ROBUST, NONPARAMETRIC MEASURES OF EXCHANGE-RATE VARIABILITY

Changes in exchange rates are well-known to possess non-Gaussian distributions, yet in testing for differences in exchange rate volatility few researchers have examined measures that do not assume Gaussian data. This paper develops a robust, non-parametric measure for comparing exchange rate variability across time. The measures are applied to monthly exchange rate data to test for differences in exchange rate variability across different monetary policy regimes. The comparisons show inconsistencies between conclusions based on the classical F-test and the robust, non-parametric equivalent. Using the robust test for changes in exchange rate variability, we do not conclude that variability has increased over time. Increases in variability do occur, but usually only in specific percentiles of the data.