This paper reinvestigates the performance of risk-based multifactor models. In particular, we generalize the Bayesian methodology of Shanken (1987b) and Kandel, McCulloch and Stambaugh (1995) from mean-variance efficiency to the ICAPM notion of multifactor efficiency. This methodology uses informative priors and provides a flexible framework to deal with the severe small sample problems that arise when estimating performance measures. We also introduce and theoretically justify a new inefficiency metric that measures the maximum correlation between the market portfolio and any multifactor efficient portfolio, which is used in conjunction with three other existing inefficiency measures. Finally, we present new empirical evidence that neither the two additional Fama-French (1992) factors nor the momentum factor move the market portfolio robustly closer to being multifactor efficient or robustly decrease pricing errors relative to the CAPM.
