"Analysisof Short-RunExchange
Rate Behavior:March1973to September1975,"InternationalFinanceDiscussionPaper
123(Boardof Governorsof the FederalReserveSystem, 1976).
15. Tests of this joint hypotheses are alternativelyreferredto as tests of foreign-
exchangemarketefficiencyor as tests for time-varyingrisk premiums,dependingmainly
on which partof the joint hypothesisthe researcheris most preparedto interpretas the
source of the rejection.Little progresshas been made towardempiricallyresolvingthe
questionof whichhypothesiscan be rejected.
new data confirmingthe existence of systematicprofitopportunities.16
Moreover, with a lengtheningperiod of floatingrates to examine and
new econometric techniques, tests on weekly, monthly, and even
quarterlychanges in exchange rates have become more powerful, and
these tests tend to give qualitativelysimilarresults.17
Beginningin June 1977, the dollar began a steep slide against most
currenciesthat continueduntil November 1, 1978,when a majoreffort
was undertakento stabilize its foreign exchange value. The weighted-
averagedollardroppedmore than 15percentduringthis period-and 13
percentin real terms. The depreciationagainstthe yen was abouttwice
as large. In these years the U.S. inflation rate rose relative to that
abroad,but U.S. short-termnominalinterestrates rose relativeto rates
abroadby about as much. Althoughinflationaryexpectationsmay have
contributedto the nominaldepreciationof the dollar, the combination
of price and interest rate developmentsin this period did not appearto
explain such largechangesin the real exchangerate.
Fellnerprovidesa case for understandingthe slide of the dollarduring
this period in terms of the sticky-pricemonetarymodel. He arguesthat
differentialsin real interestrates amongcountriesmightbe expected by
marketparticipantsto be much more persistentthan assumedearlier.18
Frankel in his work on the mark-dollarrate and others assume that
changes in long-termnominalinterestrates are a measureof changesin
inflationexpectations.19According to this line of thought, only short-
term interest rates are viewed as moving somewhat independentlyof
inflation. By extending the hypothetical duration of disequilibrium,
16. See MichaelP. Dooley and JeffreyR. Shafer, "Analysisof Short-RunExchange
Rate Behavior:March1973to November 1981," in DavidBigmanandTeizo Taya, eds.,
Exchange Rate and Trade Instability: Causes, Consequences and Remedies (Ballinger,
1983),pp. 43-69.
17. Recent econometricstudies include Lars Peter Hansen and RobertJ. Hodrick,
"ForwardExchangeRates As OptimalPredictorsof FutureSpot Rates:An Econometric
Analysis," Journal of Political Economy, vol. 88 (October 1980), pp. 829-53; Robert E.
Cumbyand MauriceObstfeld, "A Note on Exchange-RateExpectationsand Nominal
InterestDifferentials:A Test of the FisherHypothesis,"Journalof Finance, vol. 36 (June
1981),pp. 697-703;and John F. 0. Bilson, "The 'SpeculativeEfficiency'Hypothesis,"
Jolurnal of Business, vol. 54 (July 1981), pp. 435-51.
18. See WilliamFellner, "The Bearingof Risk Aversionon Movementsof Spot and
ForwardExchangeRelativeto the Dollar," in John S. Chipmanand CharlesP. Kindle-
berger, eds., Flexible Exchange Rates and the Balance of Payments: Essays in Memoty
of Egon Sohmen(Amsterdam:NorthHolland,1980),pp. 113-26.
19. Frankel,"Onthe Mark."
Fellner associates much largerchanges in the real exchange rate with
the effects of divergentmonetarypolicies on realinterestdifferentials.20
But havingdroppedthe long-terminterestrate as a measureof long-run
inflationaryexpectations, Fellner'sargumentrests on the plausibilityof
particularvalues for inflationaryexpectations rather than hypothesis
tests. We returnto the Fellnerhypothesisbelow.
Many observers began to view the 1977-78 slide of the dollar as
primarilyan adjustmentto the large and growing deficit in the U.S.
currentaccount and surplusesin Germanyand Japan.This constituted
a second phenomenonthat was not directly explainablein the context
of the sticky-pricemonetarymodel. Figure3 shows the real weighted-
averageexchange rates for these three countriesand the United King-
dom, togetherwith theircumulativecurrentaccountpositions from the
first quarterof 1973to the end of 1982. (The change in the cumulative
measure is the current account for each period.) Note that current
accountimbalanceshave been largerand more volatile from the outset
of the floatingrate periodthanthey had been duringthe BrettonWoods
period. The extenuatingfactors discussed above-greater trade inter-
dependence,inheriteddisequilibrium,and the two oil shocks-allowed
the advocates of floatingrates to remainunconcerned. But those who
had made strong claims for the self-equilibratingpropertiesof floating
exchangerates had certainlyoverstatedthe case.
The events of 1977-78had theiroriginsseveral years before. In early
1975 the U.S. current account surged into surplus as the sharp U.S.
recessionreducedimports.But with recoveryin the economy andin the
dollar, the currentaccount began to decline rapidly.At the same time,
GermanyandJapanwere in substantialcurrentaccountsurplus.Yet the
marketsreacted only after Secretaryof the TreasuryBlumenthalmade
a statement in June 1977 calling attention to the imbalances. (The
statementwas widely interpretedas a desire on the part of the Carter
administrationto see the dollardecline.)
One source of the currentaccount imbalancesand exchange market
pressureswas the relativelymoregrowth-orientedpolicies of the United
20. To illustratethis point, assumefull adjustmenttakes five years and a real rate of
returndifferentialof 2 percentis expected to prevail,on average,in the interim.Then a
displacementof the exchangeratefrompurchasingpowerparityof morethan 10percent
wouldbe warranted.Thusthe modelcould be consistentwithratherlargechangesin the
exchangeratein responseto moderatechangesin long-termnominalinterestratesrelative
to inflationrates.
Figure 3. Real Effective ExchangeRates and CumulativeCurrent
Account Positionsa
Cumulative
Real effectiveexchangerate, mark currentaccount
(index,1973:1= 100) (billionsof marks)
100
Germany , " Currentaccount

150 / " 80
10 o Real effectiveexchangerate 40
_ ,' ] ~~~~~~~~~~~~~
65 ,"oll l l
1975 1980
Cumulative
Real effectiveexchangerate, yen currentaccount
(index,1973:1= 100) (trillionsof yen)
8
150 6
Real effectiveexchangerate /
*' Currentaccount
pg ~~~~~~~~~~0
65 I V 1 I I I I -2
1975 1980
Source: Quarterly data. Federal Reserve Bank of New York, international macro data base.
a. Effective rates are weighted averages of each currency against six others (expressed in units of foreign currency
per unit of domestic currency). The seven currencies are the U.S. dollar, mark, yen, pound sterling, lira, Canadian
Cumulative
Real effectiveexchangerate,pound currentaccount
(index,1973:1= 100) (billionsof pounds)
15
UnitedKingdom
150 -10
5
Real effective exchangerate 0
1000 , 0_
Currentaccount
65 10
1975 1980
Cumulative
Real effectiveexchangerate, U.S. dollar currentaccount
(index,1973:1= 100) (billionsof U.S. dollars)
40
UnitedStates
150 _ 30
, \ Currentaccount
20
100 10
65 L l -o t
1975 1980
dollar, and French franc. Weights are normalized according to the International Monetary Fund's multilateral
exchange rate model for the seven currencies. The cumulative current account is the algebraic sum of current
surpluses and deficits from 1973:1.
States and the more anti-inflationemphasisof policies in Germanyand
Japan. Another source was U.S. energy policy, which discouraged
domesticoil productionby keepingoil pricesbelow worldmarketlevels,
therebyincreasingalreadyexpandingoil imports.At the Bonn Summit
in July 1978 these considerationslay behind a package that included
commitmentsto fiscal stimulusby Germanyand Japan,where inflation
had droppedto low levels, and commitmentsto energy price decontrol
and strongeranti-inflationpolicies by the United States. The slide of the
dollarcontinued,however, andeven steepened.21
OnNovember1, 1978,U.S. authoritiesannouncedtheirdetermination
to correct what they termed an excessive depreciationof the dollar.22
The dollarwas driven up sharply,but it was not until early 1979that it
showed some strengthon its own, and interventioncould begin to be
unwound.By then, signs of currentaccount convergencewere appear-
ing.
Marketcommentaryin 1977-78 also focused on internationaldiver-
sificationof asset portfoliosas an additionalsource of dollarweakness.
Exchange rate risk as well as expected returnreceived more attention
as potential forces in exchange markets.23Evidence on how much
currencydiversificationoccurredis skimpy. But even the data that are
availableare hardto interpretbecause changingshares of currenciesin
portfolioswere dominatedby changingcurrencyvalues ratherthan by
changes in the relativequantitiesof dollarsand other currencies.24
21. A number of efforts were made to moderate the slide. Initially intervention
operationswere undertakenmainlyby GermanandJapaneseauthorities;butas timewent
on, therewas increasedinterventionby U.S. authoritiesusingforeigncurrenciesobtained
fromswap drawingswithforeigncentralbanks.
22. The measures announcedincluded stepped up interventionbacked by greatly
expandedresourcesanda tighteningof monetarypolicyunderscoredby a hikeof 200basis
pointsin the FederalReservediscountrate.
23. It was thoughtthat MiddleEasterncountriesand some others were increasingly
spreadingtheirforeigncurrencyreservesover a numberof currenciesratherthanholding
only or mostly dollarassets in orderto reduce exposure to exchange rate fluctuations.
Privateinvestorswere thoughtto be respondingto the same incentives.
24. These valuationeffects couldbe interpretedas the resultsof attemptsto diversify
leadingto exchangeratechangesthatbroughtactualportfoliosharesintoline withdesired
shareswithoutlargenet changesin currencyholdings.Indeed,in a purefloatingexchange
rate regime(with no intervention)this is the only way diversificationcould occur in the
aggregatebecause net supplies of assets denominatedin various currencieswould be
determinedsolely by fiscal deficits. The evidence is also difficultto evaluate because
changes in exchangerates from other causes would lead to the same valuation-induced
changesin portfoliosharesif investorswere passive. Moreover,even if holdersof dollars
THE PORTFOLIO-BALANCE MODEL
The developmentof varioustheoreticalformulationsof the portfolio-

balance model was stimulatedby evidence that the uncovered interest
arbitrageconditionmightnot hold, the casualobservationthatexchange
rates seemed to be influenced by cumulative changes in the current
account, and the expressed interest of official and private portfolio
managersin exchange risk as well as the expected rate of return on
currencies.The portfolio-balancemodel has been aroundfor some time
and has antecedents in both the theory of fixed exchange rates and
domestic financialtheory.25Its distinguishingfeatureis the assumption
that interest-bearingassets denominatedin differentcurrenciesare not
perfectsubstitutesin portfolios,presumablybecauseportfoliomanagers
are risk averse and because exchange risk cannot be completely elimi-
natedby diversification.
In a simplestaticportfolio-balancemodel, money marketequilibrium
is expressed as
MX = m(r,)P,VY.
were able to shift into other currenciesor alter the relativedistributionof additionsto
reservesbecauseauthoritiesin the key financialcenterswere intervening,the shiftscould
be interpretedas a responseto expectationsof dollardepreciations-behaviorconsistent
withthe monetarymodels.
25. The portfolio-balanceapproachhas been pursuedby a numberof authors. See
LanceGirtonandDaleW. Henderson,"CentralBankOperationsin ForeignandDomestic
Assets underFixedandFlexibleExchangeRates," in PeterB. Clarkandothers,eds., The
Effects of Exchange Rate Adjustments (U.S. Government Printing Office, 1977), pp. 151-
79; Pentti J. K. Kouri, "MonetaryPolicy, the Balance of Paymentsand the Exchange
Rate," in David Bigman and Teizo Taya, eds., The Functioning of Floating Exchange
Rates: Theory, Evidence and Policy Implications (Ballinger, 1980), pp. 79-111; William
H. Branson, "Asset Marketsand Relative Prices in Exchange Rate Determination,"
Sozialwissenschaftliche Annalen des Instituts far Hohere Studien, vol. 1 (1977), pp. 69-
89;andJeffreyR. Shafer,"TheMacroeconomicBehaviorof a LargeOpenEconomywith
a FloatingExchangeRate" (Ph.D. dissertation,Yale University, 1976).Antecedentsin
the fixed exchangerateliteratureincludeWilliamH. Branson,"MonetaryPolicy andthe
New View of InternationalCapital Movements," BPEA, 2:1970, pp. 235-62; Lance
GirtonandDaleW. Henderson,"FinancialCapitalMovementsandCentralBankBehavior
in a Two-Country,Short-RunPortfolioBalanceModel," Journalof MonetaryEconomics,
vol. 2 (January 1976), pp. 33-61; and Pentti J. K. Kouri and Michael G. Porter,
"InternationalCapitalFlows and PortfolioEquilibrium,"Journalof Political Economy,
vol. 82 (May-June1974),pp. 443-67. Theframeworkwas developedin a domesticcontext
in WilliamC. BrainardandJamesTobin,"Pitfallsin FinancialModelBuilding,"American
Economic Review, vol. 58 (May 1968, Papers and Proceedings, 1967), pp. 99-122.
30 BrookingsPapers on EconomicActivity, 1:1983
equilibriumin the marketfor assets denominatedin home currencyas
= - r* - +
b(rt Se)(Bt StB*);
equilibriumin the marketfor assets denominatedin foreigncurrencyas
= [1 - - r* - +
StB* b(rt se)](Bt StB*);
andthe IS locus, includingnetforeigninvestmentincomeas anargument,
as
Y= f (StP,*/P, rt - pe,
Sr*B,*/Pt, W,/Pt),
where W denotes wealth, and W -M + B + SB*.
In this model, foreigners do not demand assets denominated in
domestic currency, and money demanddepends only on the domestic
interest rate and nominalincome. The rest of the world is large so that
its interest rates, prices, and incomes are taken as exogenously given.
One asset marketequation is redundantby Walras's law. Thus, with
home prices and exchange rate expectations given at a point in time,
three equations determinereal output, the home interest rate, and the
spot exchange rate.26What distinguishes the model from monetary
models is thatthe expected rateof changeof the exchangerate, Se, is not
constrainedto correspondexactly to the interest differential,that is,
uncovered interest paritywill not hold in general. The expression (r -
r* - Se) is the deviationfromuncoveredinterestparity.If this expression
is positive, the home currencyis said to carrya risk premium.
A dynamicversion of the portfolio-balancemodel providesa mecha-
nism for cumulativeimbalancesin the currentaccount to influencethe
real exchange rate. Over time, wealth is transferredfrom countries in
deficit to countries in surplus. Assumingthat residents have a relative
preferencefor assets denominatedin their own currency, this redistri-
butionof wealth alters the relativedemandsfor assets. The currencyof
a country in deficit falls to a point from which it can subsequentlybe
expected to appreciate,thus establishingthe riskpremium.Beyond this
common feature, this model and variants of it can exhibit a range of
26. Moregeneralformulationsof asset demandsandendogenousadjustmentof foreign

variablescan be foundin the referencescited above. But at least some of the simplifying
assumptionsor alternativeones are normallyemployedto keep the modeltractable.The
model presentedin the text is intendedmerelyto illustratethe approachas an extension
of the sticky-pricemonetarymodel.
dynamicpropertiesdependingon the degreeof asset substitutabilityand
the specific propertiesof the real side of the model-the determinants
and speed of adjustmentof goods prices, aggregatedemands, and the
distributionof demand among goods produced in differentcountries.
Hence characteristicadjustmentpaths like those shown for the more
restrictivemodels are not given.
Oneformulationof the dynamicportfolio-balancemodelincorporates
the equationsof the static model and the followingequationsof motion:
evolutionof net supplyof assets denominatedin home currencyis given
by
(B, - Bt - ) = Dt - (Mt - Mt - ) - It,,
evolution of net domestic private holdings of assets denominatedin
foreigncurrencyby
St (B,*-B,* l) = PtCt + It,,
the expectations-augmentedPhillipscurve by
Pt = g(Y, - Y) + v,
andthe currentaccount by
,= h(Y, - Y, + Str*B*/Pt,
StP*lPt, W,/Pt)
where D is the domestic government deficit, I is the central bank
intervention(the sale of foreign currency reserves), and C is the real
currentaccountbalance.In this dynamicformof the modelthe evolution
of stocks of assets is governedby the fiscal deficit, the currentaccount,
and monetary policy. A role for exchange market intervention, even
when sterilized so as to have no effect on money supplies, is also
introduced.27
27. The usual formulationsof this model ignorethe role of governmentdeficitsas a

determinantof the supplyof domesticoutsideassets. Theirimplicationsfor the exchange
rate,throughthe influenceon the IS locus, has longbeen recognized.Fora smallnet initial
positioninforeigncurrency,the effects of a currentaccountdeficiton portfolioequilibrium
will be largerelativeto the effects of a budgetdeficitof the same size because the latter
altersthe supplyof assets as well as wealthwith a partiallyoffsettingeffect. Ignoringthe
effects of budgetdeficits on asset stocks may then have some justificationbeyondjust
keepingthe modeltractable.Butwiththegrowingattentionthatfiscaldeficitsareattracting
in market commentaryon interest rate and even exchange rate developments, their
theoreticalimplicationsare worthy of more attentionthan they have received.
EXCHANGE RATE DEVELOPMENTS:

EARLY 1979 TO MARCH I983
In the periodfrom 1979to 1982the positions of the dollarand pound

as weak currencies and the positions of the mark and yen as strong
currencieswere reversed.Afterstabilizingin early 1979,the dollarbegan
a climbthat, by any measure,dwarfedall its earlierswings. By late 1982
the real weighted-averagevalue of the dollarwas at a higherlevel than
at the beginningof the float almost ten years earlier or at any time in
between. The markwas nearits all-timelow on a real weighted-average
basis andthe yen, measuredthe sameway, fell to the level thatprevailed
in the earlypartof the float.The poundappreciatedeven moredrastically
than the dollarin 1979-80but had fallen back somewhatin 1981.In late
1982the pounddroppedsharplyand the dollardeclinedslightlyfrom its
highs measuredagainstthe markand yen.
Thefactorsthathave appearedin marketcommentaryto explainrates
since 1979 have been the same as those identifiedin earlier periods,
includingthose that had been incorporatedinto the progressivelymore
generalasset marketmodels. One mightsay that, afterthe firstsix years
of floating, we had seen it all. But the large swings in interest rates
relativeto inflationreceived by far the most attention.
In retrospect, 1979appearsto have been a relativelygood yearfor the
dollar.But the sense at the time was one of apprehension.Nonmonetary
as well as monetaryfactors received considerableattention.The dollar
dropped against European currencies as the second round of large
increasesin oil prices beganat mid-year.In contrastto 1974,the United
States was much more dependent on importedoil, and there was no
longer a presumption that OPEC reserve accumulations would be
invested mainly in dollars. Developments in the oil market and the
electionof the Thatchergovernment,whichpromisedstrongermonetary
control in the United Kingdom,were associated with a sharpupswing
of the pound. The mark generally moved upward against the dollar,
alongwith the Europeancurrenciesincludedin the EuropeanMonetary
System, which was launched in March 1979. The yen was the most
markedlyweak currencyin 1979;a decliningtrendof the currentaccount
became clear, and Japan,once again, was expected to be a majorloser
as a result of higheroil prices.
Sincelate 1979,monetarydevelopments-notably erraticmovements
in U.S. interest rates-seem to have been by far the most important
influenceon exchange rates. In Septemberthe weakness of the dollar
againstthe markreappearedas earlierexpectationsof a recession in the
United States gave way to heightened concern about inflation. On
October6, 1979,the FederalReserve adoptedits new reserve targeting
proceduresdesignedto gaincloser controlover M1.Thedollarrecovered
againstthe markas interestrates rose initially,then sagged as expecta-
tions of furtherincreases in the U.S. interestrate were not fulfilled.
In 1980U.S. interestrates and exchangerates moved in tandem,first
rising throughApril, then falling until late summer, then rising again
throughthe end of the year. In 1981the dollarcontinuedto climbto still
higher levels even though interest rates did not surpass their earlier
peaks andinterestrates in Germanyandmanyothercountriesalso rose.
But by then a surplus in the U.S. current account had emerged that
contrastedwith deficits in GermanyandJapan.DecliningU.S. inflation
was also seen as a positive factor for the dollar. While the shorter-run
movementsof the dollarcontinuedto be relatedto changesin the interest
ratein 1982,the strengthof the trendcould no longerbe explainedsolely
by movements of short-termrates. And the real value of the dollar,
relativeto earlierperiods, appearedto point towardsubstantialdeficits
in the currentaccount down the road. A dropin U.S. interestrates and
the emergenceof a currentaccount deficitbroughtonly a smallreversal
of the cumulativerise in the dollar. Some observersarguethat portfolio
preferencesshifted towardthe dollaras a safe haven, given the earlier
election of a Socialist governmentin France, turmoil in Poland, and
politicaluncertaintyin Germany.
Meanwhilethe strengthof the poundin 1979and 1980has been largely
reversed since the end of 1981. It is easy to point to qualitativefactors
that would account for this-moderately lower interestrates and weak-
ening of the world oil market. But inflationin the United Kingdomhas
finally slowed decisively; growth of the money supply was slower in
1982thanearlier,and the currentaccounthas remainedin surplus.
Althoughthe qualitativebehaviorof exchangemarketsin recentyears
seems consistent with the earlieryears of floating,manywould question
whetherrelationsthat are stable over the entire periodof floatingrates
canbe identifiedandwhetherthese relationsarequantitativelyconsistent
with the theory as it now stands. Without such relations, one cannot
offer general, as opposed to ad hoc, explanationsof the behavior of

exchangerates. In the next section we look for such systematicrelations
and explore some specific questions concerningthe quantitativecon-
sistency of exchange rate behaviorusing the predictionsof the various
models of the asset market.
Empirical Evaluation of Exchange Rate Models
Duringthe past decade the developmentof the theory of exchange

rate determinationhas been heavily influenced by the events of the
floatingrate period. Models have been continuallymodifiedto incorpo-
rate importantnew stylized facts. Yet empirical confirmationof the
evolvingtheorieshas provenmoreelusive. Whenjudgedby the criterion
of out-of-sample fit, the structuralmodels surveyed above fail at all
forecastinghorizonsto outperforma simplerandom-walkmodel.28
Whenjudged fromthe standpointof a weakercriterion-consistency
with historicalexperience-various models have proven successful in
termsof tests of significanceandpredictingdirectionsof effects, although
often for only a particularcurrencyandsubsampleperiod.29This section
28. RichardA. Meese and KennethS. Rogoff comparedthe out-of-samplefit (using
actual values of the exogenous variables)of several popularstructuraland time-series
models of exchangerate determinationfor the dollar-mark,dollar-pound,and dollar-yen
exchangerates. They concludethata random-walkmodel, whichtakes today's exchange
rate as the best predictorfor all futurerates, performsas well as the structuraland time-
series models over the sampleperiodfrom December1976to June 1981.See Meese and
Rogoff,"EmpiricalExchangeRateModelsof the Seventies:Do TheyFitOut-of-Sample?"
Journal of International Economics, vol. 14 (February 1983), pp. 3-24.
29. Early evidence in supportof the flexible-pricemonetarymodels include Bilson,
"The MonetaryApproach,"and Hodrick,"An EmpiricalAnalysis." Frankelin "Onthe
Mark" presented evidence favorable to the sticky-pricemodel. Later, less favorable
results were providedby RudigerDornbusch,"ExchangeRate Economics:WhereDo
We Stand?"BPEA, 1:1980, pp. 143-85;and JeffreyA. Frankel,"On the Mark:Reply,"
AmericanEconomic Review, vol. 71 (December 1981),pp. 1075-82. For the portfolio-
balance approach,WilliamBranson, Hannu Halttunen,and Paul Masson, "Exchange
Rates in the Short-Run:The Dollar-Deutschemark Rate," EuropeanEconomicReview,
vol. 10 (December1977),pp. 303-24, providemildlysupportiveevidence that was later
qualifiedin their "ExchangeRates in the Short-Run:Some FurtherResults," European
EconomicReview, vol. 12 (October 1979),pp. 395-402. Obstfeldestimatesa structural
portfolio-balancemodel of the dollar-markrate and finds little empiricalsupportfor the
model'skey policyimplication-the abilityof sterilizedinterventionto affectthe exchange
rate. See MauriceObstfeld, "ExchangeRates, Inflation,and the SterilizationProblem:
Germany, 1975-1981," European Economic Review, vol. 21 (March-April 1983), pp.
161-89.
focuses on the short-runvolatility of exchange rates and evaluates

modelsof exchangerate determinationon the basis of the less stringent
criterionof in-sampleabilityto trackthe facts.
One of the hallmarksof the periodfollowing the BrettonWoods era
has been the highvolatilityof bothrealandnominalexchangerates. The
evidence in figure1 reveals that, even using quarterlyaveragedata that
smoothover muchof the short-runvolatility, there have been wide and
sustained swings in exchange rates over the 1970s. Table 2 reports
estimates of standard deviations of daily, weekly, four-week, and
twelve-week percentage changes in nominalexchange rates over four
subperiodsof the past decade. Volatilityof the dollar-markand dollar-
yen rates was generally relatively high in the early years of flexible
exchangerates, a consequence of the inevitablelearningperiodand the
severe oil andother commodityprice shocks thataccompaniedit. There
followed a period of relative calm in foreign exchange marketslasting
severalyears, but the past five years have been characterizedby a rising
trend in exchange rate volatility. For the dollar-poundexchange rate,
the second subperiodwas punctuatedby a weakeningof the pound in
1976,which culminatedin an InternationalMonetaryFundstabilization
programfor the UnitedKingdomin January1977.Volatilityrose steadily
for the dollar-poundrate duringthe 1970swith no respitein these middle
years.
The observationthatnominalexchangerateshave been volatile, and,
in some cases, increasingly volatile, does not in itself constitute a
condemnationof the floatingrate regime. In an efficientlyfunctioning
foreign exchange market, the exchange rate will respond immediately
and fully to new informationreceived by marketparticipants.In this
respect, a high degree of exchange rate volatilityis impliedby the asset
price characterizationof the exchange rate that serves as the common
denominatorof all the major theories. In this view, a high degree of
exchangerate volatilitywould be predictedif expectationsof the future
course of variables influencingthe exchange rate were uncertainand
subject to large revisions. Indeed, if all exchange rate changes were
relatedsolely to the adventof new andunanticipatedinformationon the
market, the exchange rate would follow a randomwalk-today's ex-
changerate would be the best predictorof all expected futureexchange
rates. If exchange rate changes were caused by new informationabout
the futurepath of economic variables,superiorto that containedin the
Table 2. Volatility of Exchange Rates, Selected Periods, April 1973
to February 1983a
Percent
Standarddeviationof change in exchange rate
Dollar-mark Dollar-yen Dollar-pound
Nor- Nor- Nor-
Period Actual malized Actual malized Actual malized
April 1973-September1974
(transitionto floating
rates and first oil shock)
Daily 0.86 0.86 0.52 0.52 0.42 0.42
Weekly 1.65 0.74 1.12 0.50 0.98 0.44
Four-weekintervals 4.52 1.01 2.73 0.61 2.19 0.49
Twelve-weekintervals 6.89 0.89 4.26 0.55 3.95 0.51
October 1974-May 1977
(relativecalm)
Daily 0.43 0.43 0.23 0.23 0.49 0.49
Weekly 0.89 0.40 0.56 0.25 1.05 0.47
Twelve-week intervals 3.49 0.45 2.56 0.33 4.49 0.58
June 1977-October5, 1979
(weakness of the dollar)
Daily 0.61 0.61 0.64 0.64 0.57 0.57
Weekly 1.25 0.56 1.50 0.67 1.14 0.51
October6, 1979-February
1983(resurgenceof the
dollar)
Daily 0.76 0.76 0.81 0.81 0.72 0.72
Weekly 1.52 0.68 1.68 0.75 1.41 0.63
Source: Authors' calculations based on New York noon bid rates, Board of Governors of the Federal Reserve
System, international macro data base.
a. Standard deviations of changes in the logarithms of three bilateral exchange rates, calculated for daily and
weekly changes and changes at intervals of four and twelve weeks. The normalized standard deviations are the
actuals divided by the square root of the number of days in the change calculated-that is, by the square roots of 5
for the weekly changes, 20 for changes every four weeks, and 60 for changes every twelve weeks. Use of this
normalization implies that, if the exchange rate exactly followed a random walk, the normalized entries for each
subperiod would be equal in large samples.
current values of those variables themselves, then the exchange rate

would be a good predictorof other variablesbut would itself be poorly
explainedeven by currentvalues of other variables.
In fact, evidence has been found that is consistent with the view that
nominal exchange rates follow approximatelya random walk.30The
resultsin table2 providesome furtherevidence of the nearrandom-walk
patternof exchange rate movements, based on the patternof exchange
ratevariabilityacrossvaryingtimehorizons.If the exchangeratefollows
a randomwalk, the varianceof the five-day,orone-trading-week,change
in the exchange rate will be five times that of the one-day change, and
the standarddeviationof the five-day change will be the squareroot of
five times that of the one-day change. Normalizedstandarddeviations
are reportedin table2; the standarddeviationsof one-week, four-week,
and twelve-weekchangesin thattableare dividedby the squareroots of
five, twenty, and sixty, respectively. For each subperiod the four
normalizedstandarddeviationswouldall be equalin largesamplesif the
exchange rate exactly followed a randomwalk. The normalizedmea-
sures, while not identical,aregenerallyclose in value, so thatthe pattern
of volatilityacross the four time aggregationsfor each subsampleperiod
suggests random-walkbehavior. There is no clear tendency for fre-
quentlyoccurringchangesin exchangerates to eitherreinforceor offset
one another over longer time intervals (positive or negative serial
correlation).
SHORT-RUN EXCHANGE RATE VOLATILITY
An assessment of whether exchange rates have been too volatile

duringthe 1970sshould include three considerations.First, attainment
of greater exchange rate stability could entail sacrificing domestic
monetarypolicy objectives. For example, the decision by the United
30. Interestin exploringwhetherexchange rates follow a randomwalk dates back
more than fifteen years to work by Poole. See WilliamPoole, "SpeculativePrices as
RandomWalks:An Analysisof Ten Time Series of FlexibleExchangeRates," Southern
EconomicJournal,vol. 33(April1967),pp. 468-78. Fora morerecentanalysissee Michael
Mussa, "EmpiricalRegularitiesin the Behaviorof ExchangeRates and Theoriesof the
ForeignExchange Market," in Karl Brunnerand Allan H. Meltzer, eds., Policies for
Employment, Prices, and Exchange Rates, Carnegie-Rochester Conference Series on
Public Policy, vol. 11 (Amsterdam:North Holland, 1979),pp. 9-57. Mussa claims that
recent exchange rate behaviorhas conformedclosely to the random-walkhypothesis.
Econometricevidenceconsistentwith(butnot directlyproving)the random-walkhypoth-
esis is providedby RichardA. Meese and KennethJ. Singleton,"On Unit Roots and the
EmpiricalModelingof ExchangeRates," Journalof Finance, vol. 37 (September1982),
pp. 1029-35.In the forecastingcontext, Meese andRogoffin their "EmpiricalExchange
RateModels"show thata random-walkmodelof exchangeratesgenerallyyields a better
out-of-samplefit thanthe forwardrateor any of the structuralmodels.
States authorities in October 1979 to abandon operating procedures

focusing on the federal funds rate in favor of reserve targetingled to
higher interest rate volatility. Asset market models indicate that ex-
changeratechangeswill be closely relatedto interestratedevelopments.
Thus the change in domestic operatingproceduresof monetarypolicy
may have heightenedexchangerate volatilityin the recent period.
A second considerationis whether nominalexchange rate volatility
has a detrimentaleffect on the realmacroeconomy.It is evidentin figure
1 thatnominalexchangeratevolatilityhas been closely associatedwith,
andperhapsoften the sourceof, realexchangeratevolatility.These real
exchangerate changesmay serve as a conduitfor transmittingthe effect
of nominalexchange rate volatilityto the real economy. For the dollar-
markand dollar-yenrates duringthe past decade the contemporaneous
correlationbetween the nominalandrealexchangerateshas been high-
0.61 and 0.84, respectively. For the dollar-poundrate, the correlationis
lower, 0.30, at least partly because rapid price inflationin the United
Kingdomduringmuch of the past decade has often roughlykept pace
with changesin exchangerates over time. The highcorrelationbetween
nominal and real exchange rates in the cases of the dollar-markand
dollar-yenrates is consistent with the notion that prices are sticky so
thatpurchasingpower paritywill not hold in the shortrun.31
At the microeconomic level it has been argued that nominal and
associated real exchange rate volatility may deter internationaltrade
and investment by greatly complicatingthe profit-maximizingcalcula-
tions of firms. The uncertaintycreated by real exchange rate volatility
could lead to misallocationof resources.Little statisticalevidence exists
to date, however, that exchange rate volatilityhas had negative effects
of this sort.32
31. Sluggish price adjustmentis not, however, a necessary conditionfor a strong

correlation between nominal and real exchange rate movements. For example, the
exchangeratemay respondto realshocksthatrequirea changein the equilibriumrelative
price of nationaloutputs. For an exampleof this type of model, see Alan C. Stockman,
"A Theory of Exchange Rate Determination,"Journal of Political Economy, vol. 88
(August1980),pp. 673-98.
32. See, for example,PeterHooperand S. W. Kohlhagen,"The Effects of Exchange
Rate Uncertaintyon the Prices and Volumeof InternationalTrade,"Journalof Interna-
tional Economics, vol. 8 (November 1978),pp. 483-511. They find little evidence of a
significanteffect of unanticipatedvariabilityin the nominalexchangerate (measuredas
the absolute differencebetween the lagged forwardand currentspot rates) on export
volumes,butidentifysome influenceon exportpricesforthe periodfrom 1965:1to 1975:4.
Therearefew empiricalstudiesfor the morerecentperiod.
The final and probablymost importantconsiderationin assessing if

nominalrates have been overly volatile is whetherexchangerates move
principallyin response to new and unanticipateddevelopments in so-
calledfundamentals- variablessuggestedby theoryas importantdeter-
minantsof the exchangerate- andwhetherthey reactto an appropriate
degree. In other words, does this volatility appear to serve a clear
economic purpose? This issue is not readily resolved because any
assessment of whether exchange rates are too volatile is necessarily
rooted in an underlyingmodel of exchange rate determinationand its
predictionof what a warrantedresponse to changes in the determinants
would be. Therefore, each model of exchange rate determinationmay
implya differentmeasureof excess volatility.Moreover,some important
economic variablesare empiricallyunmeasurable,or at best measurable
with substantial error given existing statistical techniques and data
sources. How can one accurately quantify the effect of a Thatcher,
Reagan,or Mitterrandelection on expected futuremonetarygrowthand
inflation?Orthe impactof the Iranianrevolution,the rise and decline of
OPECwealth, or the internationaldebt crisis on portfoliopreferences?
Clearly the overwhelmingimportanceof expectations in asset market
modelsandthe inherentdifficultyof empiricallymeasuringexpectational
variablesimplythat some significantsources of exchange rate volatility
remainunquantifiable.
EXCHANGE RATE VOLATILITY AND FUNDAMENTALS
It is nonethelessusefulto seejust how muchof exchangeratevolatility

can be related to past and contemporaneouschanges in the variables
suggestedas determinantsof the exchangerateby asset marketmodels.
Of course, if the exchange rate anticipates these variables,as Sims has
suggested for the interest rate, then its movements will appearpoorly
explained.33But it is worth trying to discover how important such
problems seem to be, using an atheoretical methodology. Previous
attempts at structuralmodelingof exchange rates have met with little
success, and this failureis not restrictedto the simpler,single-equation
models. Even large structuralmodels that take explicit account of the
33. ChristopherA. Sims, "Comparisonof Interwarand Postwar Business Cycles:

Monetarism Reconsidered," American Economic Review (May 1980, Papers and Pro-
ceedings, 1979), pp. 250-57.
multicountrygeneralequilibriumnatureof exchangeratedetermination,
such as the multicountrymodel (MCM)developedover the past decade
at the Board of Governorsof the Federal Reserve System, have met
with limitedsuccess in trackingandforecastingshort-runexchangerate
fluctuations.34
For this reason, an alternativeempiricalstrategyis adopted below.
No a priori assumptions are imposed about speeds of adjustment.
Althoughthe econometricapproachemployedis nonstructuralinnature,
it is possible (under assumptions equivalent to those used in single-
equationmodels)to drawsome qualitativeconclusionsaboutthe ability
of fundamentalsto explainshort-runexchangerate volatility.
In the reduced-form model developed below the change in the
logarithm of the bilateral (dollar-foreigncurrency) exchange rate is
related to changes in the logarithmof relative (U.S. to foreign)money
supplies, MI; changes in the logarithmof relative output (industrial
production);changes in the logarithmof relativeprices (consumerprice
index); changes in the short-term(three-month)nominal interest rate
differential;andthe level of the two countries'currentaccounts(or trade
balances,when monthlycurrentaccountdataareunavailable).35 All raw
data are seasonally unadjustedand seasonal dummiesare incorporated
in estimation.
The exchange rate equation is derived and estimated as part of a
vector autoregressive(VAR) system of equationsin which each of the
seven variables-the exchangerateandits six predicteddeterminants-
is regressedon its own laggedvalues andon laggedvalues of all the other
34. A summaryof the variousapproachesto exchangedeterminationattemptedwithin
the contextof the MCMis describedin PeterHooperandothers, "AlternativeApproaches
to General EquilibriumModelingof Exchange Rates and Capital Flows: The MCM
Experience," InternationalFinance Discussion Paper213 (Boardof Governorsof the
FederalReserveSystem, 1982).
35. Muchof the dataused in the followingsimulationswas providedby RichardMeese
and KennethRogoff. The exchangerates, interestrates, and money suppliesare aligned
at end-of-monthdates. This alignmentof the datahelps to capturesome of the announce-
ment effects of money and interestrateson exchangerates. Becausejoint stationarityof
the regressorsis a key assumptionunderlyingthe existenceof a VARrepresentation,most
variablesappearas firstdifferencesof logarithms.See ThomasJ. Sargent,Macroeconomic
Theory(AcademicPress, 1979),fora discussionof the conditionsrequiredfortheexistence
of VAR and MA representationsof a vector time series. Trade balances and current
accountsappearin level formbecausethe portfolio-balancemodelpredictsan association
between the change in the exchange rate and the change in private sector holdingsof
foreignassets.
variables.36Lag lengthwas selected using the likelihoodratiocriterion.
Then, to focus attentionon the importanceof unanticipatedmovements
in the determinants,the estimatedVAR system is expressed in termsof
its associated Woldor movingaverage(MA)representation.In the MA
system, the change in the exchange rate is expressed as a linear
combinationof currentandpastunanticipatedmovements(disturbances
or innovations)in the explanatoryvariables,and disturbancesfurtherin
the past aregiven progressivelyless weight.An unanticipatedmovement
in a variable in this context is defined as the forecast error from the
VAR equationfor that variable.37
The disturbance terms in the moving average representationare
typically contemporaneouslycorrelated, which makes it difficult to
provide a meaningfuldefinition of a disturbanceto any single VAR
equation.As is standardin theVARmethodology,theMArepresentation
is transformedbefore the simulationsare performedin orderto obtain
uncorrelateddisturbances.Since we wish to place an upper bound on
the component of exchange rate movements that can be explained by
innovationsin other variables, the particulartransformationemployed
assumes that all contemporaneouscorrelationbetween the exchange
rateand its determinantsreflectscausationfromthe latterto the former.
This assumptionis also commonlyused in the single-equationexchange
rate models that aboundin the literature,which assume that the regres-
sors are econometricallyexogenous.38
To impose this assumptionduringthe simulationsthat are used to
decompose the varianceof the exchange rate change, several steps are
required.First,beforesimulationthe systemofestimatedVARequations
is orderedso that the exchange rate equationappearslast. The ordering
selected for the six remainingvariablesis (fromtop to bottom):relative
output, relative prices, relative money, the U.S. trade balance, the
36. Significancetests for the variablesin the VARexchangerateequationarereported
in the appendix. They indicate the importanceonly of lagged values of the variables
becausecontemporaneouseffects are containedin the disturbanceterms.These contem-
poraneouseffects areincorporatedin the simulationresultsreportedin the text.
37. These disturbancesincludethe possible effects of omittedvariablesthat operate
directlyon the exchangerateand not throughthe otherregressors,includingthe political
or confidencefactorsthatare often saidto influenceportfoliopreferences.
38. The exogeneity specificationof popularexchangerate models has been critically
evaluatedby Glaessner.See ThomasGlaessner,"A Test of the ExogeneitySpecifications
of Modelsof ExchangeRateDetermination"(Boardof Governorsof the FederalReserve
System, 1979).
42 BrookingsPapers on Economic Activity, 1:1983
other country's currentaccount (or tradebalancefor the United King-
dom), andthe nominalinterestratedifferential.39 The covariancematrix
of the estimatedresidualsis then used to transformthe system so that,
duringsimulations,innovationsin variablesplacedhigherin the ordering
of equationsare allowed to influencecontemporaneouslythose below
them, but not vice versa. This transformationis tantamountto incorpo-
ratingcontemporaneousvalues of the six determinantsalong with their
laggedvalues in the exchange rate equationbefore simulation.40
Selection of an orderingfor the equationsof a VAR model is simply
the econometricidentificationprocedurein anotherguise. Choosingan
orderingis equivalent to imposing enough identifying restrictions to
achieve a recursivesystem in the classical econometricframework.It is
not surprising,then, that the orderinghas importantimplicationsfor the
allocationof explanatorypoweramongthe individualdeterminants,just
as the selection of exclusion restrictionsinfluencessimulationresults in
a classicaleconometricframework.Theprimaryconclusionshighlighted
below, however, are not a productof the orderingselected for the six
determinants.
VAR simulationsare generallyused to decomposethe varianceof the
forecast error of the dependent variableat differentforecast horizons
into the percentagesof the varianceattributableto each of the explana-
tory variables.41 Decompositions of the forecast error variance of the
change in exchange rates are provided in figure4 for the dollar-mark,
dollar-yen,anddollar-poundrates. The percentagethatis attributableto
disturbancesto each of the explanatoryvariablesis plottedfor forecast-
ing horizons of one to thirty-sixmonths. The part of the forecast error
variancethat is attributedto own disturbancesis denoted by the line s.
39. Thefollowingconsiderationshaveguidedthechoiceof ordering.Becauseexchange
ratesand interestrates react stronglyto contemporaneousevents, they are placedat the
end of the ordering.The placementof the priceandoutputvariablesbeforethe monetary
variablereflects the desire to allow for a reactionfunctionof the monetaryauthorities.
Placementof tradeandcurrentaccountsnearthe end of the orderingis consistentwiththe
view thatthey have littleeffect on relativepricesandoutputin the currentperiodbut that
they maybe somewhatsensitiveto contemporaneousdisturbancesin these variables.
40. More precisely, the covariancematrixof estimatedresidualsis lower triangular-
ized. See the discussionof VAR simulationmethodologyin RobertJ. GordonandStephen
R. King, "The Output Cost of Disinflationin Traditionaland Vector Autoregressive
Models," BPEA, 1:1982, especially pp. 207-15.
41. For other examplesof this type of simulation-basedvariancedecomposition,see
ChristopherSims, "MacroeconomicsandReality,"Econometrica,vol. 48 (January1980),
pp. 1-48;andStanleyFischer,"RelativeShocks,RelativePriceVariability,andInflation,"
BPEA, 2:1981, pp. 381-43 1.
Figure 4. Decompositionof the Variance in the Forecast Error
for the Change in the Exchange Ratea
Percentof the variancein theforecast error
100 _
90 _
80
70 -
Dollar-markexchangerate
60 -
50 -
40 -
S
30
20 -
_* _ - __ _
/ S~~~~~~~~~~PR
10 / -m
Tw
lo
1 6 12 18 24 30 36
(0.37) (0.66) (0.82) (0.89) (0.93) (0.94) (0.95)
Monthsahead (V*in parentheses)
100,
90
80 -
70
Dollar-yenexchangerate
60 -
50
40
30
20
tCi
10 P-R - - =- -"
, I ,
O_
l 6 12 18 24 30 36
(0.41) (0.70) (0.86) (0.92) (0.97) (1.0) (1.0)
Monthsahead (V* in parentheses)
Source: Authors' calculations based on data from the Board of Governors of the Federal Reserve System.
a. The decomposition is explained by orthogonalized disturbances to the explanatory variables. The variables are
s, percentage change in the dollar-foreign currency exchange rate; Ar, the change in the short-run interest differential;
na, the percentage change in the relative money supply; y, the percentage change in relative output; PR, the percentage
100 _
90
80 -
70-
Dollar-poundexchangerate
60
50 -
40 -
30 - 5
20 -
_A~~~~~~~~~~~~~~~~~~C
/_ ,<r~_____ _ _PR
10 Ar
m
1 6 12 18 24 30 36
(0.31) (0.53) (0.88) (1.0) (1.0) (1.0) (1.0)
Monthsahead (V* in parentheses)
change in relative prices; Cg, the German current account; Ps, the U.S. trade balance; Cj, the Japanese current
account; and 7Tk, the U.K. trade balance. The V* is the ratio of the variance in the forecast error to the variance of
the dependent variable. When V* is close to 1.0, the variance in the forecast error approximately equals the variance
of the dependent variable.
An intuitiveinterpretationcan be given to these figures:they tell what
types of (orthogonalized)disturbanceshave caused the change in ex-
change rates to deviate from its predictedvalue at various forecasting
horizons. The in-samplesimulationevidence in figure4 reinforces the
out-of-samplefitresultsreportedby Meese andRogoffthatfundamentals
help very little in forecastingexchangerates over shorthorizons.42This
is evident fromthe highpercentageof the forecast errorvariancethat is
not explained by disturbancesto the fundamentals(denoted by s) at
horizons of less than six months. At longer horizons, disturbancesto
fundamentalsaccountformoreof thedeviationof the changein exchange
ratesfrom its predictedlevel.
A more importantissue is whether the high volatility of exchange
rates during the past decade can be related to movements in the
theoreticallysuggested determinants.To addressthis issue, it is neces-
sary to obtain a decomposition of the variance of the exchange rate
change itself, ratherthan of its forecast error.An approximationto this
variancecan also be obtainedfromthe above simulations.The one-step-
ahead forecast errorbased on the MA system is simply the contempo-
raneousdisturbanceappearingin the MA representation;the two-step-
aheadforecast errorincorporatescontemporaneousand first-lagvalues
of disturbances;andso on.43As we use the MArepresentationto forecast
further into the future, each successive forecast error incorporates
progressively more terms of the MA representation.In this way, the
42. Meese andRogoff, "EmpiricalExchangeRate Models."

43. To see this, consider the following MA equation from a two-variablesystem
consistingof variablesx,, andx2,:
Xi,= u + allu,-1 + al2u,2 + **
+ a2lv11 + a22Vt-2 +
whereu andv arethe disturbancetermsfromthe firstand secondequations,respectively.

The one-step-aheadforecasterroris then
xl,,+ I - E,xl,,+ I =u,+ 1-
From the assumption of stationarity, var(u,+I) = var(u,) = (r2. Thus the one-step-ahead
forecast-errorvarianceinvolves the varianceof the contemporaneousterm of the MA
representation.Similarly,the two-step-aheadforecasterroris
Xl,t+2 - E,xl,t+2 = Ut+2 + aliu,+1 + a2lv,+l,
so that its variance is calculated from terms involving contemporaneousand lagged

disturbancesfromthe two equations.
forecast errorat very long horizonsyields a good approximationof the
entireMA representationof the dependentvariableitself.44
In figure4 the V*value reportedin parenthesesbeneatheach forecast
horizonis the ratio of the forecast errorvarianceat that horizon to the
total varianceof the change in exchange rate. At forecast horizons for
which V*is close to 1.0, the forecasterrorvarianceclosely approximates
the varianceof the exchange rate change itself. A horizon of thirty-six
monthsproved sufficientin all cases to obtaina value of V*close to 1.0.
Variance decompositions are reported in table 3 for percentage
changesin the dollar-mark,dollar-yen,anddollar-poundexchangerates
fromMarch1973to August 1982.The last row of the table indicatesthe
part of the variance of the change in exchange rates that cannot be
attributedto (orthogonalized)innovationsin the six explanatoryvari-
ables and is ascribed to own disturbances.This part of the variance
amountsto 30 to 40 percentand is invariantwith respect to the ordering
selected for the six other equationsplaced before the exchange rate in
the simulations.45
A commoncriticismof the VAR methodologyis thatit only provides
evidenceon the importanceof reduced-formdisturbanceterms(forecast
44. Because MA coefficients furtherin the past become progressivelysmallerin a

stable system, ignoringthe distant past of the theoreticallyinfinite, moving-average
representationmay not entailmuchloss of information.
45. Given that the exchange rate is placed last in the orderingfor simulations,the
orderingof the remainingsix equations has no effect on the simple dichotomyof the
varianceof the exchangeratechangeinto the partattributableto VARdisturbancesin the
six fundamentalstakentogether,andthe partexplainedby own disturbances.To see this,
consider a trivariateVAR system of equationsfor variablesx, y, and z that has been
orderedfor simulationsand whose estimatedresidualsare (fromfirstto last) ux,,UV,and
uzi. These disturbancesare serially uncorrelated,i.i.d. disturbanceterms that may,
however, be correlatedwith each other. When the covariancematrixof the estimated
residualsis orthogonalizedbefore simulation,it is equivalentto a redefinitionof the
residualsusingthe followinghypotheticalregressions:
ux, = ex,)
U= d1ux, + ev1,
U= d2ux1+ d3uYI + ezI.
Theresultingresiduals-ex,, ey,,andezi-are uncorrelatedwitheach otherby construction.
The redefinedresidualfor the last equation,ez,,is clearlynot influencedby the orderingof
the two otherequations.Hence whenthe MA equationfor z is used in simulationsto split
the explanatorypowerbetweencontemporaneousandlaggedvaluesof ez,on the one hand,
andof ex,and ev,on the other, the divisionis invariantwith respect to the orderingof the
two top equations.
Table 3. Variance Decompositionof the Change in Exchange Rates,
March 1973 to August 1982a
Percentage of varianceaccountedfor
Orthogonalizedinnovation rate rate rate
Relative output 9.1 8.6 19.4
Relative prices 13.5 9.5 10.7
Relative money 11.2 5.5 6.3
U.S. trade balance 9.1 9.0 8.4
Othercountry'scurrentaccountb 4.9 11.5 15.4
Short-terminterest differential 13.8 22.1 9.7
Unexplained(spot rate) 38.3 33.9 30.1
Sources: All U.S. data except for the trade balance are from the Board of Governors of the Federal Reserve
System. Exchange rates are noon New York spot bid market rates, aligned at end-of-month dates with the interest
rates and money supplies. Output variables are industrial production indexes from OECD, Main Ecorionoic Indicators.
Consumer price indexes for Germany, Japan, and the United Kingdom are from the Monthly Report of the Deutsche
Bundesbank, table VIII-7; Bank of Japan, Economic Statistics Monithly, table 119(1); and Department of Employment,
Emiployment Gazette, table 6.4, respectively. The Ml series for the same countries are from the Monithly Report of
the Deutsche Bundesbank, table 1-2; the Bank of Japan, Ecorionoic Statistics Monthly, table 4; and the Bank of
England Quarterly Bulletin, table 1. Data on trade balances are from OECD, Main Econonmic Indicators, for the
United Kingdom; and from U.S. Bureau of the Census, Highlights of U.S. Export anid Iniport Trade, tables E-l and
1-1 through 1978, and Sunimary of U.S. Export and Iniport Merchandise Trade through 1982 for the United States.
Current account data for Germany and Japan are from OECD, Main Ecorionoic Indicators. Interest rates for Germany,
Japan, and the United Kingdom are, respectively, the three-month interbank rate from the Frankfurter Allegenmeine
Zeitunig; the "over two-month ends" bill discount rate (Tokyo Stock Exchange) from the Federal Reserve Board
data base; and the three-month deposit rate for the local authority from the London Financial Times. The three-
month Treasury bill rate is from the Board of Governors of the Federal Reserve System.
a. The entries in the table are the percentages of the variance of the exchange rate change that can be attributed
to orthogonalized VAR disturbances to each of the explanatory variables shown. The last row reports the percent
of the variance not ascribable to disturbances to any of the six explanatory variables, which is then residually
attributed to exchange rate disturbances. The variables are presented above in the order in which they were placed
for simulation (see the text for a discussion of the importance of the ordering for simulation).
b. Trade balance for the United Kingdom.
errorsfrom the VAR equations),while the interestingeconomic issues

relate to the importanceof structuraldisturbanceterms such as money
demandor aggregatedemandshocks. The VAR modelcan be viewed as
the reduced-formrepresentationof one or morestructuralexchangerate
models, so that each VAR disturbancetermis a linearcombinationof a
numberof structuraldisturbanceterms.46Then, unless the underlying
structural model is identified, it is not possible to unscramble the
structuraldisturbancesfromthe estimatedVAR disturbances.
Although results on the importance of VAR innovations cannot
generally be translatedinto evidence on the importanceof structural
46. JohnTaylordiscusses this pointin his commentson a paperby StanleyFischerin

BPEA, 2:1981, pp. 434-38.
disturbances, some qualitativeconclusions concerningthe latter issue
may be inferredif additionalidentifyingassumptionsare made. If, as in
single-equationexchange rate models, it is assumedthat the regressors
in the exchange rate equation of the structuralmodel underlyingthe
VAR model are predetermined,that structuralmodel must be block
recursive. Then the exchange rate will not affect contemporaneously
any of the other variables in the associated VAR system, and the
disturbancetermin the VARexchangerateequationwill be independent
of any of the other disturbanceterms. In this case, the structuraland
reduced-form interpretationsof the VAR exchange-rate innovation
coincide: the innovationrepresents disturbancesto the exchange rate
equationthat are not due to structuraldisturbancesto any of the other
variablesin the model. Under these assumptions,the results in table 3
may be interpretedas evidence that 60 to 70 percent of exchange rate
volatility can be ascribed to (structuralor VAR) disturbancesto the
theoreticallysuggestedexplanatoryvariables.
Further inferences from the VAR evidence require even stronger
identifyingassumptions. As noted earlier, in the presence of contem-
poraneouscorrelationamongthe estimatedVAR disturbanceterms, the
allocation of explanatorypower among the individualdisturbancesis
influencedby the ordering, so this decompositionmust be interpreted
with more caution. In particular,the allocation of the varianceamong
reduced-forminnovationsusing VAR-based simulationswill only cor-
respond to the more meaningfulallocation of the variance among the
structuraldisturbancesif the system is fullyrecursive(andnotjust block
recursive). These results are summarizedbriefly here. For the dollar-
markrate, disturbancesto relative prices, the interest differential,and
relative money account for about 40 percentof the varianceof changes
in exchange rates. This evidence suggests the relative importanceof
asset marketvariablesand price developments. In additionto relative
price and interestratedisturbances,disturbancesin the currentaccount
are importantin explainingthe forecast errorvarianceof the dollar-yen
rate, despite having been placed below the other determinantsin the
orderingof equations. Disturbancesin the U.S. trade balance are also
an importantsource of volatility of the dollar-poundrate. The possible
channels through which the current account or trade balance could
influencethe exchangerate are discussed below.
RANDOM DATA EXPERIMENTS
From the above evidence it appearsthat a substantialproportionof

the volatility of exchange rates can be related to unanticipatedmove-
ments in the theoreticallypredicteddeterminants.One drawbackof the
VAR methodology, however, is that it requires estimation of a large
numberof coefficients relative to the small numberof monthlyobser-
vations since the adventof floating.Togetherwith the likely collinearity
among the many regressors, this implies that few of the coefficients
are very precisely estimated.
To illuminatethe empiricalimportanceof this problem,an experiment
was conductedfor comparison.In the experimentthe rate of change of
the exchange rate was regressed on its own past values and on past
values of six series that were createdusinga randomnumbergenerator.
The same numberof variables, lags, and seasonal dummies as in the
VAR simulationswere includedin the experimentsto closely mirrorthe
conditionsfor VAR estimationusing actualdata.
Table 4 compares the percentage of the variance of the change in
exchange rates explained by orthogonalized innovations in the six
explanatoryvariablesin the earlierVAR simulationswith thatexplained
by innovationsin the six randomlygenerateddata series in the experi-
ments. Based on evidence from five repetitions of the random data
experiments,it appearsthat, on average, the model based on the actual
data explains the varianceof the exchange rate change only about 8 to
16percentbetterthan the models with purelyrandomlygenerateddata.
Thusone mustconcludethatmuchof the short-runvolatilityof exchange
ratesover the past decade remainsunexplainedby the variablesempha-
sized in the models of the 1970s.These resultsreinforcethose fromout-
of-sampletests andindicatethatone can neitherexplainnorpredictwith
any confidence a substantialportionof month-to-monthexchange rate
volatility using even a very unrestrictedrepresentationof the asset
marketmodels.
MODEL RESTRICTIONS
While the VAR technique provides a way to assess whether the

variables suggested by the asset marketmodels can explain exchange
rates, it does not provide tests of the restrictions imposed by these
Table 4. Variance Decompositionof the Change in Exchange Rates, Actual Data
versus RandomlyGeneratedData, March 1973 to August 1982a
Percentage of varianceaccountedfor
Item rate rate rate
Simulationsbased on actual data 61.7 66.1 69.9
Averagefor randomdata experiments 53.4 52.9 53.9
Rangefor the experiments 48.4-59.5 49.7-58.1 43.5-67.5
Improvement gainedby usingactualrather
than randomdatab 8.3 13.2 16.0
Sources: Same as table 3.
a. Entries show the percentage of the variance of the exchange rate change explained jointly by orthogonalized
innovations in the six explanatory variables (not including own innovations) for two data sets: actual data-the six
variables suggested by the theoretical models (the same variables as in table 3); and random data-six randomly
generated series.
b. Difference between the first two rows.
models. These restrictionsare at the heart of the policy debate. They

determinewhat influencepolicies can have on the exchange rate, and
they are an importantpartof the cases for and againstresistingchanges
in exchangerates.
For example, the purchasingpowerparityassumptionof the flexible-
price monetarymodel leads to the conclusionsthat financialpolicies do
not affect the realexchangerateand thatthereis no need for authorities
to be concerned about the exchange rate anyway. Authoritiescan and
should focus their attention on pursuinga monetarypolicy consistent
withdomesticpricestability.Themodel,whichassumesa stabledemand
for money and rationalexpectations, suggests that the rightway to do
this is to follow a policy of constant money growth equal to the trend
rateof growthof realeconomic activity. If the demandfor money moves
erratically,the problemof findingthe optimalmonetarypolicy is more
complicated. But the essential insight of the flexible-price monetary
model from an internationalperspective is unaffected. The best policy
for each countryandfor the worldas a whole is for authoritiesto stabilize
prices in theirown economies as best they can.
In the sticky-price monetary model, the uncovered interest parity
condition, which follows from the assumptionsof perfect asset substi-
tutability and rational expectations, is the central relation yielding
internationalpolicy implications. Recall that uncovered interest rate
parity implies that the nonminal interest rate differentialexactly equals
the expected percentage change in exchange rates. If interest-bearing
assets are perfectsubstitutes,authoritiescannotexpect to influencethe
exchange rate by intervening in the foreign exchange market if the
interventionis sterilized-that is, if the stock of money is not allowed to
change. Sales of interest-bearingassets denominatedin one currency
for nonmoney assets denominatedin anotherhave no effect on asset-
market equilibrium.Moreover, intervention that is not sterilized is
equivalent to a domestic open-marketoperationbecause it makes no
differencewhetheran increasein money is accomplishedby purchasing
assets denominatedin domestic or foreign currency. Policies pursued
abroadalso influencemacroeconomicconditionsat home throughtheir
effects on the real exchangerateandinterestrate. So economies are not
insulated, and there are potentialadvantagesto coordinatingdomestic
stabilizationpolicies amongcountries.47
If the uncovered interest parity condition holds, unstable currency
preference can be ruled out as an importantinfluence on exchange
rates, and this reason for interventionto stabilize exchange rates can
be rejected. Shifts in preferencesby some portfolioholderswould then
be absorbedby other portfolioholders with no net effect on nominalor
real exchange rates. Similarly,wealth redistributionswould not affect
exchange rates. The market would be self-stabilizingwith respect to
disturbances of this type. At the same time, the case for allowing
exchange rates to influencedomestic macroeconomicpolicies would be
strengthened.The exchange rate mightthen provide a useful indicator
of economic conditionsandprospectsat homerelativeto those inforeign
economies.
In the portfolio-balancemodel, uncoveredinterestrateparitywill not
hold exactly. Imperfectasset substitutabilityimplies that interest-bear-
ing assets denominatedin different currencies are perceived to have
differentriskcharacteristics.Theexistence of differentialrisk-reflected
in risk premiums or discounts-allows scope for sterilized exchange
marketinterventionto influenceexchange rates, althoughits impact is
less powerful dollar-for-dollarthan that of unsterilized intervention.
47. The modelas presentedincorporatesa role only for monetarypolicies. But in an
elaboratedmodel, whichincorporatesthe influenceof fiscalpolicy on aggregatedemand,
varyingthe mix of monetaryandfiscalpolicy leads to differentrealinterestratesandreal
exchangeratesfor a given level of aggregatedemand.See the analysiswithinthe context
of a sticky-wagemonetaryframeworkby AndreBurgstaller,"Flexible ExchangeRates,
RationalExpectations,andthe Trade-offbetweenInflationandUnemployment,"Discus-
sion Paper85 (ColumbiaUniversity, 1980).
This model also incorporates additional determinantsof changes in
exchange rates, includingthe wealth redistributionassociated with a
currentaccount imbalance.The exchangerateresponsetends to reduce
the currentaccountimbalancesandfosterexternalbalance.Butif current
accounts are most susceptible to transitorydisturbancesand respond
slowly to changes in exchangerates, this adjustmentmechanismwould
contributelittle to dampingcurrentaccountimbalances.
Tests of uncoveredparityare tests of thejoint hypothesis that assets
denominatedin domestic and foreign currencyare perfect substitutes
(the absence of a risk premium)and that expectationsare rational.It is
importantfor policy purposes to determine which part of the joint
hypothesis accounts for the failureof uncovered parity to hold. If, on
the one hand,the failureindicatesthe existence of a riskfactorconsistent
with the portfolio-balancemodel, it implies a channel throughwhich
sterilized interventioncould affect the exchange rate. The inability of
researchersto relate the observed deviationsfrom uncoveredparityto
the variablesthat the portfolio-balancemodel suggests makes it more
difficultto interpretthe results as evidence supportingthe existence of
a risk premium.48Even if the failure of uncovered parity could be
interpretedin this way, the question would still remainwhether inter-
vention could improvethe allocationof risk and resourcesachieved by
the free market.
If, on the other hand, the failure of the tests is a reflection of
expectations that do not fully and immediatelyincorporatenew devel-
opmentsor overreact(or underreact)to them, it would still be necessary
to understandhowexpectationsbehaveandhowthey mightbe influenced
by official actions in order to determine whether intervention could
improve the situation. Also, from the basic principlethat it is best for
policies aimedat eliminatingmarketimperfectionsto act directlyat their
source, informationabout the source of any inefficiencyis essential to
determiningif interventionis in fact the optimal corrective strategy.
Thus if there is a case for interventionbased on the uncovered interest
paritytests, it is predicatedon knowledgeof the source of the rejection
of the condition.Therehas been little researchalong these lines.
We have already called attentionto the failure of purchasingpower
48. See JeffreyA. Frankel,"InSearchofthe ExchangeRiskPremium:A Six-Currency
Test Assuming Mean-VarianceOptimization,"Jouirnal of International Money and
Finance, vol. 1 (December1982),pp. 255-74.
54 BrookingsPapers on Economic Activity, 1:1983
parityto providean accurateshort-runcharacterizationof the behavior
of exchangeratesandrelativeprices. Fluctuationsin realexchangerates
duringthe past decade have been large and cumulative. Evidence has
also been cited from the literature indicating statistically significant
deviationsfrom the uncoveredinterestrate paritycondition. Failureof
this condition to hold is commonly interpretedas evidence supporting
the portfolio-balancemodelandcontradictingthe sticky-pricemonetary
model. But tests of uncovered paritydo not shed light on how well the
central insight of the sticky-price monetary model-that real interest
ratedifferentialshave an importantinfluenceon the realexchangerate-
accounts for the broadsweep of realexchangerate movementsover the
floatingrate period. If this mechanismleaves much unexplainedover
the past decade, it suggestsa role for the additionalfactors incorporated
in the portfolio-balancemodel to explain sustained and systematic
deviations from uncovered interest parity. To look at this question,
evidence is providedbelow on how well real interest rate differentials
explainthe broadmovementsof realexchange rates.
UNCOVERED INTEREST PARITY IN THE LONG TERM
It was noted earlier that the uncovered interest parity condition

implies that the expected percentagechange in the real exchange rate
will equal the real interest rate differentialobserved across countries.
Thus the uncoveredinterestrate parityconditionis expressed as
= r* + Se
where, as before, r is the nominalinterestrate;s is the change (over the

same horizon as the interestrate)in the logarithmof the exchange rate;
e denotes an expected value conditionalon informationavailableat time
t; and * denotes a foreignvariable.Let Q denote the real exchange rate.
By definition,
lnQe = ln Se + lnPe* - lnPe
where, as before, P is the price level. This relationholds for any horizon
of expectations.
From these relations, the following association can be derived be-
tween the realinterestratedifferentialacross countriesandthe expected
percentagechange in the real exchange rate (where again a lowercase
level variableindicatesa percentagechange):
q= (r - pe) - (r* - pe*).
If equilibriumis expected to be restoredby some date T periods in the

future, the deviation of the logarithmof the real exchange rate today
fromits equilibriumvalue, ln Q, mustequalthe compoundedreal return
differentialfrom now to date T (or any period, N, furtherin the future
than T). In logarithmicform,49
N- I
ln (Qt/Q) = - E [(rt+k -
Pe+k) -
(r*+k - P*k)] for (N - T),
k=o
where rt+k is the expected one-period nominalinterest rate for period

t + k. If the expectationstheoryof the termstructureholds, the average
expected nominalinterest rate over N years can be proxied by the N-
periodbond rate. The deviationfrom the equilibriumexchange rate can
be writtenas N times the averagereal interestdifferentialover N years,
ln (Qt/Q) = -N[(r, - t) - (r -
where r, is the average nominalinterest rate on an N-period bond with

no coupon, andp, is the averageexpected inflationrateover N periods.
This relationis examinedbelow using interestrates on U.S. bonds of
ten years' maturityand foreign bonds of about the same maturity. It
seems unlikely that portfolio managersproject current monetary dis-
equilibriumsfurtherinto the futurethanthis, so thatone can reasonably
assume that equilibriumis expected to be restored within ten years.
Predictedvalues for the real exchange rate are generatedfrom nominal
interest rates and measures of expected inflation rates under the as-
sumptionsthat the long-runequilibriumrealexchangerate, Q, has been
constantand that the uncoveredinterestparityconditionhas held over
the past decade. A comparisonof the actualand predictedvalues of the
real exchangerate providesan indicationof how well the exchange rate
mechanismin the sticky-pricemonetarymodel accounts for the broad
swings in real exchange rates over the floatingrate period.
49. This equationuses the approximationx i (1 + x) for a valueof x of the orderof

In
magnitudeof interestratesandinflationrates.
We tested two proxiesfor expected inflation.First, a centeredtwelve-
monthmovingaverageof actualinflation-that is, the inflationoccurring
at about the time expectations were formed-provides a "myopic"
measure of long-term inflationaryexpectations. This measure works
well if investors simply projectcurrentinflationexperience far into the
future. In contrast, an inflationforecast based on rationalexpectations
weights all relevant available informationon the basis of its historical
power to predictfutureinflation.The second proxy is a rationalinflation
expectationsseries providedby the VAR modeldescribedabove. A ten-
year average inflation series is constructed as the simple average of
inflationforecasts providedby the VAR system for the subsequent 120
months.S0
Table 5 shows the regressionresults and the percentageof variation
in the real exchange rate accountedfor by variationin real interestrate
differentials, R2, using each of the expected inflation proxies. Real
interest differentials calculated using the myopic expected inflation
proxy explain a substantialpart of the variationin the dollar-markand
dollar-yenreal exchange rates, but the rationalexpectations proxy has
virtuallyno explanatorypower. Moreover,bothproxies have the wrong
signforthe dollar-poundrealexchangerate,andthe rationalexpectations
proxy apparentlyexplainsa substantialshareof the variationin the real
exchange rate.
The empiricalresults clearly do not stronglysupportan explanation
for changes in exchange rates that depends on uncovered interest rate
paritywith rationalexpectations. The substantialexplanatorypower of
half of the bivariate regressions reported does, however, confirm an
importantconnection between interestrates and exchange rates even if
that connectionis very differentfor the poundthanfor the markand the
yen.
Figure5 shows how movementsof realexchangeratescorrespondto
movementsof real interestrate differentialsbased on the myopic proxy
for inflationexpectations. The real exchange rates are shown together
withthe regressionpredictionsanda prioripredictionsthatarecalculated
using a value of ln Q equal to the sample mean of ln Q and a coefficient
of 10 on the real interestrate differential,correspondingto the approxi-
50. An alternativeapproachwould be to estimatethe long-runexpected real interest

rate differentialusing the observed time-seriesprocess for short-runreal differentials.
Table 5. The Relation between the Long-TermReal Interest Rate and the Real
Exchange Rate, August 1973 to March 1982a
Myopic inflationexpectations Rational inflationexpectations
Real Real
interest interest
Dependent rate rate
variable Constant differential R2 Constant differential R2
Dollar-markrate - 1.49 - 2.74 0.42 - 1.46 - 1.20 0.04
(0.013) (0.321) (0.022) (0.563)
Dollar-yenrate - 6.04 - 2.19 0.51 - 6.04 0.72b 0.01
(0.009) (0.212) (0.019) (0.74)
Dollar-poundrate 0.66 0.35b 0.01 0.64 4.93b 0.59
(0.017) (0.421) (0.008) (0.397)
Source: Authors' least-squares regression, In Q = a + b [(r, - Inp,) - (r, - Inp, )]. Results are reported for
real interest rate differentials constructed using the myopic and rational expected inflation proxies described in the
text.
a. Standard errors are reported in parentheses.
b. The regression coefficients are of the wrong sign.
mate ten-year maturity of the interest rates.5' The figure shows no

correspondencebetween real exchange rates and real interestdifferen-
tials for 1973-74, the period of initial adjustmentto floatingexchange
rates and the firstoil shock. But after 1975,and particularlysince 1979,
the actual and predicted series for the dollar-markand dollar-yenreal
exchangeratescorrespondmoreclosely. Thea prioripredictionssuggest
that the very large real depreciationsof the markand yen against the
dollar in recent years are roughly commensuratewith movements of
interest rates and inflation rates if one is prepared to entertain the
hypothesis of myopicinflationexpectations.
The question remains,why do the rationalexpectationsproxies give
poor results? These proxies for inflationover ten years are relatively
insensitiveto currentinflationor to disturbancesin the othervariables-
the proxies for the inflationdifferentialrange over an interval of less
51. The estimatedcoefficientsreportedin table 5 are considerablysmallerthan the

a priorivalue of 10, but considerabledeviationsof the estimatedcoefficientsfrom their
a priorivalues shouldnot be surprisinggiven: (1) coupon paymentson bondsfor which
marketinterestratesare availablethatgive moreweightto near-terminterestrates in the
yield to maturitycalculationof the geometricalaverageinterestrate over the life of the
bond;(2) tax considerations,whichare ignored;(3) measurementerrorin the differential
for the realinterestrate-especially in the expectedinflationcomponent;and(4) the long-
runequilibriumrealexchangerate,Q, whichmay not be constant.
Figure 5. Long-RunUncoveredInterest Rate Parity under Myopic
InflationExpectations
in (dollarsper mark)a
-0.5
1975 1980
-6.0 = r '
=_0
. ~~Real exchangerate'
In(dollarsperyen)a
-2.5
in ( pA prioriprediction
In(dollarsperyeun)a
1975 1980
-5.5
,~~ ~ ~~~~ predictionetlA pror
~~~~
-5.0 ",^I' ,8^s^,'
.'l~~~~~~Ra
intenatona maRoedaarbse. nprdcto
6.51975 1980
a.0 . 1
Souce Athos predcroo
cacuatongarescribnpedincthen
internationalresio excase.at tx ,
bae ondtafomFde, Real
ReslervehankgofaNe Yok
a. Bt urniswr eltdb teClidxi hc 96=10
than 4 percent. This flatness may be largelya consequence of the way

the proxieswere calculated-forecasts for monthlyinflationdifferentials
tend to converge to their mean values in the firstfew years so that the
ten-yearaverageforecasts remainneartheiraveragevalues. Ideallyone
would preferto have long-runinflationforecasts that give more weight
to longer-rundevelopments, but with a sample period of less than ten
years, thereare few observationson lower-frequencyphenomena.Still,
the rationalforecasts seem justified on two grounds. First, diagnostic
tests did not suggest that longer lags should be included. Second, one
could reasonablyexpect average inflationdifferentialsover a periodas
long as ten years to correspondmore closely to averageinflationover a
comparablylong past periodthanto currentinflation.Nevertheless, the
evidence suggeststhat the marketsfor the yen and markmay have been
heavilyinfluencedby near-terminflationexperience.52
Uncovered interest paritydoes not seem to offer an explanationfor
movementsin the dollar-poundexchangerate.A numberof explanations
come immediatelyto mind, includingthe phasingout of the reserve role
of the poundfor the firsthalf of the period,capitalcontrolsin the United
Kingdombefore 1979, and the effect of oil prices and North Sea oil
developmentson marketassessments of the competitivepositionof the
United Kingdom's economy. But a priori one would have had little
reason to expect these factors to be more importantfor the pound than
similarfactors would be for the other currencies.
The strongperverserelationfor the dollar-poundexchangeratebased
on the rationalinflationexpectations proxy also calls for explanation.
As for the other currencies, the time series for this proxy is essentially
flat. Thus the perverse relation is primarilyone between the nominal
interestrate differentialand the exchange rate. This may well reflect a
greater tendency of the authorities in the United Kingdom to move
interest rates in response to pressures in the exchange market. Such
behaviorwas most evident in 1976and 1977as authoritiesin the United
52. Steve Golub, "InternationalFinancialMarkets,Oil Prices, andExchangeRates"
(Ph.D.dissertation,Yale University,1983),hasalso exploredthe relationbetweeninterest
rates and inflationby takingcumulativeex post realizedreal inflationdifferentialsas a
proxyfor rationalexpectations.He finds, as we do with our rationalexpectationsproxy,
thatpersistentdifferentialscannotexplainexchangeratesvery well. See also PeterIsard,
"An Accounting Frameworkand Some Issues for Modelling How Exchange Rates
Respond to the News," in Jacob A. Frenkel, ed., Exchange Rates and International
Macroeconomics(Universityof ChicagoPress, forthcoming).
Kingdomraised interest rates sharplywhen the pound weakened and
moved them down againwhen it recovered.
This analysis of interest rates, inflation, and real exchange rates
provides some evidence supportingthe sticky-pricemonetarymodel's
predictedrelationbetween the real interestdifferentialacross countries
and changes in the real exchange rate. But the results are not robust,
andthe relationfits best undera myopicview of expectationsthat seems
inconsistentwithrationality.Thislatterresultsuggestsa needfor further
study to determinewhetherthe apparentexpectationalanomaliesdo in
fact exist. If they do, are they specific to exchange marketsor can they
be related to behavior in other asset markets, particularlyto the term
structureof interestrates and capitalinvestmentmarkets?
THE ROLE OF THE CURRENT ACCOUNT AND TRADE BALANCE
While the VAR evidence (including the results in the appendix)

provides some supportfor the relative importanceof the trade balance
or current account in explainingfluctuationsin some exchange rates,
that effect has not been constrained in estimation to pass through a
portfolio-balancechannel. There are other hypotheses indicatingthat
currentaccounts shouldinfluencethe pathof the exchangerate. It seems
unlikely that the portfolio-balancechannel can explain the strengthof
the often-notedassociationbetween currentaccount deficitsand depre-
ciation or currentaccount surplusesand appreciation.It is implausible,
for example, that the portfolio-balancemechanismwas the cause of the
pronounceddollardepreciationaccompanyingthe U. S. currentaccount
deficitpatternin 1977and 1978.The effect of the $28 billioncumulative
U. S. currentaccountdeficiton net U. S. privatesectorholdingsof foreign
assets was swamped by concurrentinterventionof more than double
thatamount."3Moreover,in manyyears, the effect of governmentdeficit
financingoperationson wealth and relative domestic and foreign asset
stocksfaroutweighsthatof currentaccountimbalances.Currentaccount
imbalancescould also have an impacton goods marketequilibrium.But
the wealth effect of government deficits again can more than offset
wealth transfersinduced by the currentaccount, so that any potential
53. See Peter Hooper and John Morton, "Fluctuationsin the Dollar: A Model of
Nominaland Real ExchangeRate Determination,"Journal of International Money and
Finance, vol. 1 (April1982),pp. 39-56.
effects of currentaccounts on goods marketsmighteasily be offset by
other forms of asset accumulation.S4
This leads one to suspect that forces other thanthe portfoliobalance
mechanismmay be at work. One possibility that has been explored in
recent work is that unanticipateddevelopmentsin the currentaccount
or trade balance provide new informationabout a country's evolving
competitivepositionandhence aboutthe exchangeratelevel consistent
with long-run current account balance."5Thus the current account
may signal requiredlong-runexchange rate adjustmentsand so have a
significant short-run influence on the exchange rate. The practical
limitationof this version of the currentaccount-exchangerate nexus is
that it provideslittle guidanceon the difficultempiricalquestionof how
best to model the expected long-runequilibriumexchange rate.
This signalinghypothesis, favored by proponents of the monetary
model, provides no direct channel throughwhich the currentaccount
can affect the exchange rate and thus suggests that it is only current
account innovationsthat affect the exchangerate.56Some findingshave
indicatedthe importanceof currentaccountnews, althoughthere is also
evidence indicating that anticipated changes in the current account
matter.S7
54. The impact of wealth transfersinduced by the currentaccount on the goods
markethasbeen exploredby RudigerDornbuschandStanleyFischerin "ExchangeRates
and the CurrentAccount," AmericanEconomicReview, vol. 70 (December1980),pp.
960-71; and GuillermoA. Calvo and CarlosAlfredoRodriguez,"A Modelof Exchange
RateDeterminationunderCurrencySubstitutionandRationalExpectations,"Journalof
Political Economy, vol. 85 (June 1977), pp. 617-25.
55. See Hooper and Morton, "A Model"; Isard, "An AccountingFramework";or
Michael Mussa, "The Role of the CurrentAccount in Exchange Rate Dynamics,"
DiscussionPaper(Universityof Chicago,June 1980).
56. Unless wealth is incorporatedin the money demandfunction. See Pentti J. K.
Kouri,"TheExchangeRateandthe Balanceof Paymentsin the ShortRunandin the Long
Run: A Monetary Approach," Scandinavian Journal of Economics, vol. 78, no. 2 (1976),
pp.280-304;orJeffreyA. Frankel,"TheMysteryof the MultiplyingMarks:A Modification
of the Monetary Model," Review of Economics and Statistics, vol. 64 (August 1982), pp.
5 15-19.
57. Fora theoreticalmodelexploringthe importanceof the currentaccountsee Mussa,
"The Role of the CurrentAccount." FrenkelandDornbuschprovideempiricalevidence
on the role of news in exchange rate determinationin Jacob A. Frenkel, "Flexible
Exchange Rates, Prices and the Role of 'News': Lessons from the 1970s," Journal of
PoliticalEconomy,vol. 89 (August1981),pp. 665-705;and Dornbusch,"ExchangeRate
Economics." Dornbuschmodeledcurrentaccount news as the differencebetween the
actual value and the value predictedby the OECD. Evidence is providedin Bonnie E.
Loopesko, "The Role of CurrentAccountImbalancesin ExchangeRate Determination:
ARE EXCHANGE RATE RESPONSES NONLINEAR?
A number of observers of exchange markets have suggested that

exchange rates overreact to some new informationrelative to what
wouldbe expected underrationalexpectations.Overreactioncouldtake
manyforms. Testingfor it dependson havinga modelthatplaces bounds
on the appropriateresponse. Here we simply examined whether large
changesin fundamentalsare associated with morethanproportionately
large changes in exchange rates, behavior that would be inconsistent
with linearmodels.
To see whether the marketresponds more stronglyto large changes
in other economic indicators,the changein the log of the exchange rate
was regressedon the same variablesas in the earlierVAR system, and
on squaredvalues of the firsttwo lags of the determinants.The signs of
the originalobservationswere preserved, so that these termsrepresent
disproportionatelylargeor smallchangesin the same direction(positive
or negative). If the true model were linear in all the determinants,the
findingthat the squared terms are significantcould indicate a market
overreactionto recent information.
The results in table 6 indicate that there is some evidence of a
disproportionateresponse by the marketto large recent developments
in the determinantsof the exchange rate. There is no evidence of
nonlinear response for the dollar-poundrate, but some significant
nonlinearresponse terms are found for the dollar-markand dollar-yen
rates. There is no well-definedpattern to the results. All that can be
concludedis that these two exchange rates react more stronglyto large
changesin some of theirdeterminantsthanwouldbe predictedby simple
linearmodels.
At best, these results suggest an avenue for furtherresearch. More
generally,furtherempiricalresearchis requiredinto othertheoretically
suggestedsources of marketinefficiencysuch as bubbles, bandwagons,
and extraneousbeliefs about the determinantsof exchange rates."8It is
CompetingHypothesesandEmpiricalEvidence,"ResearchPaper8236(FederalReserve
Bankof New York,December1982),indicatingthatit is notjustnews, butalso anticipated
currentaccounts,thataffectthe exchangerate.
58. Fora discussionof these phenomenainthecontextof rationalexpectationsmodels,
see OliverJean Blanchard,"SpeculativeBubbles,Crashesand RationalExpectations,"
Economics Letters, vol. 3, no. 4 (1979), pp. 387-89; Oliver Jean Blanchardand Mark
Watson, "Bubbles,RationalExpectationsand FinancialMarkets"(HarvardUniversity,
January1982);andRobertP. Flood and PeterM. Garber,"MarketFundamentalsversus
Table 6. NonlinearityTests, March 1973 to August 1982a
Significancelevel of squaredvariables
Variable rate rate rate
Relativeoutputs 8.7* 90.0 14.1
Relativeprices 62.2 8.5* 44.2
U.S. tradebalance 56.8 64.5 70.0
Othercountry'scurrent
accountb 60.2 79.2 19.1
Short-terminterestdifferential 76.0 5.1* 70.0
Exchangerate 0.6** 6.6* 91.7
All nonlinearityterms taken
together 15.7 18.4 48.4
Source: Authors' calculations based on data from sources in table 3.
* Marginal significance of at least 10 percent.
** Marginal significance of at least 5 percent.
a. Marginal significance levels are reported above for F-tests of the joint significance of two lags of sqluared values
of the explanatory variables, with original signs of the observations preserved. The equation also includes six lagged
values of each explanatory variable (unsquared), so that the squared values are rough indicators of the nonlinear
response of the exchange rate to movements in its determinants.
b. Trade balance for the United Kingdom.
not only important to determine whether the exchange rate moves too
much in relation to economic variables but also why it may overreact, in
order to determine whether intervention provides the most direct means
to eliminate the problem. Attempts to address the latter issue will most
likely have to focus on the microeconomic decision process in exchange
markets.
Exchange Rate Behavior and Policy
What can be said about exchange rate policy when no existing model
is strongly supported by the data? Should exchange rates be managed
Price Level Bubbles:The First Tests," Journalof Political Economy, vol. 88 (August
1980),pp. 745-70. ObstfeldandRogoffshow thateven minimalgovernmentinterventions
can preventspeculativepricebubblesin MauriceObstfeldand KennethRogoff, "Specu-
lative Hyperinflationsin MaximizingModels: Can We Rule Them Out?" Journal of
Political Economy, vol. 91 (August 1983),pp. 675-87. McKinnonand also Dornbusch
relatethe discussionto the foreignexchangemarket.See RonaldI. McKinnon,"Floating
ForeignExchangeRates 1973-74:The Emperor'sNew Clothes," in Karl Brunnerand
Allan H. Meltzer, eds., Institutional Arrangements and the Inflation Problem, Carnegie-
RochesterConferenceSeries on PublicPolicy, vol. 3 (Amsterdam:NorthHolland,1976),
pp. 79-114;andRudigerDornbusch,"EquilibriumandDisequilibriumExchangeRates,"
Zeitschriftffur Wirtschafts- und Sozialwissenschaften, vol. 102, no. 6 (1982), pp. 573-99.
moreheavilyor even fixedonce again?And, if so, how shouldthis policy
be carriedout? Policymakersmust choose exchange rate and macro-
economic policies despite the absence of clear answers to theoretical
and empiricalquestions. What is more, there is little reason to expect
thatthe fog of uncertaintywill be dissipatedsoon.
The very fact of uncertainty-about economic behavior and about
futureshocks-needs to be kept at the centerof policy discussions. With
considerable uncertainty about key structuralrelations, policies that
avoid disastrous consequences under a broad range of models are
preferableto policies that are optimalfor a strict interpretationof one
model but would serve very badly for other plausiblemodels. In addi-
tion, flexibilityof policies is desirablewhen the world economy seems
vulnerableto largeshocks, both real and monetary.
One positive conclusion seems well established after ten years of
experiencewith floatingexchangerates:countriesarenot insulatedfrom
disturbancesor policies in othercountriesby a strongpurchasingpower
parity relation between exchange rates and relative prices. At a mini-
mum,the interdependenceof macroeconomicdevelopmentsindicatesa
need to take external developments into account when developing
strategiesfor domestic macroeconomicpolicy.
To carry the policy discussion furtherthan these generalconsidera-
tions, one must go well beyond what can be establishedaboutexchange
ratedeterminationon the basis of strongstatisticalevidence concerning
aggregativerelations.In this section moreconcreteviews are offeredon
policy issues that reflecta plausible,but not strictlyempiricallyproven,
interpretationof the evidence. These views also reflectjudgmentson a
numberof issues not addressedin the paper.Threeshouldbe madeclear
at the outset. First, it is takenfor grantedthatcumulativefluctuationsof
the real exchange rate on the orderof 10or 20 percentthatare sustained
for six months or more have importanteffects on tradeflows, domestic
economic activity, and inflationaryprocesses, even in an economy as
largeand relatively self-containedas that of the United States. It is less
clear, however, that exchange rate volatility over shorter periods has
large real effects. Second, responsibilityand accountabilityfor macro-
economic performanceareviewed as unlikelyto be shiftedfromnational
governments to some internationalbody. Policy proposals have little
chance of practicalimplementationif they requireauthoritiesto subor-
dinate national goals to internationalones. Third, it is doubtful that
controls on capital flows or the use of trade restrictions and trade

subsidies would be desirable or even effective in managingexchange
rates. We draw a sharpdistinctionbetween these policies, which seek
to supplantor frustratemarketprocesses, and centralbankintervention
in the foreignexchangemarket,whichseeks to influencemarketsthrough
the managementof the government'sown assets and liabilitiessuch as
foreigncurrencyreserves, high-poweredmoney, andnonmonetarydebt.
The majorpolicyissues canbe groupedintofourcategories:exchange
rates and monetary policy, the monetary and fiscal policy mix with
floatingexchange rates, internationalcoordinationof macroeconomic
policies, and sterilized intervention in the exchange market. After
discussingeach of these, we concludewith our views on the desirability
of returningto fixedexchangerates.
EXCHANGE RATES AND MONETARY POLICY
To the extent thatexchangeratesreflectrealinterestrates, they could

be a useful guide to monetarypolicy when money demandis unstable
and inflation expectations cannot be observed with much precision.
Using the exchange rate in this way would involve, in effect, inverting
the argumentunderlyingour predictionsof exchangeratesfromnominal
interest rates and expected inflation rates. Some inferences about
inflation expectations relative to interest rates could be made from
exchangeratemovements, althougha rigidresponseof monetarypolicy
to exchange rate developments would presume a tighter relationthan
seems to exist. The evidence on the relationbetween real interestrates
and exchange rates from our crude experiments is mixed. Monetary
factors do not seem to provide the entire story of exchange rates. And,
even if they did, allowance would have to be madefor developmentsin
other countries.
A second reason for using the exchange rate as an indicator of
monetarypolicy is that the deviationsof real exchange rates from their
long-runvalues, owingto disturbancesin a worldwith sticky prices, will
affect export- and import-competingsectors and hence aggregatede-
mand and inflationwith a lag. Thus a given monetarygrowth rate or
interest rate may be more or less restrictivedependingon whether the
currencyis strongor weak.
These considerationssuggest a supplementalandjudgmentalrole for
the exchangeratein monetarypolicy ratherthana centralor mechanical
role. But its potentialvalue as an indicatorwarrantscloser study.
MONETARY AND FISCAL POLICY MIX UNDER FLOATING

EXCHANGE RATES
Thegreaterfreedomof monetarypolicy underflexibleexchangerates

can easily contributeto a neglect of the stabilizationrole of fiscalpolicy.
A balanceof monetaryandfiscal policy is important,however, to avoid
cumulative distortions in trade and investment. There is a risk that
budgetimbalanceswill go uncheckedlong enoughto alterinternational
competitive positions and to affect capital formation. The greatest
difficultyresults from persistent budget deficits. The inflationarypres-
sures created by the fiscal impetuswill need to be containedby greater
monetaryrestraintand higher real interest rates and the currencywill
appreciate. Higher real interest rates will reduce domestic capital in-
vestment, while a stronger currency will reduce the current account
balance and generate a correspondingshift toward an inflow of funds
fromabroadon the capitalaccountof the balanceof payments.Because
of the real interestrate sensitivity of domestic saving and investment,a
capital inflow smaller than the governmentdeficit will balance supply
and demand for funds in domestic markets. Thus net capital inflows
cannotpreventdomesticcapitalformationfrombeingcrowdedout even
if internationalcapital flows are perfectly elastic with respect to the
expected rate of return. Moreover, any net flow of funds from abroad
entails an erosion of the competitive positions of exporters and those
competingwith imports.
INTERNATIONAL POLICY COORDINATION
The large fluctuationsin real exchange rates over the floating rate
period, interpretedwithin a sticky-pricemonetaryor portfolio-balance
model, suggest strongly that economies are not insulated from what
happens abroad. Consequently, even governmentsthat pursue purely
national objectives should seek close consultation and exchange of
informationon economic developments. Such consultationswould be
essential, for example, to the informeduse of the exchange rate as an
indicatorfor monetarypolicy.
In principle, macroeconomic interdependenceamong a relatively
smallnumberof largecountriesmeansthatcooperativepoliciesinvolving
internationalquidpro quo or the acceptanceof internationalconstraints
on policies should be superiorto unilateralnational policies, even if
these are informedby internationalconsultationon economic develop-
mentsandintentions.Butit is notobviousthatthe majormacroeconomic
problemsof the past ten years could have been solved simplyby more
cooperative policies. Large shocks in the world economy and the
breakdownof economic relations that had guided policy in the 1960s
overwhelmedpolicymakers.Ad hoc cooperativepolicypackages,which
are seen by all partiesto offer near-termadvantages,can be put together
within the existing internationalconsultative framework of summit
meetingsand many lower-level meetings. Still, examples of agreement
on substantivepolicy trade-offsare rare, presumablybecause cases in
whichall partiesexpect net benefitsare unusual.
Adding more structureto policy coordinationseems politically im-
practical and of questionable economic advantage. The uncertainty
surroundingthe effects of policies and the risk of shocks makethe long-
term benefits of a systematic exercise of cooperationdoubtful. Under
these conditions, governmentscan hardlybe expected to accept even
short-termconstraintson policies.
STERILIZED EXCHANGE MARKET INTERVENTION
Exchangemarketinterventionthat is divorcedfrom domestic mone-

tarypolicy has been advocatedby manyas a policy instrument.Indeed,
considerableintervention, much of it sterilized, has been undertaken
duringthe floatingrateperiod.The effects of sterilizedinterventionhave
not been explicitly analyzed in this paper, but the analysis presented
bearsindirectlyon this issue.59
59. We have not exploredhere directlythe questionof the effects of foreignexchange

marketinterventionusing publicly availabledata on official foreign exchange market
transactions.A multinationalstudy of the effects of intervention,in which the authors
participated,has recently been completed using, in part, data unavailableto outside
researchers.The resultsof this studyindicatethatthe effect of sterilizedinterventionwill
be short-lived;sterilizedinterventioncannotreversethe exchangerateeffectsof divergent
policies across countries;and the effect of coordinatedinterventionby several central
bankswill be greaterthanthe same volume of interventionby a singlecentralbank.See
"TheReportof the WorkingGroupon ExchangeMarketIntervention"fromthe Working
Groupon ExchangeMarketIntervention,a groupestablishedat the VersaillesSummitof
the Headsof State andGovernmentin March1983.
The currentstate of understandingof exchangeratebehavior(or lack
thereof) argues for a flexible and exploratoryinterventionpolicy. The
evidence on deviations from covered interest parity points to the
possibility that sterilized interventioncould have an influence on ex-
change rates, although competing explanations for these deviations
would suggest different channels through which intervention would
affect exchange rates. The strong case for not interveningis based on
the view that changes in the exchange rate reflectcurrentand rationally
forecastedfutureeconomic developmentsand policies. We find a large
component of exchange rate volatility that cannot be related to the
variables included in the VAR model. The possibility that exchange
rates are excessively volatile because of unstableportfoliodemandsfor
assets denominatedin differentcurrencies,or because of volatilemarket
psychology and bandwagons, must be taken seriously. The issue be-
comes a question of whether authoritiesare wise enough to sort out
exchangeratechangesattributableto these largelyunobservablefactors
from more fundamentaldevelopments, which may also be difficultto
observe except with a long lag. An additionalcase can be made that
sterilizedinterventionshouldalso be used to moderatedeviationsof real
exchange rates from long-runequilibriumvalues caused by disparities
in monetaryconditionsacross countries. But such a policy would carry
greatrisk of treatingone symptomwhile leavingthe underlyingproblem
unresolved. It would seem better to attack such problems at their
source-that is, by giving some weight to exchange rates in monetary
andfiscal policy.
Balancingthe possible benefits and risks of sterilized intervention,
we advocate a cautious approach.Wherethere is strongcorroborating
evidence that shifts in portfolio demandare occurringrelatively inde-
pendently of expectations, their effects on exchange rates might be
neutralized.One example would be an announcedshift of asset prefer-
ences motivated by political considerations,althoughsuch events are
likely to be rare.
Interventionthat has as its objective reducingintradayand day-to-
day fluctuationsin exchange marketsalso warrantsserious considera-
tion. Its justificationis not that short-runvolatility is very costly, but
thatby reducingshort-runvolatility, the longer-rununexplainedswings
in ratesmightbe moderated.Mostactiveparticipantsinforeignexchange
markets say that when rates exhibit great short-runvolatility, longer-
runconsiderationshave littleor no weightin formulatingtheirstrategies.
Highvolatilityengendersmoremyopicbehaviorin marketparticipants,
perhapsbecause it causes longer-termexpectations to be more loosely
held. With longer-termvolatility roughly proportionalto day-to-day
volatility, the possibilitymust be taken seriouslythat rates fluctuatefor
littleorno economicreasonfarenoughandlongenoughto haveimportant
macroeconomic, microeconomic, and random income-redistributive
effects. Officialactionto reduceshort-runvolatilitycouldthencontribute
to more stable long-runbehavior and a more stable world economy.
Such an interventionpolicy would not involve a cumulativebuildupof
largeofficialforeigncurrencypositions, norwouldit attemptto maintain
exchangerates at unsustainablelevels. Thus the risks would be small.
Fixed Exchange Rates for the Major Industrial Countries?
Should the exchange rates of major industrialcountries be fixed?

Implicitlya negative answer has already been given to this question.
Maintainingfixed exchange rates necessitates a level of macroeconomic
policy coordinationthat was never achieved on a sustainedbasis, even
underthe Bretton Woods system. National authoritiescould not place
the maintenanceof the system above nationaleconomic goals. Only if
politicalmechanismsevolved that providedfor supernationalresponsi-
bility and accountabilityfor the conduct of macroeconomicpolicy and
for internationaltransferpayments to balance nationalinterests would
a truly fixed exchange rate system be likely to endure. Once par value
changesbecome an accepted way of relievingpressuresthatbuildup in
a fixed-rate system, currencies are susceptible to speculative attack.
This was evident in the finalyears of BrettonWoods and more recently
within the European Monetary System. Domestic policies are then
deflectedfrom course in efforts to maintainthe system, with little if any
long-rungain in exchange rate stability.A flexibleexchangerate regime
does not prevent authorities from giving weight to an exchange rate
objective in the conduct of policy. It only dispenses with a rigid
commitmentto a particularrate, which has never been more than a
contingentone for sovereigngovernments.Theepisodes of mostextreme
fluctuationsof exchange rates duringthe floatingrate period occurred
when exchange rate and other externalconsiderationshad virtuallyno
weight in monetaryand fiscal policy, even so far as their effects on the
domesticeconomy were concerned. Moreover,policies to reduce vola-
tility through coordinated intervention have not been pursued with
enoughcontinuityto develop maturestrategies.Before exchangingone
extremepolicy for another,it makes sense to explore the middleground
more thoroughly.
APPENDIX
Results of VAREstimation
TESTS OF SIGNIFICANCEarereportedbelow for the VAR exchange rate

equationunderlyingthe simulationsdiscussed in the text.
Table A-1. Determinantsof Changes in Exchange Rates,

March 1973 to August 1982a
Significance level
Lagged variable rate rate rate
Relative output 77.4 55.6 42.2
Relative prices 17.7 80.1 40.7
U.S. trade balance 3.7* 62.3 30.6
Other country's current
accountb 0.3* 44.7 0.2*
Short-terminterestdifferential 4.8* 28.8 36.0
Own lags 0.7* 78.6 67.6
Statistic
RI2 0.62 0.59 0.69
CorrectedR2 0.25 0.07 0.17
Q (30)c 23.4 18.9 21.0
Source: Authors'calculations,basedon datacited in table 3.
* Marginalsignificanceof at least 5 percent.
a. Marginalsignificancelevels of the F-statisticsfor nullhypothesisthat, conditionalon lags of the othervariables
beingincludedin the equation,coefficientson laggedvaluesof the variableconsideredare not jointly significant(a
low marginalsignificancelevel impliesthe variableis a highlysignificantdeterminantof the exchangerate change).
Using a likelihoodratio test, lag lengthsof 6, 7, and 8 were selected for the dollar-mark,dollar-yen,and dollar-
poundexchangerates, respectively.
b. Tradebalancefor the UnitedKingdom.
c. The Q(k)is the Box-PierceQ statistic,whichindicateswhetherthe firstk autocorrelations of the residualsare
significantlydifferentfrom zero. Marginalsignificancelevels for Q(30)are 80 percentfor the dollar-markrate, 89
percentfor the dollar-pound rate, and 94 percentfor the dollar-yenrate. A highmarginalsignificancelevel indicates
that thereis little evidenceof serialcorrelation.It is the probabilitythat the Q statisticis at least as largeas shown
if the firstk residualsare, in fact, not autocorrelated.
Comments
and Discussion
Ralph C. Bryant: This paper contains many perceptive observations

aboutthe variabilityof exchangeratesanddoes a goodjob of summariz-
ing the recent literature. It is thus a welcome addition to the papers
presented at earlier meetings of the BrookingsPanel that focus on the
internationalaspects of macroeconomicbehaviorand policy.
The first half of the paper is a skillfulexposition of the evolution of
theoreticalanalysis and actual experience. The authors strain a bit to
force the chronology of the theory to fit the chronology of actual
developments. (It is misleading,for example, to portraythe portfolio-
balance strand of the literatureas developing after the flex-price and
sticky-pricemonetarymodels and as a response to awkwardfacts that
were not analyzed satisfactorily in those models.) As the authors
themselves point out, the assumptionsof purchasingpower parityand
uncovered interest rate parity have been shown not to hold in actual
practice. By now, therefore,it is time to retirethe flex-priceand sticky-
pricemonetarymodelsandgive themless timeon the stage. These minor
criticisms notwithstanding,the first half of the paper is insightfuland
shouldbe useful backgroundfor a varietyof readersof thisjournal.
The second half of the paper is the more important.In that part, in
which the authorsuse VAR analysis to try to "explain" exchange-rate
movements, I am skeptical of their empirical generalizations. The
discussionof policy issues at the end of the paperis sound,butsomewhat
too agnostic. My commentsthus focus on these two areas. I summarize
my doubts about the authors' VAR analysis and then offer, less cau-
tiously than the paper, some conclusions about policy attitudes to
exchange-ratevariability.
71
How does an analyst "explain" the movement of an economic
variable?At a proximatelevel, explanationrelies on causal premises
embedded in a single behavioral relation of a structuralmodel. For
example, changes in bank borrowingat the Federal Reserve discount
windoware typicallyexplainedby changesin the federalfunds rate, the
discountrate, and a scale variablefor the aggregatesize of bankbalance
sheets. For a deeperexplanation,however, one cannotrestrictattention
to any single behavioral relation. In the case of discount-window
borrowing,for example, one wants to know why the federalfunds rate
and the size of bank balance sheets behaved as they did, not merely
thatborrowingis proximatelydependenton those variables.One is thus
frequentlydrivento consideran entirestructuralmodel, or at least major
blocks of an entire model, in which manyvariablesare treatedas jointly
endogenous. To explainat thatlevel, one must derivethe reduced-form
and final-formversions of the structuralmodel.
Where do exchange rates appearin structuralmodels? If a model is
specified accordingto today's best-practicemacroeconomictheory for
open economies, exchange rates (or a weighted-averageexchange rate
servingas a proxyfor all the bilateralrates)appearas argumentsin many
behavioralfunctions-in equationsfor asset demandandliabilitysupply
that serve as components of variablesfor expected returns, as a com-
ponent of many price variables, and hence in demand and supply
functionsforgoods. Exchangeratesarealso presentinincome-statement
and balance-sheet identities. Generally speaking, the exchange rate
appearsthroughoutthe structuralequations as a variableon the right-
hand side. The exchange rate itself, however, is not a left-handvariable
in any structuralequation. The exchange rate is like goods prices and
interest rates-quintessentially endogenousin the system of structural
equations as a whole but not proximately"determined"in any single
equationor small subset of equations.
In particular,it is wrong to characterizethe exchange rate as just a
component of goods prices, or just an asset price (even though it has
manyof the attributesof an asset price). The exchangerateis not merely
the relativeprice of home andforeigngoods. It is not merelythe relative
price of home and foreign monies. It is not merely the relativeprice of
home andforeignsecurities. It is all these, and more.
To "explain" exchange rates in a meaningfulway, therefore,there is
in principle no choice but to obtain the reduced-formor final-form
equationsof some structuralmodel.'
Each behavioralequation in a structuralmodel typically includes a
stochastic error term to allow for unexpected disturbances in the
behaviorapproximatedby thatequation.Such disturbances("shocks")
have a clear conceptual interpretation.In a structuralequation for
discount-windowborrowing,for example, one can interpretthe error
term as an unexpected shift in bank demand;with clarity, one can say
sucha shock "originates"in the marketfor immediatelyavailablefunds.
On the other hand, interpretationof the errorterms in the reduced-
form equations of a model is not conceptually straightforward,as the
authors recognize. In the typical case, each reduced-formerror is a
complex combinationof many structuralerrors. Only if one knows the
structureof the model is it possible to unscramblethe reduced-form
errorsinto their complex, structuralcomponents. And of course even
when the structuralerrors are contemporaneouslyuncorrelated, the
reduced-formerrors will be correlated. In the typical case, it is not
possible to speak of a composite, reduced-formerrortermas originating
in some one sector or marketin the model.
It is true that if the structuralmodel happens to be recursive in a
convenientway, one can moreeasily derivethe reduced-formequations
of the model and find that some of the reduced-formerrors are a less
complex composite of the underlyingstructuralerrors. For example, if
the matrix of structuralcoefficients associated with the contempora-
neous values of the endogenous variables can be written in a lower
triangularform, the reduced-formerrors will have a correspondingly
recursive structure.For structuralmodels of open economies in which
the exchange rate appears throughoutthe behavioralequations as an
argument, however, this convenient recursiveness property will not
exist. The theory alone is sufficientto discountthat possibility.
Supposean analysthadavailablea plausiblestructuralmodelandhad
derived its reduced-formand final-formequations. He would then be
1. It is instructiveto askwhicheconomicvariablesin a structuralmodelarenot capable

of significantlyinfluencingthe exchangerate (that is, given best-practicetheory, do not
appear, contemporaneouslyor as lagged values, in a reduced-formequation for the
exchange rate). As a matterof theory, the list of such variablesis quite short, perhaps
even nonexistent.
ableto offera straightforward explanationof exchange-ratechanges. Ex
post he could identify unexpected disturbancesin the structural equa-
tions and interpretthem unambiguously.He could study the relative
frequenciesandintensitiesof those shocks. Andhe couldtherebyidentify
the "causes" of exchange-ratechanges and unambiguouslyparcel out
the varianceof the exchange rate amongthem.
With these points as background,now consider the VAR systems
estimatedand used by Shaferand Loopesko. Is it possible to use such
systems to assess whetherthe variablesfeaturedin asset marketmodels
can "explain" changes in the exchange rate? Can calculationssuch as
those in figure 4 and in table 3 of the paper indicate which types of
unexpected disturbancescause the exchange rate to deviate from its
predictedvalue? Is it valid to make inferences about the proportionof
the variance of the exchange rate attributableto disturbancesoriginat-
ing in the foreign exchange marketitself ratherthan being transmitted
from other sectors of the economy? I believe the answer to all these
questions is negative.
If a nonstructuralVAR system is correctly to subsume a class of
structuralmodels containing the true model, at least two necessary
conditionsmust be met. The VAR system must includeall the variables
appearingin the true model (with as many lagged values of all the
variables as appear in the true model). And the true structuralmodel
must contain a recursive patternin the contemporaneousinteractions
among its endogenous variables. The VAR systems studied by the
authorsare quite unlikelyto satisfy eitherof those conditions.
First, the authors'VAR systems omit a numberof variableslikely to
be important.For example, they contain no variablegenuinely repre-
sentingmonetarypolicy. "Money," as conventionallydefined,does not
qualify. Even the theory in the simplifiedmodels discussed in the first
half of the paper, at least when presentedcarefully, makes it clear that
the relevant money is central-bank,high-poweredmoney. The VAR
systems have no variablesrepresentingfiscal policy. Perhapsthe most
puzzling omission of all is a variable representing the quantity of
interventionin the exchangemarket,the stock of internationalreserves,
or some other type of asset stock. Even the simplestand least adequate
of the portfolio-balancetheoriesassertsthatwealth,at homeandabroad,
belongs in the VAR systems. But the authors include only changes in
wealth comingfromimbalancesin the currentaccount.
If many variablesare includedin a VAR system, the problemswith
degrees of freedomtend quicklyto become unmanageable.But that is a
difficultywith the technique, not a valid analyticalreasonfor excluding
the additionalvariables.
Another difficultywith the Shafer-LoopeskoVAR systems is their
treatment of all home and foreign variables as relative values. The
practice of using variablesthat are ratios, or differences, of home and
foreign variablesis typical of the literature;the authorshave plenty of
company. Nonetheless, a chief argumentused to justify VAR analysis
is its avoidance of arbitraryrestrictions.It is difficultto imaginea more
arbitraryassumption,which is what this practicecomes down to, than
presumingthat the behaviorparametersin the nome and foreign struc-
turalequationsare of the same magnitude.
Still another weakness of the VAR systems in this paper is their
ruthlessly bilateralapproach.In the mark-dollarsystem, for example,
only Germanand U.S. variablesare included.Nothing that happensin
the rest of Europe or in the rest of the world can influencethe mark-
dollar exchange rate except by sneakinginto one or more of the VAR
residuals.
The second necessary conditionfor a VAR system to be valid, that
the contemporaneousinteractionsamong endogenousvariablesfollow
a recursivepattern, seems also quite unlikelyto hold. As noted above,
carefully specified models of an open economy with the exchange rate
as an endogenousvariableare even less likely thanmodelsof a domestic
economy to satisfy this condition. Certainly for quarterlydata, and
probablyfor monthly data, most theories suggest that a contempora-
neous, two-way correlationwill exist between many pairsof variables.
For example, exchange rates influenceinterest rates and interest rates
influence exchange rates. Goods prices influence exchange rates, but
exchange rates probablyalso influencegoods prices.
Shafer and Loopesko acknowledge that the disturbancesto which
they refer in figure4 and in table 3 are essentially reduced-formerrors,
not structuralerrors.They note that the reduced-formerrors,or "VAR
innovations," cannotbe givena structuralinterpretationunless a pattern
of block or full recursivityexists in the underlyingstructuralmodel. But
these conditions are quite implausible, as noted above. Yet it is a
structuralinterpretation,and only a structuralinterpretation,that is of
analyticalinterest.
To the authors'credit, they reporta set of random-dataexperiments
in the paper. The results in table 4 strongly reinforce my reasons for
discountingthe results in table 3 and figure4.
Just as I doubt that the authors' VAR analysis helps to explain
exchangerates in a meaningfulway, I doubtthat it can help in assessing
whetherexchange rates move "too much." With a structuralmodel in
hand, an analyst can either refute or supportassertions that exchange
rates move too much or are out of alignment. Doing so requires
specificationof a counterfactualscenario and of normativecriteriafor
rankingactual and counterfactualoutcomes. But the analyticalproce-
duresare conceptuallystraightforward.
Withonly a VAR system of the sortestimatedin this paper,however,
one cannot get a genuine handleon such questions. The authorsthem-
selves acknowledgethis difficultyat one point in theirdiscussion. Later
in their paper, however, when they ask specificallywhether exchange
ratesmove disproportionatelyin responseto largechangesin economic
variables,they are temptedto revertto theirVAR systems andto report
a roughtest for nonlinearresponse (table6). The same basic difficulties
exist with this roughtest, andit is no morerevealingthanthe underlying
VAR results.
I reluctantly conclude that in the VAR section of their paper the
authorshave not providedany dependableevidence on why exchange
rates move as they do and whether they move excessively. What is
neededis evidenceon determinantsof exchangeratesina structuralsense.
Despite an apparentappealas a methodof avoidingthe difficultiesof an
old-fashioned structuralapproach, VAR systems are not capable of
yieldingsomethingfor nothing.If we wantto makestructuralinferences,
as we clearlydo, there is no easy shortcut.2
ShaferandLoopesko areunderstandablyconstrainedfromadvancing
policy views very differentfrom those of the Federal Reserve and the
Reaganadministration.Not being subjectto such constraints,I want to
state-incautiously-several propositionsabout the implicationsof ex-
change-ratevariabilityfor policy. These propositions seem to me a
reliableanchorfor more detailedrecommendations.
Therearetwo widelyheld, somewhatideologicalviews aboutchanges
2. The criticismsof VAR analysismadehere are developedmore systematicallyin a

recent paperby ThomasF. Cooley and StephenF. LeRoy, "AtheoreticalMacroecono-
metrics:A Critique"(Universityof Californiaat SantaBarbara,March1983).
in exchange rates. The first, most often encounteredin Europebut also
increasinglypopularelsewhere,is the minimum-variance position.Those
holdingthis view emphasizethe uncertaintyand disruptionthat may be
associated with fluctuationsin exchange rates and argue that govern-
ments should act to maintainas much stability in exchange rates as
possible. The second view, the untrammeled-market position,holdsthat
every nation shouldpursueappropriatedomestic macroeconomicpoli-
cies and then permit currency values to be freely determinedin the
exchangemarketwithoutany interventionby centralbanksandgovern-
ments.
Both of these traditionalpositions are analyticallydeficient.Variabil-
ity in exchange rates is neithergood nor bad in itself. And its presence
or absence shouldnot be a goal of nationalmacroeconomicpolicy.
The traditionaldebate between the minimum-varianceand the un-
trammeled-market positionshas a polarizedcharacterthatis misleading.
Policymakershave no compelling reason to choose between fixed or
flexible exchange rates. Since they do not have to make that choice,
furthermore,they should not make it. When discussingdomestic mon-
etary policy, economists do not have analogously polarized debates
about interest-rate variability-at least not to the same unfortunate
degree. There is little pressure on policymakers to choose between
minimum variability in interest rates and completely untrammeled
variability.Nor should there be. When debate occurs about excessive
interest-ratevolatility, there are few if any who challengethe basic case
for managedvariabilityof some sort.
A thirdview aboutexchange-ratevariability,the insulationposition,
was once very popular among economists. According to that view,
"flexible exchange rates are a means of combining interdependence
among countries throughtrade with a maximumof internalmonetary
independence"and are a "means of permittingeach countryto seek for
monetarystability accordingto its own lights, without either imposing
its mistakeson its neighborsor havingtheirmistakesimposedon it." A
sweeping version of that view asserted that flexibilityin exchange rates
bottles up policy actions and nonpolicy disturbanceswithin the nation
where they originate,therebyinsulatingother nationsfromtheirconse-
quences. A more cautious version asserted only that flexible rates
insulatenationsfrom each other's "monetary"disturbances.3
3. Milton Friedman, "The Case for Flexible Exchange Rates," Essays in Positive
Economics (Universityof ChicagoPress, 1953),p. 200.
As is now fairlywidely recognized,andas ShaferandLoopekso note,
this thirdview is also analyticallydeficient. It is simply not correct, in
theoryor practice,thatflexibleexchangerates can insulatean economy
fromdisturbancesoriginatingabroad,even frommonetarydisturbances.
There is an element of truth in the conventional view about the
insulating properties of flexible rates. For most if not all domestic
macroeconomic policy actions taken abroad and for many types of
nonpolicy disturbances,the effects spill over less into the home nation
if the home currencyis permittedto appreciatein response to external
stimulithat are expansionaryandto depreciatein responseto those that
are contractionary. But this element of truth does not constitute an
unqualifiedrecommendationin favorof moreratherthanless variability
in exchange rates. For one thing, the bufferingtendencies of variability
do not apply to all types of disturbancesoriginatingabroad.Even more
important,the bufferingtendencies associatedwith exchangerate vari-
abilityare not always beneficial.(Forexample,policymakersshouldnot
want to be insulatedfrom the rest of the world in periodsdominatedby
disturbancesoriginatingwithinthe real sectors of theirown economy.)
Still anothercommonattitudeaboutexchangerate arrangementshas
been to assert that one or anothertype of exchangerate system imposes
constraintson policy decisions. Before 1973it was widely believed that
the obligationof maintaininga par value tends to "discipline" policy-
makersand constructivelyconstraintheir choices about domestic poli-
cies. The advocates of floating,using a similarargumentto advance an
opposite policy recommendation,claimed that the decision to float
would provide independencefor domestic policies. More recently, one
frequentlyhearsthe assertionthata system of flexibleexchangerates is
a majorconstrainton domestic policies.
All these assertions and counterassertions, however, tend to be
misleading.It is neitherfixedexchangeratesnorflexibilityof those rates
that impose constraintson domestic policies. Exchangerates per se are
not the central influence. Rather, it is the openness of the economy-
the magnitudeand patternof internationaltransactions-that puts gen-
uine constraints on what can be done with domestic policies. The
autonomy of national policies, furthermore, is undermined by the
openness of the economy no matterwhat happensto exchangerates.
ShaferandLoopesko observe thatthe past ten yearswouldhave been
difficultto live throughunderany exchange rate arrangements.I would
state the point more strongly.There is no set of exchange rate arrange-
ments underwhich the 1973-75and 1979-80oil shocks would not have
had traumaticconsequences. It is impossible to imagineany arrange-
ments that would not have transmittedmajorinflationaryand contrac-
tionary impetuses back and forth amongthe majoreconomies. I doubt
that the competence and appropriatenessof domestic macroeconomic
policies in the past ten years was much influenced,positively or nega-
tively, by the exchange rate arrangementsthat actuallyexisted, or that
those policies would have been greatly improved under any other
arrangements.For the same reasons, the next ten years in the world
economy are likely to be difficultno matterwhat exchangeratearrange-
ments are in place.
Intervention in the exchange market is not an unimportantissue,
despite the pointsjust made. CurrentU.S. policy about interventionis
rigidand verges on a theologicaladherenceto the untrammeled-market
position. It would be preferablefor the United States to take a more
eclectic stancetowardinterventionfor U.S. accountandto adopta more
forthcomingposition about cooperative interventionwith foreign gov-
ernments. (The discussion of interventionby Shafer and Loopesko is
consistent with this recommendation,but they refrain from openly
criticizingcurrentU.S. policy.)
Even thoughU.S. interventionpolicy needs to be modified,it would
be wrongto expect a greatdealfromthatchangealone. Indeed,the most
importantmodificationin policy that is needed is for the United States
and foreign governmentsto refine their overly simple attitudes. Inter-
vention policy and exchange rate arrangementsare secondary rather
than primaryissues. The basic macroeconomicproblemstroublingthe
United States will remain,and cause headachesof roughlycomparable
intensity, regardless of what is done with intervention policy and
exchange rate arrangements.It is thus a mistake to allow the manner
and degree of variabilityin exchange rates to be seen as, in itself, an
issue of overridingimportance.
Rudiger Dornbusch: Various studies that have appearedin the past

year, in particularthe work of Meese and Rogoff, have documentedthe
failure of structuralmacroeconomicmodels to explain the facts about
exchange rates in a satisfactory manner.1Indeed, a random walk is
1. RichardA. Meese and KennethRogoff, "EmpiricalExchangeRate Modelsof the
Seventies: Do They Fit Out of Sample?" Journal of International Economics, vol. 14
(February1983),pp. 3-24.
shown to be preferableto structuralmodels in predictingrates. In the
field of exchange rate economics analysts have reached the point at
which the stock marketliteraturearrived some years earlier. Against
this background,Shaferand Loopesko set out to investigatewhat is left
of exchange rate economics and to determinewhat inferences can be
madefor exchange rate policy.
The paper presents a refreshingapproachthat blends marketcom-
mentary and the centerpieces of exchange rate models. It studies
episodes ratherthan tryingto explain (whereothers had alreadyfailed)
all the evidence in a uniform way. The authors immediatelydismiss
purchasing power parity as a plausible model to explain short-run
developmentsin exchange rates and appropriatelygive emphasisto the
difference in adjustment speeds between goods markets and asset
markets,to the risk premium,and to the currentaccount.
The dismissal of purchasingpower paritythat now has become well
establishedis one of the importantinsightswe have gainedfromthe ten-
year experience with floatingrates and from a more thoughtfulreview
of earlierepisodes. It was believed thatexchangerates mightnot follow
purchasingpower parityexcept duringperiodsdominatedby monetary
disturbances.Now the evidence leads to a much strongerargument:in
the case of monetarydisturbancesin particular,largesystematicdevia-
tions frompurchasingpowerparityare observed. This was dramatically
illustratedduringthe Germanhyperinflationafter WorldWar I. And it
appearswhenever monetaryinstabilitydominates:for instance, in the
experience of the United Kingdom or in the U.S. real exchange rate
since 1979.The prices of U.S. manufacturedgoods relativeto the dollar
prices of other nations are today 20 to 30 percenthigherthanthe 1973-
82 average. That representsa very strikingdeparturefrom purchasing
power parity, the reason for which is to be found in the monetary-fiscal
policy mix.
Thereare two centralconclusionsof the paper.First, the sticky-price
asset-market model provides a relatively satisfactory framework to
explain the broad patternof exchange rate movements, at least in the
cases of the dollar-yenand the dollar-mark.Second, there is no striking
evidence of exchangerateeffects throughriskpremiumsandthe current
account, although on occasion these factors may have residually ac-
counted for some of the experience. The combinationof these two
conclusions leads to a third, this one directed to policy: there is no
presumptionthat exchangerates have moved in an unaccountableway,
and for that reason an active exchange rate policy does not appear
imperative. However, because there is some evidence that sterilized
intervention might work and that deviations from equilibriumreal
exchange rates persist, the issue of whether to pursue a more active
exchange marketpolicy remainsopen.
Shafer and Loopesko offer two kinds of evidence on the success of
structuralmodels in explaining exchange rate developments. One is
evidence from a vector autoregressive (VAR) representationof the
exchange ratefocused on the macroeconomy.The VARs are estimated
for seven variables,with exchange rates placed last, and the results are
used to determinethe role of the "fundamentals"from popularmodels
in accountingfor the variancein the forecast error.The exercise shows
that the fundamentals-innovations in money, prices, output, interest
rates, and trade balances-account for only a small portion of the
forecast-errorvariance for the short-termhorizon, with the portion
accountedfor reaching60 percentonly aftertwo years or more.
The variancedecompositionthat the authorsshow is interesting,but
theirinterpretationis misleading.VARs cannothelp, except by assump-
tion, to identify the structuralerrors that give rise to exchange rate
innovations. From the VAR estimation procedure it is clear that the
innovation in the exchange rate equation is an amalgamof the error
termsinallthe structuralregressionsrepresentingthe news 2 Innovations
in prices, output, or money are simply responses to combinationsof
these primitiveinnovations. The decompositionof the variancein the
forecast errorthereforedoes not help to allocateforecasterrorvariance
to such phenomenaas shifts in money demand,transitoryfiscal expan-
sion, strikes, a blip in the wage equation,or an oil price shock. Because
most of the variablesused in the VAR representation(certainlythe trade
balance,forexample)cannotbe thoughtof as predetermined,the analysis
reallydoes not succeed in linkingchanges in exchangerates to news.
The authors conclude from their VAR analysis that fundamentals
account for only about 60 percent of exchange rate movements. That
conclusion is certainly not warranted.Forward-lookingmodels of the
exchange rate based on rational expectations show that anticipated
2. See ThomasF. Cooley and StephenF. LeRoy, "AtheoreticalMacroeconomics:A

Critique"(Universityof Californiaat SantaBarbara,March1983).
future movements in fundamentalsaffect the current exchange rate.
News relevant to these future values of the fundamentalsneed not be
correlatedwith the currentrealizationsof the particularregressorsthat
are used. The failureto explain exchange rates by structuralmodels or
by VAR analysis thus comes down to our inability to track these
expectations, not necessarilyto a failureof the models.
There are other reasons why I view this particularVAR approach
with reservations. First, the entire analysis is excessively bilateral.It
focuses only on two countriesat a time andassumes implicitlythatthere
areno importantmultilateralinteractionsthroughtradeflows or through
the world capital market. Data scarcity may have precluded a more
generous specification, but that only means the exercise cannot be
conducted meaningfully.The same is true for other importantomitted
variablessuch as wealth. It would be interestingto know, for example,
whether stock market prices play an importantrole in exchange rate
determination, as one would expect from portfolio models. I also
discount the empiricalresults because, if there are importantlinkages
between the interestrate and exchange rate, the 1979changein Federal
Reserve operatingproceduresis certain to have affected the relations
amonginnovations,interestrates, andexchangerates, which makesthe
assumptionof an unchangingstochasticprocess hardto accept.
The second kind of empiricalevidence supportingstructuralmodels
concerns the link between real interest rate differentialsand the real
exchange rate. The authors note that the real interest differential,
(r - r*), is equal to the expected rate of change of the real exchange
rate,
(1) r= r* +zq,
where q = p* + e - p is the log of the real exchange rate and p and e
arethe logs of the pricelevel andthe nominalexchangerate,respectively.
Assuming now that the real exchange rate, following a disturbance,
converges asymptoticallyto its long-runequilibrium,q, yields
(2) Aq = v(q - q)
or, combiningequations 1 and 2,
(3) - r*)iv.
q = wqi-r(r
The model therefore predicts that, when there is areal interest differentiaI
with foreignrates exceeding those at home, the relativeprice of foreign
goods will be high and falling. The model carries the assumption of
perfect asset substitution,and the evidence is offeredin figure5.
The authors express qualifiedsatisfactionwith the findingsfor the
case of the dollar-markand the dollar-yenexchange rates, though not
with the dollar-poundsterlingrate. I am not certain by what criterion
theyjudge theirsuccess, except to note thatin the regressionfor the first
two cases the linkagebetweeninterestdifferentialsandthe realexchange
rateis indeedsignificantandof the sign indicatedby the theory. But it is
also true that these regressionsperformwell only if the 1980-82 period
is included and that they show little stability in a sample for a longer
period. It is also worth noting that the coefficients of the interest
differentialreportedin the regressionsshow values of 2 to 3 for 1/v.Thus
the meanlag, (1 - v)/v, would be of the orderof one and one-halfyears.
These resultsare certainlyout of line with the conjecturethatdisequilib-
riumsin real exchange rates that persistfor long periods, say ten years,
accountfor the patternsin the real exchangerate and real interestrate.
The model for the real interestrate does well in explainingthat a rise
in U.S. interestrates shouldlead to an appreciationof the realexchange
rate. But it fails when it predictsthat the real exchange rate shouldalso
be depreciating.That has not in fact been happening,and a theory is
needed that will explain why the dollar-real or nominal-is both high
and stuck. I believe fiscal policy may providean answer. In the United
States the prospectivefull-employmentdeficitshave increasedgreatly;
abroadthey remainunchangedor even decline. The effect is a rise in
worldaggregatedemandat full employment.Moreover,althoughworld
aggregatedemand has increased, there has also been a relative rise in
thedemandfor U. S. goods becausethe fiscalexpansionthathasoccurred
in the United States has led to relativelylargerincreases in spendingon
U.S. goods, though of course there are also spillover effects abroad.
One thus expects the full-employmentreal interestrate-the long-term
rate-to increase in the world to restore the balance between world
spendingand worldfull-employmentincome. At the same time the real
exchangerateof the dollarwill appreciateto eliminatethe relativeexcess
demandfor U.S. goods.
The diagrambelow illustratesthese points. On the axes I show the
world real interestrate, r, and the relativeprice of U.S. goods (the real
exchangerate), q. Alongline I the worlddemandfor U.S. goods is equal
r
I*~~~~~~~~~~~~~
to the full-employmentsupply.Higherrealinterestratesdepressdemand
and therefore require an offsetting real depreciation to maintainfull
employment.Similarly,alongI* thereis fullemploymentabroad.Higher
interestratesreducedemandandrequirea realappreciationof the dollar
to maintainfull employmentabroad.A U.S. fiscal expansion,by raising
the demandfor U.S. goods, shifts the equilibriumscheduleof the goods
marketupwardand to the left. A new equilibriumobtains at point E'
with a higherworld interestrate and a real appreciationof the dollar.
These prospective changes in interest rates and exchange rates are
anticipatedunderrationalexpectationsand show up in higherlong-term
realinterestratesandin dollarappreciation.Theforward-lookingnature
of assets markets,however, makesrecovery muchmoredifficult.If this
analysisis correct,a move towardsmallerlong-run,not cyclical, deficits
would lead to a collapse of the dollar. The analysis emphasizes the
peculiarandcentraleffects of fiscalpolicy underflexibleexchangerates.
This is a point Shaferand Loopesko indeedrecognize, althoughthey do
not go beyond sketchingan interestingframeworkfor a more complete
investigationof fiscal policy.
The authorsare cautious in their assessment of the experience with
floatingexchange rates, and they are equallycautiousin offeringpolicy
advice. They do not conclude, for example, thatflexiblerates played an
importantpartin the deteriorationof macroeconomicperformanceand
in the growthof protectionism.Nor do they argueforcefully either for
interventionor againstit. Now only few fully committedsupportersof
floatingrates remain. But criteriaare still lackingby which to form the
judgment that flexible rates have been a bad experience, just as we
confidentlyannouncethat the BrettonWoods system was poor.
Shaferand Loopesko do not express a strongview on intervention.
They believe some intervention may be desirable in some circum-
stances. But, for instance, would they advocate using it to alter the
presentdollarexchange rate?
General Discussion
C. Fred Bergstenpointedout that, while muchof the attentionin the
paperand comments was devoted to the possibilityof excess volatility
in exchange rates, the possible misalignmentof exchange rates was of
greater concern. These two issues are frequently mixed up in formal
discussions, though they are conceptually quite distinct. In current
policy discussions much of the present rationale for intervention in
exchange markets has to do with maintainingrelative stability of ex-
change rates, or what is usually referred to as "leaning against the
wind." Yet often such interventionresultsin slowingdown the required
adjustmentprocess as rates head toward their equilibriumvalues. He
noted that the Germans had intervened in January 1983 to slow the
appreciationof their currencyagainstthe dollar. This interventionwas
in keepingwith the accepted internationalpracticeof leaningagainstthe
windbutwas the oppositeof whatwas requiredto achieve a fundamental
realignmentof currencylevels.
Georgevon Furstenbergrespondedthatwe mayknow less thansome
people thinkaboutwhatalignmentof exchangeratesis appropriate.The
Germanelection results in early 1983 would appearto have removed
uncertaintyabout the future of Germanpolicy, yet the deutsche mark
subsequentlyfell againstother currencies.Relatively high real interest
rates, which were presumedto have overvaluedthe dollar,have melted
away, but the dollar stubbornlypersists in being "overvalued." U.S.
budgetdeficits are presumedto contributeto the strongdollar,but even
larger deficits in Japan apparentlyhave no similareffect on the yen.
RobertLawrence disagreedwith the statementthat even well-informed
specialists could not predict the impact of fiscal policy on exchange
rates. He observed that Japanand the United Kingdomhave actually
tightenedtheir fiscal policies, while the United States has loosened its
policies. The consequences on the relative values of the yen, pound
sterling,and dollar have been the expected ones. EdmundPhelps also
commentedon the relationbetweenfiscalpolicy andexchangerates. An
anticipated increase in future government expenditure will raise the
anticipatedfuture value of the local currency and thus strengthenthe
currency immediately. He reasoned that this effect, together with tax
liberalizationsthathave raisedthe realrateof interest, may help explain
the currentstrengthof the dollar.
Even assumingthatthe authoritiesknewin whatdirectiontheywanted
to move exchange rates, WilliamNordhaus noted that recent studies
cast doubton the feasibilityof sterilizedintervention-buying or selling
foreign exchange without affectingeither country's money supply. He
arguedthat evidence suggests the effect of such interventionis exceed-
ingly small. This result parallelsthe historicalexperience with "Opera-
tion Twist" under the regime of fixed exchange rates in the 1960s.
OperationTwist representedan attemptto alter the term structureof
interestrates so that short-termrates would be high, to attractfunds to
the United States, while long-termrates would be kept low so as to
stimulateinvestment. However, the scale of the requiredintervention,
even to changerelativeinterestrates by a few basis points, discouraged
authoritiesfrom a large-scaleOperationTwist.

Fler 10yrs

Uploaded by

Copyright:

Available Formats

Fler 10yrs

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fler 10yrs

Uploaded by

Copyright:

Available Formats

JEFFREY R.

Floating Exchange Rates

We gratefullyacknowledgethe excellent assistance of Peter Borish and the help of

The Transition from Fixed to Floating Exchange Rates

Duringthe late 1960sand early 1970s, the system of fixed exchange

omies wouldbe relativelyinsulatedagainstmacroeconomicshocks from

Problems in Evaluating the Period of Floating Rates

The period of floating exchange rates has been marked by great

exchange rate regime. These differencesinclude the following:(1) It is

4. Capitalcontrols were pervasive at the beginningof the floatingrate period. The

Exchange Rate Behavior and the Development

The theory of exchange rate determinationhas evolved largelyfrom

THE FLEXIBLE-PRICE MONETARY MODEL

for purchasingpower parity,

where r is the nominalinterest rate, e denotes an expected value, and a

depreciation is just the difference in inflation rates between the two

exchangerateis equalto the expectedfuturespot rate, whichprovidesan alternativeand

EXCHANGE RATE BEHAVIOR: MARCH 1973 TO LATE 1975

By late 1975it was becomingclear that purchasingpower paritywas

Markper dollar(index,1973:1= 100)

Yenper dollar(index,1973:1= 100)

Poundper dollar(index,1973:1= 100)

EffectiveU.S. dollar(index,1973:1= 100)

--' ominal ">

1960 1965 1970 1975 1980

Three reasons were advancedto explainthese adjustmentsto the oil

THE STICKY-PRICE MONETARY MODEL

The principaltheoreticalresponseto these observationswas renewed

Yt = y(StP* /Pt, rt - pe);

for the expectations-augmentedPhillipscurve,

Yenper dollar Interestand inflationdifferential

Poundper dollar Interestand inflationdifferential

ultimatelyconverge to the same paths as in the flexible-pricemonetary

EXCHANGE MARKET DEVELOPMENTS:

Developments in 1976 did not seriously challenge the sticky-price

Germany , " Currentaccount

THE PORTFOLIO-BALANCE MODEL

The developmentof varioustheoreticalformulationsof the portfolio-

26. Moregeneralformulationsof asset demandsandendogenousadjustmentof foreign

27. The usual formulationsof this model ignorethe role of governmentdeficitsas a

EXCHANGE RATE DEVELOPMENTS:

In the periodfrom 1979to 1982the positions of the dollarand pound

offer general, as opposed to ad hoc, explanationsof the behavior of

Empirical Evaluation of Exchange Rate Models

Duringthe past decade the developmentof the theory of exchange

focuses on the short-runvolatility of exchange rates and evaluates

current values of those variables themselves, then the exchange rate

SHORT-RUN EXCHANGE RATE VOLATILITY

An assessment of whether exchange rates have been too volatile

States authorities in October 1979 to abandon operating procedures

31. Sluggish price adjustmentis not, however, a necessary conditionfor a strong

The final and probablymost importantconsiderationin assessing if

EXCHANGE RATE VOLATILITY AND FUNDAMENTALS

It is nonethelessusefulto seejust how muchof exchangeratevolatility

33. ChristopherA. Sims, "Comparisonof Interwarand Postwar Business Cycles:

42. Meese andRogoff, "EmpiricalExchangeRate Models."

whereu andv arethe disturbancetermsfromthe firstand secondequations,respectively.

so that its variance is calculated from terms involving contemporaneousand lagged