Forecasting VaR and ES in emerging markets: The role of time‐varying higher moments

Abstract

AbstractThis paper aims at the role of accounting for time‐varying higher moments in conditional volatility models in emerging markets. In particular, we perform a comprehensive analysis of out‐of‐sample value at risk (VaR) and expected shortfall (ES) forecasts of eight generalized autoregressive conditional heteroskedasticity (GARCH) models with alternative distributions for 10 leading emerging markets. To evaluate the forecast accuracy, we employ a battery of absolute performance backtests and two alternative loss functions in the relative performance exercises. We find that the asymmetric GARCH models with time‐varying skewness and kurtosis significantly outperform traditional GARCH‐based forecasts across quantiles across quantile levels. In particular, we explore that the superior performance of the GARCH models with improved distributions is mainly driven by their performance during crisis periods, where the traditional GARCH specifications often underestimate the tail risk in the markets.