THE BENCHMARK BETA, CAPM, AND PRICING ANOMALIES

Abstract

THE area of finance owes, in a large measure, its emergence as a scientific discipline to the advent of the capital asset pricing model (CAPM) which describes the determination of asset prices in an equilibrium framework. It is noted that in the classical CAPM of Sharpe (1964), Lintner (1965), and Mossin (1966), the structure of capital market equilibrium is presented in a highly parsimonious way, relating the equilibrium asset returns to a single risk factor, namely, the market beta. This parsimonious approach enables the equilibrium asset pricing relationship to be expounded in a simple and intuitively appealing way, thereby helping the CAPM to become one of the dominant paradigms in modern finance. Recently, however, the CAPM appears to have lost some of its earlier appeal for at least two reasons. First, Roll (1977) casts a serious doubt on the testability of the CAPM. As is well known, Roll forcefully argued that the only testable implication of the CAPM is the mean-variance efficiency of the market portfolio, and that the model is really neither testable nor applicable unless the true market portfolio is known. Because of the Roll critique, the CAPM is widely regarded as a theory without clear observational consequences. Second, researchers such as Friend and Blume (1970), Banz (1981), and Reinganum (1981) documented various asset pricing 'anomalies' unexplained by the CAPM. Friend and Blume, for instance, found that Jensen's alpha, which is supposedly a 'risk-adjusted' performance measure, is strongly inversely related to the level of beta risk. Banz and Reinganum, on the other hand, found that stock returns are negatively related to firm size even after controlling for the effect of beta risk. The Roll critique prompted several researchers to devise alternative ways of testing the CAPM taking into account the unobservability of the market portfolio. Kandel and Stambaugh (1987) and Shanken (1987), for instance, developed and tested the joint hypothesis that the true market portfolio is mean-variance efficient and that the proxy and true market portfolios are highly correlated. Their test results generally rejected the hypothesis for the assumed correlation of.7 or higher. Gibbons and Ferson (1985), on the other hand, tested a restriction derived from the temporal behavior of conditional expected returns and failed to reject a single-factor pricing hypothesis. While their findings are inconclusive so far, these researchers and others made serious attempts to mitigate the Roll critique (see footnote 1, base of next page). In contrast, however, no serious efforts were made to explain various asset pricing anomalies in terms of the unobservability of the market portfolio. In the present