Residual risk revisited

If a portfolio is not mean-variance efficient, the residual variances from its market model regressions should help explain expected returns. Previous research produced no statistically reliable residual risk premium estimates. This paper examines alternative econometric methods for measuring risk premiums, highlighting corrections for measurement error in market model parameter estimates, the use of individual securities as opposed to grouped portfolios, the comparison of the Fama-MacBeth and Black-Jensen-Scholes formulations, and methods for aggregating subperiod risk premium estimates. The results suggest measurement error corrections produce different and apparently significant residual risk effects, the Black-Jensen-Scholes formulation yields more reliable estimates, and the estimates are sensitiveto the aggregation method.