Alternative Approaches in the Analysis of International Capital Movements: A Case Study of Austria and France

In recent years significant advances have been made in the theory and econometric analysis of international capital movements. In the empirical work three basic approaches can be distinguished that essentially reflect the way that interest rates at home and abroad are treated in the underlying structural models.

Abstract

In recent years significant advances have been made in the theory and econometric analysis of international capital movements. In the empirical work three basic approaches can be distinguished that essentially reflect the way that interest rates at home and abroad are treated in the underlying structural models.

I. Introduction

In recent years significant advances have been made in the theory and econometric analysis of international capital movements. In the empirical work three basic approaches can be distinguished that essentially reflect the way that interest rates at home and abroad are treated in the underlying structural models.

The first approach follows the lines of a traditional partial equilibrium analysis. It is based on a portfolio selection model that focuses on the relative yields on financial assets in national markets and the associated flows of capital between these markets. Interest rates, an important component of relative yields, are viewed as exogenously determined. Therefore this approach implies that, while interest rates are allowed to influence capital movements, capital movements themselves do not influence interest rates. An implicit assumption in the model is that monetary authorities adopt sterilization policies so as to counteract the effect of balance of payments developments on the domestic money supply or on interest rates.

The other two approaches incorporate financial flows into a macroeconomic model of an open economy. More recent models used as a basis for empirical analysis consider a small country facing exogenously given interest rates abroad. The domestic interest rate is treated as an endogenous variable within the system of equations determining the equilibrium in the asset and money markets and satisfying the constraint imposed by the balance of payments. One approach then uses a reduced form of such a model as a basis for regression analysis. On the assumption that the model is properly specified, this approach lends itself to a straightforward estimation. However, since regression coefficients are generally complex, incorporating several structural parameters in the underlying equations, an interpretation of these coefficients is often quite difficult.

The alternative structural approach involves the estimation of the behavioral equations specified in the model. Several such contributions have been made in recent years.1 However, in most cases these were fairly highly disaggregated structural models of the economy that did not consider the estimation of capital movements as the main objective.2

An evaluation of the partial equilibrium approach and the approach based on the reduced form of the financial sector’s model is attempted in this paper. Austria and France are used as case studies. However, recent results obtained from applying these approaches to capital movements of several other industrial countries are also mentioned.

II. Partial Equilibrium Approach

This approach is based on stock adjustment analysis of a financial sector in which domestic and foreign interest rates play a major role.3 The structural form is usually expressed as a set of two equations, defining a demand for foreign assets by investors and traders at home and a demand for home assets by foreign investors and traders abroad (the latter being expressed as liabilities of home entrepreneurs to foreigners). Stocks of assets and liabilities are generally expressed in the following functional forms:

A=A(Id,If,ReR,Va,Wd,X)(1.1)
L=L(Id,If,RR,Ve,Wf,M)(1.2)

where A and L are net foreign assets and net liabilities to foreigners, Id and If are domestic and foreign interest rates, R and Re are the current exchange rate and the expected exchange rate at a given future time period, Va and Ve are appropriately specified risk variables associated with subjective expectations of domestic and foreign investors regarding the effect on returns of the projected change in the exchange rate and other factors,4 Wd and Wf are domestic and foreign wealth, and X and M are exports and imports.

The relative yield on foreign versus domestic assets (YfYd) can be expressed approximately as the difference between the premium of the expected future exchange rate over the present exchange rate and the forward premium for the same maturity.

YfYdRePFRReRR+Id  If(1.3)

where ReP is the premium of the expected over the present exchange rate and FR is the forward premium, and the interest rates are for the same maturity as the expected and forward premiums.5

Therefore, equation (1.1) indicates that holdings of foreign assets by nationals are determined by relative yields at home and abroad and risks associated with these yields, while domestic wealth serves as a budget constraint and a scale variable, and that, finally, exports fulfill the same function for traders.6 Equation (1.2) indicates that holdings of domestic assets by foreigners are also determined by relative yields and associated risks, while foreign wealth and imports (that is, exports into the country in question) serve as budget constraints and scale variables.

Most studies of capital flows in industrial countries other than the United States and Canada7 are conducted in terms of net capital movements, which are obtained by netting out assets from liabilities. Furthermore, again following the general practice, lagged adjustment is incorporated by entering variables in the current and the lagged form:

LtAt=N[Id(ti),If(ti),(FeR)(ti),V(ti),Wd(ti),Wf(ti),TB(ti)](1.4)

where i = 0, 1 is the risk involved, and TB (= XM) is the trade balance.8

Expressing relation (1.4) in a first-difference form yields a basic regression equation that is generally used in explaining international capital movements. However, a number of difficulties are encountered when the variables are translated into observable data. A major difficulty is in deriving testable expectation hypotheses that could satisfactorily describe the behavior of speculators in the system of pegged parities.9 In most cases this effort has resulted in using a set of dummy variables for speculation against the peg, while the speculation within the band is left unexplained. This approach is also followed here; instead of the variables referring to exchange rate expectations and to risk, dummy variables are used for the periods of major speculative crises. The second problem relates to the lack of quarterly data on wealth in most industrial countries. However, in a regression equation of a first-difference form, a constant can serve as a scale variable.

With these adjustments the basic regression equation becomes

Δ(LtAt)=Ct=a0+a1ΔIdt+a2ΔIdt1+a3ΔIft+a4ΔIft1+a5ΔTBt+a6ΔTBt1+a7DSP+u(1.5)

where C stands for net capital flows (inflows positive) and DSP is a dummy variable for destabilizing speculative pressures. The expected signs in equation (1.5) are as follows:

a1 > 0; a3, a5 < 0; a2, a4, a6 ≷ 0; a7,

coefficient for the speculative dummy, is expected to be positive in the case of speculative capital inflows and negative in the case of speculative outflows. As can be seen, indeterminate signs in coefficients of lagged independent variables suggest a possibility of return flows during the succeeding quarter.10 In such a case, however, the sign of the sum of coefficients in the two quarters should be the same as the predicted sign for the first quarter, that is,

a1 + a2 > a3 + a4 < 0; and a5 + a6 < 0.

In equation (1.5) domestic and foreign interest rates are taken as exogenous variables and thus independent of each other and of capital movements.11 The simultaneous relationship between domestic and foreign interest rates and capital movements would not be a problem if complete neutralization of the effect of capital flows on domestic money supply could be assumed. However, the high degree of financial integration that now exists among industrial countries makes such an assumption unrealistic. Furthermore, a simultaneous relationship between international flows of foreign exchange and domestic monetary variables, under conditions of less than complete neutralization of these flows, indicates another partial equilibrium aspect of the approach described in equation (1.5). There a change in net trade flows is taken as an indicator of the need for trade finance.12 No evaluation is made of the effect of a current surplus or deficit on domestic monetary variables and, through the resulting adjustments in the latter, on international capital movements.

In the absence of a complete neutralization of foreign exchange flows on the domestic money supply, a regression analysis based on equation (1.5) would tend to bias downward the coefficients indicating the response of capital movements to domestic and foreign financial variables.13 A causative change in the domestic interest rate would generate mutual adjustments in the capital flows and domestic and foreign interest variables, so that a capital flow response to the initial interest change would appear smaller than in the case when both interest rates are truly exogenous variables. In the limiting case of perfect capital mobility, where no autonomous change in domestic interest rate is possible, no relationship exists between interest rate movements and capital flows. Finally, a high degree of simultaneity between capital movements as the dependent variable and domestic and foreign interest rates as independent variables will also be reflected in a high degree of collinearity between the latter two interest variables. In such a case, the estimates of the interest rate coefficients will be inefficient, and the sign may even be wrong.

III. Analysis Based on a Reduced Form of a Macroeconomic Model

A macroeconomic model of the financial sector developed recently by Kouri and Porter (1974) provides the basis for an empirical analysis of capital movements in which the domestic interest rate is treated as an endogenous variable. The model has further general equilibrium properties because capital movements are considered within the context of the country’s overall balance of payments and the effect of foreign exchange flows on domestic monetary variables is explicitly taken into account.

This model is a portfolio model of a small open economy in which capital movements occur through adjustments in the financial sector. The financial sector is specified in terms of the domestic demand functions for base money and for domestic and foreign assets and of the foreign demand function for domestic assets; the supply of base money is partly a function of the flow of the central bank’s foreign assets across the border and partly of its control over domestic component of base money. The real sector of the economy is exogenous to the model, so that changes in income, prices, and the current account of the balance of payments are exogenously given, as are the wealth at home and abroad, which serve as portfolio constraints.

However, when Kouri and Porter assume perfect capital mobility as a limiting case, their approach collapses into a very simple monetary model of an open economy. Equations defining demand and supply conditions on the asset markets drop out, and the model is reduced to the following set of equations:

MOD=M(If,GNPd,Wd,ReR)(2.1)
MOSNDA+NFA(2.2)
ΔNFAC+CA(2.3)
MOS=MOD(2.4)

where, in addition to variables specified in the preceding section, MOD and MOs are demand for and supply of base money; GNP stands for income;14 CA is the net current account; and NDA and NFA are net domestic and foreign assets of the central bank. NDA is controlled by the central bank through open market operations.

In equation (2.1) the demand for money is defined as a function of the foreign interest rate, income, wealth, and the difference between the expected future exchange rate and the spot exchange rate. Since the model considers the case of a small country in a simplified version based on the assumption of perfect capital mobility, yields on assets at home and abroad are equalized at all times. Kouri and Porter translate this equality as between domestic and foreign interest rates. However, such a view is valid only within a system of rigorously fixed exchange rates. When a band between intervention points and exchange rate expectations is taken into account, a degree of variation between interest rates at home and abroad that satisfies the interest rate parity condition is still possible.15

Relations (2.2) and (2.3) are identities showing the components of supply of base money and of the change in the net foreign assets. Equation (2.4) is the condition for equilibrium in the domestic market for base money. Demand and supply of money, net foreign assets, and capital movements are endogenous variables in the model. Net domestic assets (which are determined by the central bank), the current account, income and wealth, and expected and actual exchange rates are assumed to be exogenously determined.

Upon using the first-difference transformation of equations (2.1), (2.2), and (2.4), substituting equation (2.3), and solving the system for capital movements as a dependent variable, the result is a reduced form that can serve as a basis for regression analysis. However, for reasons already specified in the preceding section, the wealth variable is dropped from the regression equation, while the change in the difference between expected and actual exchange rates is in part substituted by dummy variables. In order to test the possibility of a lag in adjustment of capital movements to independent variables, a simple one-period lag is entered for all these variables. The resulting regression equation is as follows:

Ct=a0+a1ΔIft+a2ΔIft1+a3ΔGNPdt+a4ΔGNPdt1+a5ΔNDAt+a6ΔNDAt1+a7CAt+a8CAt1+DSP+u(2.5)

The expected signs are as follows:

a1 < 0; a3 > 0; 0 > a5, a7 > – 1; a2, a4, a6, a8 ≷ 0.

Also, in the case of reflows:

a1 + a2 < 0; a3 + a4 > 0; 0 > (a5 + a6);(a7 + a8) > – 1.

The foreign interest rate and income variables show the effect on capital movements of changes in the demand for base money. The current account position (CA) and change in the net domestic assets (ΔNDA) should indicate the response of capital movements to the current account surplus or deficit and to monetary instruments affecting money supply. These two variables are particularly important in the Kouri-Porter analysis, since their coefficients should reflect the degree to which, other things being equal, addition to or subtraction from the domestic component of base money are offset by capital outflows or inflows. In the conditions of perfect capital mobility, Kouri and Porter assume complete offsetting, in which case ΔNDA and CA would have a coefficient of –1. In the present formulation of the model this would also imply no lag in adjustment of capital movements to independent variables.

Statistical estimation of the modified equation (2.5) is beset with several difficulties.

(1) The assumptions of perfect capital mobility as a limiting case provides a specific meaning to a role of the foreign interest rate variable in the estimation of international capital movements. The assumption that the domestic interest rate adjusts completely and without lag to changes in the foreign rate implies that the coefficient of the foreign rate can be used in the regression equation in estimating the interest elasticity of the demand for money (Kouri and Porter, 1974, pp. 451 and 460–61). Such an assumption, however, is questionable for two reasons:

(a) As already pointed out, the presence of exchange rate movements within the intervention band and exchange rate expectations imply that domestic and foreign interest rates would not be expected to be equal, although—if the movements of the exchange rate within the band are random—these rates could be considered to be approximately equal. In actual fact, however, even under the assumption of perfect capital mobility, the relation between domestic and foreign interest rates is much more complex. If such complexity is taken into account, perfect capital mobility is consistent with the fact that the offset coefficient is significantly different from – 1.16

(b) The small country assumption is crucial for disregarding possible adjustments of the foreign interest rate to changes in the interest rate of the country whose capital flows are analyzed. If such adjustments exist, however, the coefficient for the foreign interest rate in the regression equation will be subject to a downward bias similar to the one found in the partial equilibrium models.

(2) The major problem of the approach based on the estimation of a reduced form equation relates to the use of a change in net domestic assets (ΔNDA) as a key exogenous variable intended to show the response of capital movements to the positive or negative excess supply of money. Kouri and Porter recognize that government intervention in the domestic money supply in response to foreign exchange flows would create a simultaneous relationship between ΔNDA and capital flows as a dependent variable. Hence, in order to obtain an unbiased estimate of determinants of capital movements, they adopt as a key assumption a “neutral” monetary policy with respect to the balance of payments. This means that no attempt is made either to offset the liquidity effect of foreign exchange flows on money supply or to use monetary policy to restore an equilibrium in the balance of payments.

If this assumption is not satisfied, the seriousness of a bias in the estimate can be seen when the government’s monetary intervention is examined somewhat more closely—still assuming that the money supply is the only instrument of the government intervention. Under conditions of perfect international financial integration without an attempt at neutralization by the monetary authorities, domestically caused changes in base money will be offset, other things being equal, by an opposite change of the same size in net foreign assets—a change assumed here to occur through capital flows. However, when the degree of financial integration is so low as to permit the authorities to insulate total monetary base from foreign influences completely, a change in net foreign assets of the central bank caused by capital flows would immediately be offset through the opposite change of the same size in net domestic assets. The total monetary base will remain unchanged. In both the case of no neutralization with perfect capital mobility and the case of complete neutralization, the coefficient of ΔNDA will be –1; thus the estimation of the reduced form shown in relation (2.5) cannot distinguish between these two diametrically opposed situations. In the case of partial neutralization and imperfect capital mobility, the coefficient of ΔNDA has an upward bias (in absolute terms) that is directly related to the degree of neutralization of foreign exchange flows.

(3) If, on the other hand, the authorities assign monetary policy to the balance of payments target, any change in net foreign assets will be followed by an accommodating change in the total money supply in the same direction and aimed at restoring either the balance of payments equilibrium or the original level of foreign exchange reserves. The relationship between the change in net domestic assets and capital flows, under such circumstances, will be very weak, as will the explanatory power of the other exogenous variables (such as a change in the foreign interest rate or in exchange rate expectations) that had triggered the original flow of capital. The offset coefficient of ΔNDA will be biased toward zero.

Thus, if the reaction function of the monetary authorities is not properly specified, thereby not eliminating the simultaneity between the change in the domestic money supply and the change in the foreign assets as dependent variables, little confidence can be attached to some key coefficients in the regression equation based on the reduced form equation (2.5). The offset coefficient in the regression equation may therefore be biased, but the direction of the bias is not clear.

Argy and Kouri (1975) incorporated such a reaction function into equations estimating capital flows, when they analyzed policies of sterilization of foreign exchange flows in several European countries. A change in the domestic component of base money was expressed in terms of variables that should have reflected various targets of the monetary policy. The resulting regression equation was the following:

ΔNDAt=b0+b1Ct+b2CAt+b3RCUt1+b4T+b5DS+v(2.6)

where RCU was the rate of capacity utilization, T was a trend variable, and DS stood for quarterly seasonal dummy variables.

From equation (2.6) the targets tested were assumed to be a balance of payments constraint, reflected in the flows on the current and capital accounts, and a domestic anti-cyclical monetary policy, reflected in the capacity utilization variable. The trend variable should have reflected an upward shift in the supply of money as a result of long-term growth of income. Equation (2.6) was used in conjunction with equation (2.5) in a two-stage least-squares estimation of capital movements and ΔNDA as dependent variables.

Specification of a reaction function for monetary policy is a very complex problem. Monetary measures must be looked at within the whole set of instruments oriented toward various targets of the overall economic policy of the government. It is clear that the use of monetary policy instruments differs from country to country; furthermore, it is not consistently employed toward a specific target in any one country. Often a change in targets has a degree of regularity. For instance, countries with weak currencies, such as the United Kingdom in the 1960s, may assign monetary policy to domestic targets during periods of relative strength of the balance of payments, but may try to achieve a high degree of international harmonization of interest rates when the country is faced with a loss of foreign exchange. On the other hand, countries with strong currencies, such as the Federal Republic of Germany over the same period, may do the opposite in the pairing of targets and monetary instruments. They may use monetary policy to discourage capital inflows during periods of strength in the balance of payments and may sterilize the effect of capital flows on domestic money during periods of relative weakness in the balance of payments.

Finally, sterilization policies may well depend on the range of other policy instruments available to the authorities. A degree of harmonization of interest rates can be maintained by altering the monetary/fiscal mix over the cycle. This can be achieved by a relatively tighter fiscal policy in periods of boom and a more expansionary policy in periods of recession.17 Again this policy mix can be asymmetrical, depending on whether the country in question has a weak or strong currency.

However, one of the major problems in the econometric analysis of short-term capital movements is the proper specification of exchange rate speculation. With very few exceptions,18 both partial equilibrium models and macroeconomic models have been unable to incorporate exchange rate expectations that determine speculative behavior. In the fixed exchange rate system this would have entailed explaining the determination of forward exchange rates and of destabilizing speculation based on expectation of a change in the exchange rate parity. Such a dual formation of exchange rate expectation in the system of pegged parities that incorporates an intervention band is particularly difficult to formulate in a model having empirical applicability. In a system of freely flexible or managed flexible exchange rates, proper handling of exchange rate expectations is central to the explanation of equilibrium in the balance of payments. This inability to handle speculative forces may be a crucial flaw in the specification of the regression equation determining capital movements and may exceed the difficulties created by biases inherent in the recent treatment of both partial equilibrium and macroeconomic models.

IV. Empirical Analysis

Since the mid-1960s there have been a number of empirical studies of capital movements of industrial countries based on a partial equilibrium portfolio selection approach. Initially, most of them were limited to the analysis of capital movements of the United States.19 The analysis of capital flows of other industrial countries was pursued only in the 1970s when capital flows of the United Kingdom,20 the Federal Republic of Germany,21 France,22 Canada, Italy, and Japan 23 were examined. Because of substantial differences in the methodology and in the periods covered, a comparison of the results obtained by these studies would not be useful. However, most of these studies show estimates of a fairly low interest elasticity of capital movements, suggesting that domestic monetary policy is quite independent in these countries.24 The results of studies of financial flows for the United States, estimating domestic and foreign interest rates separately in the regression equations, point out another interesting phenomenon. These regression results suggest that, in a fixed exchange rate system, interest rates in several industrial countries adjust rapidly to changes in U. S. interest rates, while the monetary authorities of these countries can intervene in their domestic money markets without making a large impact on the U. S. interest rate.25 In the partial equilibrium portfolio approach, where interest rates are taken to be exogenous variables, this behavior suggests a downward bias in regression coefficients that understates the true interest elasticity of capital movements.

Studies of capital flows based on the reduced form of a monetary macroeconomic model are more recent (Kouri and Porter (1972, 1974); Argy and Kouri (1974); and Kouri (1975)). The cases of the Federal Republic of Germany, Italy, the Netherlands, and Australia were examined in these studies. The estimates of the offset coefficient between changes in the domestic component of base money and capital flows were low enough 26 to support a view that monetary authorities in these countries may be able to sterilize in part the effect on the monetary base of balance of payments deficits or surpluses and to maintain in this way a degree of control over the money supply.27

An empirical investigation into the determinants of capital movements in Austria and France yields further quantitative information in this area of analysis. In both cases a degree of comparison between the partial equilibrium and macroeconomic approaches is provided by using the two approaches as alternatives in the regression analysis. Since the models used in these two approaches are different, the comparison consists in examining the degree to which the econometric results are consistent with some key postulates in the two approaches rather than in directly comparing various statistics.

Austria

Early in 1959 the Austrian authorities liberalized commercial financial transactions and most capital account transactions by nonresidents, and this liberalization was extended to residents in mid-1962. Further elimination of restrictions took place in the mid-1960s. However, the domestic banking system has oligopolistic elements, causing substantial constraint on the flexibility of the interest rate as a pricing mechanism. In determining the appropriate level of domestic credit, the Austrian National Bank seems to rely less on monetary instruments, such as open market operations and minimum reserve requirements, than on gentlemen’s agreements with commercial banks that are represented on its monetary decision-making committee. Interest rates on deposits are determined by an agreement between the banks and the monetary authorities, and rates are differentiated according to the size of the depositor. There are also sectors where loans are granted at a subsidized interest rate. The bond market is regulated and is dominated by securities of the government and public agencies. Thus, one of the major problems for an empirical analysis of Austria’s international capital movements is the lack of data on a representative domestic short-term interest rate.28 Another problem is the quality of the quarterly data on balance of payments. Since the quarterly data published by the Austrian government agencies are not revised, there may be a measurement error in the statistical analysis.

The results of the regression analysis of short-term capital (STC) and total private capital (TCP) movements as dependent variables are shown in Table 1. Equations (3.1) and (3.2) test these two dependent variables by using a partial equilibrium approach, while equations (4.1) and (4.2) are based on a reduced form of a Kouri-Porter type of macroeconomic model. The symbols for the independent variables are as follows: IEU is the 90-day Euro-dollar deposit rate; Id is the Austrian domestic interest rate, consisting of a yield on new bond issues; TB is the trade balance; CA is the net current account balance; NDA is net domestic assets of the Austrian National Bank; GNP is nominal gross national product; and DS is the seasonal dummy variable, equal to + 1 and – 1 for the first and the last quarter of the year and zero otherwise. Data in parentheses below the regression coefficients are the t-values. Quarterly series used in the regression analysis cover the period from the first quarter of 1960 to the first quarter of 1971, and they are terminated just before the build-up of the speculative pressures that led to a floating of several European currencies and to a breakdown of the Bretton Woods monetary system.

Table 1.

Austria: Regression Analysis for Capital Movements1

article image

See text for explanation of symbols.

In equation (4.1), where private short-term capital (STC) is a dependent variable, CA incorporates the net current account and the balance on official and private long-term capital; in other words, CA stands for a basic balance. In equation (4.2), where total private capital (TCP) is a dependent variable, CA incorporates the net current account and net official capital movements.

Equations (3.1) and (3.2) are based on a partial equilibrium approach. They show a reasonable fit, although the explanatory power as reflected in the R¯2 is not very high. The Euro-dollar interest rate, the trade balance, and the dummy variable for the end-of-the-year window dressing are significant in both equations. The domestic interest rate is significant at the 95 per cent level (one-tailed test) only in equation (3.2), with total capital movements as a dependent variable. This is not surprising, since the available data on interest rates are basically indices of the medium-term to long-term rates.

Equation (3.2) suggests a net outflow of some S 400 million over a six-month period as a result of an increase of 1 percentage point in the Euro-dollar rate. The coefficients of a current and a lagged Eurodollar rate suggest a lag distribution that generates an outflow of S 1.3 billion over the first quarter and then a reflow of some S 0.9 billion in the following quarters. On the other hand, an increase of 1 percentage point in the domestic interest rate generates an inflow of S 2.2 billion in the first quarter and a further inflow of S 0.6 billion in the following quarter. This disparate influence of changes in the foreign and domestic interest rates may reflect the policy constraint on domestic interest rates imposed by the international mobility of capital. While changes in domestic interest rates have a negligible influence on the Euro-dollar rate, changes in the latter may force adjustments in domestic interest rates in order to reverse the initial outflow of capital. Equation (3.2) further suggests that all of the increase in the export surplus is financed through capital flows over the six-month period 29—thus indicating the importance of international financing of trade in Austria.

Equations (4.1) and (4.2), based on a reduced form of a macroeconomic model, appear to have a fairly high explanatory power. However, the coefficients of ΔNDA, which is the key argument in this approach, are not statistically significant although they have a right sign. This crucially impairs the internal consistency of the regression equations. Once ΔNDA is not significant, this approach becomes somewhat similar to the partial equilibrium analysis. The coefficients of the Euro-dollar rate in equations (4.1) and (4.2) are fairly similar to those in equations (3.1) and (3.2). However, the role of the foreign interest rate should be different in the monetary model from that in the partial equilibrium approach. In the monetary model the influence of the foreign interest rate is not exercised directly on capital movements but indirectly through the adjustment in the demand for base money at home. This situation occurs because international highly integrated financial markets and fixed exchange rates cause domestic interest rates to adjust immediately to movements in the foreign interest rate; hence, in a small country the foreign interest rate becomes, under these assumptions, an independent variable determining domestic demand for money. At the 95 per cent level (in a one-tailed test) ΔGNP is significant, while the coefficients of the current account variables indicate that the current account surplus or deficit is almost fully offset over the six-month period by capital outflows or inflows.

Thus, in the regression analysis of Austrian capital movements, the partial equilibrium approach yields results that appear internally consistent but the explanatory power of the regression equation is not very high. Furthermore, the behavior of domestic and foreign interest rates is not consistent with the assumption of exogenously given interest rates, an assumption on which this approach rests. The results of the macroeconomic approach, on the other hand, are not internally consistent, and the regression equations tested have not captured the general equilibrium adjustments implied in the model.

France

Even after the introduction of external convertibility of the French franc in 1958, the authorities maintained a rather strict control over capital transactions of French residents. These restrictions were progressively relaxed throughout the first two thirds of the 1960s, so that in 1967 the remaining controls were relatively unimportant. However, restrictions reimposed to stem capital outflows caused by disturbances in 1968 were strengthened substantially during the period of the dual exchange rate system, beginning in 1971. In view of the available data for short-term capital flow and GNP, the various series start at the beginning of 1963 and are terminated at the end of the first quarter of 1971.30

French monetary authorities have exercised a fairly tight control over the domestic money supply and the activity of the commercial banks. These controls relied heavily on quantitative ceilings imposed on bank credit and on penalty rediscount rates. Use of these monetary instruments appears to have permitted control over the banking system to be consistent with a degree of harmonization of domestic and international interest rates. Thus the deviation of the French call money interest rate from the Euro-dollar interest rate was quite small in the first half of the 1960s; even in the second half of the decade it was significantly lower than in other major industrial countries (Argy and Hodjera, 1973, p. 31, Chart 2).

Table 2 shows the regression analysis that explains French short-term and total capital movements. Symbols other than those already used in equations (3.1) through (4.2) are as follows: CMR is the French call money rate; GDP is the nominal gross domestic product; DSP is the dummy variable for destabilizing speculation during and immediately following the disturbances of spring 1968. It is assumed that the disturbance unforeseen at the beginning of the second quarter in 1968 generated speculative capital outflows for two quarters and that these were reversed throughout the following two quarters. This pattern of capital movements generated by speculation is approximated by the following set of dummy variables: –1 in 1968 (second quarter), –½ in 1968 (third quarter), +1 in 1968 (fourth quarter), in 1969 (first quarter), and zero in all other periods.31 There are 37 quarterly observations in each series.

Table 2.

France: Regression Analysis of Capital Movements

article image

See notes for Table 1.

Equations (5.1) and (5.2), based on the partial equilibrium approach, yield poor results. The only significant variables are the current domestic interest rate, the lagged current account, and the dummy variable for speculative flows. Furthermore, the domestic interest rate has a wrong sign. Such a relationship between domestic interest rates and capital flows is possible in the case of the international interest rate harmonization by the monetary authorities. A significant interest rate harmonization will generate collinearity between the two interest rates as dependent variables. Indeed, by re-estimating the equations (5.1) and (5.2) with the Euro-dollar rate as the only interest variable, the signs of coefficients of that interest rate change and become negative as expected. However, the coefficients remain statistically insignificant.

As shown in equations (6.1) and (6.2), the macroeconomic approach appears to perform better. Equation (6.2), where total private capital movements are used as a dependent variable, shows a better fit. The offset coefficient between changes in the monetary base and capital movements is close to 60 per cent over the two quarters; this appears somewhat high, particularly when compared with the offset coefficient of some 35 per cent over the same period for the short-term flows tested in equation (6.1). The offset coefficient of the current account flows is even higher, approaching 80 per cent over one quarter. The Euro-dollar interest rate is significant as a lagged variable; it suggests that an outflow of between F 500 million and F 600 million may be induced by an increase of 1 percentage point in the Euro-dollar rate. On the other hand, the change in nominal income is significant only with a lagged form in equation (6.2), but it has a wrong sign. Both dummy variables, correcting for seasonal window dressing at the turn of the year and for destabilizing capital flows in 1968, are significant.

In order to take into account the possible simultaneous relationship between changes in the base money and capital movements, a reaction function is specified for monetary policy and capital movements are re-estimated by using a two-stage least-squares estimation.

The following fairly simple reaction function was used:

ΔNDA=f(ΔNFA,RCU,T)(2.7)
ΔNFATCP+CA(2.8)

It tests the degree to which changes in base money were used to neutralize movements of foreign exchange and as an instrument of anti-cyclical monetary policy. RCU is the rate of capacity utilization in France, and T is a trend variable. Two-stage least-squares analysis was performed by using equation (2.5), a linearized form of equation (2.7), and equation (2.8). The estimates for capital movements and the changes in monetary base are shown in Table 3.

Table 3.

France: Two-Stage Least-Squares Regression Analysis (Macroeconomic Approach)

article image

See notes for Table 1.

ΔNF^A = CA + TĈP, where the circumflex indicates computed values in the two-stage least-squares estimation.

Starting first with equations (7.3) and (7.4), which estimate the monetary policy reaction function, it would appear that during the 1960s a policy of neutralization of the effect of foreign exchange flows on total money supply was pursued and that the money supply was also used sometimes as an instrument of anti-cyclical economic policy. A change in net foreign assets in equation (7.4) is significant, and it would thus appear that some three fourths of the movements in net foreign assets of the Bank of France were offset by opposite changes in the domestic component of base money. The rate of capacity utilization is only marginally significant, implying that a decrease (increase) in money supply of some F 125 million is associated with an increase (a decrease) of 1 percentage point in capacity utilization.32 The trend variable is also significant, and so are the dummy variables for seasonal changes in the first quarter of the year and for unusual increases in money supply as a result of events in 1968 and immediately thereafter in the second quarter of 1969.33

However, in comparing equations (7.1) and (7.2), obtained by a two-stage least-squares estimation, with equations (6.1) and (6.2), obtained by the ordinary least-squares method, it is obvious that the use of computed values for ΔNDA did not significantly change the regression coefficients. Thus, on the one hand, equation (7.4) suggests a systematic neutralization of some 75 per cent of the change in net foreign assets; on the other hand, the fact that a two-stage least-squares estimation of capital movements in equation (7.3) yields the same coefficients as those generated by the ordinary least-squares method suggests that such policy does not create a simultaneous relationship between changes in the domestic component of the base money and capital movements as dependent variables.

There are several explanations for this rather implausible conclusion. Argy and Kouri (1975) obtained similar results in cases where they have not obtained a good fit for their policy reaction function; they have suggested that intervention of the monetary authorities, which is necessarily rather arbitrary in timing, creates sufficient instability in the reaction function to remove this policy-induced simultaneity between ΔNDA and capital movements (Argy and Kouri, 1975). However, in the French case, variables suggesting neutralization of foreign exchange flows are significant.

The complexities of French monetary policy and of the controls of the banking system suggest that, despite an apparently good fit of equations (7.3) and (7.4), the major problem may be that of specifications. This indication is further confirmed by a low Durbin-Watson statistic in both equations, suggesting—in view of a rather simple form of the reaction function—a missing independent variable as the cause of the poor fit. Furthermore, in exploring the process of a two-stage least-squares estimation of both capital movements and ΔNDA, a certain lack of robustness was observed. This is also shown when observed and computed values in regression equations (7.3) and (7.4) are correlated, after excluding the observations for the exceptional period of four quarters of substantial speculative flows in 1968–69, for which period a set of dummy variables is used. The R¯2 statistic for TCP decreases from 0.82 to 0.61, while the R¯2 for NDA decreases from 0.85 to 0.77. It would thus appear that the capital flow equation is less robust.

V. Summary and Conclusions

Two approaches used in econometric analysis of capital movements are examined in this paper. They reflect the substantial progress that was achieved in recent years in that area, but they also raise a number of questions that still need to be resolved. The more traditional partial equilibrium approach, concentrating on the asset market, ignores important influences that are not contained in the asset demand and supply function as specified in this approach. In particular, it ignores the role of the asset and money flows in generating an interdependence between domestic and foreign interest rates. A version of the macroeconomic approach that was developed by Kouri and Porter incorporated base money into the analysis of capital movement in a model that has some important general equilibrium properties. However, the reduced form, which is used as a basis for the regression analysis and assumes perfect capital mobility as a limiting case, provides no information about the asset supply and demand functions at home and abroad that drop out in the process of solving the system. A further difficulty involves the assumption that the monetary authorities do not neutralize the effect of foreign exchange flows on domestic base money. Relaxing this assumption raises a difficult issue of the proper specification of the reaction function of the monetary authorities.

The results obtained in the statistical analysis of capital movements of Austria and France point out some of these difficulties.

In Austria, where monetary policy of the authorities is conducted through very close contact between the Austrian National Bank and the major commercial banks, there are possibilities for a degree of neutralization of foreign exchange flows. The partial equilibrium approach is useful in estimating the response of capital flows to changes in interest rates at home and abroad, although this approach cannot be used to harmonize domestic interest rates with the rates abroad. The monetary macroeconomic approach does not appear to be useful, since no significant response of capital movements to changes in the money supply was found.

In France, the partial equilibrium approach has not proved to be useful in view of a substantial harmonization of domestic interest rates with the interest rates abroad and particularly those in the Euro-dollar market. The approach based on a reduced form of a macroeconomic model yields better results and suggests that changes in the domestic component of the base money are moderately offset by induced capital movements. The attempt to specify a simple reaction function of the authorities, to eliminate the simultaneity between the changes in the domestic component of base money and capital movements introduced by the policy of neutralization of foreign exchange flows, yielded mixed results. Although the reaction function appeared to be significant, its incorporation into a two-stage least-squares estimation did not change significantly the explanatory variables of capital movements. A more complex reaction function incorporating some of the shifting policy objectives described in Section III may be called for.

Introducing current and lagged independent variables in estimating capital flows as a dependent variable permitted an evaluation of the length of adjustment in the process of offsetting changes in the domestic component of base money by induced capital movements. In the case of France, the results suggest that the offsetting process is not unduly fast. Only about 25 to 40 per cent of the change in net domestic assets appears to be offset by capital flows over the first quarter, and another 10 to 20 per cent over the second quarter. The length of adjustment in the offsetting process is an important factor in evaluating a degree of independence of the monetary policy under the fixed exchange rate system. The process of adjustment shown in the regression results for France suggests that the monetary policy of French authorities in the short run is fairly independent.

All regression results based on a reduced form of a macroeconomic model developed by Kouri and Porter (1974) suggest that conditions approaching perfect capital mobility cannot be assumed for most industrial countries. Under conditions of less than perfect capital mobility, a reduced form regression equation is not useful in tracing the adjustment process between asset and money markets at home and abroad, and government economic policy must be explicitly incorporated in the model. An econometric estimation of a complete model would avoid these shortcomings. Such a model could integrate a carefully specified portfolio analysis of the financial sector, achieved in the past through partial equilibrium analysis, within a macroeconomic model that can make endogenous general equilibrium adjustments of financial and monetary variables and can quantify government policy measures.

APPENDIX Source of Data

All data are quarterly quotations expressed in domestic currency units.

1. Euro-dollar interest rates: 90-day deposit rates (averages of end-of-week quotations)—Bank of England.

2. Austria (in billions of schillings)—first quarter of 1960 to first quarter of 1971.

  • Balance of payments: Die Österreichische Zahlungbilanz—Österreichische Nationalbank, Vienna.

  • Interest rate: Weighted average of return on fixed interest-bearing paper newly issued on the Austrian market—Institute for Advanced Studies, Vienna.

  • Net foreign assets and reserve money: International Monetary Fund, International Financial Statistics, lines 11 and 14.

  • Gross national product: Monatsberichte—Österreichisches Institut für Wirtschaftsforschung.

3. France (in millions of francs)—first quarter of 1963 to first quarter of 1971.

  • Balance of payments: first quarter of 1963 to fourth quarter of 1967—International Monetary Fund, Balance of Payments Yearbook; first quarter of 1968 to first quarter of 1971, Banque de France, Direction générale des services étrangères, Balance de paiements entre la France et l’étranger.

  • Interest rates: Call money rate—International Monetary Fund, International Financial Statistics, line 60b.

  • Net foreign assets and reserve money: International Monetary Fund, International Financial Statistics, lines 11, 14, and 16c.

  • Gross domestic product: Production Intérieure Brute, Tendances de la Conjoncture—Institut national de la statistique et des études économiques.

BIBLIOGRAPHY

  • Argy, Victor, and Zoran Hodjera, “Financial Integration and Interest Rate Linkages in the Industrial Countries, 1958–71,” Staff Papers, Vol. 20 (March 1973), pp. 177.

    • Search Google Scholar
    • Export Citation
  • Argy, Victor, and Pentti J. K. Kouri, “Sterilization Policies and the Volatility in International Reserves,” in National Monetary Policies and the International Financial System, ed. by R. Z. Aliber (Chicago, 1974), pp. 209230.

    • Search Google Scholar
    • Export Citation
  • Arndt, Sven W., “International Short Term Capital Movements: A Distributed Lag Model of Speculation in Foreign Exchange,” Econometrica, Vol. 36 (January 1968), pp. 5970.

    • 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.

    • Search Google Scholar
    • Export Citation
  • Black, Stanley W., “The Use of Rational Expectations in Models of Speculation,” Review of Economics and Statistics, Vol. 54 (May 1972), pp. 16165.

    • Search Google Scholar
    • Export Citation
  • Black, Stanley W., International Money Markets and Flexible Exchange Rates, Princeton Studies in International Finance, No. 32 (Princeton University Press, 1973).

    • Search Google Scholar
    • Export Citation
  • Bourginat, Henri, Y. Marquet, and G. Nancy, Effets tendanciels des mouvements de capitaux à court terme, Université de Bordeaux I (June 1973).

    • Search Google Scholar
    • Export Citation
  • Branson, William H., Financial Capital Flows in the U. S. Balance of Payments (Amsterdam, 1968).

  • Branson, William H., “Monetary Policy and the New View of International Capital Movements,” Brookings Papers on Economic Activity: 2 (1970), pp. 23562.

    • Search Google Scholar
    • Export Citation
  • Branson, William H., and Raymond D. Hill, Jr., “Capital Movements in the OECD Area: An Econometric Analysis,” OECD Economic Outlook: Occasional Studies (Paris, December 1971).

    • Search Google Scholar
    • Export Citation
  • Bryant, Ralph C, and Patric H. Hendershott, Financial Capital Flows in the Balance of Payments of the United States: An Exploratory Empirical Study, Princeton Studies in International Finance, No. 25 (Princeton University Press, 1970).

    • Search Google Scholar
    • Export Citation
  • Helliwell, John F., “A Structural Model of the Foreign Exchange Market,” Canadian Journal of Economics, Vol. 2 (February 1969), pp. 90105.

    • Search Google Scholar
    • Export Citation
  • Helliwell, John F., Lawrence H. Officer, Harold T. Shapiro, and Ian A. Stewart, The Structure of RDXI, Bank of Canada, Staff Research Studies, No. 3 (Ottawa, July 1969).

    • Search Google Scholar
    • Export Citation
  • Hodjera, Zoran, “Short-Term Capital Movements of the United Kingdom, 1963–1967,” Journal of Political Economy, Vol. 79 (July/August 1971), pp. 73975.

    • Search Google Scholar
    • Export Citation
  • Hodjera, Zoran, “International Short-Term Capital Movements: A Survey of Theory and Empirical Analysis,” Staff Papers, Vol. 20 (November 1973), pp. 683740.

    • Search Google Scholar
    • Export Citation
  • Kesselman, Johnathan, “The Role of Speculation in Forward-Rate Determination: The Canadian Flexible Rollar 1953–1960,” Canadian Journal of Economics, Vol. IV (August 1971), pp. 27998.

    • Search Google Scholar
    • Export Citation
  • Kouri, Pentti J. K., “The Hypothesis of Offsetting Capital Flows: A Case Study of Germany,” Journal of Monetary Economics, Vol. 1 (January 1975), pp. 2139.

    • Search Google Scholar
    • Export Citation
  • Kouri, Pentti J. K., and Michael G. Porter, “International Capital Flows and Portfolio Equilibrium,” Journal of Political Economy, Vol. 82 (May/June 1974), pp. 44367.

    • Search Google Scholar
    • Export Citation
  • Lee, C. H., “A Stock-Adjustment Analysis of Capital Movements: The United States-Canadian Case,” Journal of Political Economy, Vol. 77 (Part 1, July/August 1969), pp. 51223.

    • Search Google Scholar
    • Export Citation
  • Machlup, Fritz, Walter S. Salant, and Lorie Tarshis (eds.), International Mobility and Movement of Capital, National Bureau for Economic Research (Columbia University Press, 1972).

    • Search Google Scholar
    • Export Citation
  • Miller, Norman C., and Marina v.N. Whitman, “A Mean-Variance Analysis of United States Long-Term Portfolio Foreign Investment,” Quarterly Journal of Economics, Vol. 84 (May 1970), pp. 17596.

    • Search Google Scholar
    • Export Citation
  • Porter, Michael G., “Capital Flows as an Offset to Monetary Policy: The German Experience,” Staff Papers, Vol. 19 (July 1972), pp. 395424.

    • Search Google Scholar
    • Export Citation
  • Rhomberg, Rudolf R., “Canada’s Foreign Exchange Market: A Quarterly Model,” Staff Papers, Vol. 7 (April 1960), pp. 43956.

  • Rhomberg, Rudolf R., “A Model of the Canadian Economy under Fixed and Fluctuating Exchange Rates,” Journal of Political Economy, Vol. 72 (February 1964), pp. 131.

    • Search Google Scholar
    • Export Citation
  • Stoll, Hans R., “An Empirical Study of the Forward Exchange Market under Fixed and Flexible Exchange Rate Systems,” Canadian Journal of Economics, Vol. 1 (February 1968), pp. 5578.

    • Search Google Scholar
    • Export Citation
  • Willms, Manfred, “Controlling Money in an Open Economy: The German Case,” Federal Reserve Bank of St. Louis, Review (April 1971), pp. 1027.

    • Search Google Scholar
    • Export Citation
*

Mr. Hodjera, Senior Economist in the Research Department, has degrees from the Graduate Institute of International Studies in Geneva and from Columbia University and also has studied at Oxford University. He has been lecturer at the City College of New York, Assistant Professor at Yale University, and Visiting Associate Professor at the University of Virginia. He has contributed a number of articles to economic journals.

2

For a quite recent econometric analysis of capital movements in the Federal Republic of Germany within a structural model of the financial sector, see Artus (1976).

3

The approach was developed by Branson (1968); Bryant and Hendershott (1970); and Branson and Willett, and Miller and Whitman, in Machlup, Salant, and Tarshis (1972). For a discussion of this approach, see Hodjera (1973, pp. 700–713, and 721–28).

4

In a portfolio selection model, of which equations (1.1) and (1.2) are a particular form, risk variables Va and Ve can be seen as standard deviations based on a weighted sum of variances of investors’ subjective distributions of returns on foreign and domestic assets and of covariance between these returns. For an expression defining risk variables, see Hodjera (1973, p. 700, footnote 35).

5

Therefore it is assumed that asset holders and traders behave both as arbitrageurs and speculators in the foreign exchange markets. For further discussion of this point, see Argy and Hodjera (1973, pp. 36–39).

6

It is assumed that home exporters are providing short-term credit to foreign importers.

7

This approach is followed because complete data are lacking on total foreign assets and liabilities to foreigners. Balance of payments statistics on the net basis are used in most cases. See, for example, Hodjera (1971), Branson and Hill (1971), Porter (1972), Kouri and Porter (1974), and Kouri (1975).

8

Trade flows are entered in a net form (X – M = trade balance) in order to associate them with net flow of capital (LtAt).

9

Branson (1968, pp. 43–66), Stoll (1968), Arndt (1968), and Kesselman (1971) attempted to derive expectation functions that could be used in estimating capital movements, but they had little success where speculation within the band is concerned. In his analysis of capital movements of industrial countries other than the United States, Branson reverted to the usual practice of using dummy variables.

10

For statistical results suggesting such return flows, see Branson (1968, Ch. 4), Hodjera (1971), and Branson and Hill (1971).

11

Interest rates are entered separately and not as a differential because domestic and foreign interest elasticities of capital movements are different. This also provides a degree of comparability between the partial equilibrium approach and the approach based on a reduced form of a macroeconomic model.

12

On a reasonable assumption that each country is financing its own exports, a change in the trade balance will generate fresh capital flows needed for such finance. If the repayment period is short enough, an unchanged trade balance would not generate a need for additional net financing.

13

See Kouri and Porter (1974) on this point.

14

The symbol GNP is used for income in order to distinguish it from Yd, which stands for yield in equation (1.3).

15

In terms of relation (1.3) the condition is

IDIfReRR

Furthermore, risk indicating the standard deviation of subjective expected returns should also be considered. The larger the premium of the expected spot rate over the actual spot rate and the larger the standard deviation, the wider will be the possible discrepancy between domestic and foreign interest rates in the equilibrium.

16

Exchange rate fluctuations within the band and uncertainty associated with risk make the absolute size of the offset coefficient smaller than unity since, other things being equal, only a part of the changes in base money will be transmitted abroad through capital flows. The wider the intervention band and the more frequent the fluctuations within the band, the smaller will be the absolute value of the coefficient. If, however, there is no assumption of exogenously determined exchange rate expectations, the absolute value of the offset coefficient could be larger than unity when exchange rate expectations are influenced by changes in the supply of base money.

17

For further discussion on policy options by the authorities, see Argy and Hodjera (1973, pp. 21–28).

18

See, for example, Arndt (1968) and Black (1972; 1973). For flexible exchange rates, see Artus (1976).

24

The one exception is Porter’s (1972) study of the Federal Republic of Germany.

25

See Branson (1968, pp. 99–103). He studied the relation of the U. S. interest rates, on the one hand, and the interest rates of the United Kingdom, Canada, the Federal Republic of Germany, and Switzerland, on the other. A simple regression coefficient of changes in interest differentials between the United States and these countries on changes in corresponding non-U. S. interest rates were in all cases considerably larger than the regression coefficient on the same changes in interest differentials on the change in the U. S. interest rate.

26

Single equation estimates yielded the following offset coefficients: Australia, –0.5; Italy, –0.4; the Netherlands, –0.6; and the Federal Republic of Germany, between –0.7 and –0.8. A two-stage least-squares estimation that incorporated the policy reaction function reduced the offset coefficient for the Federal Republic of Germany to –0.5, while leaving unchanged the coefficients for the Netherlands and Italy.

27

Kouri (1975, pp. 36–37), however, claims that in the Federal Republic of Germany the sterilization policy would entail accumulation of foreign exchange of such a magnitude as to make this policy unfeasible.

28

For the Austrian domestic interest rate, a weighted average of return on fixed interest-bearing paper, newly issued on the Austrian market, is used.

29

The sum of these two coefficients in equation (3.1) exceeds unity. However, this difference is not statistically significant.

30

Another difficulty relates to a 1967 revision of the coverage in balance of payments statistics. While French balance of payments statistics beginning with 1967 cover transactions of France with the rest of the world, the series prior to 1967 cover transactions of France with countries outside the French franc area. For further discussion, see International Monetary Fund, Balance of Payments Yearbook, Vol. 20 (1963–67) and Vol. 23 (1966–70).

31

The pattern of speculative capital flows over these four quarters is approximated by the lag structure shown below.

It is assumed that the initial speculative outflow is reversed over the following two quarters and that the equilibrium was re-established by the end of the first quarter of 1969.

32

One would expect that the relation between RCU and ΔNDA is not linear. However, no simple nonlinear relation between these two variables yielded satisfactory results.

33

It should be pointed out, however, that large increases in money supply in the second quarters of 1968 and 1969 are associated with substantial outflows of foreign assets. Dropping DM from the regression equation improves its R¯2 substantially.