Evaluation of Volatility Forecasting Models in Vietnam Stock Markets

Abstract

This study aims to find the most appropriate model(s) to estimate and forecast volatility in Vietnam stock markets. Considered volatility models in this study include RiskMetrics, GARCH, EGARCH, IGARCH, FIGARCH and APARCH. The forecast performance evaluations are conducted with two Vietnam stock indices – VNI-index and HNX-index. Selected data periods is from 01 March 2002 to 30 June 2011 for VNI-index and the period for HNX-index spans from 01 June 2006 through 30 June 2011. Symmetric loss functions and asymmetric loss functions are used as basic analysis criteria. Robust conclusions are achieved with the superior predictive ability (SPA) test, the model confidence set (MCS) procedure and Value-at-Risk (VaR) forecast evaluation.

The general empirical results generated from symmetric loss functions, the SPA test and the MCS procedure demonstrate that for VNI-index, RiskMetrics and EGARCH have equally best forecast performance while for HNX-index, only EGARCH has the best. However, there are contrast findings resulted from different assessment criteria specifically with asymmetric loss functions and VaR forecast. Actually, the ranking of models is sensitive to the selected criterion. Therefore, selecting reasonable evaluation criteria is very critical and it must be established on the ultimate aims of the forecasting procedure.