Exchange rate predictive densities and currency risks: A quantile regression approach

We investigate the ability of the Fama equation to compute proper conditional densities and currency risks. Based on quantile regressions, we fit a Skewed t-distribution to estimate the conditional densities on the monetary policy of eight currency pairs. We demonstrate that the conditional densities are highly sensitive to the monetary policy stances. Then, we use the estimated conditional densities to measure the currency risks. Our results highlight that the depreciation/appreciation risks are extremely heterogeneous and that the currencies are more exposed to depreciation risks, especially during turmoils. Our findings can be used as a supplementary tool to assess whether a currency behaves as a safe-haven currency. We also investigate the relative and absolute performance of our model in forecasting densities. We find that the predictive densities are perfectly well-calibrated. Moreover, our results also demonstrate that our methodology can outperform the random walk in forecasting densities.