摘要:It has been demonstrated that consideration must be given to the effect that outliers have on the parameters of the OLS regression used to generate the usual Risk and Return profile of firms in market studies, in particular: Total Firm Risk, the CAPM β, the usual surrogate for Systematic Risk, Idiosyncratic Risk, Jensen’s α, the SPI and the TPI. We address, for the first time, an important related question: As there are two ways to deal with outliers in the OLS-regression space: Trimming and Winsorizing each of which has a possible different effect on Power; are there practical Power differentials between Trimming an Winsorizing. Results: We find that there are no differences in bi-variate tests against the usual Null between Trimming and Winsorizing over the market performance profile; as this is the case one prefers Trimming as it avoids the possibility of agenda serving replacement.
关键词:OLS-Regression outliers; Moral Hazard Selection