Bayesian tests of global factor models

I use the Bayesian approach of Barillas and Shanken (2018) to examine the mean-variance efficiency of nine global factor models in global stock returns and to conduct multiple model comparison tests. The mean-variance efficiency of each factor model is strongly rejected. In the multiple model comparison tests, the three-factor model of Asness, Moskowitz and Pedersen (2013) has the best performance at higher prior maximum Sharpe (1966) ratio multiples and significantly outperforms all the other factor models. However, in out-of-sample tests, the AMP model significantly underperforms the best performing models that can be formed among the set of all factors used by the global factor models.