Bayesian alphas predictability and short-run persistence of the UK market

Abstract

To examine the ability of the Bayesian alpha, as an alternative of the frequentist alpha, to predict the UK unit trust short-run future performance. By incorporating the returns on passive non-benchmark assets that correlate with the unit trust holding and an additional information -- the fund expense given to the prior distribution, the Bayesian alpha is expected to improve the predictability from the frequentist one. However, the results show that the difference in predictabilities of the Bayesian and frequentist measures is small. An additional test also shows that portfolios formed using either the Bayesian or the frequentist alpha as the selection criteria tend to similarly have abnormal losses in the subsequent years. While earlier studies suggest that using the Bayesian alpha can significantly improve the performance predictability in the US market, this study finds that, when applied to the UK market, the improvement is poor. The difference in the improvement between the US and the UK market might be due to the different environments of the two markets. A possible evidence that demonstrates the difference between the market environments is the fund expense that can provide a moderate short-run performance predictability in the US market but not in the UK market. The results from this study suggest that the Bayesian alpha is not suitable to be used to measure the unit trust performance in the UK.