Forecasting Stock Market Volatility

Abstract

Volatility as a measure of investment risk is widely accepted by academic researchers and industry professionals and has become ubiquitous in investment analysis. Further, it is among the few financial variables that exhibit predictable time variation. Hence, there is an extensive amount of literature describing volatility models and assessing their forecasting power. This article provides a discussion of the prominent models and compares them in a unified notation framework. The empirical analysis shows that it is hard to outperform even simple trailing variance–type models. ARCH (autoregressive conditional heteroskedasticity), GARCH (generalized ARCH), implied volatility, asymmetric, and seasonal models hardly improve forecasts despite added complexity. In this study, only momentum- and intraday data–based models improved predictive accuracy significantly.