Robust beta estimation: Some empirical evidence

Abstract

The effect of allowing for skewness and excess kurtosis in estimating market model betas is examined using the Generalized Student-t (GET) Distribution. The GET Generalized the widely used Student-t distribution by allowing for skewness as well as leptokurtosis. Using data on monthly returns of twenty-two stocks listed on the Singapore Stock Exchange, we find that the GET provides a significantly better fit to the data than the normal distribution or the symmetric Student-t distribution. Based on a small out-of-sample experiment, the GET was also found to outperform OLS and Student-t betas in forecasting ability. © 1997 JAI Press Inc. 1058-3300.