| The CAPM and the Fama and French (FF) Three-Factor Pricing Model are evaluated using out-of-sample data, free from many of the biases present in typical data sources. Data is collected from the Value Line Investment Survey (VL) allowing survivorship and look-ahead biases to be eliminated. Further, simple portfolio selection rules and the use of data from a source commonly consulted by investors minimizes data snooping concerns.;The Fama and MacBeth (1973) and Gibbons, Ross, and Shanken (GRS, 1989) methodologies are employed to test the models' ability to predict returns on beta, size, and book-to-market (B/M) sorts. Additionally, the success of investment strategies utilizing VL's timeliness rankings and momentum portfolios, formed from lists of recent winners and losers appearing in VL, are explored with the FF model. Finally, the discrete ICAPM framework of Fama (1997) serves as a foundation for an investigation into the relationship between economic fundamentals underlying the FF model and the VL ranking system.;The results are far different than those commonly reported. Cross-sectional regressions show beta to be a strong predictor of stock performance in both "up" and "down" markets, while size and B/M are not related to return variation. GRS F-tests reject the mean-variance efficiency of the VL index only when securities are divided by beta. In short, size and B/M are not anomalies. While the CAPM is not perfect, it fails because of the high (low) returns on low (high) beta stocks.;The FF model is found to offer no significant improvement over the CAPM. It increases accuracy slightly when describing returns on related portfolios but fails on all others. It, like the CAPM, cannot account for returns on beta-sorted portfolios. Further, both the VL ranking and momentum strategies are successful based on the predictions of the model. |
