Abstract: | The nature of the stochastic process generating the time path of foreign exchange rates plays an important role in dynamic theories of international financial economics. An important consideration in this stochastic process is the relationship between currency return variances and exchange rate levels. Using five years of daily data separated into quarterly intervals, this study demonstrates that currency return variances depend on exchange rate levels and the dependency is unstable intertemporally. Thus, these empirical results contradict the assumption of stable log-normal distribution and the more general assumption of constant elasticities of variances. For elasticity coefficients ordinary least squares estimates are compared to maximum likelihood estimates; the maximum likelihood estimator clearly is superior. Implications of these results for models of foreign exchange rates are discussed. |