Are Short-Horizon Equity Returns Predictable? Evidence from Large and Frequently Traded Italian Stocks

We analyze the serial correlation of the weekly returns of the MIB and MIB30 indexes over the period 1985-1997 from two different perspectives. First, we explore whether the observed serial dependence of returns is consistent with the random walk model. This has been done by testing two specifications of the model. We find that the first specification, given by the hypothesis of i.i.d. returns, is clearly rejected by the data, while the statistical evidence against the second specification, given by the more general hypothesis of uncorrelated or linearly independent returns, is much weaker, especially for the MIB30 returns. In the second part of the paper we study the possibility that a serially correlated process implied by a specific model of time-variation in expected returns could generate a degree of linear dependence similar to that observed in actual returns. We find that a combination of fast mean reversion in expected returns and small pricing errors in realized returns produces an autocorrelation pattern similar to that actually observed.