| Abstract: | This study is an analysis of the forecasting ability of adjusted and unadjusted betas. Based on the Canadian data for 252 stocks, random errors in betas are the most important reason for the poor predictive ability of individual security betas. Most of this random error is eliminated if securities are grouped into portfolios. However, further improvements in forecasting ability are gained by adjusting the security betas for bias and inefficiency. Five methods of adjusting the naive beta estimates have been tried, including two methods not tested before. These two, Vasicek's two-stage method and order-bias adjustment method, gave results generally superior to others. |
