On asset pricing models and mutual fund performance : an empirical analysis of US mutual funds
Abstract
Asset pricing models introduce the challenge of testing a joint hypothesis. This thesis tests the hypothesis of model misspecification and true alpha separately, using the testing methodology of Gibbons, Ross and Shanken (1989) on US mutual fund returns. As there is extensive research on mutual fund performance, our main motivation is to analyze which asset pricing model is the most appropriate for performance evaluation. We test seven asset pricing models on 2971 US mutual funds existing in the period of January 1999 to August 2018.
First, we use the test methodology to measure mutual fund performance under the assumption of perfectly specified factor models. We find ambiguous evidence on fund mangers’ ability to create abnormal return gross of fees. We conclude that the most comprehensive models argue for a small, but significant alpha. Net of fees, all models produce negative abnormal returns. This leads to a strong rejection of the null-hypothesis of alpha being equal to zero. Second, we interpret the same test statistics, but now under the assumption of zero abnormal return. This allows for using the GRS-test for testing asset pricing models. We find that the Capital Asset Pricing Model performs surprisingly well given its simplicity. Furthermore, we find that the models including the investment- and profitability factors are better at explaining mutual fund return, with our results indicating that the Fama French Five-Factor Model is the most correctly specified model. Lastly, we find the results to be sensitive to portfolio formation. We conclude that a sorting method based on two criteria is superior to one, and that portfolios sorted on characteristics lead to the strongest inference.