RATE
John F.O. Biison’@ twvu~ity ofcMi?tqfo
IJlQodPctron In 19% a I~MK~M mark a~ approximately 24 CmncY+
T&s
&
WM hcmmtd
cents in U.S.
bY one cent in March 1961
and then remained at 25 cm4 UN Octobca 1969. Under the Brettor: Woods agreement, the price of a Deu*he mark was faed by the Bundesbank*s commitment to PUNan dolloffered at any price above 25.225 cents and to sell drjlfars at any price below 24.75 cents. For most of tEtetwenty-year period, this was not an onelxy)uscommitment. The Federal Eepublic of Germany (hereafter, simply Germany) experienced p rsistznt surp~~!s in its balance of payments, which permitted the Bundesbankto accumulate a desired increase ?:nits stock of international reserves, while simultaneously lowering batrriers to trade in goods and capital. In addition, the early sixties were a period of prize stability in thz United States, and the German authoiities couli WV on the fixed exchangerate to enhancethe stability of their own prices and costs.. However, by the late 196Os, the importation of t. .S. inflation into Germany was beginning to be a matter of political concern. h its attempts to maintain the dollar parity, the Bundesbank increased the morqtary base at a rate that was inconsistent with domestic inflation targets. The i::onffict between the external policy of maintaining the parity with the dolla:- and the internal policy of maintz ining stable prices became particularly sevew in 1969. At the end of Gpril 1959, the &.Fndesbank purchased 2.5 billion df::lllars,an amount that can be transhlted in@y an 18 percent increase in the monetary base, in their un(;uccessfM Jefe .~s.eaf the parity. In October 1969, the German alitho+ ties dcc,ided that price stability deserved greater priority than exchange rate stability,and the exchange r,itc ~~~srevalued by nine percent.’ In retrospect, the ~cto’3ta revaluation marked the end oftbe Wentyyear period of statijity in file DM;‘$ rate* The appreciation did not ste,m the flow of dollars into the Buv~dcsbank and, in an attemp+ to find a new Wilibrium level, the rate was a...owed to float from. May to hcembcr 19’71. A e ~rt&Aly rsknwledged. alw, likt~ to thaaonkR.ichrrcl
59
fwher appre&Mn was ne@@ed at- the S@@onian meetings in December 1971, but this was also ‘not sufficient to dampen the demand for Deutsche marks. After further spec:ulatiie attacks, the Smithsonian parities were abandoned in March 1973. During the subsequent fnre years of flexible exchange r&es, the Deutsche mark has continued to appreciate against the do&u. By September 1978, the price of a Deutsche mark had risen to 52 cents, This paper is an attempt to vide an analytical history of the DM/S rate. The attempt includes nondefinitive answers to the following questions: 1. Why did the Bretton Woods syf~~rncollapse in 19691 2. Why has the DM/$ exchange rate increased so rapidly since 1969? 3. Why has the 1depreciationof the dollar tended to be erratic and unpredictable? 4. What is the role of private speculation in the determination of the exchange rate? 5. Does official intervention work? 6. Have floating rates given national governments the freedom of action they expected from them?
The answ’ers to these questions involve a combination of economic theory, institutional detail, and empirical evidence. In the first section of the paper, I shall describe a theoretical model which I believe to be a logically consistent and believeable description of the underlying determinants of the exchange rate. In the second section, the history of the DM/$ rate is reviewed in the light of this theory. Finally, some empirical tests of the model are presented,and an attempt is made to formally attribute the change in the exchange rate to the factors discussed in the previous section. A Monetary Theory of the Fmhange Rate The characteristic feature of any “monetary” model 2 is the assumptian that the demand for money is a stable function of a limited number of aggregate economic variables. In addition, monetary models invariably assume that the! money demand function is homogenous of the first degree in the nominal variables. Homogeneity is normally imposed by expressing the demand function. in relative terms. Fsr example, the quantity theory equation
)+ =K(i,
Y,. . .)
relates the demand for money (M) relative to nominal incor;ke(P Y) to the set of variables specified in the function K(i, Y, ...) , Standard aguments in the function include the nominal rate of interest, i, and the level of red inmm% Y. ‘Ihe nominal rate of in:erest is included as a measure of the holding cost Of
60
money time to other asWs, and the level of real income is inchrded as a measure of the vdume of transrtctions or as a measure of real wealth. The quantity theory formuhtion, with its emphasis on the demand for money relative to income, is particularly usefuf for explaining movements in keW4he price _of#oq$$ in terms of money. Alternatively, if the price of boYds ot interest rate is ths.fo#s of the analysis, it may be more useful to ertprez~t$e money dernmd function 311terms of the demand for money relative to bonds or ~tal wealth. The important point to note is that the particular form of the money demand functicn cannot be specified on theoretical grounds; it depends upon t8reparticular relaoive price that is under study. The faus of the preseat study is the relative price of two currencies, and it is, therefore, logical to begin with SIdemand function specifyhg the demand for one currency relative to the other. For ease of exposition, the relative money demand fun&,?- is presented in equation (2) in an explicit functional form.
In equation (2), S is the exchange rate expressed in units of Del;~scht;marks per dollar; x f: ln(F/S) is the forward premium on the Deutsche mark, and” is the forward exchange rate, The parameter K is a sh8t factor, q is th.e income elasticity of the demand for money, and e is the semielasticity relating the relative demand to the forwerd prem&rn, Finally,an asterisk denotes the United States. The left-hand side of this equation is the relative real value of the two money supplies. Formally, it is the DM value of the German money stock divided by the GM vJue of the U.S. money stock. It may be thought of as the amount of goods that WAId be purchased with the German money supply divided by the amount of ~WN~S that could be purchased with the U.S. money cnty y’:wrs, this ratio has &creased sub:rtantiAly. In supply. 0ver the pst 1960, the value of the rats WI!S0.12. By 1970, :t had increased to 0.22,.and by 1978 (3rd quarts )I, it l ad reac+led 0.46.3 The fact that t’.le German money supply can now purchase over half as mu& as the U.S. money supply is a startling statistic given the r&We size of tht: tv,o countries. The ;~oplllation of Germany in 1978 is only 28 percent of the C’.S.popuiatioq. The right=hand side : F the equation r:Istes the relative demand far the shift faci,or, K; the relative levels of real two monies to three facto! ;: tXa%, income, W/Y*; and i:he forward pr~~~ium,x. T!I latter arguments reflect the traditional presumption that the demand for r w :-myis pr.rsitivelyr&W to t!ie volume of transactions and negatively related to its
61
real income term is intended to capture move&n~ in the -vohimc cif &$&I: action:s denominated in Deutsche maiki relative to’ihe v&&j &t&&&&n denominated iindollars. This measureshould account for both the domestic and international tilansactions’demand since a major determinant of the demand for international media of exchange is the level of real income @Ithe issuing Gountry. ‘4 In addition, ielative real incomes ditso serve as& &&y ‘f&rr&&e . real wealth. For both OFthese reasons,,one would expect in h&e&% ‘&&&au real ilrcome relative to KS, real income to increase the dema&i’for I&it&he marks relative to the demand for dollars. The specific functbnal form specified in equation (2) imposes a further restriction in that it embodies the assumption thax the income elasticity of the demand for money is the same in both countries. This restriction implies that it is only the relative real income which is important in the determination of the exchange rate. The assumption is not restrictive i.n the theoretical analysis, and its empirical validity is easily tested. The second argument in the relative money demand function is the forward premium on the exchange rate. Under two assumptions, the forward premium may be interpreted as the expected value of the rate of depreciation of the exchange rate conditional upon the current information set. The necessary assumptions are that the forward exchange rate is determined by profit maximizing speculators who offer, in aggregate, an infinitely elastic suyply of forward contra:ts at a price equal to the expected future spot price. 5 The second assumption ir that speculators are rational, which implies that their expected rate of depreciation is equal to the rate predicted by the model itself. The virtue of the rational expectations assumption is that it is the only assump tion which eliminates comistent profits from foreign exchange speculation and closes the model in an internally consistent manner. The parameter E, which relates the relative demand for the two currencies to the forward premium, will be referred to as the “currency substitution” parameter, This title is in deference to the recent literature on currencysubstitution. 6 Grrency substitution models stress that real money balances may be he14 in a portfolio of currencies, and that the demand for any one currency depends upon its holding cost relative to the holding cost of available altwnatives. The existence of a number of close currency substitutes implies that the ability of any single central bank to undertake an indepcnden’tmonetary policy is very limited. Within the context of the presenl: model, a high degree of currency substitution implies that the currency substitution parameter is large. Although the relationship between the relative demand for money and the forward premium has been introduced through the concept of currency
62
~~,ut~~~~
it should be nscognited that the relationship need not only be due W mob nKwnents Out Of one currency and into the other. On the contrary, one ~~fll. ~%~isge a situ&ion in which all substitution is between money and int bm assets in response to changes in the nominal rate of interest. rWes on dollardenominated assets would lead to substiinto %W~~~~thu~ Ieriding .to an incipient bafance of paythe capitaf actiunt al a market clearing depreciation r, In MS context, equation (2) is a quasi-reduced form derived from the &uAtion of internatbnal arbitrage conditions-purcha;:ing power parity and interest rate parity -on tie cc’nventimal fomrulations of ihe money demand function in each country, If the arbitrage conditions hold, the forward premium is equal to the nominal interest rate differential and to the differential in the expected rates of inflation. The ,zremium consequently measures the opportunity cost of hoIding one currency relative to the other, or the reIittiveopptunity cost of holding either currency Mead of any other asset or commodity. Clearly only the sirnI test version of the relative money demand function is considered in the equation. In partictir, only one holding cost variable is considered; there is no direct measure of relative wealth and no consideration of risk. In addition, the equation is set in a twoauntry -Iorld so that thirdcountry influences are not considered.7 All of these exw. A.-s could, and should, be made, and I shah consider some of them in the empiric! analysis. However, for the moment, the simple form in equation (2) is s&Men*. for the purposes of the theoretic4 analysis. The only restriction imnosed in equation (2), which would, I ‘Mie~*e,be a feature ofany theory of exchange rate determination, is the imposition of degree one homogeneity between t Le relative money supplies and the exchange rate. The f ml assumption necessary to complete the model is the assumption that the money r~,arf* et is continuously in equilibrium, 8 The money market equilibrium conditbr! m&,1then be expressed as
where m = ln(M/M*); k z 1 n( ‘1, y = I n( Y/Y*) .I = 1n(S), andf- 1.n(F). Under a regime of flexible excha,rs rates, the money supplies may be assumed to be exogenous policy instruments. It is also likely th& the level of real cwtput will be independent of the cur ent level of the e.rch!angerate, althougfa it: may be related to the past histor:,! ,f prices and exc&rgc rates. If the erogeneity of va:rdexchange rate,s ze left as these variables is granted, airen the saot and the endogenous variables. The forward rate is di ;ermincd by the ratio-~1 expectations assumption:
63
ft=
_(4)
yr+1* .
where /5’is the expectations operator conditional upon the information set at period L This information set is assumed to include the structure of the model and the stochastic properties of the exogenous variabies. Using @;-_~uat&~r(3) j may be l&ten as
where rt = mt - kt - qy t and 7 = e/ lte. Equation (5) may be used to find the expected value of the exchange rate in any future period r+i.
(6) Repeated substitution of (6) into (5) for ah future periods yields the following reduced form equation for the exchange rate: (71 Equation (7) statr:s that the current spot rate is equal to the present dbunted value of the expected future leveis of the exogenous variables where the discount factor, y, is related to the size of the currency substitution parameter. If the degree of currency substitution is high, current changes in the exogenous variables will have, c~?&v?s @bus, a relatively minor influence on the current spot rate. In 1969, for example, the Bundesbank was supporting the dollar by rapidly expanding the German money supply. In retrospect, it appears that this expansion increased the excess demand for Deutzhe Marks rather than decreasing it, because the market recognized that a rapid expansion of the money supply increased the probability of an appreciation of the exchange rate. In other words, any cwreat increase in the money supply is associated with a reduction in the expected future rate of monetary growth, and the intervention had, if anything, a perverse effect. Equation (7) may also be used to explain the unpredictable nature of the exchange rate. In demonstrating this point, it is convenient to mduce the complexity of the equation by specifying a simple time series pracess to generate the vector of exogenous variables included in z. As an example, 1 assume that the change in ;Tis generated by a first order autoregressive process.
where A is the difference operator, (0 is a parameter whose value lies between
64
&nue mw, and ut is a y uncorrdated random variable. Using eq+t+on (C), t& &md future values of z mgy be found to be
phw :“,
(9) .
(10)
If 4bisposiave, predicts that the elasticity of the exchange rate eqGti0 withre8pecttozwillex unity. This “magnification” effect occurs because any increase in the cwrent ~ahe of z induces upward revisions in the expected fttu.re tiues of the variable tnd, hence, leads to an increase in the expected rate of deprsc,iation. The spot rate must depreciate immediately in order to eht’f’kinate the incipient excess supply of the home currency. This magnifkation effect should be carefully distinguished fron. the “overshooting” dynamics derived by Dornbusch (1976) ix a pIode in which commodity prices are not completely flexible. In both c;lses, the exchange rate depreciates by z greater proportional ambunt than the increase in the money supply. However, in the Dornbusch model the initial depreciation is followed by a period of appreciation. while in the model described above the exchange r.A CcPntinues to dcpreciatt: in the subsequent time periods. Equation (10) may also be used to explain the frequently stated a:sertion that the exchange rate follows a random walk. The empirical evidence in favor of the random walk model is quite strong in the DIM/$ rase. Differencing equation (10) and substituting from equation (8) yields the final form of the exchange xatz equation as an infinite order moving average process.
(11) Equation (I 1) expres~s klre current rare of depreciation .as a function of the current innovation, ut, and an exponentially weighted average of the series of past innovations. If the sriand aerm in the equation dominates the first ^ changes in the exchange rate wrrl be gr~!ual anlrl predictable. On the other hand, if the first term dominates, thk char t in the exchange rate will be serially uncorrelated and the level of the exchange rate will follow a random walk. Tht: relative importance of the two term3 depends upon the size of tb2 two parameters, y i%nd urrencq substitution parameter in the 4. The y parameter is rel. ted to t following manner: cf
r=y-*e
(12)
’
65
It is clear from equation (12) that 7 ranges over the intental fibm zero to u&y as e ranges over the interval from zero to infinity. If the two crarencies are perfect substitutes in the neighborhood of the equiliW*m ;&::i;;;, i -;;i!?5-, unity, and equation (11) predicts that the exchange rate will follow a random wa~lk.This result will be familiar to students of equity markets. If portfolio managers are profit maximizing, they will be unwilling to h&d any equities wlaich offer, after adjustment for risk, a predictably lower rate of return than the market. The equilibrium conditions which result from this behavior state that equity prices will follow a random walk. In the same way, profit maximiaiinp:international currency managers will be unwillinn’ to hold any currency w&h is expected to depreciate in terms of any other other currency, and the outcome of market dominance by these participants is the random walk model. By their nature, currencies are not as easily substituted as interest bearing assets. H#owever, the fact that exchange rates do behave in a manner that is consistent with the behavior of other asset prices suggests that the hypothesis of a %gh degree of substitutability between monies, and between monies and other ayaets, deserves to be taken seriously. This issue is closely related to the ability of the forward premium to predict future changes in the spot exchange rate. If two currencies are close substitutes, the fcrward rate will be close to the spot rate and will provide poor forecasts of the future spot rate. This point may be formally stated by deriving the ratio of the variance of the forecast errors to the variance of the spot rate. From equation (1 I), the variance of the change in the spot ral:e is
where V(U) = the variance of the innovation. Since the second component in equation (11) is equal to the conditional forecast of the current spot rate based upon the information set available in period f-1, the variance of the forecast error may ble found to be
811 the basis of (13) and (-14), the hypothetical R2 statistic that would result from a regression of the actual change in the spot rate on the change predicted by the lagged fbrward premium is R
:! = 1 - (l#“b/(
1-2&t&J)
.
(15)
Inspection of this expression reveals that the value of the R 2 statistic is negatively related ‘to y and, further, that the R2 tends to zero as 7 tends toward its limiting value of unity, In addition, the value of the R2 statistic is positively
66
related to the degree of serial correkttion in the exogenous variables.
The impartantlesson to be learned from these results 1s that the empirical geneaal,ization that forward ratea are typically more closely related to the current spnt rate tlqn the future spot mte is not evidence of market inefficiency; nor ar8 I~QG f&cast -izrorsevidence of %uffrcient” stabilizing speculation. On the conttary, the~preceding su that these characteristics attest to the existence of a lare number sf acttie currency traders whose currency preferences are dommated by mte of return considerations. E This discussion concludes the theoretical section of the paper. The theory suggests that the exchange rate, as the relative price of two monies, is determined by the current and future qsuppliesof the two monies after adjustment for changes in relative demand. If a h&h degree of currency substitution is present, the exchange rate will exhibit the random walk behavior that has characterized the behavior of other asset and cc;;unodity prices which are determined in organized spot and forward markets. In the following sections, this theoretical framework will be employed in an institutional and empirical investigation of the recent behavior of the DMlS rate. A Brief Institutional History of the DM/S Rate The previous discussion has stressed that the forward premium on a currency is related to the anticipated future changes in the Aative supply and demand for the cunency. In general, it is extremely difficult to construct a statistical model which accurately captures these expectational fac&ors. On the supply side, changes in the expected future growth of the relative supply of money are related to changes in the international monetary system which are once for all occurrences. These events include the revaluation of the DM/$ rate in 1969, the brelkdotvn of the Bretton Woods system, the Smithsonian realignment, the adoption of flexible exchange rates, and the Rambouillet accords of December 1975. On the demand side, there arc significant changes in the relative demand for the two currencies that cannot be explained by the standard arguments in money demand functbns, and it is clearly impossible to explain how these unobserved demand shifts vriil behave in the future. For these reasons. the forward premium will be considered as an exogenousvariable in the follow ing econometric analysis, In this section, an attempt will u,e made to interyret the behavior of the forward premium in the light of the main institutional developments in the international monetary system. Jn the period from 1950 to the present, the institutional arrangements governing the behavior of the DM/$ rate have passed through four formatl;lvr stages. The first soage covers the period from the earlv 1950s to late 1969,
67
which the C&s&he mark maintained a strang, stable r&ti&hip wtih the dollar within the international mont;tary arrangements of the Bretton Woods system., I:n retrospect, it is clear that the failure of the 9 percent apprecistion of the Deutsche mark in October 1969 to stem the flow of dollars into the Bundesbank marked the end of the fixed exchange rate period. .In the second stage, fmm late 1969 to March 1973, the Deutsche mark c&in-u&d to.increase in dollar value through a se&s of revaluations and temporary floats. In March 1973, the attempt to find a new par value for the dollar was abandoned. A g;eriod of “r&ctant floafing” followed. Although most of the major currencies were floating against the dollar, the object of the float, as expressed in the official anno’uncements of the times, was to find a new set of equilibrium exchange rates in order to return to a system of stable but adjustable P;arvalues. The final period, which covers the Rambouillet meeting in December 1975 to the present clay, witnessed the official acceptance of flexible exchange rates by the major countries, and the Board of Governors of the international Monetary Fund (IMF). During this period, the member countries of the IMF agreed to approach the issue: of exchange rate stability through a combination of control over underlying econo mic and financial conditions alrd exchange market intervention rather than throui:h the establishment of n new set of par values. ddng
These irrstitutional developments influence the exchange rate through two channels. OI~4cusly different rules of the game imply different reaction j:‘unctionsfor the czrltral banks in the system. Under the Bretton Woods system, the mon&ary poiicies of both countries were constrained by the demand for the respective currenl:ic:s.The Bundesbank was obliged to offer an infinitely elastic supply of Deutsch: marks in exchange for dollars in thr neighborhood of the par value, and the Federal Reserve, acting as an agent for the Treasury, was obliged *to excha.nl,:e dollars for gold at the fixed price of $ 35 per ounce to foreign central bal Iks. Under the Smithsonian agreements of December 1971, the U.S. was freed from the obligation to fix tlhe price of gold, but the Bundesbank remained committed to fixing the DM/%exchange rate+ Finally, under the flexible rate system, both countries were free to pursue independent monetary policies. The other channel through which the institutional arrangements influenced the exchange rate is through their impact on the expected rate of depreciation. I:f a country is strongly committed to a fixed exchange rate in the sense that it is willing to abdicate control over the money supply in order to maintain the rate, it is likely that; movements in the expected rate of depreciation will be stabilizing with respect to the par value. In a situation in which nonspeculative forces cause the actual rate to move below the par value, speculators will anticipate an expansionary monetary policy and the forwa.rd rate will rise above the spot rate. The expected depreciation of the rate will reduce the demand for the
68
currency and will result in an increase in the spot rate toward the par value. Gn the other hand, if it is likely that the par value will be abandoned, speculators will tend to Push the current spot zrte toward the future expected spot rate and away from the sparvalue. tbt “d~tibilizing speculation” is associated with ~~wflicts within the central bank between the external goa, of fiing the exchange rate and the intti mtrnetary targets. Such policy conflicts have ~hamW&d the lxhvbr of both the Bundesbank and the Federal Reserve during the past twenty years. As mentioned above, the commitment to fti the exchange rate was weksmed by the Cwman government during the 1950s as a means of increasing its stock of internationalreserves, liberalizing its trade policies, and contiling domestic wages and prices. However, by 1960, the persistent b&nce of payments surpluses were beginning to create an undesired increase in the German money supply. The German authorities the&l faced a choice between maintaining the parity and maintaining the domestic inflation rate at a low level. Give11the political costs of inflation in Germany, it was clear to both official and private observers that the exchange rate would be given a lower priority. The exchange rate was subsequently revalued in March 1961.
In the case of the U.S., the conflict arose between the commitment to
fm the price of gold and the demands for an expansionav domestic monetary policy. For the Bretton Woods system to work, the U.S. would have had to have a negative or very low rate of monetary growth so that the real price of gold would increase to equate the growing demand for the metal with the limited supply. However, deflationary policies were inconsistent with the Keynesian activism of postwsr economic advisors and, as a result, the quantity of dollars
increased steadily in relaiion to the relatively fared quantity of gold. g The policy coitflicts may be expressed in the terms of the model by considering two paths *or the money supply, the ftist of which is associated with the external PO&,:es of the government and the second with the internal policies. Let & equal .he rate of growth of zI under the external policy a.nd Ai equal the rate of gswth tf under the internal policy. If a probability a is associated with the acceprance of the external polic:r, then the solution for the exchange rate fsom equ;btion (7) i3
In the German case in the late 6Os, it is clear that Xi is less than AX,and thst the probability that the internal grcwth rate will be mdoited is positively related to the current change in the money supply. ~ifferent~t~~~ (14) with respect to
69
zr yields
The prlyceeding discussion suggests that the second term on the right-hand side of this equation is negative, It may also be greater than unity in absolute value. The importance of this result is that it demonstrates tha#.tintervention to support the dollar may be ineffective, even if it is successful in incrarsingthe monetary base, because the current expansion in the money supply induces expectations of greater monetary control in the future. In the German case, the end of the system was accelerated by an incffectivc policy of sterilizing the monetary consequences of the external support for the dollar. As Deutsche rn(sks were exchanged for dollars in the foreign exchange market, open market operations by the Bundesbank were undertaken to repurchase the Deutsche marks in exchange for domestic credit instruments. Although this policy may have reduced the outstanding stock of Deutsche marks in the short run, the ultimate effect was to create a capital account sur@us, which led to a further accumulation Iof dollars by the external desk of the Bundesbank. Eventually, the persistent use of this policy must lead to a situation in which all of the German monetary base is comprised of dollar denominated assets, and in which further sterilization of the monetary consequences of the balance of payments surplus is impossible. Given the policiies of the Bundesbank, the leading indicator fox this moment of truth is the ratio of international reserves to the monetary base. The value of this ratio in the last few years of the fixed rate period are given below. * o International Reserves Monetary Base
Year : Quarter 1966 196’7 1968 1969 1969 1969
:
4 4 4
:
1
:
2 3
: :
:
0.67 0.72 0.78 0.71 0.79 0.90
It is clear from these statistics that the external policy of firring the exchange rate ‘could not be maintained without permitting a rapid growth of the German money supply. The “hot money” flows into Germany in the second half of 1969 should consequently be seen as an al;curate reflection of changes in the expected future rate of growth of the German money supply. The instability in the system cannot be traced to the “exogenous activity” of private speculators but to the
70
inconsidrtent policy goals of the German government. By the third quarter of 1969, the Bundesbank could no longer attempt to offset the monetary consequences of the dollar in*Iow by domestic credit contraction, since, in the simplest possible terms, it nad run out of domestic credit instruments to sell. On the US, side, the goldpledge had never appeared to be a serious consW& on U.S. monetary policy until the late 1960s. At the end of the Second World War, the U.S. heid ovw 60 percent of the stock of gold in official hands, and some redistribution of the world goid stock was considered both desirable and inevitable. In addizion, most central banks preferred to hold their international reserves in interest bearing dollar assets, which were convertible into gold, rather than gold itself ps long as the convertibility oi these assets into gold was assured by the U.S. government. The ability of the U.S. to maintain convertibility was directly related *Dthe ratio of U.S. gold k.oldings to the monetary base, since the monetary base may be considered as the outstanding claims against the gold stock. In 1969, the ratio stood at around 40 percent. Uo*Alever, by 1964, the ratio had fallen to below 30 percent, and, by 1968, it had fallen a further 10 points to 20 percent. This decline in the gold ratio influenced the Bretton Woods system in ~WCJways. First, it weakened the confidence of foreign central banks in the ability of the 7i.S. to maintain .;M gold pledge, and hence, made them more reluctant to hold their international reserves ir: dollar denominated assets. Second, the general expansion in world liquidity resulting from the U.S. monetary expansion led to an increase in world prices z nd E steadily increasing demand for gold as its price in terms of other assets and commodities fell. Speculation that the price of gold would eventually have to be increased also added to the demand for the metal. After acccunt ir _aken of axogenous movements in demand, the most important determinant of the dollar price of gold is the supply of gold relative to the supply of dollars. This implies that maintenance of the gold pledge req~lires that the US. money supply grow at the same rate as the stock of gold. l=Iowever, a slow rate of monetary growth in the U.S. clearly conflicted with the dclmestic policy goals associated with the maintenance of full employment ano the financing of the Vietnam War. As in the German case, it is clear that the domestic objectives would be given priority over the gold pledge, The inevitable breakdown led private participants to sell dollars for gold and strong currencies. These clallars accumulated at European central banks which, in turn, passed them back to the Federal Reserve in exchange for gold. On the U.S. side, the end of the Bretto.n Woods system came with the StisPension oftheconvertibility of the dollar into gold in August f 9’7 1.. The collapse of the Bretton Woods system may therefore be traced to the failure of both central banks to recognize the monetary implications of the
71
syatm, For the syrstemto work, the U.S. would have had to tiesthe’expa&,m of the U. money supply to their holdings of gold and other intmmg&)dy tmkd assets. The significant increase in real liquidity that OcclMed durlm the posWar period would then have had to result fi-oma geneml deflation of world prices. On the Bundesbank’s side, the selfdefeating policy of sterilizing reserve flows through domestic credit contraction contributed to the speculative pressure on the Deutsche mark in iate 1969, Seen in this, light, the debate between fixed and flexible exchange rates appears to be misdirected since, given that both governments gave priority to domestic policy objectives, this is no independent tchoiceof an exchange rate regime. The key question in the second stage of the evolution of the IN/$ rate is why the appreciation of the exchange rate in September 1969 did not reduce thee German balance of payments surplus. In the fourth quarter of 1969, the Bundesbank took the opportunity presented by the appreciation to realiin the composition of its monetay base. The outcome of this action was that the ratio of international reserves to the monetary base fell from 90 percent to a litt!e over 50 percent. The Bundesbank consequently had a new stock of donlestic credit instruments with which to defend the domestic monetary &gets from the attacks associated with the U.S. balance of payments deficit. However, by the first quarter of 1971, the international reserve ratio had risen back to 85 percent and, rather than fight another losing battle in defence of the d’ollat, the Bundesbank withdrew from the market and allowed the rate ‘to float on May 5,197l. One reason for the failure of the 1969 appreciation may be found in the monetary statistics of the two countries. Over the period from 1960 to 1969, the average rate of growth of the money supply (A421in Germany was approximateIy 12 percent while the average growth rate of the U.S. money supply was approximately 7 percent. Under fixed exchange rates, these statistics imply a growth in the demand for L)eutsche marks relative to dollars of 3, percent per year. In 1970, the growth rate of the U.S. m,aney SUPPLYincmwd to over f2 percent, while ‘the growth rate of the German money SuPPlY fd to I 1 percent, Under the rem,3trictiveassumption that there was no dww in the trend in the relative demo (Id for the two currencies, these statistics imply u rate trf depreciation of the doll&r relative to the Deutsche mark of around 6 percent ;,er ~~~LIIII.The fact that the actual depreciation of the dollar Wasnot of this mognltude may he traced, in largs part, to thr- abatement of Speculative Pressure following the 1969 appreciation. These statistics imply that the primary cause al the contiaued pres~uts ,nrr the Deutsche mark after the 1969 appreciation was an increase in the growth l-ate of the LJ.S,money supply, Over the period from the end of 1949 to ihe end
72
of 1972, the rate of‘grow& of the U.S. money supply was 12.41 percent. In terms of the domestic political tituation in the U.S., the change in the growth r&e COnesponds t0 the appdntment of Arthur Bums as Chairman of the Board of Governors of the Federal F&serveSystem and to the policy of “benign negle& towards the U.S, deficit on the part of the U.S. Treasurv. ‘Ws =c!icj cuhninated in the nldon of the ~~~~~~~y at’ the dollar into gold in Aulprst-1971. An aWm@ Wa m&e to resurrect ~~33eof the features of the Bretton Nods system in the Smithsonian cements of December 1971, At the Smithsonian meet@@, new par w&c~; were established for the major currencies. HWWW, these parvdues were akm mconsistent with the economic and political ~&ties of the time. The agreanent did not contain any provisions aimed at ending the WovertibiBty of the doliar into gdd or, alternatively, of constraining the growth of the U.S. money supply. Since the U.S. money supply was growing very rapidly during this period, acceptance of the Smithsonian agreements threatened the European countries with rapid monetary growth and the inflation which would be associated with it. ‘Thisthreat was particularly apparent in the German case where the rate of monetary growth had exceeded 15 percent in both 1971 and 1972. Part of the reason for the more rapid growth of the German money supply was the belief by private participants that the inconsistency between the internal and external policies of the Bundesbank would result in a further apprecbtion of the Deutsche mark. This belief increased the demand for the German currency and hence further exacerbated th: problem caused by the rapid rate of U.S. monttary growth. The end of the Smithsonian parity finally came on February 4, 1973, when the .New York Times reported that U.S. Treasury Secretary Sc,“ultz had
recommended to (then) German Finance Minister Schmidt that Germany allow the mark to float. During the next week, the Bundesbank purchased $6 billiona figure amounting to 16 percent of’the German monetary base-and then withdrew from the m&et, By this time, it was clear to the European countries that the maintenance of fixed parities with the dollar was impossible in the light of the differential rates of monetary growth, In March 1973, Germany, France, Belgium, the Netherlands,and Denmark agreed to fix their exchange rates within a 2.5 percent margin with each other, while at the same time floating jointly inst tha dollar. The motivation behind the adoption rate system was not based upon any idea that flexible fixed exchange rates as a basis for a properly functioning system. On the contrary, flexible ates wl:re reluctantly possible response to the inflationary monetary policies of
of a flexible exchange rates were sllperior to international monetary accepted as the only the U.S.
The first stage of the reluctant flo.at period was characterized by several
73
large swings in the DM/% rate. From the March 1373 value of $0.35, the Deutsche mark rapidly appreciated to $0.43 in July 1973, and then Tellback to SO.36 in January 1974. In 1974, the rate rose to $0.41 in April, feff to $038 in August, and then rose to $0.41 in April, fell to $038 in August, and then rose 10 $0.44 in February 1975. These cycles confirmed the worst fears of flexible (exchange rate critics concerning the dynamic instabilify of unreguhited exch&nge rates. However, the issue is not as clear-cut .as a simple investigation’of the tirk series of the exchange rate would suggest. In the first place, flexible exchange rates were adopted at a time of great uncertainty and instability in the international monetary system. In addition, there are nonspeculative factors which may have been responsible for the sharp movements in the rate. The July 1973 rxovery of the dollar, for example, was related to the resumption of intervention by the Group of Ten (G-10) through an increase in the swap network. Currency swaps between the U.S. and the European countries permitted the Federal Reserve to borrow foreign currencies at short notice in Order td Sk\PPOfi the dollar. In effect, the European countries gave the U.S. some shortgerm control over the level of their monetary bases. As a second example, the OPEC oil increase led to an increase in the official demand for dollar denominated reserves. This increase in the demand for international reseTyes does not directly increase the demand for U.S. dollars, since official reserves are typically held in Treasury Bills or Eurocurrency deposits rather than demand or time deposits in the United States. The increase dild however, hav: an expansionary effect OR the monetary base of the countries concerned. In the case of the Bundesbank, the stock of international reserves increased by 37 percent between the first quarter of 1972 and the first quarter of 1974. This increase was associated with an increase in the monetary ba,se of 41 percent over the same period. A further factor which may have contributed to the instability in the exchange rate during the “reluctanl float” period was the uncertainty surrounding the future of the system, 1~1, the “Outline of Reform,” published by the IMF in June 1974, the Committee: of Twenty presented guidelines for the reform of the international monetary system, which included provisions for a return to a system of “stable hut adjustable par values.“1 1 The prospect of a :-:
74
tb WUa*n esfned by flexibleexchange rates in the early Years of the fkrat, th e authorities in the major countries were in fact adopting a more positive attitude tow&s the permanent adoption of a flexible exchange rate sYam* Aftho@ the rates: had been volatile, there had not been any major disruption to trade and capital flows, and, indeed, some controls on capital flows had k@n-*im@edIn addition, although the rates had been volatiie, the trend had bWt Small far moSt of tie m&x cumncieg - the U.K. and Italy being the @Or ex
75
this system arose out of choice rather than nec&sity, it is a vafid model for judging the efficacy of flexible rates as an international monetary system. The principle advantage of flexible exchange rates is the freedom it gives to conduct an independent monetary policy. During the period from 1975 to 1978, the U.S. and Germany have indeed followed quite ctifkmt monetary policies, and the exchange rate had adjustedto offset the exttinai onrrlse~uen&zs of these policies, The U.S. has followed )an expanskary m~netaty @liw designed to increase the level of output and employment, while the Germarb authorities have followed a tight monetary policy aimed at control over the domestic rate of inflation. The predictable differences between the domestic policy stance of the ttvo countries has resulted in a large averagerate of depreciation of the dollar, but it has also been associated with a decline in the va& ability of the rate of inflation around the trend. A part of this reduction in exchange rate uncertainty has been associated with a significant decline in the intensity of the conflict between the internal and external goals of monetary policy. This is particularly so in the case of Germany, where international reserve movements have been a primary cause of fbctuations in the monetary base a.nd where the appreciation of the exchange rate does not impose a large political cost. In the U.S., the discussion of the importance of the depreciation of the exchange: rate has been misdirected by a widespread tendency to attribute the depreciation of the dollar to the current account deficit, Under flexible exchange rates, a current account deficit must be matched by an offseking capital account surplus through variation in the exchange rate. As a fitst approximation, one can therefore say that the balaircc of payments position of th6 IJ.S. reflects an excess demand for goods and services matched by an excess sup~plyof interest bearing assets and money. It is only the excess supply of money which results in an overall incipient balance of payments deficit which is eliminated by a depreciation of ,i..e dollar. In the absence of an excess supply of money, it would be possible to have a large current account deficit which is associated with a capital account surplus without having any influence ott the exchange rate. This analysis suggests that the primary cause of a current account deficit under flexible exchange rates is an excess supply of interest bearing assets, a phenomenon which is generally associated with a large government deficit. The cornbination of a current account deficit and a depreciating exchange rate is therefore a consequence of a combination of an expansionary rrronetary policy with an expansionary fiscal policy, 13
To conclude and summrize this section, Table 1 presents some statistics relating to the efficiency of the exchange rate system in each of the periods previously discussed. The f’irst statistics relate to the variability in the stock of
76
tmld93
685
t3.25
om
Qm
2&8
(L.98)
2a2
<4M
bp14J2t4
1369
31A9
dAl
(9m
su
(t&34)
2J4
t1ao
7):1-7Sd
038
6.26
329
(33.7)
d.13
(33.X)
3.16
(269)
7&t-m3
3Jf
451
*IO&o
a391
431
(13.7)
IDI
(098)
77
internatiomal reserves. For both the U.S. and Germany,the greatestwwiabUtyin the stock of internaltiortalreserves was experienced d@n@the 7&l to 72:4 period when the Bretton Woods system was breaking up. The large variationin the stock of international reserves is associated with the attempts by both central banks to defend unbelievable external objectives. In terms of stability in the stock of international reserves, the best period is clearly that of the mature float when, as one would expect, the formsi adoption ‘of flexible wchange rates freed the two coumtries from the necessity of intervening in the foreign exchange market, The second and third columns present the average rate of depreciation and the average forecast error of the forwa:A rate* Although the mature float period has witnessed the largest averagerate of depreciationof the dollar, it is clear that the early float period was characterized by a far greater degree of uncertainty. This conclusion emerges both frolrn the standard deviation isf the qot rate and from the standard deviation of the forecast error. Two points should be made about the forlecast errors. First, there is no evidence that the forward rate is a biased forecast of the spot rate, since the mean forecast erroris typically bmall in relation to the standard deviation. On the other hand, it is also clear that the forward rate iprovidesan extremely poor forecast of the future spot rate so that the variance of the forecast error is comparable to thevariance of the change in the spot rate itself. The association between the increase in the average rate of depreciation and the decline in the variance of the rate is consistent with the theoretical model if the substitutability of the currencies has declined since i975. This would be the case if much of the “hot money” has already moved out of the dollar by this time. It is possible that the current holders of dollars have less opportunity to move out of the currency in response to its depreciation. The final statistics iw Table 1 give the average value of the covered interest rate differentia! between the threemonth U.S. Treasury Bill rate and the German Call Money Rwte. Because these instruments differ, one would expect that the covered yield differential would not be constant through time, However, large changes in the covered interest rate differential which persist for extended periods of time are most likely to reflect the presence of restrictions on capital flows. This index is therefore used to reflect the success of the system in permitting cap.$al market integration. On this indeh, the mature float period is clearly superior to the earlier periods, since the average value of ihc cr!vered yield differential is approximately one-third of the average values for the earlier periods. In conclusion, flexible exchange rates do appear to have provid:.3 the benefits that “heir supporters claimed of them. These benefits include increased
78
CoRtrtnt OVQt&e mQl_ supply, greater fi%Rdomto conduct f-t policy, less control over trade and capital flows, and a reatinable degree of exchange rate st&ility. It ig liktrly that th& stability will increase as government authorities learn how flexible e9x_hange rates work, and as they transmit clear s&m& to the rrrasketplace concerning the appropriate value of the exchange rate. There is clearly a need for government authorities to understand and advertise the implie&i@tid-f the exe mte of th& monetary and fiscal policies, since this a&&y will aot:Wy exp&tatW~ (decrease the variance of the subjective Uribution of exchange rate changes) and reduce exchange mte uncertainty. An EmpirIcalAnaIyarsO of the MonetaryApproach
The two primary assumptions of the theoretical model described in the second section of the paper were thst a stable demand function exists for the demand for Deutsbhe marks relative to dollars and that expectations were rational. Ihe outcome of this theory, as specified in equation (1 l), is the following relationship: (18) which relates the forecast error of the forward rate to the innovations in the exogenous variables. In general, a very strong test of the rational expectations approach would be to relate the deviation of the actual spot rate from the lagged spot rate to the unanticir.&edchanges in the relative money supplies, relative rest incomes. and other exogenous determinants of the demand for the two currencies. If the forecasting equations could be represented by sinlple time series models, :t would be possible to strengthen the tests by specifying bounds for the size of the coefficient relating the innovation in the exogenous variable to the unanticipated change in the exchange rate. An initial investigation of this type of test demonstrated that no s&l&ant relationdkip could be found between unr.nticipatedchanges in the exogenous variables and the unanticipated change it* the exchange rate. The Tact that no other studies have been published or distrikuted which have Zound a significant relationship suggests that my failure to fu*d a relationship is not entirely due to specific weaknesses in the approach that I adopted. The hilure must therefore ba traced to one or more of three factors: a) the relative money demand function is not stable; b) expectations are not rational; or c) the equations osed to forecast the brture values of 2he exogenous variables are not correct. Of these 1hree possibkities, I believe that the tkrd is the most plausible reason for the failare of the pure tests af the model. During a period of marked
instability and gre&er structura%change, it il; extremely difficult to construct a simple regression or time series equation which accurately descrRes the evolu-
19
tion of the money supply, the level of real income, and the host of other L&m which induce changesin the relative demand for the two currencies. Among the ‘latter factors, it is important to note the extremely rapid technical advance in the American banking industry which has resulted in a large decrease in the demand for money. r 3 For these reasons, the empirical tests will be primarily concermd with examiningthusstability .A?the relative demand for money fumtion. The starting point is equation (3), which may be solved for the exchange rate to yield st ==mt - k, + qyt + ext.
(1%
The following steps must precede the estimation of thb relationship: .Y’/ze choice of II ttite period. After an initial examination of the data, 1 decilded to estimate the equation using quarterly data over the period from ! 963: 1 to 1978: 3. The inclusion of the last seven years of the fuced rate period was based upon the following considerations. FLst, I was interested in testing whether the equation was equally applicable to both the fixed and the floating rate period and, in particular, if it was capable of tracking the breakdown in the Bretton Woods system in 1969. Second, the inclusion of the earlier period increased the ilumber of observations and increased the variance of the exogenous variables. This is particularly true in the case of the Gannany money supply growth relative to the U.S.: In the period firm 1963 to 1969, the differential rate of monetary growth was approximately 4 percent while, from 1970 to 1977, the differential was less than 2 percent. The longer sample period also allows a mcrc: accurate estimate of the r;;ffectof the forward premium on the exchange rate, since the icrgest values of the premium occurcd in 1969 before the imminent appreciation of the Deutsche mark. Cltoic:eo$ exogePto[tsvurlclbles.From the discussion in the text, it is clear that the relative money supplies were an endogenous variable during the fixed exchange rate period. One approach to this problem is to rewrite the d;lpendent variable as (s-nz)t, the relative serpply of the two currencies,as is done ky Girton and Roper (1977) in their excellent study of the Canadian economy under fixed and flexible exchange rates. The difficulty with this approach is that it presumes a unitary elasticity between the relative money supplies and the exchaage rate, and thus asserts one of the most important testable hypotheses of .the monetary approach. In this study, it is assumed that it was not the exchange rate but the relative money supplies that were “fixed” by the Bretton Woods system, since the exchange rate was free to vary within a one percent margin. In ariy case, the results do support the use of a composite endogenous variable and the Girton-Roper approach will be used in tne later part of the tests.
80
The endogen&y of the faNvardpremium poses more serious problems. The m&r probkrn is that the assumption that the forward premium is exogen~usprevents any analysis of the &inamicsof the adjustment of the spot rate to Changes in the exogenousvariablea,since these dynamics are embodied in h$atmt problem concerns the sihultaneous It-a&oni the amption that this variable is exogenous. b-se of the thmetlcal correlation between the residual in the e equation and the value of the forward premium. Clearly, if the errors in the equation a 1, then the error and the forward premiumafen ely corm&ted, and the estimated coefficient understates the true coefl%ient. At the other extreme, if the residuals were serially independent in the fmt differences, then there would be no correlatir$n under rational expectations between the cunent valve of the residual and the forward premium. In fact, the results presented below suggest a small positive serial correlation in the residuals, This implies that the estimates of the e parameter underestimate the true v&e. However, it is likely that movements in the exogenous variables dominate the determination of the forward premium so that the extent of the bias is small. The specification of the shift jiwtor, k. Theory and a preliminary
investi@on suggested that the relative demand for Deutsche marks could be increasing overtime. The main technical innovations in the U.S. banking industry that have occurred over the past twenty years have resulted in large economies of scale in thrt holdings of dollars. In addition, the stability of the Deutsche mark was not considered secure is the immediate postwar years because of the hyperinflation followiq the First World War. Finally, many money demand studies for the U.S. havr: suggested that the demand for money in the U.S. has been decliniq relatively rapidly during the 1970s. Among the factors responsible for this decline, which has occurred primarily in corporate money holdings”has been the increased use of external deposits and the availen@‘id
ability of overnight repurchase agreements to large corporate depositors.14 In order to aklow for a trend in the relative demand for the two currencies, and also to allow for the trend to be changing over time, the shift factor was specified to be a quadratic function of time. In audition, allowance is made for first order autocorr&&+on in the residuals.The final form of the shift factor is consequently k ,=k@ X1t f h# where ut = put-1
+
cm
+ q,
wt.
.4!10wancefor distributed lags. In domestic money demand functions, the demand for real money balance is generally specified to be a distributed lag f2 and n~~i~~~~ i~t~~e3t rates. ~~t~~~ of current and past levels of r
81
the behavior #ofexchange rates suggestsVW short adjustment ?ags,the lity of grad1ua.I adjustment was allowed far by the introduction of polynomW distributed lags on each of the exogenous variables.The distributed lag coef’ficients werle constrained to lie along a fourthdegree polynomial with the last coefficient, which corresponds to the 13quarter lag, constrainedto zero. The jnstifkation for these lags may be based upon costs of adjwtment or cws of information, or they may be considered as forecasting equkions’for the Mure values of the exogenous variables. After these steps had been taken, the following regression equation was derived:
where ut ==put.1 + wt,
S,/,i= Qj#0 i + aj,2 i2 + aj,3 i3 + ~i,4 19,and
On the basis of the theoretical model, lthe estimates of this system of equations are preceded by the following presumptions concerning the estimates of the important parameters. First, the homogeneity postulate predicts that the sum of the coefficients on the relative money supply variable would not differ signifiantly frlom unity. In addition, the description of the monetary approach suggests that the transmission of a change in the money supply into the exchange rate is fairly rajpid. Second, the approad’2 suggests that the sum of the coefficients on the relaGve real income term will be significantly negative. This hyp0thesi.s conflicts with the conventional wisdom which asserts that higher real income depreciates the exchange rate by inducing a larger demand for imports and a deficit on the current account. While the monetary approach does not dispute the proposition that.the income elasticity of the demand for imports is positive, it relies on a posit:&e,income elasticity of the demand far money lto assert that higher real income results in an overall incipient balance of payments surplus, Finally, the mlonetary approach postulates a positive relationship between the forward premium and the level of the exb.;\an$erate. Through the interest rate parity condition, this postufate implies that higher domestic interest rates lead, cetmiriu paribus, to a depreciation af the @.change rate. This postulate may be contrasted with the theory that higher interest rates appreciate the exchange rate bq’ creating an incipient capitnl fkw into the domestic economy. The major difference in this case is in the -viewof how the intz~national monetary system works. The monetary apprcach, with 11s emphasis on integratecl world capital markets, views L lterest rate differentials
82
as reflec:tionsof differencea in expected rates of depreciation. The alternative view, based on segnented q=GtaImarkets, views nominal interest rates as market clear@ prices between assets that are not perfect substitutes. It is important to note that the tests of the model are not based upon the .~bbiwny of the gilation to fit the hi&to&alcjata,It is clearly the case that any ..~ equathr wi$h ho time trend var&bIesand three polynomial distributed lags will acc&nt for BiUge hactioa 6f the variance in almost any economic time series, The tests of the model must therefore be based on the size and sign of the estimatti coefficients and on the statistical significance of other plausible determinantsof the exchange rate. The ordinary least squares estimates r>fequation (18) are presented in Table 2. In general, these resllts provide strong support for the monetary theory. Changes in the money supply have a rapid and direct influence on the exchange rate and the sum of the coefficients on the relative money supply variable, 1.02, is very close to the hypothesized value of unity. In addition, the hypothesis that increases in real income lead to an appreciation of the exchange rate is also strongly supported by the estimated coefficients: The results yield an estimate of the long-run income elasticity of the demand for money of 2.4743. The puzzling aspect of these results is the shape of the distributed lag relating relative real income to the exchange rat:: There is no statistically significant relationship between these variables until after a lag of six quarters. The implausible shape of the distributed lag does suggest that the relative income variable may be picking up the influence of other omitted variables on the exchange rate. The most likely candidate for an omitted variable is the level of relative prices, which is likely to respond to relative real income with a lag if there are costs to price adjustment, and which is likeiy to have an immediate effect on the exchange rate. The influence of relative prices Gn the exchange rate is examined below. One aspect of the results reported in Table 2 that b worthy of note is and significant relationship between the forward premium and the spot exchange rate. In this case, the distributed lag ti~llows a plausible humped pattern with the strongest effect being felt after three to four quarters. It is also notic’eable that the zero lag coefficient is both smaller in size and less significa.nt than that of early lagged values. This result may be due to the srmwltaneclus equations bias which arises from the assumption that this variable is exogenous. If the nominal rate of interest is 19 percent per annum, the results in Table 2 imply an interest elasticity of the demand for money of approximately -0.65, which is considerably larger than estimatg:s derived from conventional studies of the demand for money using aggregate price indices. It is also true that the speed of adjustment of the exchange rate to changes in the exogenous variables
83
0
1
032w
am
+03486’ (2JJD
2
+2a49* QS3)
3
W9)
+3ms* 8-m
5
*a279 (033)
+2927* 0.19)
6
bl32 @.17)
+2.711* 4.92)
7
90.0413 (054)
+2397* 056)
8
O&S73 (ofi)
+2.012* QJS,
9
a&6%6 1063)
+I1583 (1.73)
4
to.8836
10
+1.140 (MS)
11
+0.709 (1.03)
12
94258 (OAl)
*320 (oJ6)
lm69 (1 Al)
26.241 (2.93)
is ocrtmrjderablymore rapid than wt suggeskd by domestic money demand s&&s. Tltesc: featureaof the results may help to explain which protracted deviatio? fro& purchasing power psrity occur in response to monetary shocks. aSp”t .Qf tirere@ts t.hat requjres comment is the size and .,@I$Ne end *arriaa&. The@ v&b&-s suggest that the demand
d23llarswas ,inct at a rate of approximately M ~aeent .pa~year &1963,5.6 percent per year in 1970,and 9.2 percent per yeat at the ad of the *apIe p-e&xi, In the early part of the sample, the trend in relative money demand was probably due almost entirely 30 the increase in the demand for money in &many. &cause of the hyperinflation after the Fira World War, the stability of the Deua:,che mark after the Second Wtirld War was uncertain, and it is plausible that the demand for the currency would increase as its strength became rnod’ewidely recognized. The appreciation of the Deutsche mark in March 1961 undoubtedly also contributed td the desirability of the currency, and the prospect of a further appreciation in the late 1960s could only result in a further increase in demand. nwdi
&Uivc
to
It would, however, be wrong to attribute ah of the trends in relative money demand to increases in the de&ability of the Deutsche mrk. In the U.S., Golldfeld (1976)hasnoted a secular decline in the demand for U.S. dollars which cannot be attriiuted to the traditional arguments in money demand functions. The combination of controls on interest payments on demand and time deposits and. a competitive banking system have led to financial innovations which have decreased the demand for money as it is conventionally defined. One of the earliest manifestations of this dmelopment was the growth of offshore dollar maurketsin London, Zurich, and Luxembourg. Domestically, the growing use of credit cards- a development which effectively transfii.3 cash management from the individual to a specialized corporation -the provision of automatic overdraft
facilities, the introduction of NOW accounts, and other financial innovations effectively permit the Mividual to economtie on the use of money. Although such financial developments have been rapid in the retail banking market, they have been even more important for the large corporate customers of commercial banka. Garth and Pak (1979) suggest that the introduction of repurchase agreements, by which banks purchase deposits from !ileir customers overnight, may, by itself, be able to account for the major portion of Goldfeld’s “miss% money .” If a financial innovation allows individuals or corporations to reduce thl:ir holdings of demand deposits, the actual. reduction will take place when the
individual attempts to exchange the excess money balances for goods and services or interest bearing assets. In aggregate, these actions imply an incipient balance of payments deficit for the county+; Le., an excess supply of dollars
85
matched by an excess demand for goods and interest bearing assets. This d&i& is eliminated by a depreciation of the .exchange rate which reduces the”real purchasing power of the domestic currency and, hence, increases the nominal demand for it. Although it is clearly unsatisfactory to represent this change in the relative demand for the two currencies in terms of a time trend, it is almost impossible to ‘find adequate proxies for the introduction -e;n;d dif+aioti of ~the financial innovations that are taking place during this s&n@ p&W. @I that the preceding; discussion is attempting to establkh is that the size and fnlp&t&nceof the time trend is not implausible in the light of the financial developments within the two countries. In the remainder to this section OPthe paper, an attempt is made to tighten up the exchange rate equation by imposing some restrictions suggested by the original results and by considering some additional explanatory variables. After a “best” equation is found, the sources of change in the DM/$ rate will be discussed. The most glaring omission from the equation estimated above is the level of relative prices. The hypothesis that the exchange rate is “determined” by relative prices is the fundamental contribution of the purchasing power parity approach to exchange rate determination. The problem with t!:“; approach is that it is difficult to reconcile with the generally accepted view that exchange rates, as asset prices, ladjust more rapidly to changes in the demand and supply of money than do commodity prices. In his influential synthesis of the purchasing power parity and asset market approaches to exchange rate determination, Dornbusch (1!)76) demonstrated that the slow adjustment of commodity prices leads to overshooting in the exchange rate. In his model, increases in domestic prices are associated with an appreciation of the exchange rate, since higher prices increase the demand for the domestic currency. In order to test the Dambusch model, the ratio of the consumer price index in Germany to the consumer price index in the U.S. is included as an additional independent variable. As with the othler exogenous variables, the possibility of a distributed lag relationship is allowed for by constraining the distributed lag coefficients to follow a fourth-degree polynomial with the thirteenth lag coefficient set equal to zero. The results from this procedure did not offer any support for the Dornbusch model: The hypothesis that all of the coefficients were zero could not be rejected by a standard F-test. l5 The conclusion to be reached on the basis of these results is that relative prices have no explanatory power over exchange rates after direct account has been taken of the conventional arguments in the relative money demand function.
The next hypothesis to be tested is that the relationship betwc:en the relative supply of money is both proportional and instantaneous. To test this
86
#the following
The estimates of this eqWiOn system are presented in Table 3. Under the null hypotheai$ the value of the~-!Mistic is 135 y which is well below the 5 percent critiicalvalueOf 2.88 with 3 degrees of freedom in the numerator and 50 degrees of fmxhn in the denominator. On the basis of this evidence, it is consequently not pogsible to reject the hypotnesis that there is P proportional and instantaneous r&tiOntip between the relative money supplies and the exchange rate. ace Qhe Constrained form of the relative money demand fkction also eliminates an)’ simuhneous equations bias which may result from the assumption that the rdative money supply is exogenous under fmed exchange rates, this form will be retained for the remainder of the empirical investigation For the fmal test of the model, the importance of other measures of the relative opportunity cost of holding money is considered. In the estimates presented above, the opportunity cost variable was assumed to be *he threemonth forward premium on the Deutsche mark. The motkation for this choice is that the premium represents the differential yield which is most clearly due to the difference in the currency denomination of the assets. It is, however, important to recognize that the premium will differ from other measures of short-term interest rates. Doo!ey (1977) has presented convincing evidence that the German program of capital controls did induce a divergence between the domestic interbank borrowing rate in Germany and the Ewrocurrency rate on Deutsche mark denominated assets. In the absence of interest rate parity, there is no theoretical presumption concerning the choice of the most appropria:e measure of the relative opportunity cost of tile two currencies, consequently, the issue must be addressed through an emptical investigation. On the issue of short versus long rates of interest, money demand studies in the U.S. have provided some empirlcialevidence that the entire term structum of interest rates must be considered in the assessment ofthe opl’ortunity cost of holding money. Hamburger (1977) argues that the failure of Goldfeld’s money demand function to predict money demand during the 19 Xs is due, in part, to the fact that the function does not include long term-interest rates. In addition, I-Ieller and ahn (1978) represent the term structure of
87
_:,
.._3
.-,
CoYtnllibd~nmte8dttle~lht8 1%3zl-1973:3
2
3
*2333’ Q43)
4
+23w cw
5
Q
7
03310’ (4A2)
8
l1JfM* a201 MA32 a $3)
9
10
11
231174
QdJ6)
88
@mat ratok~4th a polynomial &&ion in the maturity and find that all of the parameteraof the filnction have estimated coefficients which are sign&antly ot
eonI
zero sN%xwd@
to istandard
tests.
~
In CMb to- i@eiIlIM3~~ ,tlpsc tes@, ,two a&Iitional time series were ~;.._ --~ i- .“..
@becovered interest rate differential, the &man call an bdex of shozt-term domestic interest rates in Ml ate was employed as an index of sh~rt-temdon&tic interest rates in the U.S. Long-term government bond rates were used as indices of long-term interest rates in the two clountries.The additional variableawere incorporatedinto the relative money demand function in the same way as the precedingvariables.In other words, the distributed lag coefficients were constrained to Ile along a fourthdegree polynomial, and the coefticient on the thirteenth lagged value was constrained to zero. The tests involved conventional F-k&s on the null hypothesis that the coefficients on the additional variablesairerqual to zero. The results suggests that it is not possible to reject this hypothesis. When the covered interest rate differentiaI and the long-term interest rate differential are entered joixitly into the equation, the F-statistic on the significance of the additional coefficients was F(6,47) = 1.47, which is less than the 5 percent critical value of 230 for these degrees of freedom. When the two additional variables were entered separately, the F-statistic on the covered interest rate differential coeffkients wasF(3, 30) = 1.84 and the P-statistic on the lowterm interest rate differential was .F(3, 50) = 1.76. Both of these statistics lie beneath the 5 percent critical vdue of 2.80 for the relevant degrees of freedom.The conclusion to be reached on the basis of these results is that the forward premium does appear to be a suffkien; measure of the relative opportunity cost of holding the two currenciesdespite deviatians from purchasingpower parity and movements in the term structureof interest rates. This section concludes with an analysis of the predictive power of the model and a discussion of the sources of change in the DM/$ exchange rate. III Figure 1, the actual and predicted values of the exchange rate arc plotted against time. It is clear from the figure that the model captures ah of the major turning points in the history of the DEI/$ rate. In patiicular, the predicted values oscillate around the pax value cluring the fixed exchmge rate period, decline coincidentally with the actual rate during the turbulent period from the end of 1969 to the beginning of 1973, level off during the early float period, and then res~rne a trend rate of depreciation during the mature float period. Clearly, the largest prediction errora occur durLg the 1974-75 period following the increase in oil prices by the OPEC cartel. QveraIl, htiwever, the predictive power of the model is quite satisfactory. Without reliance upon lagged dependent variables cr
89
. ; : “_, ~_ ;
\ \ :
--
90
..
.
alar,the nioild a~~~untsrfar 98 percent of the error of
rate aver a fifkmyear period with a stadni
‘Ihc predic&as of the model may be usefMly divided into long-term aud shortcorn order to understand the long-term factors af. m’ QCU@$M mqk: q+n& :$b dollar, the may be :&tr&ted to each of the four exogning to the ,end of the sample pciod is reproducedbeIow: ’ f
j
Time Trend
-90.24%
ReUive Money Supply
+47.17%
Relative Real Income
-14.47%
ForwardPremium
-14.07%
Slllll
-71.61%
These statistics should be interpreted in the following way. In the period from 1963:l to 1978:3, the Deutsche mark has appreciated by approximately 72 percent against the dollar. Holding all other things constant, the trend in the relative demand for the two currencieswould have resulted in an appreciation of approximately 90 percent. On the other hand, the German money supply has typically grown at a more rapid rate than the U.S. money supply so that this factor, ceterfs pcvlbus, would have resulted in a depreciation of the Deutsche mark of 47 percent over the sample period. Compared to these major factors, the other variables have a relatively minor impact on the long-run movement in the exchange rate, although they may, of course, be responslble for substantial short-term fluctuations. TO examine the cause of the short-termfluctuations in the rate, Figure 2 plots the change in the exchange rate which may be attributed to each of the four components, For purposes of comparison, the relative money supply series is adjusted for the trend in relative money demand. Figure 2 dramatically illustrate the speculative pressurr! against the dollar in 1969 and the associated increase in the German money supply in support of the parity. The extent of the necessary increase was mitigated by an increase in the demand for dollars associated with the Vietnam war boom in the ‘U.S.,which increased real income in the U.S. relative to Germany. The rapid appreciation of the Deutsche mark in the 1970 to 1973 period is associated with both a decline in the relative money
91
Figure 2
92
and a relative expansion in German red in the DeuWhe IILafkduring the floating with the trend adjusted e floatig mte period, ,the S‘ reiatiive~t&I&many h;&ve , and this increase, ddllat duringt&
brs beun to determine whether a stabIerelative dermmdfun&ion relative to dollars can be established. This quest lM&Bmet On the po&ve aide, ail of the arguments inthefbnc&athat monetary theory did appear to have a t cm the axck~ Me, and the signs of the coefflcietlts on w62rrcorrect. In addition, the model predicted the m*r turning +tita in the &end ia the exchange rate. On “h negative tie, the most &nit& in the demand fbnction was a quadrat&@me tre&. thistrendmaybemhMdto~andASan&i wt8 is
has the doPar cl
93
factor here, I believe, is the very rapid rate of financial innovation in the U.S., which haa led to a decline in the demand for dollars. Among the important financial innovations, one can point to the growth of the Eurocurrency market, the growth of nonbank financial intermediaries in the U.S., the introduction of credit cards, automatic overdraft facilities, N0W accounts and overnight repurchase agreements. These developments allow a far greater volume of “liquidity” to be prclduced from a given monetary base and’hence, if the dem‘andfor liquidity is relatively stable, to an excess supply of money. As a counter to this tendency, the German money supply has generally grown at a more rapid rate than the: U.S. money supply. As far as the trend in the rate is cancerned, movements in relative real income and interest rate differentials have played a minor role. Why has the dollar tended to move in an erratic and unpredictable fashion? The theoretical analysis for this behavior was based upon the conjecture tha; the two currencies are close substitutes. This implies that the current spot rate is extremely sensitive to movements in the expected rate of depreciation. This conjecture was borne out by the empirical results. If, for example, the exogenous variables exhibited extremely strong serial correlation - 4 = 0.9 then the estimated value of E of 23.07 leads 1:othe prediction *%attie fraction of the variance in the spot rate that is predicted by the forward premium would be 0.18. In fact, there is extremely little serial correlation in the relative money suppfiy and the relative real income series, so that the poor forecasts of the Iforward rate are entirely consistent with the theory and the em@cal evidence. What is the role of private specula*tion in the determination of the exchange rate? As in all organized asset markets, the effect of speculation is to discount future information into current prices. This implies that the exchan.ge rate will adjust to changes in the exogenous variables at the moment whrenthe change is anticipated, rather than when it actually takes place. Why does official intervention seldom seem to work? The theory and evidence suggest that intervention will work if it increases the current and expected future levels of the monetary base. During this sample period, intervention by the Bundesbank was ineffective for two reasons, First,,the effect on the Icurrent monetary base was small because of the sterilization operations of the domestic desk of the Bundesbank. Second, to the extent that intervention diid /increase the monetary base, this increasle induced expectations that the fixed parity would be abandoned so that a more stringent monetaqr policy could blefallowed in the future. What are the prospects for a return to fmed rates? Since most of the major countries appear content with flexible r,ates as an international monetary
94
) this question should be ted in teams of the prospective stability 6~lh%r.My view is that the prospects for a stable or appreciating dollar are #)odA was mentioned above, there is no evidence that the rate of monetary groMh in the US . is greater than in Germany. Instead, the depreciation has been due to a rapid decline in the demand for dollars relative to Deutsche marks. It is extremely d at th& rapid decline can continue, and it is more likely thatitmbe the batlking innovations which originated in the U.S. are exported to the Eu countries. In the &st quarter of 1978, the U.S. money supply amounted to 43 percent of U.S. gross national product. At the same time, the C&mm money supply amounted to 62 percent of German gross national product. This large difference between the velocity of circulation between the two countries reflects a soph&tication of &heAmerican fiiancial system. My hunch is that as this sophistication is exported through American branch banks and financial institutions, the demand for Deutsche marks will d&e, and the dollar will begin to appreciate. This analysis is, of course, based upon the current rates of monetary growth in the two countries. Have floating rates delivered the goods? The answer to this questior‘.is positive. Since 1975, the U.S. and Germany have been able to pursue quite different domestic policies and the monetary consequences of these policies have been absorbed by the exchange rate. The large current account deficit is a reflection of the fiscal deficit of the U.S. it is certainly the case that this policy independence would have been impossible if the two countries had kept to the rules of the Bretton Woods system. As far as the uncertainty of a flexible rate system is concerned, the period since 1975 has witnessed a significant reduction in the short-run volatility of the exchange rate and in the variability in the stock of international resemes. Finally, the evidence from the covered interest rate differential calculations supports the view that flexible rates have permitted a reduction in controls over capital flows. clearly
The possibilities for a contmuation of these trends depends crucially upon the reaction of the U.S. authorities to the depreciation of the dollar, If this depreciation is correctly seen as n consequence of excessive monetary growth relative to demand, then a eontractionary monetary policy would reverse the depreciation without losinrr the benefits that have been gained from flexible exchange rates. On the other hand, if the depreciation is seen as an “external” problem to be fought by the imposition of capital controls and shortterm intervention in the fore’ exchange market, a return to the uncertainty of the early floating rate period appears likely.
95
lJlrbr*Ap~&~:
.”
._
_‘,
: .I_
‘, /(
Au data +re taken from the Sep$snbm I978 vepsrsOn ,of the hi&rnc
ti& F&w&d Shtistics da@ tape issued by the InternationalMuwtary Fund, Wash&@m, D.C. Subsequent revisions for the tl&d quarter of: 1978 wme taken from t+e Janwry 1979 edition ,of,.lntepatic& ,ISq~$@.~~t~~i~s,~~~ n&k de&t= the vadabks u’&din the$gt in hrmq of :$h$‘l?$h!.i&e~:~~%~~ htrpmmritiodRewwes = Foreign&s&s of the Cent& Bad @3&.I31.’ /, : Mont&y Base= Reserve Money (line 14). j Money Supply = Money (seasonally adjusted) (line 34b) plus Quasi Money (line 3,s). Short-Term In&rest Rate (U.S.) = 3-month-Treamy Bill Rate @ne QOd). ShortXerm Interest Rate (Germany) = Call Money Rate Qine 6Qb). Lo~-Tlmn Interest Rate (U.S.) = Long Term Govemment Bond Yield (line 6 1.). Long-‘I&m Interest Rate (Germany) = Public Authorities Bond Yie1.d(line 6 1). Price Levei = Consumer P.riceIndex (line 64). Real hmme = Gross National Product, 19’75prices (line 99a r). Exchange Rate = Market Rate (end of period) (line Je). Forward Rate = 3month Forward Exchange Rate (line b).
5.
A+I@,+ of at&i&i li&attsm&d io oslt the.hypoth& .thit #he f-ad rate dr a~ unbiaud pa&$&of-Ihe fimtuwqwt fYIbrSee, for axamph, Dbn md f.Mch 0977), C!omdi arid Diet&h (1978), G&dulyiad Dralbjr(1976)&vkh (1978), mudStockman (1978y,
6.
In jMddwv
7.
BdBtenbowg snd Sch8dia (1978) co@&uct and temt l muIticount#y modeI of re&ive money demamid,
a+ Girtua rpndRoper (1976), King, Ratnmn, and Wilfti (1977Y and Milea(1978).
8. 9,
Foi an ax&lent d&a&m Cha 1).
of tlltegdd p&lema of the Bxetton Woo& ystem,aw Coomk 11976,
10. 11.
lMP (i974). The Committee of Twenty co&ted of OIMmember md two u(lodltca tirn eacJ~ of the twenty constitueadea that droorcr an exwutive director of the Jnternatio~I MoMauy Fund.
12.
See the hMnutioaal Monetmy Fund, Amnmf Rcpart, (1977).
13.
See GddfNd (1976) for the fimt evidence that the cozwentianallmoney demand function was oveqmdi&ng money demmid &sing the 197Ui1.
14.
GardaaadPek(1979) offer some convincing evidence that the failure of the tndition~ inmY demandhmction to predict lhe demand loo money in the 19700 is due to the wideslpterduk of
mjwdw 15.
n@emenb, ktwwn banb and theiz oorpants cwtonwa
ate rctual vrluw of the F~taiMc ~28 135, w&h i8 wedI below the 3 percent cdticd value of 288 J~x3 degmeaof fmdom in thn nunwmti and 50 dqrem of fmdom fn fhe den~minah.
97
Rderences Barre, R.J.
(111978) A stochastic equilibrium model of an open economy under flexible exchange rates. (&arfe@ Journal uf &anomies, 92: 147-64. ,.
B&n, J. F.O. (11978)
:‘:- j
;. ~(
Rational expectations and the exdhange rate. T!ze economics of exchange rates: selected s&dies, cds. J.A. Frenkel a&I H,G. Johnson. Reading, Mass.: Addisr+Vilesley .
Bilson, J. ?.O. and Levich, R.M. ( 1977) A test of the forecasting efficienciy of the ,forward exchange rate. (Manuscript, New York University). Brillembourg, A. and Schadler, S. A simple portfolio model of multilateral exchange: rate deter(1977) mination. (Manuscript, International Monetary Fund). Chrystal,, K.A. (1977)
The demand for international media of exchange. American Economic Review, 67: 840-50.
Coombs, CA. (1976)
The arena of interncationalfinwrce.New York.: Wile:y.
Cornell, W.B.and Dietrich, J.K. The efficiency of’ the market for foreign exchange under (1978) floating exchange rates. Review of Economics and WatL+ tics, 60: Dooley
, M.P. (1977)
Don?burich, R. d 1976)
11
I-20.
A note on interest parity, eurocurrencies and capital controls. International Finance Discussion APapers,Board of Governors of the Federal Reserve System.
Expectations and exchange rate Idym!mics.JourrzaFof Politieal Economy, 84:
116
I-76.
98
Emminger,-0. (1’977)
nt3 &Mar& in thb ConjZictBedwetg Internal and hYxtemcrl egutiibrium, 194h7.5, Princeton, N,J. : Princeton University Press.
Frenkel, J.A. 1 ,:‘, ’ ”
U’P~$): I_
monetary approwh to, the exchange. rate: doctrinal aspec’ts -gkJ e:mpiricaI evidence, SclrndinaviatiJournal of Economics, :A
78:
200-24,
Frenkel, J.A. and Clements, K.W. &eh.ange rates in the 1920s: a monetary approach. (Manu(1977) script, University of Chicago). Garcia.,G. and Pak, S. Some clues in the case of the missing money. Ametical~ (1979) Economic Review: Papers and Proceedings. forthcoming.
Giddy, I.H. and Dufey, G. The random behavior of flexible exchange rates: implications (1975) for forecasting. Journal of International Business Studies, 6: l-32. Girton, L, and Roper, D. (1976) Theory and implications of currency substitution. hternational Firaance Discussion Papers, Board of Governors of the Federal Reserve System, Number 56.
(1977):
Goldfeld, :_I.M. (1976)
Ai monetary model of exchange market pressure applied to post-war Canadian experience. Ameriun Economic Review, 67: 53748.
The case of the missing money. Brookings Papers on Economic Activity, 3 : 683-730.
Hamburger, M.J. Th,e behavior of the money stock: is there a puzzle? Journal (1977) @Monetary Economics, 3: 265-88.
99
Hodrick, R,J. (r97@
An empticai analysis of the monetary approach to ,&e -dater& n&ration of the exchangF ra&,; ~~-:&x&*&&~~, &I$&& ruta:
selected &w&es; eds, J.A,. Ftinkeland
‘II;6 Johnson.
Reading, Mass.: Addison-Wesley. International Monetary Fund ( f 974) International monetary reform: documents of the cotimittee of twenty. Washington, DC.: International AMonetaryFund.
(1977)
1.077 Annual report. W~~shington,D.C.: Intemattional Mone-
tary Fund. King, D.T., Putnam, B.H., and Wilford, D.S. (1977) A currency portfolio approach to exchange rate determination. (Manuscript, Federal Reserve Bank of New York). Levich, R.M. (1978)
Tests of forecasting models and market efficiency in the international money market. :me economics of exchange rates: selected studies. eds. J.A. Frenkel and H.G. Johnson. Reading, Mass.: Addison-Wesley.
Miles, Ed.
(1978)
Mussa, M. (1976)
Currency substitution, flexible exchange rates and monetary independence. American Economic Review,. 68: 428-36The exchange rate, the balance of payments, and monetary
and
fscal policy under a regime
of controlled
ScandinavianJournal of Economics, 78; 22948.
floating.
&ifj#+t, T., ind ‘Wallace,N. (1973)
Rationat expectations and the dyrmwics of hyperinflation. ~nternutfofa&?konomic Review, 14: 32840.
Stwkman, AC (4 508) ‘, “: .&i&i, *fqmatSan and forward exchange rates. Tlze economics -of ~e&cfzan& Wei: selected studies. eds. J.A. Frenkel and H.G. Johnson. Reading, Mass.: Addison-Wesley.
101