End-of-Day Stock Trading Volume Prediction with aTwo-Component Hierarchical Model

We present a hierarchical model approach for predicting the end-of-day stock trading volume (total daily volume). It effectively combines two sources of information: the trading volume already accumulated from the beginning of the trading day to the time of the prediction, and the historical daily trading volume dynamics. The hierarchical model assumes an ARMA–GARCH model for the daily volume series and a Gaussian multinomial distribution for the volume distribution within the trading day. This model is an extension of the stable-seasonal model in Chen and Fomby [1999] used for tourism prediction. We also present out-of-sample prediction performance, with a comparison to several simpler approaches, using real volume series, where the improvement is quite significant (30% to 40%). TOPICS:Statistical methods, information providers/credit ratings, exchanges/markets/clearinghouses