An Empirical Analysis of Value-at-Risk Forecasting Models

Abstract

We perform a large-scale empirical evaluation of Value-at-Risk forecasting models, including several that incorporate intraday data. Our analysis spans numerous equities on two distinct exchanges, covering the global financial crisis period as well as the more stable period that immediately followed.Studies of this nature have been performed before, although generally not on such a large scale. The main appeal of the present paper is two-fold: first, we utilize recent advances in econometric theory to ensure our forecast evaluation procedure is much more powerful than procedures in other comparable studies, and second, we introduce a new class of Value-at-Risk forecasting models that employ an intraday-based proxy. Our analysis leads to several recommendations regarding the choice of Value-at-Risk forecast models. Of particular interest: we find strong evidence to support the use of intraday data.