A note on capital mobility.

Moosa, Imad A.

I. Introduction

Empirical work on capital mobility, using the Feldstein-Horioka [5] saving-investment correlation as a measurement criterion has generally produced two results. The first result is that even when the econometric problems (particularly the endogeniety of domestic saving) are addressed, the Feldstein-Horioka finding of low capital mobility seems to hold up [7, 1084]. The second result is that the surprising finding of low capital mobility is obtained even if the data set extends well into the 1990s as was found by, inter alia, Dar, Amirkhalkhali, and Amirkhalkhali [4]. This finding is not consistent with the conventional wisdom that capital mobility has increased at an accelerating rate since the early 1970s. For example, Frankel and MacArthur [7, 1084] list the following stylized facts: (i) the degree of capital mobility is high; (ii) it is higher for industrial countries; and (iii) it has been rising since the 1950s and particularly since 1973.(1) The saving-investment correlation as a criterion does not only indicate that the degree of capital mobility is low across the board, but also fails to show that it has been increasing over time, thus refuting the stylized facts (i), (ii) and (iii).

Economists have for long recognised the contradiction between the casual-observation-based conventional wisdom that capital mobility has reached a high level and the empirical evidence which indicates the contrary. It seems, however, that this contradiction is due to a combination of conceptual, methodological and econometric issues. The objective of this note is to shed some light on these issues and to present empirical evidence on capital mobility based on real interest parity.

II. Some Conceptual and Methodological Issues

The literature on capital mobility and market integration seems to have traces of ambiguity, confusion and lack of clarity, leading to contradictory conclusions. Frankel [6, 27] makes this point clear by asserting that "the aggregation together of all forms of capital has caused more than the usual amount of confusion in the literature on international capital mobility." It is possibly the case that economists reaching the opposite conclusions about capital mobility might refer to different things i.e., different concepts and measurements of capital. For example, the word "capital" may imply "micro" as opposed to "macro" capital, "net" as opposed to "gross" capital, "portfolio" as opposed to "physical" capital,(2) and "short-term" as opposed to "long-term" capital. Furthermore, it is not really clear whether or not economists regard capital mobility and market integration as the same thing or if one is a necessary condition for the other.

It may be the case that the controversy raised by the Feldstein-Horioka work is due to their use of capital mobility in a macroeconomic sense whereas most of the studies are based on capital in a microeconomic sense. And while capital moves in all directions, it could be the case that net capital flows are not significant, and that it is net capital movements which Feldstein and Horioka were concerned with. This is because net capital flows, which constitute the financial counterpart to the transfer of real resources through payment imbalances, arise only when saving and investment are not matched within individual countries. On the other hand, gross capital flows can be mutually offsetting across countries.

In this note we are concerned with financial or portfolio (gross) capital flows. An issue that needs some clarification is the relationship between capital mobility and market integration, two concepts that are often used interchangeably. A careful reading of the literature, as manifested by Goldstein, Mathieson, and Lane [8] and Frankel [6], leads to the following propositions:

1. Macroeconomic capital mobility is not a necessary condition for market integration.

2. Equalization of real interest rates across countries is not a necessary condition for market integration.(3)

3. Market integration is conducive to capital mobility, or at least it leads to increasing "potential" capital flows.(4)

4. Capital mobility is a sufficient condition for market integration, or at least that a high level of capital mobility indicates a high level of market integration.

5. Equalization of real interest rates does not preclude the effect of a shortfall in domestic saving on domestic investment projects.

The literature also reveals a diversified menu of measurement criteria. Apart from the saving-investment correlation, capital mobility and market integration can be measured by deviations from the international parity conditions: covered interest parity (CIP), uncovered interest parity (UIP) and real interest parity (RIP). While Frankel and McArthur [7, 1087] assert that CIP should be the measure of capital mobility in the sense of the degree of integration of financial markets, they argue against the use of RIP for this purpose because it can be invalidated by imperfect integration of goods markets. It seems, however, that the choice of CIP as a criterion for measuring capital mobility is motivated by the desire to reconcile the findings of empirical tests with the conventional wisdom that capital has become highly mobile. This is because (conventional) RIP tests, such as Mishkin [12], have shown significant deviations from this parity condition and thus failed to confirm the observed high degree of capital mobility. Tests of CIP, on the other hand, have indicated only insignificant deviations from the parity condition as shown, for example, by Taylor [14].(5)

It is arguable, however, that measures of capital mobility that are based on rates of return (such as CIP, UIP and RIP) are superior to saving-investment correlations because the former require no assumption about the exogeniety of saving, nor about the sensitivity of investment to the real interest rate. It is also arguable that CIP is only a measure of capital mobility in a narrow sense, as it is concerned with short-term, and not long-term, financial capital only.(6) But for capital mobility as a general concept, RIP seems to be a more appropriate criterion.

If this argument is accepted then one is confronted by the task of reconciling the results obtained from RIP tests and conventional wisdom. Several reasons can be presented to explain why the results of these tests have not been favourable to the RIP hypothesis, and thus to the hypothesis of high capital mobility. The first is that most of these studies used data that did not cover the 1980s. For example, Mishkin [12] used a sample covering the period 1967:2-1979:2. Secondly, since the RIP model specification involves expectational variables, measurement errors could be responsible for the failure of RIP. Thirdly, it is possible that the tests used are not powerful enough to confirm RIP.(7) Although this statement is not specific it can be substantiated by the observation that studies using more powerful tests than the conventional ones on more recent data, as demonstrated by Cavaglia [3], have produced results that are supportive of RIP. Moreover, Goodwin and Grennes [9] put forward the existence of non-traded goods and transportation costs as explanatory factors for the failure of RIP and, therefore, the apparent incompatability between the equalization of real interest rates and market integration.

III. Model Specification and Empirical Evidence

The RiP hypothesis can be tested in a univariate framework by detecting, or otherwise, mean reversion in ex ante real interest rate differentials. The RIP condition can be derived by combining the hypotheses of covered interest parity, ex ante purchasing power parity, and unbiasedness of the forward rate as a forecaster of the market's expectation of the future spot rate. These conditions are given respectively by

[Mathematical Expression Omitted]

[Mathematical Expression Omitted]

[Mathematical Expression Omitted]

where [f.sub.t] is the logarithm of the one-period forward exchange rate, [s.sub.t] is the logarithm of the spot exchange rate, [i.sub.t] is the domestic nominal interest rate, [Mathematical Expression Omitted] is the logarithm of the spot exchange rate expected to prevail at time t + 1, [Mathematical Expression Omitted] is the expected change in the logarithm of the domestic price index from time t to t + 1, and the asterisk denotes the corresponding foreign variables. Combining equations (1) and (3), we obtain uncovered interest parity, which is given by

[Mathematical Expression Omitted].

Combining equations (4) and (2) and rearranging, we obtain

[Mathematical Expression Omitted].

Equation (5) implies that if the Fisher closed condition is valid both at home and abroad, then the nominal interest rate differential adjusts fully to the expected inflation differential, maintaining the constancy within and equality across countries of ex ante real interest rates. Therefore

[Mathematical Expression Omitted]

where [Mathematical Expression Omitted] is the domestic (foreign) real interest rate expected to prevail in period t + 1. Under the assumption that the actual (ex post) real interest rate realised at time t + 1 differs from the expected real interest rate by a zero mean stationary error, the expectation formation mechanism may be represented by

[Mathematical Expression Omitted]

and

[Mathematical Expression Omitted]

where [Mathematical Expression Omitted] and [Mathematical Expression Omitted]. If the ex post real interest rate is stationary, this may be interpreted as indicating that the ex ante real interest rate is constant over time. Thus

[Mathematical Expression Omitted]

where [Mathematical Expression Omitted] is a zero mean stationary process.(8) In general, RIP holds as a long-run equilibrium condition if the real interest differential is mean-reverting over time.

I will initially test for mean reversion in the real interest rate differentials of five countries vis-a-vis the U.S. These countries are Australia, Belgium, Canada, Germany and the U.K. The choice of these countries was determined by the availability of data covering a long period of time and extending well into the 1990s. The sample consists of 86 quarterly observations covering the period 1972:1-1993:2. Data were obtained from Datastream (IMF series).

Three different tests are used for this purpose: (i) the Phillips-Perron [13] (PP) test of the null hypothesis of unit root; (ii) the Kwiatkowski-Phillips-Schmidt-Shin [10] (KPSS) test of the null hypothesis of stationarity; and (iii) the Brock-Dechert-Scheinkman [1] (BDS) test of the null hypothesis that the interest rate differential is an independently and identically distributed (iid) variable with zero mean and constant variance. The results of these tests, which are presented in Table I, are highly supportive of RIP as a long-run equilibrium condition.(9) First, the four PP statistics Z([[Tau].sub.[Mu]]), Z([[Tau].sub.[Tau]]), Z([[Phi].sub.2]), and Z([[Phi].sub.3]) reject the null hypothesis of unit root in the interest [TABULAR DATA FOR TABLE 1 OMITTED] rate differentials. Second, the KPSS statistics [[Eta].sub.[Mu]] and [[Eta].sub.[Tau]] cannot reject the null hypothesis of stationarity, thus confirming the interpretation of the PP test results that the rejection of the null of unit root implies stationarity of the interest rate differentials. Third, the BDS statistic, which is based on the correlation dimension, does not reject the null hypothesis that the differentials are iid.(10)

Table II. Testing Mean Reversion in Real Interest Differentials
(1972-93)

 Z([[Tau].sub.[Mu]]) Z([[Phi].sub.2])

Australia/U.K. -8.29 22.82
Canada/U.K. -8.24 24.24
U.K./Germany -6.47 21.12
Belgium/Germany -5.52 11.42
Canada/Germany -5.58 13.76

Two points may arise out of these results. The first point is that testing the RIP hypothesis with the U.S. being the reference country may bias the results since the U.S. was the most volatile economy during a large part of the period under study and has always enjoyed the most open markets. The second issue arises from the consensus view that the degree of capital mobility changed over time as regulations waned and institutions evolved. Therefore, the results do not show the time specific effects. To deal with these two points, the RIP tests are repeated for country combinations excluding the U.S., and for combinations including the U.S. over two sample periods, 1972-79 and 1980-93. Given the space constraint, the results of two tests only are reported (the Phillips-Perron Z([[Tau].sub.[Mu]]) and Z([[Phi].sub.2]) tests).

Table II reports the results of testing RIP over the whole sample period for five country combinations excluding the U.S. The results still support RIP as the null of nonstationarity is rejected in all cases. Table III presents the results of testing RIP vis-a-vis the U.S. over two periods, the first of which ends in 1979:4.(11) The results show that only in the case of Belgium did RIP not hold in the earlier period which gives some support to the proposition pertaining to the gradual nature [TABULAR DATA FOR TABLE III OMITTED] of increasing market integration. One reason why the results are not strongly supportive of this proposition is that even in the 1970s market integration had reached a high level, propelled by the removal of capital controls by the U.S., Germany, Canada and Switzerland; by the process of financial innovation; and by the recycling of OPEC surpluses to developing countries [6, 28]. In fact, by the second half of the 1970s, economists started talking about the world financial system as characterised by perfect capital mobility. It is perhaps the case that conducting these tests over the 1950s or the 1960s would produce evidence indicating lack of integration and capital mobility as represented by the failure of RIP.

IV. Conclusion

The evidence presented in this paper lends some support to the RIP hypothesis. This is attributed to the application of some appropriate tests to a sample of data that extends well into the 1990s, thus capturing the full effect of recent developments which have been conducive to market integration and capital mobility. Even if we allow for the distinction between market integration and capital mobility and interpret the results of the RIP tests to be favorable to the former only, then the conclusion is that an environment that is conducive to capital mobility has been created. If we combine this formal evidence with the casual observation that capital flows have been large (as indicated by the volume of trading in the foreign exchange market, for example) we have no alternative but to question the validity of the empirical results indicating low capital mobility.

The evidence presented in this paper can be interpreted as implying that the mobility of capital has become so high that disparities in real interest rates cannot persist and that in the long run the real interest differential reverts to a mean value of zero. Such evidence is obviously in contrast with the Feldstein-Horioka type of results, but it is consistent with conventional wisdom that capital mobility has become very high.

Imad A. Moosa La Trobe University Bundoora, Victoria, Australia

1. Goldstein, Mathieson, and Lane [8, 1] point out that there has been a sharp expansion in the scale of net and gross capital flows among industrial countries, and that the easing of capital controls and the liberalization of financial markets have brought about a growing integration of domestic and offshore markets.

2. The distinction between portfolio and physical capital pertains to the distinction between portfolio and direct investment.

3. Equalization of real interest rates across countries is, however, a positive indicator of market integration.

4. For example, Frankel [6, 27] asserts that "the increased degree of worldwide financial integration since 1979 is identified as one likely factor that has allowed such large capital flows to take place over the past decade".

5. While the relaxation of capital controls has significantly reduced deviations from CIP as implied by existing empirical evidence, there are other reasons for the failure of UIP and RIP. Failure to find support for UIP could be due to errors in measuring the expected change in the exchange rate or the presence of a risk premium. Even if UIP holds, RIP may not hold because of the failure of ex ante PPP.

6. CIP is unlikely to hold in long-term capital markets because of the unavailability of long-term forward contracts.

7. For example, Frankel [6, 29] asserts that RIP "has not held any better in recent years than it did in the past". However, he based his assertion only on Mishkin [12] and Caramazza et al. [2].

8. The assumption of stationary expectation errors is less stringent than the rational expectations requirement of white noise.

9. In these tests interest rates are not adjusted for differences in national tax rates although it is arguable that when economic agents pay taxes on interest income, the net of tax real interest rate will be more economically meaningful [11, 196]. The main reason why taxes are not allowed for in this study, and all other studies of RIP except Mark [11], is the unavailability of time series on comparable national tax rates. This is evident in the procedure used by Mark who applied a single point value of the corporate (not income) tax rates prevailing in 1977 to the time series of nominal interest rates. In my view, this procedure is faulty for it introduces measurement errors and cannot serve as an empirical representation of the sound theoretical proposition that it is more plausible to use net of tax rates. In any case, there was no significant difference between the results based on gross and those based on net of tax rates: both rejected RIP.

10. The BDS statistics were calculated for an embedding dimension, m, of 2 while the distance between two m-histories, [Epsilon], was fixed at a value equal to half a standard deviation. Trials with several other combinations did not change the results significantly.

11. There are two reasons for using 1979:4 as a split point. The first is that Frankel [6, 27] refers to 1979 as the year since which increased integration is identified as one likely factor that has triggered large capital flows. The second reason is that it divides the full sample in such a way as to leave the sub sample sizes reasonably suitable for empirical testing.

References

1. Brock, W. A., W. D. Dechert and J. A. Scheinkman. "A Test for Independence Based on the Correlation Dimension." Social Systems Research Unit, University of Wisconsin, 1987.

2. Caramazza, F., K. Clinton, A. Cote and D. Longworth. "International Capital Mobility and Asset Substitutability: Some Theory and Evidence on Recent Structural Changes." Technical Report 44, Bank of Canada, 1986.

3. Cavaglia, Stefano, "The Persistence of Real Interest Differentials: A Kalman Filtering Approach." Journal of Monetary Economics, June 1992, 429-43.

4. Dar, Atul, Saleh Amirkhalkhali and Samad Amirkhalkhali, "On the Fiscal Policy Implications of Low Capital Mobility: Some Further Evidence from Cross-Country, Time-Series Data." Southern Economic Journal, July 1994, 169-90.

5. Feldstein, Martin and Charles Horioka, "Domestic Saving and International Capital Flows," Economic Journal, June 1980, 314-29.

6. Frankel, Jeffrey. "Quantifying International Capital Mobility in the 1980s," in International Finance: Contemporary Issues, edited by Dilip K. Das. New York: Routledge, 1993.

7. ----- and Alan MacArthur, "Political vs. Currency Premia in International Real Interest Differentials." European Economic Review, June 1988, 1083-121.

8. Goldstein, Morris, Donald Mathieson and Timothy Lane. Determinants and Systematic Consequences of International Capital Flows, IMF Occasional Papers, No. 77, March 1991.

9. Goodwin, Barry and Thomas J. Grennes, "Real Interest Equalization and the Integration of International Financial Markets." Journal of International Money and Finance, February 1994, 107-24.

10. Kwiatkowski, Denis, Peter Phillips, Peter Schmidt, and Yongcheol Shin, "Testing the Null Hypothesis of Stationarity Against the Alternative of a Unit Root: How Sure are We that Economic Time Series Have a Unit Root?" Journal of Econometrics, October/December 1992, 159-78.

11. Mark, Nelson, "Some Evidence on the International Inequality of Real Interest Rates." Journal of International Money and Finance, June 1985, 189-208.

12. Mishkin, Fredric, "Are Real Interest Rates Equal Across Countries? An Empirical Investigation of International Parity Conditions." Journal of Finance, December 1984, 1345-58.

13. Phillips, Peter and Pierre Perron, "Testing for a Unit Root in Time Series Regression." Biometrika, June 1988, 335-46.

14. Taylor, Mark P., "Covered Interest Parity: A High-Frequency High-Quality Data Study." Economica, November 1987, 429-53.