Modeling International Financial Markets

A large number of issues in the field of international economics depend on one crucial question: Is the worldwide financial market integrated, or is it made up of a number of segments with imperfect capital mobility among them? To try to decide this issue, some macroeconomists have derived clues from macroeconomic variables. Feldstein and Horioka,[1] for example, have examined the behavior of national savings and investment, testing for complete segmentation. But high correlations found between savings and investment could be compatible as easily with full integration of capital markets. A second macroeconomic avenue is based on the correlation among consumption rates across countries. Under integration and certain other conditions, national consumption rates ought to be correlated perfectly. In fact, the correlations are very low (in many cases lower than for national production rates). But the correlations may be low because financial markets, although integrated, are incomplete. Evidence from the composition of aggregate portfolios of national investors also indicates that portfolios are not diversified worldwide nearly as much as they should be in a fully integrated world. This fact is called the "home-equity bias." In principle, the best way to decide the matter of integration versus segmentation is to look at prices in financial markets. If similar assets - or similar dimensions of risk - traded in different places do not receive the same price, then full integration does not exist. Furthermore, data on prices in the financial market are plentiful, more so than macroeconomic data. However, the approach based on financial-market prices necessarily relies on some model of the relationship between expected rerum and risk. Deviations from the pricing model act as "noise" in the data, and prevent any clear empirical conclusion on the issue of segmentation. Therefore, it is important to have a model that provides a decent degree of explanation of the worldwide cross section of asset returns at any given time. Here I discuss two aspects of such a model on which some progress has been made recently. The Pricing of Exchange Risk Prima facie, it would seem that randomly fluctuating exchange rates could be a cause of market segmentation, since investing abroad brings returns that are subject to exchange risk, whereas investing at home does not. Exchange risk is important mainly because deviations from Purchasing Power Parity (PPP) - for example, movements in real exchange rates - cause investors who consume in different places to adopt different attitudes toward the same securities. However, currency-based financial instruments provide a way of hedging exchange risk, at a price. The hedge is considered perfect when nominal and real exchange rates are correlated perfectly.[2] It is imperfect, but empirically nearly perfect, when they are not.[3] As a result, equity portfolios should be well diversified in equilibrium despite the presence of exchange risk. Since it is believed that in fact they are not, the model is rejected in principle. It is not inconceivable that the model nonetheless would provide some indication of the pricing of exchange risk. In view of our goal of determining whether "similar" assets are priced similarly, controlling for the price of exchange risk is certainly important. In a recently published paper,[4] Solnik and I show empirically that exchange risk receives a statistically significant price, and therefore constitutes a dimension of risk for each security (measured by the covariance of each security with exchange rates) that must be taken into account. The degree of fit of the model is not yet sufficient, however, to provide evidence of segmentation. The Time-Varying Nature of Risk and Returns The insufficient degree of fit of a pricing model conceivably could be caused by the fact that the first and second moments, corresponding to expected return and risk, move over time. …