Content uploaded by James Robert Lothian

Author content

All content in this area was uploaded by James Robert Lothian

Content may be subject to copyright.

A History of Yen Exchange Rates

James R. Lothian*

Fordham University

Graduate School of Business Administration

New York, N.Y. 10023, USA

January 19, 1991

Forthcoming in Japanese Financial Market Research,

William T. Ziemba, Warren Bailey and

Yasushi Hamao, eds., Amsterdam: Elsevier, 1991

Abstract

The history of Japanese exchange rates, though short by British or

American standards, is exceedingly rich, both from the standpoint of

variation in the data and in the institutions governing exchange rate

arrangements and Japanese monetary conditions. This paper reviews that

history and traces the evolution of yen-dollar and yen-sterling exchange

rates from 1874 until the present, comparing their behavior to that of the

dollar-sterling rate. It shows the relationships of all three nominal

exchange rates to indexes of purchasing power parity, and it investigates the

links among exchange-rate regimes, exchange rates themselves and other

macroeconomic variables. Two conclusions emerge: (1) Purchasing power parity

— at least in relative form — held remarkably well for the yen over the

longer run. (2) The variability of real yen exchange rates under the current

float does not, in fact, differ greatly from the often substantial and

largely self-reversing movements observed historically.

* I would like to thank Cornelia McCarthy for comments on and assistance in

the preparation of this paper, and Kyung-Won Lee and Dario Werthein for their

computational help. Most important is the debt that I owe to the late John

Metcalf who stimulated my interest in Japanese monetary issues and directed

me to what had then been a largely unexploited body of data. This paper

extends and in several places draws upon my earlier paper "A Century Plus of

Yen Exchange Rate Behavior," written under the sponsorship of the Japan-U.S.

Center for Business and Economic Studies of New York University and published

in Japan and the World Economy.

1

1. INTRODUCTION

The history of Japanese exchange rates, though short by British or

American standards, is exceedingly rich, both from the standpoint of

variation in the data and in the institutions governing exchange rate

arrangements and Japanese monetary conditions. In the approximate century

and a quarter that followed the Meiji restoration of 1867, Japan experienced

three episodes of floating exchange rates, two episodes — one very brief — on

the gold standard, a period of heavily managed floating rates during the

interwar years, the dollar peg of the Bretton Woods era, and the

exceptionally severe inflation and accompanying substantial yen depreciation

of the World War II years.

In this paper I review that history. I trace the evolution of yen-

dollar and yen-sterling exchange rates from 1874 until the present and

compare the behavior of both to that of the dollar-sterling rate. I go on to

analyze the relationship of all three to indexes of purchasing power parity

and to examine the links among exchange-rate regimes and the behavior of

exchange rates themselves and of other important macroeconomic variables.

2. HISTORICAL DESCRIPTION OF EXCHANGE-RATE BEHAVIOR

Following the Meiji restoration, Japanese trade with the rest of the

world increased dramatically. From a base of essentially nil at the time of

Admiral Perry's first visit in 1853, exports rose to approximately 15% of GDP

shortly after the turn of the century and to over 20% of GDP by the 1920s.1

For most of the nineteenth century, however, Japan was not on the gold

standard. From 1867 to 1878, the Japanese monetary system was effectively a

system of fiat currency and floating exchange rates. From 1878 until 1897,

when Japan did finally adopt gold, Japan both de jure and de facto was on a

silver standard. The result was a continuation of floating rates relative to

the gold standard world.2

This flexibility in exchange rates enabled Japan to avoid the deflation

that prevailed in Britain, America, and other countries on gold during these

years. We can see this in Table 1.3 Over the subperiod 1874 to 1887 Japanese

wholesale prices showed virtually no net change, earlier inflation being

offset by later deflation. Over the later subperiod 1888 to 1896, they

actually rose by 35%. In the United States, wholesale prices declined by

slightly over 40% between 1874 and 1887 and by another 21% between 1888 and

1896. In Britain, the deflation followed a largely similar pattern,

cumulative declines in wholesale prices of 41% and 11% in the two subperiods,

respectively.

In the foreign exchange market, the yen depreciated relative to both

the dollar and sterling. For the full period 1874 to 1896, the increase in

the yen-dollar rate averaged 3.15% per year and the increase in the yen-

sterling rate, 3.03% per year. In both instances, the declines were more or

less in line with the cross-country differences in inflation rates, but in

neither case was the offset exact. In real terms, the yen appreciated

somewhat against both currencies, while the dollar fell slightly against

sterling. We can see this in the figures presented in the last three columns

of Table 1 showing the average annual percentage changes in real exchange

rates and in the plot of the real yen-dollar and real pound-dollar rates in

2

Figure 1.4

Purchasing power parity, therefore, held tolerably well as a first

approximation. But for Japan in particular it was only that. One

possibility for the slack in the relationship between changes in nominal

exchange rates and the differential in inflation rates is measurement error

in the Japanese price data, an overstatement of the levels of wholesale

prices in the later relative to the later years of the period. An

alternative is a shift in the equilibrium real exchange rate, resulting

perhaps from increased productivity in the Japanese tradeable goods sector.5

In addition to these trend-like movements, both yen real rates

exhibited fairly sizable year-to-year variability, well in excess of the

variability of dollar-sterling, particularly in the subperiod from 1888 to

1896. This variability, however, had no obviously adverse effects on other

real variables, either real trade flows or real output. Exports and imports

increased substantially as already noted. Rates of growth of Japanese real

income and industrial production were comparable to or higher than the rates

in Britain and America. Between 1885 and 1900, real income in Japan

increased at an average annual rate of 3.1% versus 2.35% in the United

Kingdom and 2.79 in the United States. Over the longer period 1874-1896,

industrial production increased by 4.75% per year in Japan versus 1.96% per

year and 4.44% per year in the United Kingdom and the United States

3

respectively.6

Table 1

Rates of change of wholesale prices, nominal exchange rates and real exchange

rates in Japan, the United States and the United Kingdom, 1875-1989

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Wholesale prices Nominal rates Real rates

S)))))))))))))))))Q S)))))))))))))))))Q S))))))))))))))))Q

Period Japan U.S. U.K. ¥/$ ¥/£ £/$ ¥/$ ¥/£ £/$

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

1875-1896 1.43 -2.77 -2.30 3.15 3.03 0.12 -1.07 -0.71 -0.36

1875-1887 -0.27 -3.10 -3.03 2.16 2.06 0.10 -0.66 -0.69 0.03

1888-1896 3.91 -2.30 -1.24 4.56 4.41 0.15 -1.65 -0.74

-0.91

1897-1914 2.67 2.15 1.58 0.17 0.34 -0.17 -0.34 -0.74

0.40

1915-1940 3.76 0.54 1.96 2.86 1.62 1.24 -0.36 -0.17 -0.19

1915-1921 10.58 5.10 9.90 0.35 -2.91 3.26 -5.13 -3.59 -1.55

1922-1928 -5.21 -5.81 -9.77 0.51 3.72 -3.21 2.69 1.20 1.49

1929-1940 3.29 -1.76 1.27 5.69 3.03 2.66 0.64 1.01 -0.37

1941-1953 40.24 5.92 6.74 34.13 32.37 1.76 -0.18 -1.13 0.94

1954-1973 1.37 2.16 3.24 -1.42 -2.11 0.69 -0.62 -0.24 -0.38

1974-1989 2.85 5.71 9.74 -4.24 -6.23 1.99 -1.37 0.67 -2.04

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Source: See endnote 3.

Note: Figures are continuously compounded per cent per annum rates of

change.

The difference between Japanese and U.S. and U.K. price behavior in this

last quarter of the nineteenth century appears to a large extent to mirror

international developments rather than domestic monetary policy actions in

Japan. Throughout these years, countries were continually shifting from

silver to a gold standard. At the same time that this shift in the relative

demands for the two precious metals was taking place, discoveries of silver

in the United States were increasing its supply. The result, as Irving

Fisher (1911) documented, was an upward trend in the price levels of the

countries that remained on silver and a downward trend in the price levels of

countries on gold.

When Japan did make the switch to gold in 1897, price movements in the

gold-standard world had begun to reverse. Gold discoveries, coupled with the

introduction of improved methods of refining, led to a more than doubling of

the world's gold stock from 1890 to 1914. (Friedman and Schwartz, 1963, p.

137). This in turn increased growth in money supplies and thus caused prices

to rise. In Japan, the inflation exceeded that of the rest of the world.

Given the fixed nominal exchange rate, the yen therefore depreciated against

both the dollar and sterling in real terms. Government borrowing abroad and

the drawing down of the gold balances received as an indemnity payment

following the war with China in 1893 and 1894 are the likely reasons for this

disparity in price behavior.7 These enabled Japan to insulate itself from the

operation of international gold-standard forces and thus to pursue

inflationary policies domestically.

In 1914, with the outbreak of war in Europe, the worldwide gold standard

4

broke down, most countries blocking its workings by placing embargoes on gold

exports, some leaving gold completely. Japan took the former route until

1917, at which point it left gold officially. Unlike Britain and America,

Japan remained off the gold standard throughout the 1920s, not returning

until the beginning of 1930, and then only to leave a scant two years later.

The twenties were not, however, years of completely free floating yen

exchange rates. An initial attempt to peg the yen against the dollar (1919

to 1920) was followed by controlled depreciation and then two planned, but

aborted, moves to return to gold. The first attempt failed in the wake of

the Great Kanto Earthquake of 1923 and the large trade deficits that

resulted. The second, after a year of preliminary stabilization of the yen,

was abandoned following the financial panic of March 1927 (Takagi, 1989).

The years 1932 to 1934 were marked by a more or less free yen float, but the

five years that followed by a peg of the yen to sterling.

What stands out during the 1915-1940 period is the volatility of real

exchange rates. Judged in terms of both the subperiod-average rates of

change shown in Table 1 and the yearly standard deviations shown below in

Table 2, it was at record highs, far greater than under the gold standard

and, in the main, even greater than under the post-Bretton Woods float. The

only sense in which this is not the case is for the period viewed as a whole.

The average annual rates of change of all three real rates over the years

1915 to 1940 were in fact less than the respective averages for the years

prior to 1914. In each instance, protracted movements of the real exchange

rate in one direction during WWI and the years immediately thereafter were

very nearly offset by subsequent protracted movements in the opposite

direction. Over the longest period, therefore, purchasing power parity again

appears to have held, despite substantial departures for long periods in

between.

With World War II came the disruption of international transactions,

breaks in the official yen exchange-rate data and substantial increases in

price levels around the world. In Japan, the inflation was severe. Between

1940 and 1949 Japanese wholesale prices increased by a multiple of 127, or at

a continuously compounded average annual rate of increase of 53.8%. Relative

to American and British wholesale prices, this translated into (continuously

compounded) cumulative increases in excess of 400%.

In 1948, when official data for yen exchange rates become available, its

value relative to the dollar had fallen from the 4.35 yen per dollar rate in

place in 1940 to 160 yen per dollar. By 1950, in the face of continued

strong inflation, it reached 361 yen per dollar, roughly the rate maintained

for the remainder of the Bretton Woods era.

Then in 1971, in the face of monetary excesses in the United States, the

reserve-currency country, the Bretton Woods system broke down and the current

float began. Since then the yen has shown a trend-like nominal appreciation

against the dollar and sterling, a similar real appreciation against the

dollar, though somewhat surprisingly, not sterling, and a series of

alternating sharp shorter term real appreciations and depreciations against

both currencies.

The monetary part of the picture during the early years of the float can

be divided into three episodes, all of which show evidence of links to U.S.

policy. In the mid 1970s, Japan in part as a spillover from policy in the

United States, experienced both high money growth and inflation (Darby and

Lothian, 1983b). In its aftermath, however, Japanese policy became

5

considerably tighter than policy in the United States and remained so longer.

Despite renewed expansionary effects emanating from the United States via the

balance of payments, Japan therefore escaped the double-digit inflation that

plagued America and Britain at the start of the last decade.8 In the latter

part of the 1980s, Japan appears to have again been led into expansive policy

as the Bank of Japan, along with the U.S. Federal Reserve, sought to halt the

depreciation of the dollar.

Of particular interest in these episodes is the pattern of volatility of

real exchange rates and of inflation. Over time, fluctuations in inflation

have become more muted in Japan while fluctuations in the real yen- dollar

rate have remained substantial. This, coupled with the longer term downward

trend in the yen-dollar real exchange rate have been a source of increased

skepticism about both purchasing power parity as an equilibrium condition and

the functioning of the floating rate system.

3. PPP, REAL EXCHANGE RATES AND EXCHANGE-RATE REGIMES

The historical overview highlights several sets of important issues. One

has to do with the purchasing power parity relationship. The other centers

around the links between exchange-rate regimes, exchange-rate variability and

the behavior of other macroeconomic variables. The specific question that

arises with regard to PPP is whether the tendency for nominal exchange rates

to move in line with relative price levels that is apparent in the longer

term comparisons presented in Table 1 is a behavioral phenomenon or simply a

statistical quirk, the spurious correlation that can arise between two

trended series.

The theoretical rationale for PPP and modified PPP relationships is as a

macroeconomic equilibrium condition. In the simplest theoretical models,

which ignore the effects of real variables like productivity and differences

in relative prices of traded and non-traded goods, absolute PPP holds. It is

the open-economy analogue of the classical closed-economy neutrality

proposition as in in the monetary-approach models of the type developed in

Frenkel and Johnson (1976). Expressed in log form, the absolute PPP

relationship is

pt - et = pt* , (1)

where p, and p* represent the logarithms of the price levels in the home

country and the foreign country respectively, e represents the logarithm of

the nominal exchange rate (the price in the home country's currency of a unit

of the foreign country's currency), and t is an index of time.

There is now abundant empirical evidence suggesting that to the extent

that purchasing power parity holds, it does so only over longer time periods,

and that even then there may be disturbances to the relationship that are

highly persistent in their effects. Two general classes of models have

evolved to explain this phenomenon. In one, which is an extension of the

monetary-approach model, such deviations are purely transient, the result of

sluggish adjustment of prices to monetary shocks. (See, e.g., Dornbusch,

6

1976). In the other, which assumes instantaneous adjustment of goods prices

and thus the price level, deviations from PPP can be permanent, the result of

real shocks that affect the equilibrium real exchange rate (See Stockman,

1980). Models of this class preserve the long-run neutrality (or super-

neutrality) of money but see absolute PPP as a highly special case that would

only exist in the limit, in situations in which real influences were of no

practical significance. Relative PPP does, however, potentially fare better

in these models. Real shocks have one-time effects on levels. As the time

period lengthens, the effect on rates of change therefore progressively

diminishes.

Extending the concept of neutrality in another direction, Stockman (1983)

demonstrated that within the context of an equilibrium model similar to the

one developed in his earlier (1980) paper, and as intuition might suggest,

the (nominal) exchange-rate regime should have no effect on the equilibrium

behavior of real variables, including that of the real exchange rate.

Like continuous PPP, this insight does not appear to carry over to the

actual data. Real exchange rates, as Stockman went on to show and as

evidence in Mussa (1986) confirms, have been more variable under floating

rates than fixed rates during the post-WWII period. The behavior of other

real variables, however, appears to be invariant across the two regimes

(Baxter and Stockman, 1989; Baxter, 1991).

The major difficulty that arises in interpreting these results is that the

regime is essentially an endogenous variable. The choice of regime very

likely is influenced by factors that, in turn, affect monetary and price-

level behavior, and also by the behavior of real variables, including the

real exchange rate itself.9 How real exchange rates behave under different

regimes, as well as the choice of the regime, may be two aspects of the same

general question.

Below I examine these issues, the time-series behavior of PPP and the

relationships between exchange-rate regimes and the variability of real

exchange-rates. I begin with the analysis of real-exchange rate variability

since it is more heavily descriptive and thus serves as useful introduction

to the time-series analysis of PPP that follows.

3.1 REAL-EXCHANGE-RATE VARIABILITY AND EXCHANGE-RATE REGIMES

The data in Table 2 showing standard deviations of the log real exchange

rates and their first differences for various subperiods appear fully

consistent with the hypothesis of greater variability under floating than

fixed exchange rates. Standard deviations for the current floating rate

period, for the interwar period and the Japanese nineteenth century float are

indeed greater than the standard deviations for the Bretton Woods and gold

standard periods in most instances.10

This, however, does not appear to be the full story. One hint that more

is involved than a simple fixed-floating dichotomy is provided by experience

during the Japanese float in the latter decades of the nineteenth century.

In this episode there is a marked difference in variability — particularly in

the variability of the differenced data — between the subperiods 1875-1887

and 1888-1896. In the first of these subperiods, Japanese real exchange

rates were noticeably less variable than in the second and not appreciably

more variable than the real pound-dollar rate. Quite interestingly, this

7

first subperiod saw much greater stability of Japanese inflation rates than

the second.

Table 2

Standard deviations of real exchange rates, 1875-1989

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Period Countries

S)))))Q S))))))))))))))))))))))))))))))))))))))))))))))))Q

Levels of logs Differences in logs

S))))))))))))))))))))Q S)))))))))))))))))))))))Q

JA/US JA/UK UK/US JA/US JA/UK UK/US

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

1875-1896 8.80 7.71 4.95 8.77 6.95 4.77

1875-1887 8.31

6.29 5.89 7.90 5.26 5.28

1888-1896 8.22 7.74 3.47 10.38 9.24 4.15

1897-1914 5.50 7.93 6.31 5.52 3.82 3.95

1915-1940 17.67 18.03 12.46 9.14 11.88 10.53

1921-1928 8.72 8.83 4.07 8.83 9.62 5.57

1929-1940 14.38 16.69 13.39 9.31 13.07 14.20

1954-1973 5.43 5.03 3.89 4.37 4.45 3.23

1974-1989 13.64 9.12 13.71 10.15 10.00 10.82

S))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Note: Figures are multiplied by 100 to convert to per cent terms.

A more serious problem is posed by the interwar years. This was a period

of managed floats alternating with pegs of various sorts, over various time

spans, depending upon the currency. Real exchange rates, however, were even

less stable during the interwar years than under the current float. Hence,

if there is a behavioral-type relationship between the degree of exchange

rate flexibility and the variability of real exchange rates it is certainly

not monotonic. Alternatively, it may be that the relationship is largely

statistical, economic conditions that give rise to variability in real

exchange rates also strongly influencing the choice of regime.

To investigate these issues further, I computed standard deviations of the

changes in the log real exchange rates and of the three countries' inflation

rates for five-year periods. I then used these as the observations in a

series of dummy-variable regressions. These regressions took the general

form:

Fxj = (0 + (1 DFIX + (2 DIW + (3 DWW + ,j , (2)

where Fx is the standard deviation of variable x, DFIX is a dummy taking the

value 1 for fixed-rate periods (the gold-standard years and the years of

greatest stability under Bretton Woods) and 0 otherwise, DIW is a dummy

taking the value 1 for the interwar period and 0 otherwise, DWW is a dummy

taking the value 1 for the two world wars and 0 otherwise, the (s are

8

coefficients to be estimated, , is the error term, and j is an index for the

period. Table 3 contains the results of these regressions.

Under the hypothesis that the regime per se is the determinant of real-

exchange-rate variability, (1, (2 and (3 should all be negative, and (2 and (3

should each be less in absolute value than (1. The data, in general, do not

support these predictions.

Table 3

Regressions to analyze the variability of real exchange rates and inflation

rates across regimes

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Countries (0 (1 (2 (3

R2SEE

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Fxj = (0 + (1 DFIX + (2 DIW + (3 DWW + ,j,

)Log real exchange rate

S))))))))))))))))))))))Q

JA/US .081 -.042 .002 .086 .294 .068

(3.517) (-1.311) (0.048) (2.216)

JA/UK .086 -.060 .032 .134 .413 .081

(3.152) (-1.577) (0.665) (2.899)

UK/US .087 -.051 .031 .015 .490 .036

(4.534) (-2.145) (0.886) (.778)

Inflation rate

S)))))))))))))Q

JA .051 -.018 .062 .182 .386 .098

(1.554) (-.395) (1.056) (3.256)

US .052 -.023 .054 .040 .395 .039

(3.917) (1.264) (2.279) (1.798)

UK .043 -.010 .069 .048 .378 .045

(1.840) (-.402) (2.113) (1.631)

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Source: Observations are standard deviations of annual data for non-

overlapping five-year periods.

Note: Fx is the standard deviation of the variable x, DFIX is a dummy for

fixed-rate periods, DIW is a dummy for the interwar period, DWW is a dummy

taking the value 1 for the two world wars and 0 otherwise. Figures in

parentheses are t statistics.

In all three instances, DFIX has a negative sign, which is consistent with

the view that the variability of real exchange rates is low under fixed

rates. But only in the case of the pound-dollar rate is the difference

statistically significant. For the two yen rates only DWW is significant.

Much more important, in all three cases there appears to be little difference

9

statistically between the interwar period and periods of relatively free

float: DIW is insignificant and slightly positive in all three instances.

Certainly, there is no evidence of lower variability then as would be the

case if the degree of variability of real exchange rates were directly

related to the degree of flexibility of nominal exchange rates.

The inflation-rate regressions provide a partial clue to what underlies

these results. For all three countries, DWW is significantly positive and

for Japan especially so. For the United States and the United Kingdom, so

also is DIW. DFIX is negative in all three instances but in none of the

regressions is it significant.

Both bodies of data, therefore, show somewhat similar temporal patterns.

and, in the case of Japan, a similarity that is particularly pronounced. The

close association observed between movements in nominal and real exchange

rates since the advent of floating exchange rates in the early 1970s is,

therefore, clearly not a general phenomenon. Over the long span of years

covered by these data, variability in inflation rates very often has

accompanied variability in real exchange rates. This continued association

between the two suggests the need for a common explanation.

3.2 PPP: Evidence from tests of cointegration

The implications of the theoretical models of exchange rate determination

reviewed above translate into competing hypotheses about the nature of the

disturbances that affect the PPP relationship.

To illustrate, let us first amend equation (1) to include a disturbance

term and also, since it will be of use empirically, consider an alternate

version written in terms of the real exchange rate.

As the empirical counterpart of equation (1), we get:

pta = a + pt* + ut , (3)

and as the real-exchange-rate equation,

qt = aN + utN , (4)

where pta / pt - et, qt / (et - pt + pt*), p and p* now represent logarithms of

price indexes rather than price levels, a and aN are normalizing constants,

and u and uN are error terms.

Particularly well suited to analyzing the processes governing these errors

are the tests of cointegration developed by Engle and Granger (1987).

According to Engle and Granger, two series which themselves have to be

differenced n times to be stationary, are said to be cointegrated of order n-

1 if some linear combination of their nth differences is stationary.11

Cointegrated variables have the property that even though, for example, the

levels of both may be subject to drift, there is some linear combination of

the two that is not. Tests for cointegration are therefore essentially unit-

root tests applied to the errors in these equations, or to suitably

differenced versions thereof.

10

To see the connection between the tests and the hypotheses about exchange

rate behavior suggested by theory, let us first consider the (instantaneous)

equilibrium model. Since in this model both goods prices and the nominal

exchange rate adjust fully within the period to monetary shocks, deviations

from PPP are solely the result of real shocks. These real shocks both have

permanent effects and provoke full within-period adjustment. Accordingly,

the equilibrium model predicts: non-stationary of the uts (or utNs) and hence

non-cointegration of pta and pt*; the necessity of differencing to achieve

stationarity; and an extremely rapid pattern of adjustment.

In the monetary overshooting model, real shocks play no role. Deviations

from PPP are solely the result of lagged adjustment to monetary shocks. In

the long run, the adjustment of nominal exchange rates and the price level to

such shocks is complete, but in the short run the exchange rate overshoots to

maintain covered interest rate parity, while the price level adjusts slowly.

The overshooting model, therefore, predicts: stationarity of the uts and,

hence, cointegration of the levels of pta and pt*; and a pattern of adjustment

that corresponds to that of the price level.

A third more general class of models admits the possibilities of both real

shocks influencing the equilibrium exchange rate and sluggish adjustments of

prices (see e.g. Mussa, 1982). The predictions of these models are,

therefore, a mixture of those of the other two: stationarity of the errors,

at least after differencing or adjustment for a deterministic trend in the

real exchange rate; and a pattern of adjustment to residual errors that again

mimics that of the price level following a monetary shock.

To test these hypotheses, I followed two related procedures. One was the

two-step method outlined by Engle and Granger (1987); the other was simply to

apply unit-root tests in the context of simple univariate models of the real

exchange rate.

In the two-step procedure, the first step was to estimate the

cointegrating regression based on equation (3):

pta = a + b pt* + ut , (5)

where a and b are the coefficients to be estimated and where the slope

coefficient b, which in theory should be unity, is included as an allowance

for measurement error (Taylor, 1988).12

The second step uses variants of the Dickey-Fuller (1979) test to examine

the stationarity of the uts.13 The equations underlying these tests took the

general form:

K

)ut = $1 ut-1 + E $k-1 )ut-k + vt. (6)

k=1

Here the parameter of interest is $1, the coefficient on the level of the

lagged error term from the cointegrating regression. A negative and

statistically significant value of $1 leads to rejection of the hypothesis of

non-stationarity. This, in turn, implies that pa and p* are cointegrated and

hence is evidence in favor of long-run PPP. In one variant of the test, the

11

coefficients on the lagged differences of the errors were constrained to

zero. This is referred to as the DF (Dickey-Fuller) test. In the other, no

such constraint was imposed. This variant of the test is referred to as ADF

(augmented Dickey-Fuller) test.14

The alternative procedure, in which I examined the stationarity of the

real exchange rate directly, is essentially a test of cointegration subject

to the constraint that b, the slope coefficient in the cointegrating

regression, is unity. Given this constraint, the two steps collapse into one

and the following general equation serves as the basis for our tests:

K

)qt = :0 + :1 qt-1 + E 8k )qt-k + vt . (7)

k=1

Again the focus of the tests is on the coefficient of the lagged level, a

value of :1 significantly less than zero providing evidence in favor of the

hypothesis of stationarity of the real exchange rate. In the DF tests, the

8k again were assumed to be zero; in the ADF tests, the number of such

coefficients to be included in the regression was chosen empirically.

To investigate the influence of real variables and other factors that

might cause the real exchange rate to undergo permanent shifts, I conducted

two further series of tests. In the first, I estimated a variant of (7) that

included a deterministic time trend as an additional regressor. In the

second, I allowed for a stochastic trend by substituting log differences of

the real exchange rate data in place of the log levels used initially. In

both instances, I conducted unit-root tests similar to those used above.

The equations underlying these additional tests took the respective forms:

K

)qt = :0 + :1 qt-1 + :2 t + E 8k )qt-k + vt , (8)

k=1

and

K

)qt - )qt-1 = :0 + :1 )qt-1 + E 8k ()qt-k - )qt-k-1) + vt . (9)

k=1

Table 4 contains the results of the cointegration tests based on the two-

step procedure, both the DF and ADF t-like statistics and the Durbin-Watson

statistic from the cointegrating regression. Table 5 contains the statistics

for the analogous unit root tests for the real exchange rates. In each

instance in Table 4, we are able to reject the hypothesis of no cointegration

between pta and pt*. The same is true for the third test proposed by Engle and

Granger based upon the Durbin-Watson statistic. Correspondingly, we are able

to reject the hypothesis of non-stationarity of the real exchange rate, or

put another way, of no cointegration given the constraint of a unit

coefficient in the cointegrating regression linking pta and pt*.

Estimated speeds of adjustments to equilibrium are rather lengthy in both

instances.15 For the pound-dollar the estimated half lives range from 2.1

12

to 2.3 years based on the results reported in Table 4 and from 2.2 to 2.9

years based on those reported in Table 5. These are not the rapid speeds of

adjustment envisioned in the equilibrium models, but they are not out of line

with the estimated speeds of adjustment of price levels to monetary shocks in

the United States and the United Kingdom.16

Table 4

Tests for cointegration between the logarithms of exchange-rate-adjusted

price levels: 1875-1989

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Countries b DW 1-$1 DF 1-$1ADF(K) K

sb s$1 s$1

)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

JA/US 1.164 .540*.714 -4.486*.741 -4.663*2

(.019) (.064) (.077)

JA/UK 1.091 .553*.715 -4.421*.747 -3.392** 2

(.014) (.065) (.075)

US/UK 1.016 .474** .779 -3.495** .723 -3.583** 3

(.014) (.063) (.077)

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Notes: The slope coefficient in the cointegrating regression is denoted by

b and its standard error by sb. The coefficient in the test regression of

the lagged residual from the cointegrating regression, ut-1, is denoted by $1

and its standard error by s$1. DW is the Durbin-Watson statistic from the

cointegrating regression, DF is the Dickey-Fuller test statistic and ADF(K)

is the augmented Dickey-Fuller test statistic from a regression with K lagged

values of the differenced residuals. One, two, and three asterisks denote

significance at the .01, .05 and .10 levels, respectively in this and the

next table.

For the yen rates, the situation is similar when we confine our attention

to Table 4. The estimated half lives range from 2.1 to 2.4 years for the

yen-pound and from 2.1 to 2.3 years for the yen-dollar. The picture changes,

however, when we look at the results reported in the top panel of Table 5.

For the yen-dollar the estimates range from 3.7 to 4.2 years, and for the

yen-pound from 3.0 to 3.7 years.

The difference in the two sets of estimates for the yen exchange rates, I

believe, reflects the imposition of the constraint b=1 that is implicit in

the real-rate formulation. For the pound-dollar the constraint appears to be

reasonable. The estimate of b reported in Table 4 is virtually unity and

reversing the order of the variables in the cointegrating regression does not

alter this result. In the cointegrating regressions for the two yen rates,

in contrast, the estimates of b are both somewhat removed from unity, 1.16

and 1.09 for the yen-dollar and yen-pound, respectively. This, coupled with

the visual impression of a long-term drift in the yen-dollar rate that one

gets from Figure 1, raise questions about the importance of other than

13

transient shocks.

One possibility is outright measurement error, particularly for the price

series, but perhaps also for the early exchange rate data since these were

derived as midpoints of the yearly highs and lows published by the Bank of

Japan. The other obvious possibility is that real variables are affecting

these real exchange rates. As already noted, a number of researchers have

used differences in productivity growth in the United States and Japan to

account for the drift in the yen dollar rate over the post!WWII period (see

Table 5

Unit root tests of real exchange rates: 1875-1989

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Countries 1-:1 DF 1-:1ADF(K) K

s:1 s:1

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Levels K

S))))) )qt = :0 + :1 qt-1 + E 8k )qt-k + vt

k=1

JA/US .847 -2.987** .829 -2.887** 2

(.051) (.059)

JA/UK .792 -3.727* .828 -2.789** 2

(.056) (.062)

UK/US .786 -3.364* .729 -3.443*3

(.064) (.073)

Levels with trend

S))))))))))))))))Q K

)qt = :0 + :1 qt-1 + :2 t + E 8k )qt-k + vt

k=1

Countries 1-:1 :3 DF 1-:1 :3 ADF(K) K

s:1 s:3s:1 s:3

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

JA/US .741 -.0009 -4.040* .668 -.0012 -4.308*2

(.064) (.0003) (.077) (.0004)

JA/UK .646 -.0013 -4.903* .663 -.0012 3.847** 2

(.072) (.0004) (.088) (.0005)

UK/US .788 .0000 -3.291*** .733 .0000 3.277*** 3

(.065) (.0002) (.082) (.0002)

Differences

S))))))))))Q K

)qt - )qt-1 = :0 + :1 )qt-1 + E 8k ()qt-k - )qt-k-1) + vt

k=1

Countries DF ADF(K) K

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

14

JA/US -9.808* -7.704*3

JA/UK -10.750* -7.586* 2

UK/US -9.394* -6.026*3

S)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))Q

Notes: The symbols s:1 and s:3 represent the standard errors of the

regression coefficients :1 and :3. DF is the Dickey-Fuller test; ADF(K)

is the augmented Dickey-Fuller test based on a regression using n lagged

differences of the dependent variable.

Marston, 1986; and Yoshikawa, 1990). In light of the rapid growth of the

Japanese traded goods sector prior to WWII, it is plausible to believe that

productivity differences may have been important then also. Given the

somewhat low power of unit-root tests, we may therefore be incorrectly

rejecting the hypothesis of non-stationarity.

To investigate this possibility, I ran the additional tests based on

equations (8) and (9). In the middle panel of Table 5, I show the results of

the tests in which a deterministic trend is included in the regression. In

the bottom panel, I show the results for the first differences in q. In

both cases, we can reject non-stationarity, and generally at high levels of

significance. Real yen rates may be subject to permanent influences — either

stochastic shocks or forces that follow a trend-like pattern — but there is

a decided tendency to return to equilibrium otherwise. Over long periods PPP

appears to hold at least for rates of change, and perhaps also for levels,

albeit with the possible need for adjustment for a dterministic trend in the

real exchange rate.17

4. CONCLUSIONS

For the study of exchange-rate behavior, Japanese historical experience

offers a nearly ideal laboratory. Over the 115 years for which data are

available, the yen has floated with much greater frequency than either

sterling or the dollar, while the Japanese price level has been subject to

considerably greater variability than either the British or the American

price level.

Despite this volatility, however, purchasing power parity — at least in

relative form — has held remarkably well for the yen over the longer run.

Using an expanded body of data relative to that in my earlier (1990) study,

I can always reject the hypotheses of non-trend stationarity and non-

difference-stationarity of real yen exchange rates at high levels of

significance. Relaxing the constraint of a unit cointegrating factor between

the exchange-rate-adjusted Japanese price level and the foreign price level,

I can do the same for the hypothesis of non-stationarity of the absolute

levels.

These results stand in sharp contrast both to casual impressions gained

from experience under the current float and to much scholarly evidence

derived from it. The difference, I believe, is due in the main to the

difference in the spans of data. Given the existence of long-lived

deviations of exchange rates from PPP, a long historical series is a virtual

necessity if we are to distinguish persistent, but transient, deviations from

permanent ones.

Viewed from the perspective of this study, the continued emphasis by the

15

Bank of Japan and by business economists on PPP as a macroeconomic

equilibrium condition no longer appears anomalous.18 The fluctuations in the

real yen-dollar rate over the past decade that made it seem such turn out to

not differ much from the often substantial and largely self-reversing

movements observed historically. Relative to the fluctuations in the years

surrounding World War II they are, in fact, rather small.

Not directly identifiable are the factors producing such fluctuations.

For the sterling-dollar rate they appear to have been largely transitory in

their impact. For the two yen rates, both permanent and transitory factors

appear to have mattered. Productivity-related influences may well have been

important over much of the sample period. In addition, the close association

between variability in yen real exchange rates and variability in the

Japanese inflation rate suggests a major role for monetary influences.19

16

REFERENCES

Abuaf, Niso and Phillipe Jorion, 1990, Purchasing power parity in the long

run, Journal of Finance, 45, 157-174.

Baillie, Richard T. and Tim Bollerslev, 1989, Common stochastic trends in a

system of exchange rates, Journal of finance 44, 167-181.

Bank of Japan, Statistics Department, 1966. Hundred year statistics of the

Japanese economy (Bank of Japan, Tokyo).

Baxter, Marianne, 1991, Business cycles, stylized facts and the exchange rate

regime: Evidence from the United States, Journal of international money

and finance 10, forthcoming.

Baxter, Marianne and Alan C. Stockman, 1988, Business cycles and the exchange

rate system: some international evidence, Journal of monetary economics,

23, 377-400.

Darby, Michael R., 1983, Movements in purchasing power parity: The short and

long runs, in Michael R. Darby, James R. Lothian, et al., The inter-

national transmission of inflation (University of Chicago Press for the

NBER, Chicago, IL).

Darby, Michael R. and James R. Lothian, 1983a, British economic policy under

Margaret Thatcher: A midterm examination, Carnegie Rochester conference

series on public policy, 18, 157-208.

Darby, Michael R. and James R. Lothian, 1983b, Conclusions, in Michael R.

Darby, James R. Lothian, et al., The international transmission of

inflation (University of Chicago Press for the NBER, Chicago, IL).

Darby, Michael R. and James R. Lothian, 1989, International transmission

afloat, in Michael D. Bordo, ed., Money, history and international

finance: Essays in honor of Anna J. Schwartz (University of Chicago Press

for the NBER, Chicago, IL).

Dickey, David A. and Wayne A. Fuller, 1979, Distribution of the estimators

for autoregressive time series with a unit root, Journal of the American

statistical association 74, 427-431.

Diebold, Francis X., Stephen Husted and Mark Rush, "Real exchange rates under

the gold standard, unpublished manuscript, 1990.

Dornbusch, Rudiger, 1976, Expectations and exchange rate dynamics, Journal

of political economy 84, 1161-1176.

Enders, Walter, 1988, ARIMA and cointegration tests of purchasing power

parity, Review of economics and statistics 70, 504-508.

Enders, Walter, 1989, Unit roots and the real exchange rate before World

War I: The case of Britain and the United States, Journal of international

money and finance 8, 59-73.

Engle, Robert F. and Clive W.J. Granger, 1987, Cointegration and error

correction: Representation, estimation and testing, Econometrica 55,

251- 287.

Engle, Robert F. and Byung Sam Yoo, 1987, Forecasting and testing in

cointegrated systems, Journal of econometrics 35, 143-159.

Feinstein, C.H., 1972, National income, expenditure and output of the United

Kingdom, 1855-1965 (Cambridge University Press, Cambridge).

Fisher, Irving, The purchasing power of money (Augustus Kelley, New York).

[a reprint of the 1922 second revised edition of the 1911 original]

Frenkel, Jacob A., and Harry G. Johnson, 1976, eds, The monetary approach to

the balance of payments (University of Toronto, Toronto).

17

Friedman, Milton and Anna J. Schwartz, 1963, A monetary history of the United

States: 1867-1960 (Princeton University for the NBER, Princeton, NJ).

Friedman, Milton and Anna J. Schwartz, 1982, Monetary trends in the United

States and the United Kingdom (University of Chicago for the NBER,

Chicago, IL).

Fuller, Wayne A.,1976, Introduction to statistical time series (John Wiley

and Sons, New York, NY).

Gandolfi, Arthur E. and James R. Lothian, 1983, Price behavior and the demand

for money, in Michael R. Darby, James R. Lothian, et al., The inter-

national transmission of inflation (University of Chicago Press for the

NBER, Chicago, IL).

Hakkio, Craig S. and Mark Rush, 1989, Market efficiency and cointegration:

an application to the sterling and deutschemark exchange markets, Journal

of international money and finance 8, 75-88.

International Monetary Fund, International financial statistics, various

issues.

Lockwood, William W., 1954, The economic development of Japan: Growth and

structural change, 1868-1938 (Princeton University, Princeton, NJ).

Lothian, James R., 1986, Real dollar exchange rates under the Bretton-Woods

and floating-rate systems, Journal of international money and finance 5,

429-448.

Lothian, James R., 1990, A century plus of Japanese exchange rate behavior,

Japan and the world economy, 2, 47-70.

Lothian, James R., Michael R. Darby and Michael Tindall, 1990, Buffer stock

models of the demand for money and the conduct of monetary policy, Journal

of policy modeling, 12, 325-345.

Marston, Richard, 1986, real exchange rates and productivity growth in the

United States and Japan, NBER working paper no. 1922.

McNown, Robert and Myles S. Wallace, 1989, National price levels, purchasing

power parity and cointegration: A test of four high inflation economies,

Journal of international money and finance 8, 533-545.

Mitchell, B. R., 1975, European historical statistics, 1750-1970 (Columbia

University Press: New York, NY).

Moulton, Harold G., 1931, Japan: An economic and financial appraisal

(Brookings, Washington, DC).

Mussa, Michael, 1982, A model of exchange rate dynamics, Journal of political

economy 90, 74-104.

Mussa, Michael, 1986, Nominal exchange rate regimes and the behavior of

real exchange rates: Evidence and implications, in Karl Brunner and

Allan H. Meltzer, eds., Carnegie-Rochester conference series on

public policy 26 (North Holland, Amsterdam).

Nakamura, Takafusa, 1983, Economic growth in prewar Japan (Yale University

Press, New Haven). [ a translation by Robert A. Feldman of Senzenki nihon

keizai seicho no bunseki (An analysis of economic growth in prewar Japan)

IWANI SHOTEN publishers, Tokyo 1971].

Ohkawa, Kazushi and Henry Rosovsky, 1973, Japanese economic growth (Stanford

University, Stanford, CA).

Ohta, Takeshi, 1983, Exchange-rate management and the conduct of monetary

policy, in Paul Meek, ed., Central bank views on monetary targetting

(Federal Reserve Bank of New York, New York, NY).

18

Savvides, Andreas, 1991, Real exchange rate variability and the choice of

exchange rate regime, Journal of international money and finance, 9,

forthcoming.

Shinjo, Hiroshi, 1962, History of the yen (Tokyo, Kinokuniya Bookstore Co.,

Ltd. for the Research Institute for Economic and Business Administration

of Kobe University).

Stockman, Alan C., 1980, A theory of exchange rate determination, Journal of

political economy, 88, 673-698.

Stockman, Alan C., 1983, Real exchange rates under alternative monetary

systems, Journal of international money and finance 2, 147-166.

Stockman, Alan C., 1988, Real exchange rate variability under pegged and

floating nominal exchange rate systems: An equilibrium theory, University

of Rochester, Working paper no. 128 .

Sturm, Frederick W., 1990, The yen, the dollar and the pull to parity,

Fuji Securities, Inc., Perspectives, March 30.

Suzuki, Yoshio, 1988, Prospects for the future of the international monetary

system: A Japanese perspective, paper presented at the New York University

conference on international financial markets.

Takagi, Shinji, 1989, Floating exchange rates in interwar Japan, unpublished,

International Monetary Fund.

Taylor, Mark P., 1988, An empirical examination of long-run purchasing power

parity using cointegration techniques, Applied economics 20, 1369-1381.

Taylor, Mark P., and P.C. McMahon, 1988, Long-run purchasing power parity in

the 1920s, European economic review, 32, 179-197.

United States Department of Commerce, 1973, Long term economic trends (United

States Government Printing Office, Washington, DC).

Yoshikawa, Hiroshi, 1990, On the equilibrium yen-dollar rate, American

economic review 3, 576-583.

19

1. Data for Japanese nominal exports and imports came from Ohkawa and

Rosovsky (1973). I divided these figures by the estimates of nominal GDP in

Nakamura (1983) to convert them to ratio form.

2. The silver yen became legal tender in May of 1878. See the discussions

of this period in Moulton (1931) and Shinjo (1962) and the references cited

therein.

3. The exchange rate data for the period ending in 1965 are annual series

for the yen relative to the currencies of the other three countries as

reported in the Bank of Japan's Hundred Year Statistics of the Japanese

Economy. For the years 1880 to 1914, these are midpoints of the range

between the reported yearly high and low exchange rates for each currency;

for 1874 to 1879 and for the years after 1914, they are yearly averaged data.

These series were used directly to compute yen-other country real exchange

rates and to derive the sterling-dollar real exchange rate.

The published figures for 1894 contained what appears to be two errors.

I corrected the yen-dollar rate using the alternative estimate of Nakamura

(1983, p. 34) and derived the yen-sterling rate using that corrected figure

and the estimate of the dollar-pound exchange rate for 1894 in Friedman and

Schwartz (1982).

Observations for yen exchange rates for the years 1941 to 1947 were not

reported by the Bank of Japan. For these years, I used Swiss quotes of yen-

dollar rates from Abuaf and Jorion (1990) that Phillipe Jorion graciously

provided me, and the dollar-sterling figures reported in Friedman and

Schwartz (1982) to fill in the missing observations for the yen-sterling and

pound-dollar rates and to derive the yen-pound rate.

I updated the nominal exchange rate series for the years after 1965

using the basis of the annual average yen-dollar and dollar-pound rates

reported in the International Financial Statistics (IFS) and derived the yen-

pound rate from these figures.

The data for wholesale prices came from a variety of sources: for the

United States,the U.S. Department of Commerce's Long Term Trends for the

years 1873 to 1970, and the IFS thereafter; for the United Kingdom, European

Historical Statistics for the years 1873 to 1975 and IFS thereafter; and for

Japan, the Bank of Japan's Hundred Year Statistics for the years 1873 to 1965

and the IFS thereafter. These subseries were linked either by regression or

by multiplying the earlier series by the ratio of the overlapping

observations. The resultant series were then rebased to 1980.

4. I define the real exchange rate as the ratio of the nominal exchange

rate (the price of a unit of the foreign currency) divided by the ratio of

the home-country price index to the foreign-country price index.

5. This explanation has been widely applied to explain the behavior of the

yen-dollar rate in the post-WWII period. See, for example, Marston (1986)

and Yoshikawa (1990).

6. Japanese real income (GDP) and industrial production data came from

Nakamura (1983); U.S. and U.K. real income (NNP) data from Friedman and

NOTES

20

Schwartz, (1982); U.S. industrial production from U.S. Department of Commerce

(1973); and U.K. industrial production from Feinstein (1972).

7. See the discussion of this episode in Lockwood (1954, pp. 36-37) and the

references cited therein.

8. This short-run spillover of expansive U.S. policies to Japan is evident in

data for Japanese balance of payments and high-powered money growth. Japan's

official settlements surplus increased sharply in 1977 and remained high in

1978 in the face of substantial U.S. official settlements deficits in these

years. Japanese high-powered money growth in this environment rose from 8.1%

per year on average in 1976 and 1977 to 13.9% per year in 1978.

See Darby and Lothian (1989) on the difference between long-run and short-

run price behavior among OECD countries since the advent of floating rates

and Ohta (1983) for a discussion of Japanese policy during the late 1970s and

early 1980s.

9. Stockman (1988) attributes the greater stability of real exchange rates

under fixed rates to government actions in the goods and capital markets that

affect both the price level and the nominal exchange rate. Mussa (1986), in

contrast, attributes the variability under floating exchange rates to the

non-instantaneous adjustment of goods prices.

Savvides (1991) using a simultaneous model estimated for a group of 39

developing countries over the period 1976 to 1984 studies the relationship

between the exchange-rate regime and the behavior of real exchange rates. He

finds no independent effect of the regime on the variability of real exchange

rates.

10. The definition of the gold-standard period varies with the countries

being compared. For Japan it begins in 1897 and for the United States in

1879.

11. See the discussion of cointegration in Engle and Granger, 1987. I apply

this technique to long-term time series data for Japan, the United States,

the United Kingdom and France in my earlier (1990) paper. Unlike the data

used here those data excluded the WWII and immediate postwar years.

Other applications to exchange rate data are contained in Abuaf and Jorion

(1990), Baillie and Bollerslev (1989), Diebold, Husted and Rush (1990),

Enders (1988, 1989), Hakkio and Rush (1989), McNown and Wallace (1989),

Taylor (1988) and Taylor and McMahon (1988).

12. As I point out below, an estimate different from unity may also be an

indication of omitted variables that affect the equilibrium real exchange

rate.

13. Since there is no way a priori to choose the ordering of the variables,

I also ran the reverse set of regressions and conducted the corresponding

tests for cointegration. These resulted in no appreciable change in the

results reported below.

21

14. Significance levels for the t-statistics used in these tests are those

of Engle and Yoo (1987).

15. These estimated adjustment speeds are derived from the coefficients on

the lagged level terms in equations (6) and (7). The estimated half lives

of adjustment are ln(.5)/ln(1-$1) and ln(.5)/ln(1-:1) for the two equations

respectively.

16. For the United States and the United Kingdom, the lag before the effects

of a monetary shock become apparent in prices is often described as being on

the order of two years. Full adjustment — including overshooting — appears

to take longer, however. See Darby and Lothian (1983a) for a discussion of

this issue and for estimates of the adjustment process for the United Kingdom

that takes overshooting into account.

17. A similar phenomenon appears to characterize money-price relationships.

(See Gandolfi and Lothian, 1983; and Lothian, Darby and Tindall, 1990).

18. See Yoshio Suzuki (1988) of the Bank of Japan's research department for

an analysis based on long-run purchasing power parity of movements in the yen

exchange rates under the current float. Frederick W. Sturm (1989) of Fuji

Securities presents a similar analysis of recent movements in the yen-dollar,

DM-dollar, and dollar-pound exchange rates.

19. Evidence suggesting that this is also the case under the current float

is presented in my (1986) study of the real dollar exchange rates of 11 OECD

countries.