A Monetary Approach to the Exchange Rate: Doctrinal Aspects and Empirical Evidence

Author(s): Jacob A. Frenkel

Source: The Scandinavian Journal of Economics, Vol. 78, No. 2, Proceedings of a Conference
on Flexible Exchange Rates and Stabilization Policy (Jun., 1976), pp. 200-224
Published by: Wiley on behalf of The Scandinavian Journal of Economics
Stable URL: http://www.jstor.org/stable/3439924 .
Accessed: 26/11/2014 03:40

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Wiley and The Scandinavian Journal of Economics are collaborating with JSTOR to digitize, preserve and
extend access to The Scandinavian Journal of Economics.

http://www.jstor.org

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
A MONETARYAPPROACHTO THE EXCHANGE
ASPECTSANDEMPIRICAL
RATE:DOCTRINAL
EVIDENCE
JacobA. Frenkel*
Tel-Aviv,Israel
Universityof Chicago,Chicago,Illinois,USA and Tel-AvivUniversity,

"What, then, has determinedand will determine the

value of the Franc? First, the quantity, present and
prospective, of the francs in circulation. Second, the
amount of purchasingpower which it suits the public to
hold in that shape."
Keyne8(Introductionto French edition, 1924, xviii).

Abstract
This paper deals with the determinantsof the exchange rate and develops a
monetaryview (or more generally,an asset view) of exchangerate determination.
The firstpart traces some of the doctrinaloriginsof approaches to the analysis
of equilibrium exchange rates. The second part examines some of the empirical
hypotheses of the monetary approach as well as some featuresof the efficiency
of the foreignexchange markets. Special emphasis is given to the role of expecta-
tions in exchange rate determinationand a direct observablemeasureof expecta-
tions is proposed. The directmeasure of expectationsbuilds on the information
that is contained in data from the forwardmarket for foreignexchange. The
empiricalresultsare shown to be consistentwith the hypothesesof the monetary
approach.

Introduction
This paper deals withthe determinants of the exchangerate. The approach
that is takenreflectsthe currentrevivalof a monetaryview,or moregener-
allyan assetview,oftheroleoftheratesofexchange.1Basically,themonetary
approachto the exchangerate may be viewed as a dual relationshipto the
monetaryapproachto the balance of payments.These approachesemphasize
the role of moneyand otherassets in determining the balance of payments
* I am indebted to John Bilson forcomments,suggestionsand efficientresearchassistance.
In revising the paper I have benefitedfromhelpfulassistance fromR. W. Banz and useful
suggestionsby W. H. Branson, K. W. Clements,R. Dornbusch, S. Fischer, R. J. Gordon,
H. G. Johnson,M. Parkin, D. Patinkin and L. G. Telser. Financial support was provided
by a grant fromthe Ford Foundation.
1 This view has been forcefullyemphasized by Dornbusch (1975, 1976, 1976a). See, too,
Frenkel and Rodriguez (1975), Johnson (1975), Kouri (1975) and Mussa (1974, 1976).
For an early incorporationof monetaryconsiderationsin exchange rate determinationsee
Mundell (1968, 1971).

Sand. J. of Economica 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 A monetary
approachtotheexchangerate 201

whenthe exchangerateis pegged,and in determining the exchangerate when

it is flexible.
Being a relativepriceoftwo assets (moneys),the equilibriumexchangerate
is attainedwhenthe existingstocksof the two moneysare willinglyheld. It is
reasonable,therefore, that a theoryof the determination of the relativeprice
of two moneyscould be statedconveniently in termsof the supplyof and the
demandforthesemoneys.
The renewedemphasison theroleofthesupplyofand thedemandformoneys
and assets as stocksin contrastwiththe circularflowapproachto thedeter-
minationof the exchangerate (that gained popularitywith the domination
of the Keynesianrevolution),revivesthe basic discussionof the Bullionist
controversy culiminatedin the early 1800's and led to the developmentsof
the "Balance of Trade Theory"and the "InflationTheory"of the determina-
tion of the exchangerate (Ricardo, 1811; Haberler,1936; Viner,1937). Re-
miniscenceof that controversy can be tracedto presenttimesin the various
discussionsand interpretation of the purchasingpower parity doctrine.It
may be arguedthat the long experiencewiththe gold standardand withthe
gold exchangestandardmay have led to the retrogression of the theoryof
flexibleexchangerates (Wicksell,1911,p. 231; Gregory,1922,p. 80).1
The firstpartofthepapertracessomeofthedoctrinaloriginsofapproaches
to exchangerate determination. Its purposeis to providesome perspective
into the evolutionof the theory.2The main emphasisof the paper lies in its
secondpart wherewe examinesomeof the empiricalhypothesesof the mone-
tary approach.In that part we analyze the role of expectations,we describe
a directmeasurethereof, examinetheefficiency oftheforeign exchangemarket
duringthe Germanhyperinflation, and provide some evidence which sup-
portsthe asset view of exchangerate determination.

I. A Doctrinal Perspective to Exchange Rate Determination

I.1. The PurchasingPowerParityDoctrine:
The Natureof Equilibrium
The purchasingpowerparitydoctrine(in its absoluteversion)statesthat the
equilibriumexchangerate equals the ratio of domesticto foreignprices.The
relativeversionof the theoryrelateschangesin the exchangerate to changes
in priceratios.Many of the controversies
aroundthat doctrinerelateto the
1 It is of interestto note that the introductionof the "liquidity preference"schedule which
emphasizes the role of asset markets and characterizes much of the Keynesian revolution
in macroeconomic analysis of the closed economy, did not carry over to the popular
versions of the Keynesian theoriesof the balance of payments. The Keynesian analysis of
the balance of payments emphasizes the circular flow of income, the foreigntrade multi-
plier and, in its popular version, ignores to a large extent the role of money and other
assets. For a notable exception, see Metzler (1968).
2 An analogous doctrinal perspective of the evolution of the monetary approach to the
balance of payments under fixed exchange rate is contained in Frenkel & Johnson (1976)
and Frenkel (1976). For a furtheranalysis see Myhrman (1976).

Scand. J. of Economic8 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
202 J. A. Frenkel

questionof choiceof properpriceindicesto be used in computingthe parity.'

One extremeview arguesthat the properpriceindex shouldpertainto traded
goods only(Angell,1922; Bunting,1939 Heckscher,1930; Pigou, 1930; Viner,
1937), whileaccordingto the otherextremeviewtheproperpriceindexshould
coverthe broadestrangeof commodities(Hawtrey,1919,p. 109; Cassel,1928,
p. 33).2
Those who advocate the use of traded goods index emphasizethe role of
commodityarbitragewhilethose who advocate the broaderpriceindex em-
phasize the role of asset equilibriumas determining the rate of exchange.If
the role of the exchangerate is to clear the moneymarketby equatingthe
purchasingpowerof the variouscurrencies, thenthe relevantmeasureshould
be a consumerpriceindex.3Proponentsofthisviewrejecttheuse ofthewhole-
sale priceindexsinceit givesan excessiveweightto tradedgoods (Ellis, 1936,
pp. 28-9; Haberler,1945,p. 312 and 1961,pp. 49-50).
The two viewsdifferfundamentally in the interpretation
ofthe equilibrium
exchangerate. The commodityarbitrageview goes even furtherin arguing
that no aggregateprice index is relevantand only individual commodity
pricesshouldbe analyzed:
"Foreignexchangerateshave nothingto do withthewholesalecommodity
price
levelas suchbut onlywithindividualprices"(Ohlin,1967,p. 290).
The equilibriumexchangerate reflectsspatial arbitragefromwhichnon-
tradedgoods are excluded:
"Patently,I cannotimportcheapItalianhaircutnorcan Niagra-Falls
honeymoons
be exported"(Samuelson,1964,p. 148).
The assetviewtakesit forgrantedthattheoperationofcommodity arbitrage
equates the pricesoftradedgoods and emphasizesthatifthedoctrineonlyap-
plies to tradedgoods,then:
"the purchasing powerparitydoctrinepresentsbut littleinterest... (it) simply
statesthatpricesin termsofanygivencurrency, ofsamecommodity mustbe the
same everywhere ... Whereasits essenceis thestatement
thatexchangeratesare
the index of the monetaryconditionsin the countriesconcerned"(Bresciani-
Turroni,1934,p. 121).

In fact since the exchangerate linksthe purchasingpowervalues of moneys

in termsofthe broaddefinition ofthe pricelevel,one mayimaginea situation
in whichall tradedgoods possessthe same price,whenexpressedin common
currency, but the exchangerate is in disequilibrium:
I See Viner (1937) and Johnson (1968).
2 It might be of interestto compare these discussions with those concerningthe range of
transactionsand prices that is relevant forthe quantity theoryof money. The latter ranged
fromsuggestionsto include transactions in assets and prices of securities to suggestions
to include only what is definedas national product.
8 On the relevance of the consumer price index as a measure of the purchasing power of
money see Marshall (1923, p. 30) and Keynes (1930, p. 54).

Scand. J. of Economies 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 A monetary
approachtotheexchangerate 203

"The equilibrium to whichtheforeign exchangemarkettendsis an equilibrium of

thepricelevel... If thecurrency unitsoftwocountries be considered
in termsof
foreigntradeproductsonly,thentherateofexchangebetweenthetwocurrency
unitswillapproximate closelyto theratiooftheirpurchasing powerso calculated...
But thatis not theconditionof equilibrium ... It is to thepricelevelin general,
ofhometradeproductsas wellas foreign tradeproducts, thattherateofexchange
mustadjust" (Hawtrey,1919,p. 109).
To completelydivorcethe determinationof exchangeratesfromconsidera-
tions of commodityarbitrage,one could even go furtherin developingan
argumentforusingpriceindicesof non-tradedgoods only.Such an argument
was advanced by Graham:
"Strictlyinterpreted
then,pricesof non-internationally
tradedcommoditiesonly
shouldbe includedin the indiceson whichpurchasing powerpars are based"
(Graham,1930,p. 126,n. 44).
Furtherpursuitofthat idea leads to the use ofthe priceofthe least traded
commodity-thewage rate parity-advocated by Rueffin 1926 (reproduced
in Rueff,1967),and similarviewscan also be foundin Hawtrey(1919,p. 123)
and Cassel (1930,p. 144). The wagerate approachwas extendedto the concept
of productioncost paritiesadvocated by Hansen (1944, p. 182), Houthakker
(1962,pp. 293-4) and Friedmanand Schwartz(1963,p. 62).
Whateverthe price index used for computationsof parities,the question
remainsof distinguishing betweenan equilibriumrelationshipand a causal
relationship.Most authorsrecognizedthat pricesand exchangerates are de-
terminedsimultaneously.A minority,however,argued that there exists a
causal relationshipbetweenprices and exchangerates. While Cassel (1921)
claimedthat the causalitygoes frompricesto the exchangerate,Einzig (1935,
p. 40) claimedthe opposite.
1.2. The Asset View
Since in generalbothpricesand exchangeratesare endogenousvariablesthat
are determinedsimultaneously, discussionsof the link betweenthemprovide
littleinsightsinto the analysisof the determinants of the exchangerate. The
originalformulation of Cassel (1916) was statedin termsof the relativequan-
titiesof money.The formulation was then translatedinto a relationshipbe-
tweenpricesvia an applicationofthe quantitytheoryofmoney.Conceptually,
however,it seemsclear that the role of pricesin Cassel's computationof the
equilibriumexchangerate servesonlyto proxythe underlying monetarycon-
ditions.The determination of exchangerates does not seem to rely,directly
or indirectly,
on the operationof arbitragein goods.
In retrospectit seems that the translationof the theoryfroma relation-
ship betweenmoneysinto a relationshipbetweenprices-via the quantity
theoryof money-was counterproductive and led to a lack of emphasison
the fundamentaldeterminantsof the exchangerate and to an unnecessary
amountof ambiguityand confusion.It is noteworthy that the originatorsof
Scand. J. of Economics 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
201 J. A. Frenkel

the theory(althoughnot in its presentname)-Wheatley (1803) and Ricardo

-stressed themonetarynatureoftheissuesinvolvedas wellas theirrelevance
of commodityarbitrageas determining the equilibriumrate:
"In speakingof the exchangeand the comparative value of moneyin different
countries,we mustnotintheleastreferto thevalueofmoneyestimated in commo-
ditiesin eithercountry.
The exchangeis neverascertained
by estimatingthecom-
parativevalueofmoneyin corn,clothoranycommodity whateverbutbyestimat-
ingthevalueofthecurrency ofonecountry, in thecurrencyofanother"(Ricardo,
1821,p. 128).
When consideringmoneysfor the purpose of determiningthe exchange
rate,the relevantconceptis that of a stockratherthan of a flow.These con-
ceptswerean integralpart ofmonetarytheory:
"We mayconsequently thinkofthesupply(ofcurrency) as we thinkofthesupply
ofhouses,as beinga stockratherthantheannualproduce... (and) ofthedemand
forcurrency as beingfurnishedby the abilityand willingnessof personsto hold
currency"(Cannan,1921,pp. 453-4).
Indeed, as Dornbusch(1976) puts it "The exchangerate is determined in the
stock market". It is this conceptionof money as a stock that resultedin
Keynes' perceptivestatementthat was quoted at the start of this paper.
The asset view of exchangerate determination became the traditionalview
as witnessedby Joan Robinson:
"The traditional
viewthattheexchangevalueofa country's in anygiven
currency
situationdependsuponthe amountof it in existenceis thusseento be justified,
providedthatsufficientallowanceis madeforchangesin theinternaldemandfor
money"(Robinson,1935-6,p. 229).

I.3. The Role of Expectations

A natural implicationof the asset approachis the special role expectations
play in determining the exchangerate. The demandfordomesticand foreign
moneysdepends,like the demandforany otherasset, on the expectedrates
of return.Thus it may be expectedthat currentvalues of exchangeratesin-
corporatethe expectationsof market participantsconcerningthe future
courseofevents.This notioncan accounttherefore, forlargechangesin prices
resultingfromlargechangesin expectations.This was indeedthe logicalargu-
ment used by Mill (1864, p. 178) in accountingforthe sharp changein the
priceof billsoccuringwiththe newsof Bonaparte'slandingfromElba.
The specificrole of expectationsin determining the exchangerate has been
also emphasizedby Marshall(1888), Wicksell(1919, p. 236), Gregory(1922,
p. 90) and Einzig (1935, p. 120). If the foreignexchangemarketis efficient
-as many otherasset marketsappear to be-then currentpricesshouldre-
flectall available information.
Therefore, an expectationof monetaryexpan-
sion shouldbe reflectedin the currentspot exchangerate since asset holders
willincorporatethe anticipatedreductionin the real rate ofreturnon the cur-
Scand. J. of Economic8 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 A monetary
approachtotheexchange
rate 205

rencyin theirpricingof the existingstocks.This notionwas clearlystatedby

Cassel:
"A continuedinflation
... willnaturallybe discounted
to a certaindegreein the
presentratesofexchange"(Cassel,1928,pp. 25-26).

Similarly:
"The international
valuationofthecurrency will,thengenerally
showa tendency
to anticipateevents,so to speak,and becomesmorean expression oftheinternal
value thecurrencyis expectedto possessin a fewmonths,or pherhapsin a year's
time"(Cassel,1930,pp. 149-50).
The empiricalanalysisin Section II will be concernedwith details of the
roleof expectationsin determining the exchangerate.
Prior to concludingthis sectionit should be emphasizedthat its purpose
has notbeen to arguethat "It's all in Marshall".On the contrary,
a rereading
of the writingsof some of the eminentclassicaland neo-classicaleconomists
revealsthe greatneed forsupplementing theirgeneralconceptionswitha de-
tailed analysisof the transmission mechanisms.On theotherhand,however,
it is importantto gain perspectiveand to recognizethat someofthe general
conceptionsand framework ofanalysishave alreadybeen developedby earlier
generationsof economists.It is appropriate,therefore, to view the recent
revivalofthe monetaryapproachas a naturalevolutionratherthan a revolu-
tionarychangein views.'

II. Empirical Evidence: A Reexamination of the German

Hyperinflation

The foregoingdiscussioncontaineda doctrinalperspectiveto the assets ap-

proach to the determination of exchangerates. In this sectionwe develop
further someofthe theoreticalaspectsand reexaminethe determinants ofthe
exchangerate duringthe Germanhyperinflation in lightof theseconsidera-
tions. That episodeis of special interestsince it providesan opportunity to
examinethe assets approachto a situationin whichit is clearthat the source
of disturbancesis monetary.Furthermore, duringthe hyperinflation domestic
1 There are, of course, fundamentaldifficultiesin definingthe nature of the various junc-
tions of intellectualunderstanding.As noted by H. G. Johnson: "the concept of revolution
is difficultto transferfromits originsin politics to otherfieldsof social science. Its essence
is unexpected speed of change, and this requires a judgement of speed in the context of a
longer perspective of historicalchange the choice of which is likely to be debatable in the
extreme" (Johnson, 1971, p. 1).
A casual reading of many of the popular textbook versions of balance of payments
theoriessuggests,however,that the conditions,outlined by Johnson,fora rapid propaga-
tion of a new theorymay be satisfied:"the most helpfulcircumstancesfora rapid propaga-
tion of a new and revolutionarytheoryis the existence of an established orthodoxywhich
is clearly inconsistentwith the most salient facts of reality,and yet is sufficiently
confident
of its intellectualpower to attemptto explain those facts,and in its effortsto do so exposes
its incompetencein a ludicrous fashion" (Johnson, 1971, p. 3).

Scand. J. of Economics 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
206 J. A. Frenkel

(German)influenceson the exchangerate dominatethose occurringin the

possibleto examinetherelationship
restof the world.It is therefore between
monetaryvariablesand the exchangerate in isolationfromotherfactors,at
homeand abroad,whichin a morenormalperiodwouldhave to be considered.

I.1. Moneyand theExchangeRate

Priorto a moreelaborateanalysisit may be instructiveto examinethe as-
sociationbetweenthe Germanmoneystockand its relativepricein termsof
foreignexchange(i.e., the exchangerate). This associationis shownin Figure
1 whichdescribesthe time seriesof the monthlylogarithmsof the German
money supply and the mark/dollar exchangerate for the period February
1920-November1923 (data sourcesare outlinedin the Appendix).As evident
fromFig. 1 the two timeseriesare closelyrelated.A highsupplyof German
marksis associatedwithits depreciationin termsofforeignexchange.
This relationshipcan be examinedfurtherby estimatinga polynomialdis-
tributedlag of the effectsof the moneysupply on the exchangerate. The
estimatesreportedin Table 1 pertainto (i) the effectsof currentand lagged
values of the money supply on the currentlevel of the exchangerate and
(ii) the effectsof currentand laggedvalues oftheratesofchangeofthe money
supplyon the currentrateof changeof the exchangerate. The estimatesof
the distributedlags forthe equationof the ratesofchangerevealthatthecur-
rentrate of changeof the exchangerate dependsonly on the currentrate of

DATE
2002

2005 _

2008 _

2011

2102 _ M
2105 _ . LOG MONEY

2108

2111

2202 _

2205 _

2208 -

2211
_ LOG EXCHANGE RATE
2302 -

2305

2308 _

2311 -

Fig. 1.

Scand. J. of Economics 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 approachtotheexchangerate 207
A monetary

Table 1. Moneyand theexchange lag model.Monthly

rate:polynomialdistributed
data: February1920-November1923

Lag structure

0 1 2 3 4 5 Sum

1.032 - 0.019 - 0.248 -0.009 0.348 0.468 1.572

(0.140) (0.099) (0.127) (0.082) (0.107) (0.125) (0.177)
Dependent variable: Log Exch.
R2 = 0.996, s.e. c 0.379, D.W. = 2.05, e = 0.775, au = 0.617

0.975 0.114 -0.186 -0.136 0.052 0.168 0.987

(0.147) (0.121) (0.130) (0.101) (0.189) (0.203) (0.357)
Dependent variable: A Log Exch.
R2 = 0.895, s.e. = 0.390, D.W. = 2.27

Note: In the polynomial distributed lag equation, a fourthdegree polynomial with the
sixth lag coefficientconstrained to zero, was employed. The first equation relates the
logarithmof the exchange rate to currentand past levels of the logarithmof the money
supply. The second equation relates the percentage rate of change of the exchange rate to
currentand past percentage rates of change of the money supply. e is the final value of
the first order autocorrelation coefficient.An iterative Cochran-Orcutt transformation
was employed when firstorderserial correlationin the residuals of the regressionequations
was evident. a. given in the Table is the standard errorof the regressionequation when the
autoregressivecomponent of the error is included. All of the other statistics are for the
transformedmodel. Standard errorsare in parentheses below the coefficients.

monetaryexpansion.Furthermore, an accelerationof the rate of monetary

expansion induces an equi-proportionate contemporaneous accelerationin the
rate at whichthe currencydepreciates.None of the laggedvariablesare sta-
tisticallysignificant.The distributedlags on the level of the exchangerate
also show a unit elastic contemporaneous effect.In this case, however,some
lagged values exerta significant effecton the rate of exchange.The sum of
the coefficientsis 1.57, i.e., duringthat periodthe elasticityof the exchange
rate withrespectto the moneystockexceededunity.' The magnification ef-
fectof moneyon the exchangerate is consistentwiththe predictionof Dorn-
busch'smodel (1976a) as well as withthe predictionof the rationalexpecta-
tions modelsof Black (1973) and Bilson (1975). If the moneysupplyprocess
is generatedby an autoregressive scheme,expectationswill multiplythe in-
fluenceof currentchangesin the moneystocksince these changesare trans-
mittedinto the futurethroughthe autoregressive scheme.2

1 The firstequation in Table 1 should be interpretedwith some caution since the high
first order autocorrelation coefficientmay reflect a misspecification.Its purpose is to
provide a preliminarydescription of the relationship between money and the exchange
rate. A more detailed analysis follows.
2 Previous empirical work emphasizing monetary considerations in the analysis of the
German exchange rate during the hyperinflationinclude Graham (1930), Bresciani-
Turroni (1937) and more recently,Tsiang (1959-60) and Hodgson (1972).

Sand. J. of Economic8 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
208 J. A. Frenkel

II.2. TheBuildingBlock8oftheMonetary
Approach
The foregoinganalysisindicatedthe close associationbetweenmonetaryde-
velopmentsand the exchangerate. In thissectionwe outlinethe majorbuild-
ing blocksof the monetaryapproachto the exchangerate. Since in what fol-
lows we apply the frameworkto examine data pertainingto the German
hyperinflation,the followingpresentationis simplifiedconsiderablyby ignor-
ing developments in the restof the world.
Considerfirstthe demandforreal cash balancesmd as a functionofthe ex-
pectedrate ofinflationa*:
md g(n*);aga*<0 (1)
The formulation in eq. (1) reflectsthe assumptionthat duringthe hyperin-
flation,changesin the demandformoneyweredominatedby changesin in-
flationaryexpectationsso that the effectsof changesin outputand the real
rate ofinterestmay be ignored.
The supplyof real balances is MIP whereM denotesthe nominalmoney
stockand P "the" pricelevel (we bypassforthe momentthe questionofwhat
is theappropriatepricelevel). Equatingthe supplyofmoneywiththe demand
enablesus to expressthe pricelevel as a functionofthe nominalmoneystock
and inflationaryexpectations:
P = M/g(n*); aP/SM > 0, aP/ln*> 0. (2)

The elasticityofthe pricewithrespectto a* shouldapproximatethe (absolute

value of the) interestelasticityof the demandformoney,and in the absence
of moneyillusionthe elasticityof the pricewithrespectto the moneystock
shouldbe unity.
The secondbuildingblock of the theorylinksthe domesticpricelevel with
the foreignpricelevelP* throughthe purchasingpowerparitycondition:
P = SP* (3)
If the purchasingpowerparityconditionholds we can substituteeq. (3)
into (2) to get a relationshipbetweenthe exchangerate, the moneystock,
inflationaryexpectationsand the foreignpricelevel. Since duringthe German
hyperinflation it is justifiableto assumethat P* is practicallyfixed(as com-
pared withP), we can normalizeunits and defineP* as unity.Thus the ex-
changerate can be writtenas
S = M/g(nc*);bS/M > 0; aS/an*> 0. (4)

that theimplicationthataS/az*> 0 is in conflictwithsome

It is noteworthy
ofthetheoriesofexchangeratedetermination. It shouldbe possible,therefore,
to discriminateamong alternativetheoriesby examiningthe empiricalrela-
tionshipbetweenanticipatedinflationand the exchangerate. A popularana-
Scand. J. of Economice 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 A monetary
approachtotheexchange
rate 209

7r*

7 CL

MO MO M
_I -, S Pa

Fig. 2.

lysisof this relationshipgoes as follows:a higheranticipatedinflationraises

the nominalrate ofinterestwhichinducesa surplusin the capital accountby
attractingforeigncapital,and therebyinducesa lowerspotexchangerate(i.e.,
an appreciationof the domesticcurrency).A variantof this approachwould
arguethat the higherrate ofinterestlowersspending,and thusinducesa sur-
plus in the balance of paymentswhichleads to a lowerspot exchangerate. A
thirdvariantwouldreachthe same resultby emphasizingthe implicationsof
the interestparitytheory(whichis discussedin the next Section). Accord-
ingly,a higherrate of interestimpliesa higherforwardpermiumon foreign
exchangeand if the risein the forwardexchangerate is insufficient to induce
the requiredpremiumon foreignexchange,the spot exchangeratewillhave to
fall (i.e., the domesticcurrency willhave to appreciate).Whateverthe route,
the above analysespredicta negativerelationshipbetweentherateofinterest
and the spot exchangerate whileequation (4) predictsa positiverelationship.
The alternativetheorypresentedhereemphasizesthe role of asset equilib-
riumin determining the exchangerate and its implicationsare illustratedin
Fig. 2. A risein anticipatedinflationfrom74 to a* lowersthe demandforreal
balances, and given the nominalmoneystock,asset equilibriumrequiresa
higherpricelevel. Since the domesticprice level is tied to the foreignprice
via the purchasingpowerparity,and sincethe foreignpriceis assumedfixed,
the higherpricelevelcan onlybe achievedthrougha risein the spotexchange
rate fromSOto S1 (i.e., a depreciationofthe domesticcurrency).
The foregoing analysisdid not specifythe cause of the risein the domestic
rate ofinterest.It is necessaryto emphasizethat the exact analysisof the ef-
fectsof a changein interestratesdependsupon the sourceofthe disturbance.
The presumption,however,is that duringthe hyperinflation the source of
disturbanceswas monetary.The higherinterestrate may be thoughtofas re-
sultingfroma rise in the rate of monetaryexpansionwhichis immediately
15-764816 Scand. J. of Economics 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
210 J. A. Frenkiel

incorporated(via the Fishereffect)in inflationary expectations;to complete

the experiment,it is possibleto imagine(at least analytically)an acceleration
in the monetarygrowthrate whichdoes not instantaneously affectthe mone-
tarystock.
The resulthas an intuitiveappeal in that easy moneyinducesa deprecia-
tionofthe currency. It emphasizesagain the possibleconfusionthatmay arise
fromviewingtheinterestrate as an indicatoroftightor easy monetarypolicy.
The traditionalexpectationof a negativerelationshipbetweeninterestrates
and the exhangerate may, however,be reconciledwith the asset approach
if it emphasizesthe shortrun liquidityeffectof monetarychanges.Thus, in
the shortrun,a higherinterestrate is due to tightmoneywhichinducesan
appreciationof the currency(Dornbusch, 1976a). During hyperinflation,
however,the expectationseffectcompletelydominatesthe liquidityeffect
resultingtherefore in a predictedpositiverelationshipbetweenthe rate of in-
terestand the priceof foreignexchange.
The previousdiscussionemphasizedthe role of expectationsabout future
eventsin determining the currentvalue of the exchangerate. A major diffi-
cultyin incorporating the roleof expectationsin empiricalworkhas been the
lack of an observablevariablemeasuringexpectations.Thus, forexample,in
analyzingthe demand formoneyduringthe hyperinflation, Cagan (1956) in
his classicstudyconstructeda timeseriesof expectedinflationusinga speci-
fictransformation of the timeseriesof the actual rates of inflation.The con-
ceptualdifficultywithsuchan approachstemsfromthefactthatexpectations
are assumedto be based onlyon past experienceand the choiceofthe specific
transformation used to generatethe seriesof expectationsis to a largeextent
arbitrary.The thirdbuildingblockof the approachinvolvesthe choiceofthe
measureof expectationsthat is appropriateforempiricalimplementation. In
what followswe proposea directmeasureof expectationswhichis then in-
corporatedin the analysisof the determination of the exchangerate.
The fundamentalrelationshipthat is used in derivingthe marketmeasure
of inflationaryexpectationsrelieson the interestparitytheory.That theory
maintainsthatin equilibriumthepremium(or discount)on a forwardcontract
forforeignexchangefora given maturityis (approximately)relatedto the
interestrate differential accordingto:
F-S ._.
S z (5)

whereF and S are theforwardand spotexchangerates(thedomesticcurrency

priceof foreignexchange),respectively, i the domesticrate of interestand id
the foreignrate of intereston comparablesecuritiesforthe same maturity.
Evidence available for various countriesover various time periods suggest
that this parity conditionholds (Frenkel& Levich (1975a, 1975b); foran
earlyanalysisof the 1920's see Aliber(1962)). Although,due to lack ofdata,
Scand. J. of Economics 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 A monetary
approachtotheexchangerate 211

no comparablestudyhas been done on the periodof the Germanhyperinfla-

tion, it is assumedthat the parityconditionhas been maintained.Further-
more,it is reasonableto assume that duringthe hyperinflation most of the
variationsin the difference betweendomesticand foreignanticipatedratesof
inflationwere due to anticipateddomestic (German) inflation.It follows,
therefore, that the variationsof the forwardpremiumon foreignexchange
(F - S)/S, may be viewedas a measureof the variationsin the expectedrate
ofinflation(as well as the expectedrate of changeof the exchangerate).

II.3. The EfficiencyoftheForeignExchangeMarket

Priorto incorporating the forwardpremiumas a measureof expectations,it
is pertinentto explorethe efficiency of the foreignexchangemarketduring
the turbulenthyperinflation period.Evidenceon the efficiency ofthat market
will supportthe approachof usingdata fromthat marketas the basis forin-
ferenceon expectations.
If the foreignexchangemarketis efficientand ifthe exchangerate is deter-
minedin a similarfashionto otherasset prices,we shouldexpectthe behavior
in that marketto display characteristics similarto those displayedin other
stock markets.In particular,we shouldexpect that currentpricesreflectall
available information, and that the residualsfromthe estimatedregression
shouldbe seriallyuncorrelated.
To examinethe efficacyof the marketwe firstregressthe logarithmof the
currentspot exchangerate,log St, on the logarithmof theone-month forward
exchangerate prevailingat the previousmonth,log Ft-,.

log St = a + b log Ft-, +u (6)

The expectationis that the constanttermdoes not differsignificantly

from
zero, that the slope coefficient
does not differsignificantly
fromunityand
that the errortermis seriallyuncorrelated.Since data on the GermanMark-
Pound Sterling(DM/?)forwardexchangerateare availableonlyforthe period
February 1921-August 1923, eq. (2) was estimatedover those31 months.
The resultingordinary-least-squaresestimatesare reportedin eq. (6') with
standarderrorsin parenthesesbelowthe coefficients.

log St =-0.46 + 1.09 log Ft-, (6')

(0.24) (0.03)
1?2= 0.98, s.e. = 0.45; D.W. = 1.90

As can be seen,the constanttermdoes not differ significantlyfromzeroat the

95 percentconfidencelevel (althoughit seemsto be somewhatnegative),the
slope coefficient
is somewhatabove unity(at the 95 percentconfidencelevel)
and, most importantly, the Durbin-Watsonstatisticsindicatesthat the re-
sidualsare not seriallycorrelated.The factthatthe slope coefficient
is slightly
Scand. J. of Economic8 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
212 J. A. Frenkel
DATE

2101

2104

2107 _
_ - LOG EXCHANGE RATE (SPOT)
2110

2201_ v

2204 _

2207

2210

2301 LOG FORWARDRATE ( )

2304 _

2307

Fig. 3.

above unitymay be explainedin termsof transactioncostsor in termsof the

Keynesianconceptof normal-backwardation (Keynes,1930,vol. II, p. 143).1
Fig. 3 describesthe monthlytimeseriesplot of the logarithmsof the spot
exchangerate and the forwardrate prevailingin the previousmonth.The
generalpatternrevealsthat typically,whenthe spot exchangerate rises,the
forwardrate lies belowit whilewhenthe spot exchangerate fallsthe forward
rate exceeds it. This patternis also suggestedby the fact that the elasticity
ofthe spotratewithrespectto the previousmonth'sforwardrateis somewhat
above unity.
To explorefurtherthe implicationsof the efficient markethypothesiswe
examinewhetherthe forwardexchangerate summarizesall relevantinforma-
tion. In an efficientmarketFt-, summarizesall the information concerning
the expectedvalue of St that is available at periodt-1. Specifically,one of
the itemsofinformation available at t-1 is the stockofinformation available
at t-2, and if the marketis efficient, that information will be containedin
Ft-2. If howeverFt-, summarizesall available information includingthat
containedin Ft-2, we shouldexpectthat adding Ft-2 as an explanatoryvari-
able to the right-hand side of (6) will not affectthe coefficientof determina-
tion and will have a coefficient that is not significantlydifferent fromzero.
Eq. (6") reportsthe resultsof that regression:
log St =-0.45 + 1.10 log Ft, -0.006 log Ft-2 (6")
(0.26) (0.08) (0.08)
R2 = 0.98, s.e. = 0.46; D.W. = 1.91
The resultsin (6") supportthe efficient
markethypothesis.
1 The joint hypothesis that the constant term is zero and that the slope coefficientis
unity is rejected at the 95 percent confidencelevel.

Sand. J. of Economico 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 A monetary
approachtotheexchangerate 213

To examinefurtherthe stabilityof the regressioncoefficients duringthe

various phases of the hyperinflation we divided the sample into two parts:
"Moderate" hyperinflation and "severe" hyperinflation, where the latter
characterizedthe last eightmonthsof the sample period.A Chow test was
performed on the estimatesof equations (6') and (6") to test the equalityof
each and everycoefficient of the two sub-period'sregressions. This procedure
showedthat the hypothesisthat the regressioncoefficients do not differbe-
tweenthe two sub-periodscannotbe rejectedat the 95 percentlevel.
The resultsreportedin thissectionprovidesupportto the notionthat dur-
ing the hyperinflation expectationsmay have behaved "rationally"in the
Muthsense.In fact,it shouldnot be surprising that even duringthe turbulent
periodofthehyperinflation theefficientmarkethypothesiscannotbe rejected.
It standsto reasonthatthelargervariabilityofexchangeratesincreasestherate
of returnfromand the amountof resourcesinvestedin accurateforecasting.'
II.4. Prices,Moneyand Expectations
In this sectionwe examinethe empiricalcontentof the firstbuildingblock
of the monetaryapproachto the exchangerate by usingthe marketmeasure
of inflationaryexpectationsin estimatingeq. (2). Log-linearizing
eq. (2) and
addingan errortermyields:
logP = a + b,log M + b2loga* + u (2')
One of the difficultiesof using the double-logformis that duringsome
months,earlyin the sampleperiod,the forwardpremiumon foreignexchange
was negative(reaching-0.8 percentper monthin early 1921) reflecting the
initial expectationthat the price rise has been temporary,that the process
will reverseitselfand priceswillreturnto theirpreviouslevels.Sincethelog-
arithmof a negativequantityis not defined,the independentvariable was
transformed fromn* to (k+a*) whichhenceforth is referred
to as a. Thus,the
estimatedequationwas:
logP =a'+b' log M+b' logn +u. (2")
A maximumlikelihoodestimationof the coefficients in (2") along with the
value of k resultedin a value of k rangingbetween0.9 and 1.1 percentper
month.2For ease of exposition,in what followswe set the value ofk at 1 per-
centper month,and thusthe coefficient b2= b'x*I(l +a*).
1 For an application of the "rational"
expectations hypothesisto the German hyperinfla-
tion see Sargent & Wallace (1973).
2
While the anticipated differencebetween the domestic and the foreignrates of inflation
that is proxied by the forwardpremiumn* may be negative, the nominal rate of interest
may not. In principle,a relevant variable in the demand formoney is "the" nominal rate
ofinterestwhichfrom(5) is equal to a"*+ i*. Our estimateofk, therefore,is not unreasonable
in proxying "the" foreignnominal rate of interest.Since the purpose of this section is to
indicate, in somewhat general terms,the implicationof the firstbuilding block ratherthan
to provide a precise estimate of the parameters of the demand for money, the procedure
that was followed seems justifiable. In a separate paper we apply the market measure of
expectations to a more detailed reexamination of the functionalformand the estimates
of the demand formoney duringthe German hyperinflation(Frenkel, 1975).

Scand. J. of Economic8 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
214 J. A. Frenkel

Table 2. Prices,moneyand expectations.

Monthlydata: February1921-August
1923
Estimated equation: Log P = a + b1log M + b. log nr+ u

Dependent
variable Constant log M log n S.e. R2 D.W. e u

LWPI -5.983 1.021 0.497 0.193 0.996 1.88 0.525 0.227

(0.611) (0.041) (0.061)
LWIG - 5.423 0.997 0.293 0.144 0.998 1.95 0.909 0.366
(0.779) (0.041) (0.053)
LWHG - 4.923 0.983 0.374 0.187 0.996 1.79 0.889 0.429
(0.925) (0.051) (0.069)
LCOL -7.215 1.073 0.236 0.125 0.998 2.13 0.729 .0182
(0.450) (0.029) (0.044)
LWAG - 10.027 1.103 0.194 0.254 0.992 2.06 0.373 0.274
(0.749) (0.051) (0.074)

Note: LWPI =log wholesale price index, LWIG = log imported-goodsprice index, LWHG =
log home-goods price index, LCOL =log cost of living index, LWAG =log wage index.
Standard errorsare in parentheses below each coefficient.e is the final value of the auto-
correlation coefficient.An iterative Cochran-Orcutt transformationwas employed to
account for firstorder serial correlationin the residuals. s.e. is the standard error of the
equation and au is the standard errorof the regressionwhen the autoregressivecomponent
of the erroris included.

It may be recalledthat in postulatingeq. (2) we bypassedthe questionof

the appropriateprice deflator.To a largeextent that questionis empirical
and dependson the theorythat underliesthe derivationof the demand for
money.The presumption, however,is that ifthe aggregatedemandformoney
is dominatedby householdbehavior,thenthe relevantpriceindex shouldbe
the consumerpriceindex (the cost of living).Table 2 reportsthe resultsof
estimatingeq. (2w)usingalternativepricedeflators.Judgedby the goodness
of fit it seemsthat the cost of livingindex is the most appropriatedeflator
although,strictlyspeaking,such a comparisonis insufficient sincethe various
equationsin Table 2 differ in the dependentvariable.
Judgedas a wholeit seemsthattheresultsin Table 2 are consistentwiththe
firstbuildingblock of the monetaryapproach to the exchangerate. In all
cases the elasticityof the pricelevel withrespectto the moneystockis close
to unityand the elasticitywithrespectto a is positiveas predictedfromthe
consideration ofthe demandformoney.

II.5. The PurchasingPowerParity

The secondbuildingblockofthe theoryof exchangerate determination
is the
purchasingpowerparity.The highstatisticalcorrelationbetweenpricesand
the exchangerate typicallyobservedduringperiodsofmonetarydisturbances
led to the developmentof the purchasingpowerparitydoctrine.Figs. 4, 5
Scand. J. of Economics 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 approachtotheexchangerate 215
A monetary

DATE
2002 -

2005 f
2008 - i_ LOG WHOLESALE PRICE INDEX
_008
2011

2102 _

2105 _

2108 -

2111 - - LOS EXCHANGERATE

2202

2205 -

2208 -

2211

2302

2305 _

2308 _

2311

Fig. 4.

DATE
2002

200S_
_ - LOG COST OF LIVING INDEX
2008 _

2011

2102 _

2105 _

2108 _

2111

2202 -

2205

2208-
2211_E
2211 LOG EXCHANGE RATE

2302 -

2305 _

2308 _

2311

Fig. 5.

Sand. J. of Economic8 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
216 J. A. Frenkel
DATE

2002

2005_

2008

2011

2102 ..-LOG WAGE

2I 05

2108 _
2111

2202-

2205

2 208
LOG EXCHANGE RATE
22110l

2302

2305

2308

2311

Fig. 6.

and 6 show the relationshipsbetween(DM/$) exchangerate and the whole-

sale priceindex,the cost of livingindex and the wage rate index, and Figs.
7-8 show the relationshipsbetweenthe percentagechange of the exchange
rate and prices.These relationshipscorrespondto the various price indices
advocated forthe computationof purchasingpowerparities.'In view of the
observedhighcorrelationamongpricesand the exchangerate,even the skep-
tics agreedthat the doctrinemay possess an elementof truthwhen applied
to monetarydisturbances(e.g., Keynes (1930, p. 91), Haberler (1936, pp.
37-8), Samuelson(1948,p. 397)).
Table 3 reportsthe resultsof estimatesof the purchasingpowerparityfor
the alternativepriceindices.The estimatedequationsare derivedfromequa-
tion (3) by log-linearizing
and assumingthat duringthat periodP* couldbe
viewed as being fixed.As can be seen this buildingblock of the theoryof
exchangerate determination also standsup ratherwell.In all cases the elasti-
citiesof the exchangerate withrespectto the variouspriceindicesare very
closeto unity.It also seemsthat whilethe cost oflivingindexmay be the ap-
1 In addition to pure monetarydisturbancessome sharp turningpoints in the path of the
exchange rate (and its rate of change) can be attributedto political events whichcreated
facts and affectedexpectations. The followingoutlines some of the critical events that are
reflectedin the sharp turningpoints (based on Tinbergen (1934)). August 29, 1921: Murder
of Erzberger; October 20, 1921: League of Nations decision concerningpartition of Upper
Silesia-renewed disturbances. On that date there was the Wiesbaden agreementbetween
Rathenau and Loucheur concerningdeliveries in Rind. April 16, 1922: Rapallo Treaty
with Russia. June 24, 1922: Murder of Rathenau leading to a heavy depreciation of the
mark. January 11, 1923: Ruhr territoryoccupied by French. End of February 1923: Be-
ginningaction to support the mark. April 18, 1923: Collapse of supportingmeasures.

Sand. J. of Economic8 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 approachtotheexchangerate 217
A monetary
DATE

2102 -

2105

2108- .- _PERCENTAGE CHANGE IN

SPOT EXCHANGE RATE
2111

2202

2205 -

2208 *

2211

2302 _

2305
PERCENTAGE CHANGE
2308- IN WHOLESALE PRICE INDEX

Fig. 7.

propriatedeflatorin estimatingmoneydemandfunctions, the wholesaleprice

index performs best in the purchasingpowerparityequations.'
There remains,however,a questionof interpretation. Does the doctrine
specifyan equilibriumrelationshipbetweenpricesand exchangeratesor does
it,in addition,specifycasual relationships
and channelsoftransmission.While
the highcorrelationbetweenthe variouspriceindicesand the exchangerate
is of someinterestin describingan equilibriumrelationshipor in manifesting
the operationof arbitragein goods (dependingon the priceindex used), they
are oflittlehelpin explainingand analyzingthe determinantsof the exchange
rate.

DATE

210 2

2105

2108 PERCENTAGE CHANGE IN

7SPOT EXCHANGE RATE
2111

2202

2205

2208 _

2211

2302 -
2305

2308 PERCENTAGE CHANGE

COST OF LIVING INDEX

Fig. 8.

1 It is of interestto note that duringthe recentfloat (1973-74) the percentagedeviation of

the wholesale price index frompurchasing power parity exceeded that of the consumer
price index (Aliber (1976)).

Sand. J. of Economics 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
218 J. A. Frenkel

Table 3. Exchangerateand prices.Monthlydata: February1921-August 1923

Estimated equation: log S = a + b log P + u

Independent
variable Constant log P s.e. B2 D.W. e au

LWPI 0.146 1.006 0.124 0.998 2.01 0.356 0.135

(0.114) (0.010)
LWIG -0.219 1.058 0.208 0.996 2.01 0.269 0.216
(0.177) (0.017)
LWHG -0.383 1.031 0.215 0.995 2.09 0.471 0.241
(0.244) (0.022)
LCOL 0.115 1.076 0.273 0.993 1.97 0.499 0.325
(0.311) (0.030)
LWAG 4.415 0.887 0.350 0.988 1.94 0.889 0.767
(0.788) (0.070)
LWAG 2.682 1.074 0.360 0.987 1.66 0.471 0.414
2SLS (0.310) (0.038)

Note: LWPI =log wholesale price index, LWIG - log imported-goodsprice index, LWHG =
log home-goods price index, LCOL =log cost of living index, LWAG =log wage index.
Standard errorsare in parenthesesbelow each coefficient.e is the final value of the auto-
correlation coefficient.An iterative Cochran-Orcutt transformationwas employed to
account for firstorder serial correlationin the residuals. s.e. is the standard error of the
equation and au is the standard errorof the regressionwhen the autoregressivecomponent
of the erroris included. To allow fora possible simultaneousequation bias due to the endo-
geneity of the various prices the above equations were also estimated using a two-stage
least squares procedure with the percentage change in the money supply and the money-
bond ratio as instruments.None of the coefficientswas significantlyaffectedexcept for
the equation using LWAG as the independentvariable. The 2SLS estimates are reported
in the last line of the Table.

11.6. The Determinants

oftheExchangeRate
The two buildingblocks analyzed in the previoussectionsprovide the in-
gredientsto the estimationof the determinants of the exchangerate. Given
the foreignpricelevel the purchasingpowerparitydeterminesthe ratioP/S.
Giventhe nominalmoneystockand the state of expectations,the pricelevel
is determinedso as to clearthe moneymarket.These two relationships imply
the equilibriumexchangerate.We turnnow to the estimationofthe emprical
counterpart of eq. (4). Log-linearizing
and addingan errortermyields
log S = a' +b1log M +b1loga +u (4')

whereas before,a 1 + *. The estimatesare reportedin eq. (4") withstandard

errorsbelowthe coefficients:
log S -5.135 + 0.975 log M + 0.591 logo (4")
(0.731) (0.050) (0.073)
R2 = 0.994;s.e. = 0.241;D.W. = 1.91.

As is evidenttheseresultsare fullyconsistentwiththe priorexpectations.

Sand. J. of Economics 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 A monetary
approachtotheexchange
rate 219

The elasticityof the exchangerate withrespectto the moneystockdoes not

differsignificantlyfromunity (at the 95 percentconfidencelevel) whilethe
elasticityof the spot exchangerate with respectto the forwardpremiumis
positive.The orderof magnitudeof the latterelasticityis similar(in absolute
value) to theinterestelasticityofthe demandformoney.'In comparisonwith
the polynomialdistributedlag of Table 1 it is seen that the standarderrorof
equation (4f) is significantlysmaller.It is also noteworthythat the lower
elasticityof the exchangerate withrespectto the moneystockis consistent
with the intuitiveexplanationprovidedto the high elasticityof Table 1.
Thereit was arguedthat the magnification effectwas due to the role of ex-
pectations.Indeed as is shownin eq. (4f) whenexpectationsare includedas
a separatevariable,the homogeneity postulatereemerges, the magnification
effectof the moneystock disappearsand thusindicatingthat the equations
in Table 1 mighthave been misspecified.

11.7. Conclusions,
Limitationsand Extensions
The foregoinganalysis examinedthe empiricalrelationshipsamong money,
prices,expectationsand the exchangerate duringthe Germanhyperinflation.
Concentrating on that periodprovidedthe opportunity to isolate empirically
some of the key relationshipsrelevantto exchangerate determination. In
particular,special attentionhas been given to simultaneousroles played by
expectationsand by monetarypolicyin determining the exchangerate. The
empiricalresultsare consistentwiththe monetary(or the asset) approachto
the exchangerate.
It shouldbe emphasizedthat the monetaryapproachto the exchangerate
doesnotclaimthat the exchangerate is determinedonlyin the money(orthe
asset) marketand that only stock considerations matterwhileflowrelation-
shipsdo not. Clearly,the exchangerate (likeany otherprice)is determined in
generalequilibriumby the interactionof flow and stockconditions.In this
respectthe asset marketequilibriumrelationshipthat is used in the analysis
may be viewedas a reducedformrelationshipthat is chosenas a convenient
framework.
Concentration on the periodofthe hyperinflationhas, however,someshort-
comings.First it does not provideany insightinto the exchangerate effects
of real disturbanceslike structuralchanges(see forexample Ballasa (1964);
Hekman (1975)). Second, and probablymore important,the rapid develop-
ments occuringduringthe hyperinflation preventeda detailed analysis of

1 Recall that due to the transformationon the independentvariable the (average) interest
elasticityof the exchange rate b2 is b'n*l(1 +?r*) wheren* is the average forwardpremium.
Over the sample period the average a* was about 6.2 percentper month,yieldingtherefore
an estimate of about 112 as the interestelasticity. This estimate of the elasticity is con-
sistent with the estimates in Frenkel (1975) as well as with the predictionsof the various
models of the transactions demand for cash.

Scand. J. of Economic81976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
220 J. A. Frenkel

Table 4. Correlationmatrix:prices,exchangerate and money.Monthlydata

February1920-November1923

LWPI LWIG LWHG LCOL LWAG LEXC LMON

LWPI 1.000 0.9986 0.9985 0.9959 0.9969 0.9992 0.9933

LWIG 1.000 0.9956 0.9934 0.9947 0.9968 0.9942
LWHG 1.000 0.9945 0.9949 0.9987 0.9892
LCOL 1.000 0.9984 0.9956 0.9875
LWAG 1.000 0.9960 0.9927
LEXC 1.000 0.9850
LMON 1.000

Note: LWPI =log wholesale price index, LWIG =log import-goodsprice index, LWHG =
log home-goods price index, LCOL = log cost of living index, LWAG = log wage index,
LEXC =log exchange rate index, LMON =log money supply index.

the channelsof transmission of disturbancesamongthe varioussectorsin the

economy.For example,it mightbe usefulto examinethe exact patternand
chronologicalorder by which monetarydisturbancesget transmittedinto
changesin the various priceindices. The monthlydata used in the present
paper do not permitsuch a detailed analysissince most of the dynamicsof
adjustmentoccur withinthe month.The extentof this phenomenonis re-
flectedin Table 4-a correlation matrixofthevariousvariablesforthemonthly
data over the entireperiod(February1920-November1923).To gaininsight
into the morerefineddetails of the adjustmentprocess,it may be necessary
to analyze the period usingweeklydata. A preliminary examinationof the
weeklydata suggeststhat the variouspricesdo differin the details of their
timepaths but at thisstageno conclusiveevidencecan yet be offered.'
Althoughthe monthlydata do not revealthe detailsofthe adjustmentpro-
cess, theydo reveal some systematicrelationshipsamongthe coefficients of
variationofthe (logarithms ofthe) variablesas reportedin Table 5. As is seen
in Table 5, the coefficient
of variationof the moneystockis about 0.15 while
the coefficients of variationcorresponding to the various price indices are
about twiceas large-about 0.30. In this respectall priceindices (wholesale,
imported-goods, home-goodsand cost of living)displaya commonbehavior.
A thirdgroupof variablesincludesthe variousexchangerates (spot and for-
ward) and the wage rate. All of thesevariablesdisplaya similarcoefficient of
variation-about 0.40. The interestingphenomenaare that the extent of
variationsin the various exchangerates (and in the wage rate) exceeds the
extentof variationsin the variouspriceswhichin turnexceedsthe variation
in the moneystock.Furthermore, in view of the wage rate approachto the
exchangerate (some of the doctrinaloriginsof whichwerementionedin Sec-
tion I), the associationamong variationsin the wage rate and the various
1 An inspection of Figs. 7-8 reveals that changes in the cost of living index
lag behind
changes in the wholesale price index which are closely related to changes in the exchange
rate.

Scand. J. of Economic8 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
 approachto theexchangerate 221
A monetary

Table 5. Summarystatistics:prices,exchangerate and money.Monthlydata

February1921-August1923

Standard Coef.of
Variable Mean Variance deviation variation

LMON 15.3567 5.0948 2.2571 0.1469

LWPI 10.0477 9.5613 3.0921 0.3077
LWIG 9.8978 8.6178 2.9356 0.2965
LWHG 10.3210 9.1950 3.0323 0.2938
LCOL 9.4337 7.9337 2.8166 0.2985
LWAG 7.0522 7.6616 2.7679 0.3924
LEXC 8.6235 10.3370 3.2151 0.3728
LSPO 8.6235 10.3370 3.2151 0.3728
LFOR 8.6853 11.0123 3.3184 0.3820

Note: LMON =log money supply index, LWPI =log wholesale price index, LWIG =log
import-goodsprice index, LWHG log home-goods price index, LCOL =log cost of living
index, LWAG =log wage index, LEXC =log (DM/$) spot exchange rate index, LSPO =
log (DM/2) spot exchange rate index, LFOR =log (DM/I) one month forwardexchange
rate index.

exchangerates deservesa special notice.While a detailedanalysisof the im-

plicationsof Table 5 is beyondthescopeofthepresentpaper,it is ofinterestto
notetheassociationamongtheexchangeratesand thepriceoflaborservices-
the commoditywhichmay be mostnaturallyclassifiedas a non-tradedgood.

Appendix: Data Sources

Data on the DM/$ exchangerate (spot) are taken fromGraham(1930) and

fromInternationalAbstractof EconomicStatistics1919-30. London: Inter-
nationalConference of EconomicServices,London, 1934.
The one-monthforwardexchangerate (DM/Y,)as well as the (DM/f) spot
rates are fromEinzig (1937). The primarysourceforthis data is the weekly
circularpublishedby the Anglo-Portuguese Colonialand OverseasBank, Ltd.
(originallythe London branch of the Banco Nacional Ultramarinoof Lis-
bon). The rates quoted are thoseof the Saturdayof each week,but in cases
wherethe marketwas closed, the latest quotation available prior to that
Saturdayis used.
Data on moneysupplyare fromGrahamand HistoricalStatisticsas well
as some interpolations.Prices and wages are fromHistoricalStatisticsand
someinterpolations and primarysources.
OutstandingTreasury-Bills are fromGraham.

Scand. J. ofEconomics 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions
222 J. A. Frenkel
References
Aliber, R. Z.: Speculation in the foreignex-
changes: The European experience, 1919-
1926. Yale Economic E88ays, Spring 1962.
Aliber, R. Z.: The firm under fixed and
flexible exchange rates. Scandinavian
Journal of Economic878, No. 2, pp. 309-
322, 1976.
Angell, J. W.: International trade under in-
convertible paper. QuarterlyJournal of
Economic 36, 309-412, 1922.
Balassa, B.: The purchasing power parity
doctrine: a reappraisal. Journal ofPoliti-
cal Economy 72, No. 6, 584-96, 1964.
Bilson, J. F.: Rational expectations and
flexible exchange rates. Unpublished
manuscript,Universityof Chicago, 1975.
Black, S. W.: International money markets
and flexible exchange rates. Princeton
Studie8 in InternationalFinance, No. 32.
Princeton University,1973.
Besciani-Turroni,C.: The purchasingpower
parity doctrine. Egypte Contemporaire,
1934. Reprintedin his Saggi Di Economia,
1961, Milano, pp. 91-122.
Bresciani-Turroni,C.: The Economic8of In-
flation.Allen & Unwin, London, 1937.
Bunting, F. H.: Purchasing power parity
theory reexamined. Southern Economic
Journal 5, No. 3, 282-301, 1939.
Cagan, P.: The monetary dynamics of
hyperinflation. In M. Friedman (ed.),
Studie8 in theQuantityTheoryof Money.
University of Chicago Press, Chicago,
1956.
Cannan, E.: The application of the theore-
tical apparatus of supply and demand to
units of currency. Economic Journal 31,
No. 124, 453-61, 1921.
Cassel, G.: The present situation of the
foreignexchanges. Economic Journal 26,
62-65, 1916.
Cassel, G.: "Comment", Economic Journal
30, No. 117, 44-45, 1920.
Cassel, G.: The World'8MonetaryProblem8.
Constable and Co., London, 1921.
Cassel, G.: Po8t-WarMonetaryStabilization.
Columbia University Press, New York,
1928.
Scand. J. of Economics 1976
This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM
All use subject to JSTOR Terms and Conditions
Cassel, G.: Money and Foreign Exchange
after1919. Macmillan, London, 1930.
Dornbusch, R.: "Discussion". American
Economic ReviewPapers and Proceedings,
147-151, 1975.
Dornbusch, R.: The theory of flexible ex-
change rate regimes and macroeconomic
policy. Scandinavian Journal of Econo-
mics 78, No. 2, pp. 255-275, 1976.
Dornbusch, R.: Capital mobility, flexible
exchangerates and macroeconomicequili-
brium. In E. Claassen and P. Salin (eds.),
Recent lumes in InternationalMonetary
Economics. North-Holland, 1976.
Einzig, P.: World Finance, 1914-1935.
Macmillan & Co., New York, 1935.
Einzig, P.: The Theory of Forward Ex-
change.Macmillan, London, 1937.
Ellis, H. S.: The equilibrium rate of ex-
change. In Explorations in Economics.
Notes and Essays contributedin honorof
F. W. Taussig. McGraw-Hill, New York,
1936.
Frenkel, J. A.: Adjustmentmechanismsand
the monetaryapproach to the balance of
payments. In E. Claassen and P. Salin
(eds.), Recent Issues in International
Monetary Economics. North-Holland,
1976.
Frenkel, J. A.: The forwardexchange rate,
expectations and the demand for money
during the German hyperinflation.Un-
published manuscript, University of
Chicago, 1975.
Frenkel,J. A. & Johnson,H. G.: The mone-
tary approach to the balance of pay-
ments: essential concepts and historical
origins. In J. A. Frenkel and H. G.
Johnson (eds.), The MonetaryApproach
to the Balance of Payments. Allen &
Unwin, London and University of Tor-
onto Press, Toronto, 1976.
Frenkel, J. A. & Levich, R. M.: Covered
interest arbitrage: unexploited profits?
Journal of Political Economy 83, No. 2,
325-338, 1975a.
Frenkel, J. A. & Levich, R. M.: Transac-
tions cost and the efficiencyof inter-
 A monetary
approachtotheexchange
national capital markets. Presented at
the Conferenceon The Monetary Mech-
anism in Open Economies, Helsinki,
Finland, August, 1975b.
Frenkel, J. A. & Rodriguez, C. A.: Portfolio
equilibrium and the balance of pay-
ments: A monetary approach. American
Economic Review 65, No. 4, 674-88, 1975.
Friedman, M. & Schwartz, A.: A Monetary
History of the United States, 1867-1960.
Princeton University Press, Princeton,
1963.
Goschen, G. J.: The Theoryof the Foreign
Exchanges. lst ed. London, 1861; 2nd ed.
London, 1863; 4th ed. reprinted,1932.
Graham, F.: Exchange Prices and Produc-
tion in Hyper-Inflation:Germany,1920-
23. Princeton University Press, Prince-
ton, N.J., 1930.
Gregory, T. E.: Foreign Exchange before,
duringand afterthe War. OxfordUniver-
sity Press, London, 1922.
Haberler, G.: The Theory of International
Trade. William Hodge and Co., London,
1936.
Haberler, G.: The choice of exchange rates
after the war. American Economic Re-
view 35, No. 3, 308-318, 1945.
Haberler, G.: A surveyof internationaltrade
theory.Special Papers in International
Economics 1 (July 1961). International
Finance Section, Princeton University.
Hansen, A. H.: A briefnote on fundamental
disequilibrium. Review of Economics and
Statistics26, No. 4, 182-84, 1944.
Hawtrey, R. G.: Currency and Credit.
Longmans, Green and Co., 1st ed. 1919,
4th ed. 1950, London.
Hekman, C. R.: Structural change and the
exchange rate: an empirical test. Un-
published manuscript, University of
Chicago, 1975.
Heckscher, E. F., et al.: Sweden, Norway,
Denmark and Iceland in the World War.
New Hawen, 1930.
Hodgson, J. S.: An analysis of floatingex-
change rates: The dollar sterling rate,
1919-1925. Southern Economic Journal
39, No. 2, 249-257, 1972.
Houthakker, H. S.: Exchange rate adjust-
ment. Factors Affectingthe U.S. Balance
This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM
All use subject to JSTOR Terms and Conditions
rate 223
ofPayment. JointEconomic Committee,
87th Congress,2nd Session, December 14,
1962, pp. 289-304.
Johnson, H. G.: The Keynesian revolution
and the monetarist counterrevolution.
American Economic Review 61, No. 2, 1-
14, 1971. Reprinted in Ch. 7 in H. G.
Johnson, Economic8and Society.Univer-
sity of Chicago Press, Chicago, 1975.
Johnson, H. G.: Theory of international
trade. In International Encyclopedia of
the Social Science&.The Macmillan Com-
pany and Free Press, 1968.
Johnson, H. G.: World inflation and the
international monetary system. The
ThreeBank8 Review,No. 107, 3-22, 1975.
Keynes, J. M.: A tract on monetaryreform.
1st ed. 1923, French Edition, 1924; vol.
IV in The Collected Writingsof J. M.
Keynes. Macmillan, London, 1971.
Keynes, J. M.: A Treati8eon Money. Vol. I.
Macmillan, London, 1930.
Kouri, P. J. K.: Exchange rate expecta-
tions, and the short run and the long run
effects of fiscal and monetary policies
under flexible exchange rates. Presented
at the Conferenceon The MonetaryMech-
anism in Open Economies, Helsinki, Fin-
land, August 1975.
Lursen, K. & Pedersen, J.: The GermanIn-
flation 1918-1923. North-Holland, Am-
sterdam, 1964.
Marshall, A.: Memorandum to the Effect8
which Differencesbetweenthe Currencies
of DifferentNations have on International
Trade, 1888.
Marshall, A.: Money, Creditand Commerce.
London, 1923.
Metzler,L. A.: The process of international
adjustment under conditions of full em-
ployment: a Keynesian view. In R. E.
Caves and H. G. Johnson(eds.), Readings
in International Economic8, pp. 465-86.
Irwin, Homewood, Ill., 1968.
Mill, J. S. Principles of Political Economy.
5th ed. Parker & Co., London, 1862.
Mundell, R. A.: Monetary Theory. Pacific
Palisades, Goodyear, 1971.
Mundell, R. A.: International EconomiC8.
Macmillan, New York, 1968.
Mussa, M. L.: A monetary approach to ba-
Sand. J. of Economica 1976
224 J. A. Frenkel

lance of payments analysis. Journal of Rueff,J.: Balance of Payments. Macmillan,

Money, Credit and Banking 6, No. 3, New York, 1967.
333-351, 1974. Rueff,J.: Les fondements philosophiquesdes
Mussa, M. L.: The exchange rate, the ba- 8yst8meseconomiques.Payol, 1967.
lance of payments and monetary and Samuelson, P. A.: Disparity in postwar
fiscal policy under a regime of controlled exchange rates. In S. Harris (ed.),
floating. Scandinavian Journal of Econo- Foreign Economic Policy for the United
mics 78, No. 2, pp. 229-248, 1976. States, pp. 397-412. Harvard University
Myhrman, J.: Experiences of flexible ex- Press, Cambridge, 1948.
change rates in earlier periods: theories, Samuelson, P. A.: Theoreticalnotes on trade
evidence and a new view. Scandinavian problems. Review of Economics and
Journal of Economic-878, No. 2, pp. 169- Statistics46, No. 2, 145-154, 1964.
196, 1976. Sargent, T. J. & Wallace, N.: Rational
Ohlin, B.: Interregionaland International expectations and the dynamics of hyper-
Trade. Revised ed., Harvard University inflation. InternationalEconomic Review
Press, Cambridge, Mass., 1967; 1st ed. 14, No. 2, 328-50, 1973.
1933. Taussig, F. W.: InternationalTrade. New
Pigou, A. C.: Some problems of foreignex- York, 1927.
changes. Economic Journal 30, No. 120, Tinbergen, J. (ed.): International Abstract
460-472, 1920. of Economic Statistics 1919-30. Inter-
Ricardo, D.: The high price of bullion. national Conferenceof Economic Servi-
London, 1811. In Economic Es8ay8 by ces, London, 1934.
David Ricardo. Edited by E. C. Conner. Tsiang, S. C.: Fluctuating exchange rates in
Kelley, New York, 1970. countrieswithrelativelystable economies
Ricardo, D.: Reply to Mr. Bo0aquet'spracti- IMP StaffPapers 7, 244-273, 1959-60.
cal observationon thereportoftheBullion Viner, J.: Studies in the Theory of Inter-
Committee.London, 1811. In Economic national Trade. Harper and Bros., New
Es8ay8 by David Ricardo. Edited by E. C. York, 1937.
Connor. Kelley, New York, 1970. Wheatley, J.: Remarks on Currency and
Ricardo, D.: Principle8ofPolitical Economy Commerce.Burton, London, 1803.
and Taxation. London, 1821. Edited by Wicksell, K.: The riddle of foreign ex-
E. C. Connor. G. Bell and Sons, 1911. changes. 1919. In his SelectedPapers on
Ringer, F. K. (ed.): The German Inflation Economic Theory. Kelley, New York,
of 1923. Oxford Press, New York, 1969. 1969.
Robinson, J.: Banking policy and the ex-
changes. Review of Economic Studie8 3,
226-29, 1935-36.

Sand. J. of Economica 1976

This content downloaded from 192.231.202.205 on Wed, 26 Nov 2014 03:40:22 AM

All use subject to JSTOR Terms and Conditions