Testing the Duo-factor-model of Return and Volume

Abstract

Recent theoretical work by Lo and Wang (2000) shows that a multi-factor assetpricing model not only imposes factor restrictions on stock returns but on trading volume as well. We explicitly test their theoretical result using individual stock return and turnover data from NYSE and AMEX from 1962 to 1996. We introduce a recently developed consistent statistic by Bai and Ng (2001a) to determine the number of factors in a duo approximate multi factor model for return and turnover. While we find that the duo-factor model captures a great deal of common variation of trading volume, the data rejects a model restriction that excess return and turnover should have the same number of systematic factors. Using the duo-factor-model, we decompose excess return and turnover into systematic and idiosyncratic components. Our empirical work discovers a significant increase in the variation of idiosyncratic turnover through time, analogous to the discovery of a noticeable increase in firm level volatility by Campbell, Lettau, Malkiel and Xu (2001). We also find significant co-movement between volatility and turnover at the systematic levels. Our findings support the view that trading volume is not purely random but driven by trading activities associated with macroeconomic and firm news.