Event-study analysis by using dynamic conditional score models

Abstract

This article considers the most important information technology (IT) products in order to perform an event-study analysis of the out-of-sample predictability of IT stock returns. We define two subperiods, the estimation and forecast windows, for each IT product that are separated by the product release date. We investigate whether post-release-date returns can be predicted by using data on pre-release-date returns. We use static one-step-ahead density forecasting. We compare the forecast performance of autoregressive moving average (ARMA) plus generalized autoregressive conditional heteroscedasticity (GARCH) and quasi-ARMA (QARMA) plus Beta-$$t$$t -EGARCH (exponential-GARCH). QARMA plus Beta-$$t$$t -EGARCH belongs to the family of dynamic conditional score (DCS) models. We find that the in-sample statistical performance of DCS is superior to that of ARMA plus GARCH for most of the IT stocks. We also find that the out-of-sample density predictive performance of ARMA plus GARCH is never significantly superior to that of DCS. However, the predictive performance of DCS significantly dominates that of ARMA plus GARCH for several IT products. We undertake a Monte Carlo value-at-risk (VaR) application of our results to Windows 95 of Microsoft.