Article by: Marcel P. Visser
Published by: Korteweg-de Vries Instute for Mathematics, University of Amsterdam
Date: 14 Oct 2008
“This paper decomposes volatility proxies according to upward and downward price
movements in high-frequency financial data, and uses this decomposition for forecasting
volatility. The paper introduces a simple Garch-type discrete time model that incor-
porates such high-frequency based statistics into a forecast equation for daily volatil-
ity. Analysis of S&P 500 index tick data over the years 1988–2006 shows that taking
into account the downward movements improves forecast accuracy significantly. The
R2 statistic for evaluating daily volatility forecasts attains a value of 0.80, both for
in-sample and out-of-sample prediction.”
Full article (PDF): Link