Forecasting stock market volatility with regime-switching GARCH-MIDAS: The role of geopolitical risks

Segnon, Mawuli; Gupta, Rangan; Wilfling, Bernd

Abstract

We investigate the role of geopolitical risks in forecasting stock-market volatility for monthly horizons-ahead within a robust autoregressive Markov-switching GARCH mixed-data-sampling (AR-MSGARCH-MIDAS) framework. Our approach accounts for structural breaks through regime-switching and allows us to disentangle short- and long-run volatility components. We conduct an empirical out-of-sample forecasting analysis using (i) daily Dow-Jones-Industrial-Average returns, and (ii) monthly sampled geopolitical risks and macroeconomic variables over a time span of 122 years. We find that the impact of geopolitical risks as explanatory variables for stock-market volatility forecasts for monthly horizons hinges crucially on the specific prediction model chosen by the forecaster. After capturing the non-stationarities in the data via an MSGARCH framework, we do not find significant forecast accuracy improvements through the inclusion of geopolitical risk indices.