THE ADJUSTMENT OF BETA FORECASTS

Abstract

THE BETA COEFFICIENT of the market model has gained wide acceptance as a relevant measure of risk in portfolio and security analysis. An essential prerequisite for using beta to assess future portfolio risk and return is a reasonable degree of predictability over future time periods. If the portfolio manager cannot predict future beta coefficients, the applicability of this phase of modern capital-market theory is somewhat restricted. Attempts to predict betas using extrapolative models have met with only limited success, especially for individual securities. Blume [1] and Levy [2] found that single security beta coefficients of one period were not good predictors of the corresponding betas in the subsequent period. However, as portfolio size was increased, the stationarity of extrapolated betas improved significantly. A major problem for both single security and portfolio betas was the tendency for relatively high and low beta coefficients to overpredict and underpredict, respectively, the corresponding betas for the subsequent time period. Thus, forecasting accuracy grew progressively worse as beta levels departed significantly from the