A UTILITY-BASED COMPARISON OF SOME MODELS OF EXCHANGE-RATE VOLATILITY

When estimates of variances are used to make asset allocation decision s, underestimates of population variances lead to lower expected utili ty than equivalent overestimates: a utility-based criterion is asymmet ric, unlike standard criteria such as mean squared error. To illustrat e how to estimate a utility-based criterion, we use five bilateral wee kly dollar exchange rates, 1973-1989, and the corresponding pair of Eu rodeposit rates. Of homoskedastic, GARCH, autoregressive and non-param etric models for the conditional variance of each exchange rate, GARCH models tend to produce the highest utility, on average. A mean square d error criterion also favors GARCH, but not as sharply.