Based on the errors-in-variables-free approach proposed by Brennan et al. [J. Financial Econ., 1998, 49, 345--373], we investigate the competing explanatory capabilities of alternative multi-factor models when examining various asset-pricing anomalies using Japanese data for the period 1978--2006. We find that turnover and book-to-market (BM) ratio are the two major characteristics that significantly explain the average stock returns. A further sub-period analysis reveals that the turnover effect is significant only before 1990, but cannot be explained by any multifactor models. In contrast, the BM premium is significant only after 1990, and can be explained by the Fama--French three-factor model. Thus, the results suggest that asset-pricing anomalies documented in the literature are not universal, and may be different across different markets.