تحرک سرمایه بین المللی در دراز مدت و کوتاه مدت: ما هنوز می توانیم از داده های صرفه جویی در سرمایه گذاری یاد بگیریم؟

The idea to learn about international capital mobility from saving and investment data remains appealing. Our approach is based on VAR methods and overcomes some of the problems associated with saving–investment regressions when the data are non-stationary. We propose a new measure of long-run capital mobility that can be easily calculated as a by-product of the estimation procedure of a cointegrated VAR. In an application to historical US and British data, we find long-run capital mobility to have been remarkably stable over the century whereas variations in the mobility of capital primarily seem to have affected short-run capital flows.

In a world with perfect capital mobility, a country can always run current account deficits if its desire to consume and invest cannot be funded domestically. This basic insight provided the motivation for the seminal paper by Feldstein and Horioka (1980) in which the authors found very high saving–investment correlations for a large cross-section of OECD countries. Their result has long been perceived as a puzzle and constitutes a challenge to the view that world capital markets are well integrated. In the presence of perfect capital mobility, investment should go where it yields the highest real returns, whilst consumption should depend only on the permanent value of income, not on contemporaneous investment decisions. Even though it has recently been questioned on theoretical and econometric grounds, the basic intuition behind the Feldstein–Horioka approach—that it is possible to make inference about international capital mobility (ICM) from saving and investment data alone—remains appealing. In this paper, we propose a new approach to measuring capital mobility based on saving and investment data. Our approach overcomes some of the most important problems of simple saving–investment correlations. Not only does it take account of the fact that, under the assumptions of standard open-economy models, the two variables are likely to be cointegrated, but it also allows us to distinguish clearly between short-run and long-run capital mobility. The measure of short-run capital mobility is a suitably adjusted correlation, similar to the one suggested by Feldstein and Horioka, whereas the measure of long-run ICM is based on Johansen’s (1988) procedure for estimating the cointegrating space. In an application of our method to historical US and British data, we find long-run capital mobility to have been remarkably stable over the century whereas variations in the mobility of capital seem primarily to have affected short-run capital flows. Obstfeld and Taylor (2002) have argued that if capital mobility could be summarized in one single number, the time path of this measure would have described a U-shape over the course of the 20th century: a period of increasing capital mobility up to the beginning of world war I, then a period of low capital mobility in the interwar period, followed by a gradual recovery of capital mobility in the second half of the 20th century. Our formal results confirm this narrative evidence but they also show that countries’ ability to smooth consumption over long time periods (i.e. long-run capital mobility) has varied less than their ability to implement short-run changes in domestic investment independently from changes in savings (i.e. short-run capital mobility). It is not within the scope of this paper to attempt to survey the huge literature on the Feldstein–Horioka finding (for a recent survey see Coakley et al., 1998 or Obstfeld and Rogoff, 1995). But we will briefly sketch the theoretical and econometric developments that have led to an emerging consensus in the profession that simple saving–investment correlations are not very informative with respect to ICM. We will also survey how the empirical literature has responded to this challenge. It is interesting to note that formal theoretical rationalizations and most empirical explorations of the saving–investment correlation aim at explaining the time series behaviour of the two variables, whereas the original study by Feldstein and Horioka (1980) emphasized the cross-sectional result. The present paper is in line with the bulk of the literature and confines analysis to the time series properties of saving and investment. There are three challenges to the Feldstein–Horioka approach of measuring capital mobility that have proven to be particularly powerful, presumably because they explain the saving–investment correlation in the framework of intertemporal models of the current account: • persistentshocks; in a small open economy, total factor productivity shocks that are sufficiently persistent can create positively correlated impulse responses of saving and investment even though capital mobility is perfect ( Obstfeld, 1986, Obstfeld, 1995 and Mendoza, 1991). • globalshocks; in response to global shocks there is no scope for consumption smoothing through the current account and saving and investment move together ( Baxter and Crucini, 1993). • non-stationarityofsavingandinvestment; if shocks to total factor productivity (TFP) are permanent, TFP will be a process that is integrated of order one. The non-stationarity in TFP induces non-stationarity in output, consumption, saving and investment. At the same time, the intertemporal budget constraint in such models requires the current account to be stationary. As a result, saving and investment cointegrate and the Feldstein–Horioka regression will not have any implications for capital mobility as it is likely to pick up this cointegrating relationship ( Coakley et al., 1996). Some authors have suggested more policy-based rationalizations of the saving–investment correlation. For example, Artis and Bayoumi (1992) have investigated the possibility of government targeting of the current account. Taking stock of the problems associated with simple saving–investment correlations, the recent empirical literature has switched to more sophisticated econometric methods. Vector autoregressions are now probably the most popular technique used in the field. They allow one to decompose saving and investment into permanent and transitory components and to assess their response to various types of shocks. We will summarize some of the recent contributions. Ghosh (1995) has used an intertemporal model to derive a desired current account from observed data. He finds that the desired current account tracks the actual current account reasonably well, hence providing evidence in favour of perfect capital mobility. Moreno (1997) has suggested interpreting the degree of short-run divergence in the impulse responses of saving and investment as a measure of capital mobility. Sarno and Taylor (1998) used the structural VAR approach pioneered by Blanchard and Quah (1989) to decompose saving and investment into permanent and transitory components. They find that transitory components of UK/US saving and investment are more highly correlated than changes in the permanent components. They claim that this finding is consistent with the presence of frictions in international capital markets. Only if innovations are permanent does investment flow abroad and the link between saving and investment is loosened. If, however, shocks are transitory, then the cost of investing abroad might be too high due to market frictions and a high correlation between saving and investment emerges. However, their results are supportive of the notion that capital mobility has increased in the 1980s: they report short-run correlations between saving and investment for the period 1979–1994 that are significantly lower than for the 1955–1979 period. Plausibly the closest precursor to this paper is Taylor (2002) who suggests to interpret univariate persistence measures of the current account as measures of capital mobility and relates them to a vector error-correction model in savings and investment. Taylor shows that under general dynamic specifications, current accounts will not follow an AR(1) and that univariate adjustment coefficients may not capture the true persistence of the current account. One innovation of our paper relative to Taylor’s is that we suggest a measure of the speed of error correction that compounds the dynamic interaction of savings and investment into one single index of current account persistence, irrespective of the dynamic structure of savings and investment. We interpret this persistence index as a measure of long-run capital mobility. The reminder of the paper is organized as follows: Section 2 presents a simple model of current account dynamics based on intertemporal optimization. These models were first applied to current account dynamics by Sachs (1981). Section 3 discusses the classical Feldstein–Horioka regression. In Section 4, we suggest a new measure of long-run ICM which is easily calculated as a by-product of Johansen’s (1988) procedure for the estimation of the cointegrating space. Section 5 applies our insights to a unique set of long-run historical data from the UK and the US. Section 6 concludes.

Recent theoretical and econometric developments have cast doubt on the usefulness of saving–investment correlations as a measure of international capital mobility. Still, the suggestion by Feldstein and Horioka of making inferences about international capital mobility from saving and investment data remains appealing. After all, the theory does suggest that investment should flow where it yields the highest real return and that saving depends on the intertemporal consumption decision alone. In this paper, we have proposed making inference about international capital mobility in the framework of a bivariate cointegrated VAR of saving and investment. We have argued that the long-run adjustment process in a cointegrated system is informative about capital mobility and that it gives rise to a measure of long-run capital mobility that can be calculated easily as a by-product of Johansen’s (1988) procedure. The measure has the advantage of being a standardized index of international capital mobility that is between 0 and 1. The standard errors of this index can be calculated and it becomes possible to compare capital mobility over time and between countries. We have applied our insights to the unique Jones–Obstfeld–Taylor data set of historical saving and investment rates for the US and the UK. In the US and the UK, long-run capital mobility over the century seems to have been remarkably stable and current account positions remarkably persistent. The first world war appears as the major disruption to long-run capital mobility in this century but in both countries long-run sustainable current account positions were restored soon after the war. Obstfeld and Taylor (2002) have argued that if capital mobility could be summarized in one single number, the time path of this measure would have described a U shape over the course of the 20th century. Our findings provide more formal evidence for this view but they also allow us to show that variations in capital mobility over the century have largely been reflected in changes in the short-run saving retention coefficient whereas long-run capital mobility has been remarkably stable and apparently fairly high throughout the whole century. In particular, for the two countries examined here, we cannot reject the hypothesis that long-run capital mobility was as high in the post-WWII period as under the classical gold standard. Our findings suggest that short-to-medium term frictions in international capital markets are the prime impediment to capital flowing freely across national borders. They seem to provide empirical support for a recent literature that models impediments to capital mobility as an adjustment cost in the accumulation process of foreign assets that makes short-term capital transfers particularly costly.