Since the introduction of a floating exchange rate for the yen in 1973, pressures on the yen and official intervention in the foreign exchange market have passed through a number of distinct phases. Although the direction of intervention in each case was relatively clear—that is, it was consistent with the intention of moderating the speed of exchange rate movements—evaluation of the response of the magnitude of intervention to pressures on the exchange rate is more difficult to effect. The purpose of this paper is to explore certain methods which might be brought to bear on this essentially quantitative problem using developments in Japan over the period March 1973–October 1976 as a reference. The paper also offers evidence refuting the view that foreign exchange market intervention has aimed at establishing a target level for the yen, rather than at simply moderating the speed of exchange rate change. It is not intended to focus on possible actions by the authorities other than purchases and sales in the foreign exchange market that might also influence exchange rate developments.

Abstract

Since the introduction of a floating exchange rate for the yen in 1973, pressures on the yen and official intervention in the foreign exchange market have passed through a number of distinct phases. Although the direction of intervention in each case was relatively clear—that is, it was consistent with the intention of moderating the speed of exchange rate movements—evaluation of the response of the magnitude of intervention to pressures on the exchange rate is more difficult to effect. The purpose of this paper is to explore certain methods which might be brought to bear on this essentially quantitative problem using developments in Japan over the period March 1973–October 1976 as a reference. The paper also offers evidence refuting the view that foreign exchange market intervention has aimed at establishing a target level for the yen, rather than at simply moderating the speed of exchange rate change. It is not intended to focus on possible actions by the authorities other than purchases and sales in the foreign exchange market that might also influence exchange rate developments.

Since the introduction of a floating exchange rate for the yen in 1973, pressures on the yen and official intervention in the foreign exchange market have passed through a number of distinct phases. Although the direction of intervention in each case was relatively clear—that is, it was consistent with the intention of moderating the speed of exchange rate movements—evaluation of the response of the magnitude of intervention to pressures on the exchange rate is more difficult to effect. The purpose of this paper is to explore certain methods which might be brought to bear on this essentially quantitative problem using developments in Japan over the period March 1973–October 1976 as a reference. The paper also offers evidence refuting the view that foreign exchange market intervention has aimed at establishing a target level for the yen, rather than at simply moderating the speed of exchange rate change. It is not intended to focus on possible actions by the authorities other than purchases and sales in the foreign exchange market that might also influence exchange rate developments.

To determine the response of intervention to exchange market conditions, it is necessary to investigate the reverse mechanism—the influence of intervention on the exchange rate. Accordingly, the paper also deals with the applicability of various theories of exchange rate determination, including the asset market equilibrium theory, to short-run developments in the yen exchange rate.

I. Definition and Measurement of Intervention by Bank of Japan

Two proxies have been used for intervention in the spot exchange market: changes in gross international reserves and changes in the Foreign Exchange Fund accounts maintained by the Bank of Japan. The latter series is published monthly in the table “Demand and Supply of Funds in Money Markets,” 1 which summarizes major elements influencing the stock of base money. It is generally held that there is no official intervention in the Tokyo forward market.

While Japan’s official external liabilities are small and relatively stable, its gross official reserve movements from month to month include, as they do in other countries, the effects of transactions conducted outside the foreign exchange market. These transactions are comprised of receipts of earnings on official holdings (only 5 per cent of reserves is held in the form of gold) and receipts arising from certain U. S. military purchases of Japanese goods and services. Exact figures are not published, but extra-market receipts are reported to average $150 million-$160 million per month.2 The Foreign Exchange Fund series is the counterpart of the Ministry of Finance’s account that is used for all transactions in the foreign exchange market. For the calculations here, the series has been converted into U. S. dollars at the average of the spot exchange rates for the current end of the month and the preceding end of the month. Aside from possible differences in the precise method of exchange rate conversion, the factors accounting for the differences between the two series are changes in official deposits of foreign exchange with Japanese commercial banks and receipts of foreign exchange outside the market. Unlike official reserve movements, changes in the Foreign Exchange Fund data exclude extra-market transactions and include changes in official deposits of foreign exchange with Japanese commercial banks (the so-called hidden reserves).3 In the early 1970s, large deposits were made with commercial banks, but the period under review, since the introduction of a flexible exchange rate for the yen in February 1973, has been marked by withdrawals of these amounts. Since July 1975, however, both intervention measures have corresponded closely, and any deviations have been well within the margin arising from possible errors in reproducing the method of exchange rate conversion. Consequently, any imputed changes in official deposits since July 1975 are small and of doubtful accuracy.

Data (compiled by the Bank for International Settlements) that combine net foreign liabilities of Japanese commercial banks with gross official reserves have been used elsewhere as a basis for analyzing the appropriateness of reserve movements. While this approach may have a certain validity in assessing targets for reserve levels, the fact that the authorities exercise a significant degree of influence over commercial bank external positions does not mean that changes in these positions should automatically be included in the concept of intervention. What should be included, if indeed it could be measured, is the effect of government intervention in modifying commercial bank borrowing and lending abroad from what it would otherwise have been.4 That this often works in the opposite direction to actual changes in Japanese commercial bank positions is suggested by the lack of a consistent relationship between signs on changes in the net commercial bank position and on either intervention proxy (reserve or exchange fund changes). (See Table 1.)

Table 1.

Japan: Intervention and Exchange Rate Changes, March 1973-October 19761

article image
Source: Bank of Japan, Economic Statistics Monthly, various issues.

In each period shown, net monthly intervention (measured as the change in the Foreign Exchange Fund of the Bank of Japan) was unidirectional, except for a small amount in July 1975. Reserves and exchange rate changes were also generally unidirectional in the same periods, but with more exceptional months—roughly one month per period for the exchange rate.

II. Different Intervention Rules and Frequency of Observation

In the tests conducted in this paper, a distinction has of necessity been made between the two intervention strategies—countering erratic fluctuations and leaning against the wind. The former strategy is more closely identified with intervention on a day-to-day and week-to-week basis and has been widely accepted, while the latter strategy implies resistance to market forces of longer duration (longer, say, than one month) and is more controversial.5 Both strategies are explicitly set out in Guidelines (1) and (2) of the “Guidelines for Management of Floating Exchange Rates” agreed upon by the Committee on Reform of the International Monetary System and Related Issues in June 1974:6

(1) A member with a floating exchange rate should intervene on the foreign exchange market as necessary to prevent or moderate sharp and disruptive fluctuations from day to day and from week to week in the exchange value of its currency.

(2) Subject to (3)(b),7 a member with a floating rate may act, through intervention or otherwise, to moderate movements in the exchange value of its currency from month to month and quarter to quarter, and is encouraged to do so, if necessary, where factors recognized to be temporary are at work. Subject to (1) and (3)(a),7 the member should not normally act aggressively with respect to the exchange value of its currency (i.e., should not so act as to depress that value when it is falling, or to enhance that value when it is rising).

In the declaration following the Rambouillet conference in November 1975 the former strategy received more emphasis: “At the same time our monetary authorities will act to counter disorderly market conditions or erratic fluctuations in exchange rates.” 8

In a recent document “Surveillance Over Exchange Rate Policies” detailing the International Monetary Fund’s responsibilities under Article IV of the proposed Second Amendment of its Articles of Agreement,9 one factor that would indicate the need to examine the appropriateness of exchange rate policies is the existence of “protracted large-scale intervention in one direction in the exchange market.” 10 Guidance in this document concerning which factors intervention by member governments should respond positively to is given in Principle B, “A member should intervene in the exchange market if necessary to counter disorderly conditions which may be characterized inter alia by disruptive short-term movements in the exchange value of its currency.” 11

Although both strategies could be captured by frequent observation—say, by use of daily or weekly data—data for the intervention proxies were available on a net monthly basis only. Hence, only the existence and form of intervention by the Bank of Japan that was consistent with leaning against the wind has been tested for. A priori, the existence of such a strategy is conceivable, as monthly exchange rate movements have been sustained in one direction for extended periods spanning virtually all of the past four years. (See Chart 1.)

Leaning against the wind was first proposed as an exchange rate rule by Paul Wonnacott (1965) in his review of the Canadian experience with a flexible rate. The rule may be formalized as:

I=a1S˙(1)
S˙=a2I+Σib2iXi;
a2I<Σib2ixi(2)

That is, intervention I responds automatically to the rate of exchange rate change S, but its effect in moderating appreciation or depreciation of the rate is (substantially) less than the effect of all other influences Xi on the exchange rate. Exchange rate movement is therefore moderated rather than completely offset by the policy response. This intervention rule is closely related to that of gliding parities based on previous market rates, in which intervention points are set around a moving average of previous levels of the exchange rate.12 Both schemes aim at moderating the speed of exchange rate movement, although the dampening influence is likely to be more pronounced for the gliding parity rule. Variants of the latter rule were found by Kenen (1975), in a series of simulations, to compare less than favorably with rules based on equilibrium rates, but to compare quite favorably with rules based on other objective indicators such as the basic balance of payments and changes in, and the level of, reserves. The gliding parity rule was found to be clearly superior to the other objective indicators except when exchange rate expectations were regressive to parity. The close relationship of leaning against the wind to this scheme suggests that it is also likely to possess desirable properties, relative to other objective indicator systems, in the present circumstance of generalized floating. There is the qualification that Kenen’s results appear to be sensitive to the length of the moving average.

Chart 1.

Japan: Intervention and Exchange Rate Movements, March 1973–October 1976

A04ct01
1 Change from the end of the preceding month.

“Leaning against the wind” is also advocated in Branson (1976) as a means of reducing exchange rate overshooting in response to monetary disturbances. Branson draws a parallel between this policy and existing policies in domestic asset markets. In contrast to the definition of leaning against the wind employed in this paper, however, Branson’s definition is based on movement about an equilibrium rate, rather than on simple exchange rate change from the preceding period. “First, permit the rate to move following the equilibrium value that balances the current account or basic balance, depending on whether the country is naturally a net lender. Second, intervene to the extent possible to dampen movements away from this value.” 13

III. Response of Intervention to Exchange Rate Changes

The simplest form of reaction function consistent with leaning against the wind would relate net intervention to the change in the yen exchange rate over the same period. Possible refinements would include allowance for lagged response to rate movements in previous months, for variations in foreign exchange market volume, for differences in the definition of both the exchange rate and intervention variables, and for differences in functional forms. An alternative hypothesis is that intervention has been directed toward maintaining a particular target level of the exchange rate. Both leaning against the wind (in its various forms) and the alternative hypothesis of targeting the exchange rate have been tested; the results are set out in Table 2, with the relevant notation appearing in Table 3.

Table 2.

Econometric Results: Intervention Equation1

article image

See Table 3 for notation. The t-ratios are shown in parentheses; R¯2 is the correlation coefficient adjusted for degrees of freedom; D−W is the Durbin-Watson statistic and H is the D−W test modified for the presence of a lagged endogenous variable.

Serial correlation is not rejected at the 5 per cent significance level.

Different forms of exchange rate target have been assumed. In equation (7), it is ¥ 300 per U.S. dollar; in equation (8), it is a rate set to maintain purchasing power parity between Japanese and U.S. wholesale prices from a 1970 base; in equation (9), the target moves with relative monthly changes in U.S./Japanese wholesale prices.

Table 3.

List and Definitions of Variables1

article image

All data have been taken from Bank of Japan, Economic Statistics Monthly, various issues, and from International Monetary Fund, International Financial Statistics, various issues.

The first observation on the outcome of the empirical tests is that, of the two intervention proxies—reserves and Foreign Exchange Fund changes—the latter was more closely related to exchange rate developments. (In Table 2, consider equation (4) as against equation (11). Equations (3)—(11) are presented only in Table 2.) This result is consistent with the widely-held view that the exchange fund proxy is conceptually closer to the official measure of intervention, especially in 1973 and 1974, when substantial differences in the two series reflected changes in official deposits with commercial banks. The second, and perhaps most important, observation is that the nominal exchange rate change vis-à-vis the U. S. dollar provided a better explanation of intervention than the change in the effective rate (according to the International Monetary Fund’s multilateral exchange rate model) and was considerably more significant than the deviation from various hypothetical target levels for the yen exchange rate. Three forms of deviation from target were tested: first, the percentage deviation from an assumed target of ¥ 300 (yen) per U. S. dollar; second, after Artus (1976), the percentage deviation from a target set to maintain constant Japanese/U.S. wholesale relative prices in dollar terms from a 1970 base; and third, a target that moved with monthly changes in the Japanese/U.S. wholesale relative prices.14 None of these variables proved significant at the 5 per cent level (See equations (7), (8), and (9), respectively.) This result for equation (7) was unaltered when estimation was confined to the period after January 1974, during which ¥ 300 per U. S. dollar has been the approximate “floor.” 15

Other aspects of the econometric results are the significance of the lagged endogenous variable, indicating response to exchange rate developments in previous months, and the rejection of various forms of non-linearity in response (using logarithmic transformations not reported in Table 2). The size of the market, as measured by the total volume of spot transactions (delivery today or tomorrow) on the Tokyo foreign exchange market, also proved to be highly significant in determining the magnitude of the intervention response. (See equation (4) in Table 2.) That this variable captures not only the response to scale factors, such as the growth of the Tokyo foreign exchange market in recent years but also the influence of speculative activity, was indicated by the addition of a further variable—the acceleration of the exchange rate—to equation (4). It was found that this variable and the market volume variable are closely related and do not assume separate significance. Equation (5), in which the acceleration of the exchange rate replaces market volume, is, however, less significant than equation (4).

The results of regression analysis on monthly data for the period following the February 1973 realignment therefore support the view that leaning against the wind to moderate market pressures has been a basic characteristic of Japanese intervention behavior. The most satisfactory equation obtained (equation (4), in which all variables are significant at better than the 1 per cent level) suggests, however, that this form of response is tempered by other considerations. First, policies are somewhat gradually modified and intervention reacts to the rate of exchange rate change in previous months; one third of intervention reflects the change in preceding months. Second, some proportionality is maintained between intervention and the total volume of spot transactions in the Tokyo market. Third, although it has not been specifically incorporated in equation (4), examination of residuals suggests that the largest and sharpest exchange rate fluctuations are generally accompanied by more than the estimated amount of intervention. (See Chart 2 for “actual” intervention and intervention estimated from equation (4).) In quantitative terms, this equation yields the result that a 1 per cent exchange rate change is accompanied by intervention amounting to $160 million in the current month and leads to intervention amounting to $80 million in subsequent months.16 The magnitude of the leaning against the wind response is, therefore, approximately $240 million for each 1 per cent movement of the exchange rate.

Chart 2.

Japan: Actual and Estimated Intervention, March 1973–October 1976

A04ct02

In Table 4, estimates of intervention from equation (4) are compared with actual intervention for each phase over the period March 1973 to September 1976. For instance, intervention in the period January–September 1976 was the largest for all periods of net monthly intervention in one direction since 1973. However, in 1976 the excess of actual intervention over that predicted on the basis of behavior in the 1973–76 period is small—in particular, it is smaller than the residual in the preceding period of support for the yen. Intervention in 1976 may therefore be viewed as broadly consistent both in direction and magnitude with previous adherence to leaning against the wind.

Table 4.

Japan: Actual and Predicted Intervention, March 1973–October 19761

(In billions of U.S. dollars)

article image

Changes in Foreign Exchange Fund; positive figures indicate net purchases of dollars.

From equation (4); see also Chart 2.

IV. Simultaneity Considerations

In the preceding analysis, no account was taken of the possible simultaneity problem arising from the use of monthly, rather than instantaneous, changes; the amount of exchange rate change in any month may already be affected by the intervention in that period, and may, therefore, give a distorted view of exchange market pressures.17 Although this would appear to argue for the use of very short-term (for instance, daily) observations to ensure a greater degree of recursiveness (which would also permit testing of smoothing behavior by the authorities), there are several reasons why it would probably create even greater problems for the elimination of simultaneity bias. First, daily or weekly exchange rate movements would reflect varied and transitory influences to a greater degree, in contrast to the more regular impact of monthly economic data and developments.18 Second, the psychological response of the market to efficiently executed intervention, as opposed to the actual effect of intervention on the excess supply of, or demand for, foreign exchange, or on asset markets in general, is likely to be more pronounced in the very short run. These problems would probably remain to some extent with monthly data, but with daily and weekly data there is reason to believe that they would be overwhelming.

Before proceeding further with the estimation it might be useful to consider another possible source of bias in estimating the intervention reaction function. For the sake of simplicity, let the intervention response be shown by the stable curve I in Figure 1 and the impact of intervention on the exchange rate be shown by the curve S. In the first instant within the period of observation, an appreciation s1 of the exchange rate corresponding to a shift of the curve S to S1 leads to intervention i1, the effects of which partly offset, by s2, the initial appreciation. A second-round intervention of i2, by which the authorities react to oppose the exchange rate effects of their own initial intervention, would seem unnecessarily self-defeating behavior; therefore, it could be argued that the process would stop before i2. Similarly, for depreciation corresponding to a shift of the curve S to S2, intervention would not proceed past the first round. The outcome of this would be that successive pairs of observed points corresponding to shifts of the curve S would lie on the curve î rather than on I, resulting in overestimation of the true magnitude of the intervention response.

This problem arises from the nature of the curve I, in that it represents the behavior of a single agent—the central bank. The perversity of a second-round response such as i2 would not be apparent if I represented the net effect of actions by a large number of market participants, with individuals not necessarily contradicting their immediately prior actions. The key to resolving the problem is the knowledge or lack of knowledge by the authorities of the impact of their intervention on the exchange rate. If the slope of the S curve is known to them, then the authorities may, as is shown in Figure 2, make an estimate of the new equilibrium and proceed directly to it via i1 and s2. If the nature of the S curve is not known, then convergence to the intersection of I and S1 will also occur, as shown in Figure 3, since it will not be possible for the authorities to dissociate higher-order effects s2, s3, s4, … from subsequent exogenous shocks of the type s1 The authorities will therefore have no option but to gradually converge on equilibrium through actions that are, in part, mutually offsetting.19

The upshot is that, in either case, whether or not the central bank knows the extent of the effect of its intervention on the exchange rate, estimation of the curve I will not be biased by lack of full convergence.

V. Intervention and Other Exchange Rate Determinants

In order to provide for possible simultaneous bias in the estimation of the intervention function, the role of other possible determinants of the yen exchange rate is investigated. Because of the relative novelty of generalized floating, empirical testing of the alternative theories of the exchange rate has not yet been very conclusive.20 To increase the likelihood of detecting bias, a broad class of alternative hypotheses of the short-run dynamics of the yen exchange rate is introduced. The discussion in this paper is brief and the testing is confined to reduced-form equations, as the focus of this paper is the nature of the policy reaction function. The empirical results nevertheless do have some implications for the relevance of exchange rate theory in the Japanese case.

The first approach is an essentially monetarist one, along the lines of the presentation in Dornbusch (1977). Purchasing power parity holds as a long-run equilibrium condition, equating the prices of traded goods PT at home (country 1) and abroad (country 2)

PT2=S1PT121(12)

Monetary equilibrium also requires that the demand for real balances L be equal to the nominal stock of money M deflated by the price level P

M1/P1=L1;M2/P2=L2(13)

Substituting equation (13) into equation (12) and assuming that an increase in the overall price level in the long run is fully reflected in traded goods prices

S1=(M2/M1)(L1/L2)(14)

and, taking logarithms and differentiating with respect to time,

S˙1=(M˙2L˙2)(M˙1L˙1)(15)

where part of the money supply corresponds to the effects of unsterilized intervention on the monetary base

S˙1=(M˙2L˙2)(M˙1+IL˙1)(16)

This is a model of long-run exchange rate determination. For monthly movements in the rate, the role of expectations becomes dominant. It may be assumed that expectations are formed rationally on the basis of these determinants and are therefore revised in the short run in response to various indicators of perceived relative monetary ease or restrictiveness in the two countries.

In the short run, the asset equilibrium approach to exchange rate determination also stresses the role of interest rate differentials, as goods markets are believed to adjust slowly. Under assumptions of mobility and extensive capital markets, interest parity requires that the domestic interest rate r1 plus the forward premium II1 be equal to the foreign interest rate r2—that is,

r1+II1=r2(17)

where

II1=f1S1S1(18)

Given that f1 = E(S1), and substituting equation (18) into equation (17), we get

E(S1)S1S1=r2r1(19)

and, with perfect foresight, this is equivalent to

S˙1=r2r1(20)

It may be noted that with less than perfect foresight over the period examined or in the presence of restrictions on capital mobility, the relationship specified in equation (20) will not hold exactly.

A third and more traditional approach embodies less than complete substitutability between external and domestic asset flows.22 In this case, demand and supply schedules for foreign exchange ($D and $S) are identifiable, and the corresponding excess demand or supply determines movements in the exchange rate

S˙=a($S$D)23(21)

Foreign exchange is demanded both for transactions and speculative purposes, with speculative demand depending on the expected change in the exchange rate

$D=bVcE(S˙)(22)

The flow supply of foreign exchange stems from the balance of payments outcome in the preceding period and from intervention. The current account balance T is determined primarily by relative prices P and incomes Y, and the capital account balance is determined by interest differentials r and expected exchange rate changes E(Ṡ).

$s=T(Y,P)+K[r,E(S˙)]I(23)

Substituting equations (22) and (23) into equation (21), we get

S˙=a{T(Y,P)+K[r,E(S˙)]I+cE(S˙)bV}24(24)

and, with expectations of future exchange rate movements formed rationally, the exchange rate is then related positively to the actual and expected values of r and negatively related to actual and expected values of Y, P, V, and I.

It may be noted that although equations (16) and (20) focus on financial variables and equation (24) focuses on real variables, the distinction between the reduced forms is not clear cut. Real variables would enter as arguments of money demand in equation (14) and would enter similarly, although more indirectly, in equation (18). Monetary influences would also be important for most of the variables in equation (24). This suggests that, as an initial test for simultaneity in the intervention function, all of these variables should be included in the equation determining the exchange rate. Further, as the object of the exercise is not prediction, ex post variables may be used—for instance, the balance of payments outcome subsequent to the actual month of intervention. Use of such variables assumes that the foreign exchange market has at its disposal sufficient information to form a reasonably accurate view of what the balance of payments will be two or three months ahead. It also assumes that exchange rate effects on trade are subject to a time lag in excess of two or three months. Given the fact that Japanese trade is largely denominated in dollars, this latter assumption would seem to be justified. Similarly, forward trade indicators such as export letters of credit and import approvals might contribute to the formation of very short-run expectations.

The choice of the appropriate time derivative for capturing movements in variables that would signal revisions in expectations is an important one. Partly as a result of empirical testing and partly on the basis of a priori considerations, money stock, income,25 and price differentials were expressed as accelerations (current month’s difference less preceding month’s difference between domestic and foreign rates of increase), while interest rates were expressed as the change in the differential (current month’s domestic/foreign differential less preceding month’s domestic/foreign differential) and the current balance of payments was expressed in the terms of monthly changes. Lags in these variables, reflecting either timing of information releases or delayed adjustments, were set purely on empirical grounds.

Estimation using a broadly inclusive equation and two-stage least squares did not establish any systematic role for the intervention variable in determining changes in the yen exchange rate 26

It=1,339(3.1)+139.3S˙t(3.2)+0.1218Vt(2.9)+0.3403It1(3.0)R¯2=0.47;H=0.98(25)
S˙t=0.0499(0.1)+0.0001It(0.2)+0.885r˙t(2.3)0.387r˙t1(1.0)(26)+0.7757r˙t2(2.3)0.0029M¨t1(0.0)0.2604M¨t2(0.7)0.0996M¨t3(0.4)0.2396Y¨t1(1.4)0.2112Y¨t2(1.0)0.2425Y¨t3(1.5)0.1356P¨t1(0.7)0.2662P¨t2(1.2)+0.0732P¨t3(0.4)+0.011B˙t+1(1.1)+0.022Bt+2(2.4)6.932D1(3.5)2.685D2(2.5)R¯2=0.50;DW=1.71

No simultaneous bias in the ordinary least-squares estimation of the reaction function was therefore detected, and the coefficients of the intervention equation (25) accordingly differ very little from those of their single-equation counterpart (4).

A few comments may also be made on equation (26) per se. This equation is only moderately satisfactory, explaining some 70 per cent (unadjusted R2 = 0.71) of monthly changes in the yen/dollar exchange rate.27 Nevertheless, both the expected current account balance of payments variable and interest rate variable emerge as significant at the 5 per cent level. The significance of the interest rate variable provides some support for the substitutability hypothesis, although the existence of sizable covered interest differentials for various periods over the last three years suggests that the actual extent of substitution between foreign and domestic assets is limited. Coefficients of other variables (including forward trade indicators) were of relatively low significance, but for the most part they had the expected signs.

A further approach, after Artus (1976), in which an equation expressing short-term capital flows as a function of the elements of expected yield is solved for the exchange rate, was also applied to the Japanese data.28 The failure of this equation to attribute significance to the capital-flow variable also indicates limitations on the response of Japanese capital flows to yield incentives. Tests of the purchasing-power-parity (PPP) hypotheses were also conducted using both monthly wholesale price series and monthly export price series over the period March 1973–November 1976. Regression results, however, indicated an insignificant role for the wholesale price index (WPI) over that period, while the results for the export price index were contrary to PPP.29

VI. Conclusions

The primary objective of the analysis conducted here has been to demonstrate the nature of foreign exchange market intervention in Japan since the advent of a floating rate for the yen. Empirical testing has established highly significant regularities in the response of intervention by the Bank of Japan to market developments over the period examined (March 1973–October 1976). These may be summarized as follows:

(1) There is clear evidence that Japanese intervention behavior has been consistent, both in direction and magnitude, with leaning against the wind. It has sought to moderate movement in the yen/dollar exchange rate in either direction by providing sustained support to either the dollar or the yen—in each case for periods exceeding two months and up to eight months (as in 1976). No evidence was obtained to support the view that intervention has been directed to a target level for the exchange rate, or that intervention has responded more closely to movements in the effective yen exchange rate.

Over this period, each 1 per cent monthly movement in the exchange rate was generally accompanied by close to $240 million net intervention.

(2) Intervention has responded not only to exchange rate movement in the current month but has also responded significantly to movements in preceding months. Of the $240 million overall response, about $160 million has taken place in the same month as the exchange rate movement and about $80 million has taken place in subsequent months.

(3) The stronger the underlying pressures in the foreign exchange market (as indicated by an acceleration of the exchange rate or an increase in the spot transactions volume), the more intense has been the intervention response.

(4) An evaluation of Japanese intervention in 1976 based on these empirical relationships leads to the conclusion that it was consistent, both in direction and magnitude, with previous intervention to moderate rather than to prevent exchange rate movement—that is, with leaning against the wind at the rate of about $240 million for every 1 per cent change in the rate.

Empirical tests designed to reveal simultaneous bias in these single-equation estimates of the intervention response did not detect any bias of this sort. In the course of this testing, it was found that changes in the domestic/foreign interest differential and expected changes in the outcome of the current balance of payments were significant factors in determining exchange rate movements, lending further support to the asset market equilibrium theory of exchange rate determination. On the other hand, the short-run version of the purchasing-power-parity view of exchange rate determination was not supported by the Japanese data for this period.

BIBLIOGRAPHY

  • Amano, Akihiro, “International Capital Movements: Theory and Estimation,” in International Linkage of National Economic Models, ed. by R. J. Ball (Amsterdam, 1973), pp. 283328.

    • Search Google Scholar
    • Export Citation
  • Artus, Jacques R., “Exchange Rate Stability and Managed Floating: The Experience of the Federal Republic of Germany,” Staff Papers, Vol. 23 (July 1976), pp. 31233.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bank of Japan, Economic Statistics Monthly, various issues.

  • Bergsten, C. Fred, “The Next Steps in International Monetary Reform,” in Hearings of the Subcommittee on International Economics, Joint Economic Committee, U. S. Congress, 94th Congress, 2nd Session (Washington, October 18, 1976), pp. 215.

    • Search Google Scholar
    • Export Citation
  • Branson, William H., “‘Leaning Against the Wind’ as Exchange Rate Policy,” paper presented at the Conference on Exchange Market Uncertainty, Geneva, November 28, 1976.

    • Search Google Scholar
    • Export Citation
  • Cooper, Richard N., “IMF Surveillance Over Exchange Rates: A Wider View,” in The New International Monetary System, ed. by Robert A. Mundell and Jacques J. Polak (Columbia University Press, 1977), pp. 6983.

    • Search Google Scholar
    • Export Citation
  • Dooley, Michael P., and Jeffrey R. Shafer, “Analysis of Short-run Exchange Rate Behavior, March 1973 to September 1975,” International Finance Discussion Paper No. 76, Board of Governors of the Federal Reserve System, February 1976.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dornbusch, Rudiger, “The Theory of Flexible Exchange Rate Regimes and Macro-economic Policy,” in Flexible Exchange Rates and Stabilization Policy, ed. by Jan Herin and others (Boulder, Colorado, 1977), pp. 12343. This paper also appeared in Scandinavian Journal of Economics, Vol. 78 (1976).

    • Crossref
    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, Annual Report of the Executive Directors for the Fiscal Year Ended April 30, 1977 (Washington, 1977).

  • International Monetary Fund, International Financial Statistics, various issues.

  • International Monetary Fund, International Monetary Reform: Documents of the Committee of Twenty (Washington, 1974).

  • Kenen, Peter B., “Floats, Glides and Indicators,” Journal of International Economics, Vol. 5 (May 1975), pp. 10751.

  • Malinvaud, Edmond, Statistical Methods of Econometrics (Chicago, 1966).

  • “Rambouillet Declaration Pledges Recovery Without Renewed Inflation,” IMF Survey, Vol. 4 (November 24, 1975), p. 350.

  • Schadler, Susan, “Sources of Exchange Rate Variability: Theory and Empirical Evidence,” Staff Papers, Vol. 24 (July 1977), pp. 25396.

  • Williamson, John, “Exchange-rate Flexibility and Reserves Use,” in Flexible Exchange Rates and Stabilization Policy, ed. by Jan Herin and others (Boulder, Colorado, 1977), pp. 195207. This paper also appeared in Scandinavian Journal of Economics, Vol. 78 (1976).

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wonnacott, Paul, The Canadian Dollar 1948–62 (Toronto, 1965).

  • “Yen Is Not ‘Cheap,’” Oriental Economist, Vol. 44 (October 1976), pp. 69.

*

Mr. Quirk, Senior Economist in the Asian Department, is a graduate of the University of Canterbury, New Zealand.

An earlier version of this paper was presented at the meetings of the Western Economic Association held in Anaheim, California on June 20–23, 1977.

1

Bank of Japan, Economic Statistics Monthly.

2

See “Yen Is Not ‘Cheap,’” Oriental Economist, Vol. 44 (October 1976), p. 6. Although one may regard an increase in reserves that corresponds to transactions outside the foreign exchange market as nonintervention, it is more correct to regard the increase from an opportunity viewpoint—that is, as withholding exchange from the market.

3

∆FER ≃ ∆R − XM + ∆OD, where FEF is the net Foreign Exchange Fund position, R is gross official international reserves, XM represents the extra-market transactions, and OD is the outstanding level of official deposits with commercial banks. A positive sign denotes a net increase in assets.

4

Analysis of the respective contributions of capital controls and other factors to the determination of capital flows in the 1970s has been made more difficult by the transition to flexible exchange rates and by variations in the intensity with which controls have been implemented. For the results of estimation, using data for the period 1961–70, in which the onset or removal of capital restrictions is depicted by dummy variables, see Amano (1973).

5

As pointed out in Cooper (1977, p. 77), adoption of these strategies, described as “smoothing and braking,” implicitly acknowledges that “we do not know what the equilibrium exchange rate over any time period is, but it allows for the likelihood that the ‘market’ does not know either.”

6

See “Outline of Reform,” Annex 4, International Monetary Reform: Documents of the Committee of Twenty (Washington, 1974), pp. 33–37.

7

Guidelines (3)(a) and (3)(b) provide guidance in the situation where either the member or the International Monetary Fund, respectively, considers that adherence to Guidelines (1) and (2) would result in a departure of the exchange rate from “reasonable estimates of the medium-term norm.”

10

Ibid., p. 132.

11

Ibid., pp. 131–32.

12

Leaning against the wind is given by It = a(StSt−1), while intervention consistent with gliding parities based on previous market rates is of the form It=b(StΣiaiSti)

14

The latter two tests were also conducted in terms of the Japanese/industrial country export price relative, but these variables were even less significant.

15

Instances at which the exchange rate was at or near this floor were January 1974, August 1974, December 1974, December 1975, and, more approximately, November 1976.

16

The response estimated from equation (5), in which the acceleration of the exchange rate replaces market volume (the two variables are closely related and do not have separate significance), is virtually the same,

It = −64 + 99t + 54(t − Ṡt-1) + 0.595It-1

with the following long-run multipliers:

μ(S˙t)=9910.595=244
μ(S˙tS˙t1)=5410.595=133
17

The asymptotic biases in coefficients estimated by single-equation methods in these circumstances are well known. See, for example, Malinvaud (1966), pp. 507–509.

18

The fact that exchange rate observations over very short time intervals follow a random walk does not preclude the existence of an observable relationship between the rate and its determinants. What it would suggest, however, is that the determinants are either numerous and diverse in their movements, possessing no systematic effect in sum, or, what is less likely, that the individual determinants are themselves nonsystematic. A study by Dooley and Shafer (1976) of exchange rate movements in major industrial countries concluded that the random walk hypothesis could not be rejected for daily observations of the yen rate (over the period March 1973–September 1975).

19

Although convergence may not be completed within the period of observation, the observed points are as likely to be displaced above as below the curve I. The stability of this process depends on the relative slopes of curves I and S. As long as the slope of I does not exceed that of S in absolute terms—that is, as long as the authorities do not consistently overreact—convergence toward equilibrium will occur.

20

See Schadler (1977), pp. 265–67 for a summary of empirical evidence, and Dornbusch (1977) and Williamson (1977) for recent discussions of theoretical aspects of flexible exchange rates.

21

S1, the exchange rate of country l’s currency, is expressed as units of the currency of country 2 (currency 2) per unit of its own currency (currency 1); an increase in S1 therefore denotes an appreciation of currency 1.

22

For a recent exposition in this vein, see Williamson (1977).

23

In equilibrium, which is not guaranteed in the short run, = 0 and $S = $D, implying overall payments balance.

24

Another method of deriving an expression similar to equation (24) would be to state directly the condition for flow equilibrium in the balance of payments by setting the right-hand side of equation (23) equal to zero and to solve for the exchange rate.

25

Industrial output changes were used as a proxy for monthly income changes.

26

See Table 3 for notation.

27

Several reasons may be advanced for the unexplained residual. First, the period reviewed is the first one of generalized floating and, while quite short, it has been characterized by large structural and cyclical shifts in the world economy. Second, the influence of restrictions on capital flows over this period has not proved amenable to modeling. Use of dummy variables has been precluded by the fact that not only the capital controls themselves have varied but their implementation has varied as well. However, there is evidence that the adjustment of controls to changing circumstances in the overall balance of payments takes place over a period considerably longer than one month. It is therefore unlikely that exclusion of this factor would result in substantial bias in estimations using monthly changes.

28

The following equation was obtained using two-stage least squares in tandem with an intervention function of the same form as equation (25). (See Table 3 for notation.)

S˙t=0.0994(0.3)0.0003Kt(0.8)+0.7196r˙t(2.1)+0.1145r˙t1(0.4)+0.0004It(0.4)0.1485P˙t(0.7)7.825D1(4.0)3.263D2(2.5)+0.0241S˙t1(0.3)R¯2=0.42;H=0.31

A similar equation, based on sensitivity of short-term capital stocks, rather than flows, to expected yield, in which the acceleration of the exchange rate is the dependent variable and is replaced by P, attributed significance to the same independent variables.

29

The results were (using the Cochrane-Orcutt adjustment for serial correlation of errors):

S=0.29(4.8)0.07P(0.3)0.002T(1.7)R¯2=0.02;DW=1.92
S=0.17(6.6)+0.33PE(2.6)0.001T(0.7)R¯2=0.14;DW=1.71

where P is the ratio of the Japanese to the U. S. wholesale price index, PE is the ratio of the Japanese export price (expressed in dollars) to that of the industrial countries, S is the exchange rate, and T is a trend term.