CAPM Reconsidered: A Robust Finite Sample Evaluation

In this paper, the conventional test of the Sharpe-Lintner version of the Capital Asset Pricing Model (CAPM) are reconsidered. The CAPM is formulated as a Seemingly Unrelated Regression (SUR) system with an adding-up restriction. A statistical framework for conducting the conventional tests is provided. A Wald statistic that takes into account the adding-up restriction and that is robust to heteroskedasticity and/or autocorrelation of unknown form is used to test the CAPM. The robust Wald test with asymptotic critical values rejects the CAPM for most subsamples. Monte Carlo simulations are then performed to obtain the finite sample critical values. Our empirical results show that the CAPM is consistent with the data when the tests are based on finite sample critical values.