Robust informational tests on the CAPM

Abstract

The paper demonstrates the existence of an independent informational content in the Capital Asset Pricing Model (CAPM) that financial analysts are not fully utilizing in their forecast-generating mechanism. This existence is discovered by regressing actual values of five-year firm earnings growth against financial analysts ex-ante forecasts and simulated ex-ante forecasts generated by the CAPM. Regressions are run over a cross-section of firms for each of four adjacent five-year horizons: January 1982–1987; 1983–1988; 1984–1989; 1985–1990. In three out of four test periods, the coefficient of the CAPM forecasts is significantly positive. This is essentially the same experiment with the same results as obtained previously, with one exception: a diagnostic analysis and corrective procedures are performed and results are generated that are heteroscedasticity-robust.