Additional Evidence On The Abatement Of Errors In Predictin

A study is performed to provide additional evidence regarding the abatement of prediction error (PE) through diversification. The analysis suggests that diversification is an effective means of abating PE in the case of rank-ordered portfolios when significant security beta nonstationarity fails to exist in the estimation period and randomly selected portfolios. It also is shown that diversification does not serve to reduce PE to insignificant levels for rank-ordered portfolios when there is significant security-beta nonstationarity in the estimation period. An attempt also is made to provide evidence on the source or sources of the regression tendency and on the effectiveness of the order bias model in eliminating order bias. The analysis shows evidence of improved reliability, suggesting that the regression tendency is a function of both order bias and security-beta nonstationarity and that the order-bias model eliminates this tendency only for the control group.