Effects of Central Bank Intervention in the Foreign Exchange Market (2024)

Monetary model

Whether of the variety that relies on instantaneous purchasing power parity or not, the monetary models, by assuming perfect substitutability between domestic and foreign assets and perfect capital mobility, imply that intervention policy not affecting either the domestic or the foreign money supply will have no influence on the exchange rate. This can be shown formally by looking at the equation that Bilson (1979), for instance, derives for the exchange rate. With s indicating the natural logarithm of the exchange rate (the domestic currency price of foreign exchange), m (m^*) the natural logarithm of the domestic (foreign) money supply, y(y^*) the natural logarithm of domestic (foreign) real output, and fd the forward discount on domestic currency, Bilson derives

$S=k+m-m^{*}-\eta (y-y^{*})+∊·\mathit{fd}\left(1\right)$

where k denotes a constant, η the real income elasticity of demand for money, and ∊ the interest semi-elasticity of demand for money. When domestic and foreign interest-bearing assets are perfect substitutes, the intervention policy (combined with complete sterilization) does not alter any of the variables on the right-hand side of (1). This is clearly true for the money stocks and real incomes and can be shown to be true also for the forward discount. When there is perfect substitutability between domestic and foreign assets, the forward discount will equal the expected rate of depreciation of the home currency. This, in turn, is a function of expected future monetary policies and expected growth of real income, provided that economic agents are rational. Formally, with rational expectations based on the model underlying (1), one can derive²

$S_t=\frac{1}{1+∊}\underset{j=0}{\overset{\infty }{\Sigma }}{\left(\frac{∊}{1+∊}\right)}^j{E}_t·Z_t+j\left(2\right)$

where $Z_t=m_t-m_t^{*}-k-\eta (y_t-y_t^{*})$ and where the notation E_t+j indicates the expectation formed at time t of the value of the variable X at the time t + j.

By explicitly expressing the current value of the exchange rate as a function of future policy variables, equation (2) points to the possibility that intervention might be effective because it alters current expectations about the future course of the economy. Thus, if a current sale of foreign exchange by the central bank signals future monetary contraction, then the home currency will appreciate immediately even if the intervention is fully sterilized so that the money supply did not change in the current period. This constitutes the only possible channel through which intervention-cum-sterilization could be effective within this framework.

Portfolio balance model

In contrast to the monetary model, the portfolio balance model³ does not assume that domestic and foreign interest-bearing assets are perfect substitutes. The reasons for imperfect substitutability may be many, but some of the most quoted include exchange risk, differential political and default risks, imperfect information about foreign assets, and government regulation of international capital flows.

In its simplest small-country version, the model can be formalized in terms of equations (3)–(6).

$M=m(i,i^{*}+\lambda )·W\left(3\right)$

$B^H=b(i,i^{*}+\lambda )·W\left(4\right)$

$S·F=f(i,i^{*}+\lambda )·W\left(5\right)$

$W=M+B^H+S·F\left(6\right)$

where

As it stands, this model assumes that domestic residents hold foreign securities as well as domestic money and securities. The latter two are assumed to be nontraded. It is also assumed that the foreign rate of interest is taken as a datum by domestic residents (the “small-country assumption”). As pointed out by Dooley and Isard (1979 b), this model is underdetermined in the sense that it can be used to determine the domestic interest rate and either the spot rate S or the expected rate of depreciation λ, but not both unless an additional relationship is introduced. If, for instance, expectations were assumed to be static so that λ = 0, then the exchange rate could be determined as a function of the three asset stocks and the foreign rate of interest. A policy of intervention that leaves the domestic stock of money unchanged would then have the effect shown by

${\frac{\mathrm{dS}}{\mathrm{dF}}|}_{\mathrm{dm}=0}=-\frac{m_i}{F·(m·b_i-b·m_i)}<0\left(7\right)$

Thus, a purchase of foreign bonds for domestic bonds would lead to a depreciation of the domestic currency unless the demand for domestic money was completely interest-inelastic or unless domestic bonds were considered perfect substitutes for foreign bonds. Alternatively, if the current spot rate were determined elsewhere, the portfolio balance model could be solved for the expected rate of depreciation as a function of asset stocks and wealth.

In general, S and λ will be determined jointly once some hypothesis concerning the future spot rate is formulated.⁴ Given that our knowledge of what determines expectations is fairly limited, it may make a great deal of difference to econometric application of the portfolio balance model whether it is solved for the spot rate or the expected rate of depreciation. This point will be elaborated in the discussion of empirical evidence in Section II.

Balance of payments model

The notion that the degree of substitutability between domestic and foreign assets depends on the length of time agents can take to adjust their portfolios may seem plausible if one draws an analogy with conventional consumer demand theory. One way to capture this idea is to specify the reallocation of portfolios in response to changes in interest rates as stock adjustment processes leading to international capital flows in response to interest differentials.⁵ These capital flows would, together with the other items in the balance of payments, determine the exchange rate in such a manner that balance of payments equality would be maintained at all times. The resulting model, referred to here as the balance of payments model, thus uses the balance of payments equality as an equilibrium condition in the foreign exchange market.⁶ Letting CA stand for the current account surplus, NCI for the private net capital inflow, and ΔRES for the purchase of foreign exchange reserves by the domestic monetary authorities (all three of which are dependent on the current spot exchange rate), we can write

$\mathrm{CA}+\mathrm{NCI}+\mathrm{\Delta RES}=0\left(8\right)$

as the condition that implicitly determines the spot rate.

Within this framework, where private capital flows result from a stock adjustment process, it is evident that intervention by the central bank would always succeed in influencing the exchange rate. Whether or not a temporary intervention (corresponding to a permanent exchange of foreign securities for domestic securities by the central bank) would have any lasting effect on the exchange rate would depend on the degree of substitutability between domestic and foreign assets. Under perfect substitutability, once the process of stock adjustment in the demand for foreign assets has run its course, we would be back in the world of the monetary model with no long-run effect on the exchange rate. With imperfect substitutability, the effect derived from the balance of payments model would, over time, approach that of the portfolio balance model with imperfect substitutability between domestic and foreign assets. As far as the effectiveness of intervention is concerned, therefore, a low speed of adjustment in asset markets has the same short-run implications as imperfect substitutability between assets. In the longer run, the only factor that matters is the degree of substitutability.

Evidence from exchange rate equations

Monetary models. Even though sterilized intervention should not influence the exchange rate in the monetary model because of the perfect substitutability assumption, two recent empirical studies have included intervention measures in estimated exchange rate equations based on this approach. In Tullio (1979), the difference between the Swiss franc-lira exchange rate quoted for bank notes in Zurich and the rate quoted for cable transfers was related to the relative supplies of money in the two countries and to the relative levels of income and expected prices, as in the monetary model. The null hypothesis was that the exchange rate on bank notes, being less subject to official intervention, would react faster to the underlying determinants than the rate on cable transfers. This conjecture did receive support from the results. The amount of intervention by the Bank of Italy was also included in the equation, and the estimation indicated that a purchase of foreign exchange had a significant positive effect on the spread between the exchange rates on bank notes and cable transfers. Tullio interpreted this result as showing the ability of the Bank of Italy to slow down a required depreciation of the lira by sterilized sales of foreign exchange. Another interpretation of this result is, however, that such sales are viewed as temporary and likely to be reversed, leading to an expansion of the money supply in the future. The expectation of an expansionary monetary policy then results in depreciation of the value of bank notes in anticipation of the actual change in policy.

In Suss (1980), a three-equation model of exchange rate, interest rate, and intervention behavior was estimated for four industrial countries—France, the Federal Republic of Germany, Japan, and the United Kingdom—relative to the United States. The level of the exchange rate was explained by the amount of central bank intervention in addition to the variables in the monetary model—relative money supplies, income levels, and interest rates. Intervention was assumed to depend in part on changes in the exchange rate and in part on the desire to maintain a fixed relationship between international reserves and imports. For all countries, the results indicated statistically significant “leaning against the wind” policies of reserve management. But, paradoxically, the coefficient of the intervention variable in the exchange rate equation indicated in each case that a purchase of foreign exchange by the domestic central bank would lead to appreciation of the home currency. For three of the four countries, this coefficient was statistically significant. As in the previous case, a possible interpretation of this result¹⁴ is that intervention systematically gives information about future monetary policy, so that a purchase of foreign exchange for domestic currency is interpreted as a temporary easing of monetary conditions and, hence, generates expectations of future contraction. The expected future contraction then leads to an immediate appreciation of the home currency.¹⁵

Portfolio balance models. The portfolio balance model suggests that the exchange rate should depend not only on the supplies of domestic and foreign monies but also on domestic and foreign nonmonetary assets. The latter assets matter because their relative supplies modify yield premiums owing to exchange risk and political or default risk. The exchange rate will then be affected for the reasons explained in Section I. In estimated exchange rate equations based on the portfolio approach, the relative size of the coefficients on domestic assets and foreign assets will show the quantitative effects of an open market purchase of foreign assets for domestic assets—that is, of a sterilized intervention in the foreign exchange market.

Recent attempts at implementing the portfolio model empirically include Artus (1976, 1981), Branson and others (1977, 1979), Dooley and Isard (1979 a), Frankel (1980), Hooper and Morton (1980), Isard (1980), and Martin and Masson (1979). Of these nine studies, only the first five find any support for the portfolio balance model and, hence, for the proposition that sterilized intervention has a significant effect on the exchange rate. In Branson and others (1977), the level of the dollar/deutsche mark rate over the period August 1971 to December 1976 was explained by the money supplies in the two countries as well as their respective net foreign asset positions. The signs of the estimated coefficients of all four variables were those predicted by the theory. Although the level of significance was too low to warrant strong conclusions, using the point estimates as a guide, the results suggest that sterilized intervention is effective. According to Table 7 in the Branson and others (1977) paper, a $1 billion purchase of foreign assets by the Bundesbank, leaving the German money supply unchanged, would lead to a depreciation of 0.185 cent per deutsche mark. The low significance level of the estimates and the high serial correlation of the residuals do, however, raise some doubts as to the robustness of these conclusions. These doubts are confirmed in the authors’ 1979 update of their original study. When the estimation period was extended to March 1978, two of the four coefficients changed signs, and the only variable with the expected sign and high statistical significance was the U. S. money supply.

Like Branson and others (1977), Artus (1976) studied the dollar/deutsche mark rate and estimated simultaneously an exchange rate equation and a reaction function of the authorities in order to avoid simultaneity biases. His sample period was April 1973 to July 1975, and as a consequence of using the rate of change of the exchange rate as the dependent variable, the net foreign asset position was replaced by the current account as the variable capturing relative asset supplies. This variable was highly significant in the estimations, and the estimated coefficients of the model implied that a once-and-for-all DM 1 billion purchase of foreign assets by the Bundesbank would depreciate the mark by 1.13 per cent in the first month if it were completely neutralized by a corresponding open market sale of domestic assets. Subsequent feedbacks would lead to a further 0.48 per cent depreciation for a permanent effect of 1.61 per cent.¹⁶ In Artus (1981), the current account and the interest rate differential between the United States and the Federal Republic of Germany were again significant explanatory variables in an exchange rate equation covering the period extending from the fourth quarter of 1973 to the fourth quarter of 1979, suggesting that German and U. S. assets are not perfect substitutes.¹⁷

Dooley and Isard (1979 a) also investigated movements in the deutsche mark/dollar rate. The sample period was May 1973 to June 1977, and the results supported the existence of a risk premium and imperfect substitutability, in that the estimated model outperformed the forward rate as a predictor of the future spot rate. However, the estimates of the parameters of the model took on plausible values only by imposing prior information on the size of these coefficients.

In contrast to the studies mentioned previously, the other empirical applications of the portfolio approach found virtually no support for the underlying theoretical models. Frankel (1980) and Isard (1980) both attempted to explain the dollar/deutsche mark rate. In each case, the crucial variable measuring relative asset supplies was either not significant or took the unexpected sign. The same negative results emerged from equations in Martin and Masson (1979) explaining the U. S. dollar/Canadian dollar rate and the U. S. dollar/yen rate, as well as from equations explaining the value of the dollar in terms of a West European basket (Martin and Masson) or in terms of a basket of the industrialized countries’ currencies (Hooper and Morton (1980)).

Before discussing some possible explanations for the differences in findings between the studies just reviewed, the results of the study by Quirk (1977) should be mentioned in the present context. Quirk set out to study the reaction of the Bank of Japan to exchange rate changes. The main purpose of this article was to see to what extent interventions by the Japanese authorities could be explained by exchange rate movements. In order to take account of the possibility of simultaneity between intervention and the exchange rate, however, Quirk also estimated an equation in which the exchange rate was made a function of the amount of intervention. After reviewing various models of exchange rate determination, Quirk adopted an eclectic formulation that allowed for effects of, inter alia, intervention, changes in interest rate differentials, relative growth rates of money supplies, and changes in Japan’s current account position. The dependent variable was the monthly change in the dollar/yen rate for the period March 1973 to October 1976. The regression results failed to show any effect of intervention on the exchange rate once the influences of interest rates, money supplies, and the current account were taken out. These results can thus be taken to imply that intervention with full sterilization is ineffective. Some caution is warranted, however, as either the interest rate differential or the current account balance may respond to asset supplies and, therefore, transmit the influence of intervention indirectly.

The contrast between the results of Artus and Branson and others, on the one hand, and Quirk, on the other, is interesting and puzzling in view of the a priori expectation that substitutability between dollar assets and deutsche mark assets would be higher than that between dollar assets and yen assets. On this view, intervention should have been less effective in the Federal Republic of Germany than in Japan.

There are two major explanations for the differences in conclusions between the various studies reviewed previously. One is the very difficult problem of specifying exchange rate expectations correctly. The various papers differed in this respect, and the results may have been correspondingly affected. The importance of treating this issue properly is illustrated clearly in the paper by Dooley and Isard (1979 a), where it is suggested that most of the actual fluctuations of exchange rates have been unexpected.

The second source of difference between the studies is the measurement of relative asset supplies denominated in different currencies. Data on these are not readily available, and the constructed series may differ between studies. Resulting disparities are due to different treatments of valuation changes in official holdings of foreign assets and to differences in the treatment of central bank intervention in the construction of the stock of foreign assets owned by the private sector.

While every empirical implementation of the portfolio balance model is going to be hampered by the problem of data availability, the importance of correctly specifying exchange rate expectations may be reduced if, instead of trying to explain exchange rate movements, attention is turned toward the determinants of interest rate differentials and risk premiums. Studies that follow this route are reviewed next.

Evidence pertaining to the existence and determinants of foreign exchange risk

A major reason for imperfect substitutability between domestic and foreign assets could be the existence of nondiversifiable foreign exchange risk. An indication of such risk is a deviation of the forward exchange rate from the expected future spot rate that constitutes a risk premium. A number of recent studies have attempted to identify this premium. The procedure most commonly followed starts with the definition of the ex ante risk premium given in

$p_t^{\mathrm{ex\; ante}}=E_t{S}_{t+1}-F_t\left(9\right)$

where F_t stands for the one-period forward rate observed at time t. By invoking the assumptions of rational expectations and an efficient foreign exchange market, the actual spot rate at t + 1 is equal to the expected spot rate plus a random error u as in

$S_{t+1}=E_t{S}_{t+1}-u_{t+1}\left(10\right)$

where u_t is “white noise.”¹⁸ Using equations (9) and (10), the ex post risk premium p can be written

$p_t={S_{t+1}-F_t=E}_t{S}_{t+1}-{F_t+u}_{t+1}=p_t^{\mathrm{ex\; ante}}+u_{t+1}\left(11\right)$

Equation (11) implies that tests of hypotheses about the unobservable ex ante risk premium can be carried out using the observable ex post risk premium. Under the null hypothesis of a zero ex ante risk premium, the ex post premium should follow a white noise process. In particular, it should have a zero mean and be serially uncorrelated. Early tests for these properties failed to reject the hypothesis of a zero ex ante risk premium or, in other words, failed to reject the hypothesis that the forward rate is an unbiased forecast of the future spot rate and that the forecast error is serially uncorrelated.¹⁹ These studies took as the alternative hypothesis the existence of a constant risk premium and looked only at a single forward rate maturity.²⁰ More recent studies by Hansen and Hodrick (1980) and Meese and Singleton (1980) that consider the possibility of a risk premium that varies over time reject the hypothesis that the ex ante risk premium is zero.

Taking the results of the last two studies as the current state of the art, the question remains, of course, what the determinants of the risk premium are. The portfolio balance model implies (see, for instance, Dooley and Isard (1979 b)) that the outstanding stocks of “outside” assets denominated in each currency as well as the distribution of wealth between, and the portfolio preferences of, the residents of the two countries should be important explanatory variables. Frankel (1979) tested this hypothesis for the dollar/deutsche mark case for the sample period extending from the first quarter of 1973 to the fourth quarter of 1978 and failed to find any statistically significant relationship of the hypothesized type. Dooley and Isard (1979 b) obtained essentially similar results, although they did not test directly for the existence of a risk premium. They found that relative asset stocks can explain very little of the difference between the future spot exchange rate and the current forward rate, suggesting, on the one hand, that variations in risk premiums cannot explain a very large proportion of the variation of exchange rates and, on the other hand, that most of the movements in spot rates have been unexpected.

Interest rate determination in domestic and Eurocurrency markets

Even if exchange risk is not present, substitutability between domestic and foreign assets may be imperfect because of political risk differentials (as a result, for example, of the possibility of future controls on international capital movements). Dooley and Isard (1980) tested this proposition by investigating the determinants of the differential between Euromark interest rates and comparable rates within the Federal Republic of Germany. Their hypothesis was that in the absence of actual capital controls and political risk, this differential would be zero. By adjusting for the effects of capital controls by means of dummy variables, they were able to isolate statistically significant influences of German and non-German wealth variables on the interest rate differential. These effects were consistent with their portfolio balance theory of the determinants of this differential. Their measure of the supply of outside assets denominated in deutsche mark failed to be significant in the regression,²¹ however, a fact that could be rationalized by the opposite influences of political and foreign exchange risks. The implication of this lack of significance for the effects of intervention is that sterilized operations in the foreign exchange market do not influence the differential between Euromark interest rates and domestic rates and therefore, presumably, are also incapable of affecting the exchange rate.

In another recent study of the behavior of differentials between domestic and Euromarket interest rates, Giddy and others (1979) studied the rates on dollar-denominated deposits and loans. They hypothesized that, owing to temporary market imperfections, U.S. rates would react less quickly than Euromarket rates to changes in credit conditions. These market imperfections were ascribed to regulatory restraints, institutional and perceptual factors, and oligopolistic market conditions prevailing in the United States but absent in the Euromarkets. The empirical evidence was consistent with the underlying hypothesis and suggests that tight credit policies in the United States (as a result, for example, of interventions in the foriegn exchange market) would temporarily raise Euromarket rates relative to domestic U. S. rates and might²² through this channel succeed in affecting the foreign exchange value of the dollar. According to the hypothesis advanced by Giddy and others (1979), this influence would only be temporary and would disappear as soon as the market imperfections in the United States became nonbinding.

Determinants of covered interest rate differentials

If foreign exchange risk is either absent or not influenced by intervention policy, the portfolio balance model implies that a necessary condition for fully sterilized intervention to have some effect on the exchange rate is that it influence the covered relative yield on domestic and foreign assets.²³ It is possible²⁴ to derive a well-defined relationship between the covered interest differential and variables like the outstanding stocks of “outside” assets denominated in each currency, their distribution between countries, and the distribution of wealth between countries. In this framework, intervention would modify the supplies of assets available to the public and, hence, would affect the covered interest differential unless the assets used in the intervention operations were essentially perfect substitutes. Genberg (1981 a) tested this latter prediction by regressing the covered interest differential between the United States and the Federal Republic of Germany, and between the United States and Japan, on asset stocks, wealth, and intervention variables for the period extending from the third quarter of 1974 to the first quarter of 1980. On balance, the evidence from these regressions gave very little support for the prediction of the portfolio balance model concerning movements in covered interest differentials. This was particularly true for the Federal Republic of Germany. Applied to the intervention issue, these results indicate that the German authorities can expect to see very minimal effects of any intervention in the foreign exchange market as long as intervention is prevented from influencing the money supply. The effect of intervention on the exchange rate may be somewhat more pronounced in Japan, but comparing the estimates obtained from dividing the entire sample into two subperiods, the relationships involved seem to be unstable over time and may therefore not be useful to policymakers.