Misspecification testing and robust estimation of the market model: estimating betas for the FT-SE industry baskets

Abstract

Using daily data on the London Stock Exchange's FT-SE industry baskets, this paper subjects the market model to a set of rigorous specification tests designed to assess its statistical adequacy in the face of possible model non-linearity and autocorrelation, heteroscedasticity and, in particular, non-normality of residuals. It then utilizes robust methods of estimation to provide valid estimates of the model's parameters in the face of such misspecification. It is found that robust estimation usually provides betas that are lower than those estimated by conventional ordinary least squares.