We use time-series properties of changes in share turnover with GMM tests to examine the number of factors that drive changes in conditional expected time-varying volume. In
the three ten-year subperiods between 1966 and 1996, we detect no more than two factors. These findings support the hypothesis that changes in conditional time-varying expected volume can be represented parsimoniously as a function of a few factors.
However, whether we detect one factor (as Tkac (1999) suggests) or two factors (as Lo and Wang (1998) indicates) depends on which sample period we use, on whether we sort
portfolios by turnover or returns betas, and on whether we measure those betas relative to the equal- or the value-weighted index.