It is well-known that cross-sectional tests of the CAPM are problematic. The market indexes used in empirical tests are likely to be inefficient ex ante, which could lead to spurious results even in the absence of sampling errors. This problem has led many to express serious doubt on the testability of the CAPM. In this paper I show that the CAPM is indeed testable. This paper builds on the seminal paper by Kandel and Stambaugh (1995) and proposes a two-step procedure for testing the CAPM. The first step uses a simple combination of the coefficients of determination from both Ordinary Least Squares and Generalised Least Squares estimations. This step tests whether the index used in the empirical test is efficient and whether there are no omitted factors. The second step tests the hypothesis that the efficient index is the market portfolio. The two-step approach enables testing the CAPM regardless of whether the true expected return generating process is a CAPM or a non-CAPM. © 2014 Elsevier B.V.