Market and financial performance as related to idiosyncratic risk and the effect of outlier screening in market studies

We tested whether the information derivable from the CAPM as viewed through the OLS one-stage two-parameter linear regression, hereafter OLS-filter, could be enriched by analyzing the residuals of the OLS-filter. This was Roll's (1988) conjecture; such OLS-residual information is termed Idiosyncratic Risk [IR]. Further, recognizing that the generation of IR information requires the use of the OLS-filter, we examine the effect of attending to the assumptions of the OLS-filter, in particular screening for outliers, on the market analysis of traded firms. The outlier screens used were: Trimming, Box-Plot CC-Screens, and Mahalanobis relational screen. We used three groups of firms: the Old Economy and two analytic benchmarks: IPOs opening during the accrual period and New Economy firms also traded during the accrual period. The accrual period was the Internet Bubble Build-up period from 1 January 1994 to 31 December 1999. The results suggest that all market studies should report, in detail, their approach to satisfying the assumptions of the OLS-filter; this is critical as there are dramatic differences between the screened and non-screened results. [PUBLICATION ABSTRACT]