Do VIX and Trading Volume Subsume Incremental Information above GARCH-Type Models for the Volatility Forecast?

Abstract

The VIX index and trading volume (VO) can subsume information of the future volatility in financial markets, and therefore, have been commonly used as volatility forecasting instruments. Previous studies have identified superior VIX- and VO-based forecasts compared to various GARCH-type models, while contradictory evidence for the outperformance of GARCH-based forecasts rises controversies about the optimal forecast. To evaluate whether VIX and VO may subsume incremental information over GARCH-type models, this paper evaluates the volatility forecast efficiency of VIX, VO and GARCH.

Latest daily VIX, VO and SPX during 01/01/2006-17/07/2017 are sorted from Bloomberg. To access the forecast efficiency, Mincer-Zarnowitz (MZ, 1969) regression, the forecast encompassing test and root-mean-square error (RMSE) are to be applied. The present empirical result indicates incremental information in VIX, while VO seems to subsume little or no additional information over GARCH, which contradicts previous findings of Kambouroudis and McMillan (2016).

Recent concerns about the influence of financial turmoil on the volatility forecast have generated a considerable body of research; however, most previous investigations of the volatility forecast efficiency have not addressed this issue. Hence, this paper is motivated to identify the effect of 2008 financial crisis on the volatility estimates. Current findings confirm a negative crisis dilution effect occurred in 2012, especially for VO-based predictions. The identified dilution effect demonstrated a negative influence of turmoil on volatility forecasts due to the disruption of the economic condition during the post-crisis period. Further work is recommended on validating this issue in various financial markets, such as foreign exchange markets.