This paper re-examines the tests of the Sharpe-Lintner Capital Asset Pricing Model (CAPM). The null that the CAPM intercepts are zero is tested for ten size-based stock portfolios and for twenty five book-to-market sorted portfolios using five-year, ten-year and longer sub-periods during 1965-2004. The paper shows that the evidence for rejecting the CAPM on statistical grounds is weaker than the consensus view suggests, and highlights the pitfalls of testing multiple hypotheses with the conventional heteroskedasticity and autocorrelation robust (HAR) test with asymptotic P-values. The conventional test rejects the null for almost all sub-periods, which is consistent with the evidence in the literature. By contrast, the null is not rejected for most of the sub-periods by the new HAR tests developed by Kiefer et al. (2000), Kiefer and Vogelsang (2005), and Sun et al. (2008).
