An empirical test on the forecast ability of the bayesian and blume techniques for infrequently traded stocks

Abstract

This paper tests the forecast ability of different methods to estimate systematic risk. We address the issue in a small market where stocks are infrequently traded. We used the Blume technique and the Vasicek technique compared with two naïve techniques for different sample time periods (sample sizes) and different frequencies for data collection. We tested all the models using standard betas and betas adjusted for infrequently traded stock according to Scholes and Williams methodology. This study was carried on single stocks listed in the Portuguese stock exchange (BVL) instead of stock portfolios. We concluded that adjusted betas using either the Baysian model or the Blume technique produce a better result than unadjusted betas, but it is not clear whether the former produces consistently better results than the latter. We also found empirical support for the convergence phenomenon of betas of individual stocks towards one when they are either unadjusted or adjusted for infrequent trading