Does Introducing Futures Markets Affect Currency Variance Forecasts? Evidence from Asian Markets

Abstract

We investigate how currency volatility forecasts change after nations introduce markets for currency futures. We study three Asian markets--South Korea, India, and China. For each exchange rate, we estimate both machine learning and GARCH models, and compare their ability to forecast realized volatility during two periods: one period directly before and another directly after the introduction of futures. We find that a naive historical volatility forecast typically outperforms both in-sample and real-time GARCH forecasts, and that machine learning models outperform their GARCH counterparts. For both GARCH and machine learning models, the mean absolute percentage error and root mean squared percentage error always increase after the introduction of futures. This latter finding indicates that realized volatility is more difficult to forecast in the aftermath of futures introduction.

Keywords: Exchange Rates, forecasting, Futures, realized volatility, machine learning