Estimating Market Model Betas: A Comparison of Random Coefficient Methods and Their Ability to Correctly Identify Random Variation

When estimating market model betas using random coefficient methods, the rather fine distinction between significance or insignificance, as argued in recent studies, overlooks two important factors. First, the maximum likelihood method has not been tested in comparison to the generalized least squares approximation. Second, and more importantly, the ability of these methodologies to correctly identify a known random coefficient process has not been examined in the context of the market model. Using both simulations and subsequent empirical tests, this study shows that, for reasonable levels of variation in beta, neither method can consistently identify a random coefficient process. These results suggest that the nominal level of significant random coefficients previously observed could be indicative of a much more predominant phenomenon.