The cross section of expected stock returns revisited

Abstract

We review and extend two important empirical financial studies: Fama and MacBeth [1973] and Fama and French [1992]. Fama and MacBeth [1973] sort stocks on the New York Stock Exchange into 20 portfolios based on their market [beta]. They test for, and conclude that, [beta] does in fact explain the cross-sectional variation in average stock returns for the 1926-1968 period. After we replicate the results in their study we extend their work to the most current data. The coefficients and t-statistics for five-year sub-periods exhibit roughly the same properties during the last half of the century as they did during the period originally studied. Fama and MacBeth report statistically significant results for their overall period (1935-1968) as well. When we run the same test on the all the data currently available (1935-1998) we find that the t-statistics are lower, instead of higher, than they were for the 1935-1968 period. We run several variations on the Fama and MacBeth [1973] paper. For example, we vary the exchanges (NYSE, AMEX, and/or NASDAQ) and indexes (value-weighted or equally-weighted) employed. We also study the effect of using robust (least absolute deviation) regressions instead of ordinary least squares. In all cases, the results are similar to those described above. Fama and French [1993] show that, when size is controlled for, market [beta] does not explain the cross-sectional variation in returns for the 1963-1990 period. They find that two other variables, size (market equity) and book-to-market equity, combine to capture the cross-sectional variation in average stock returns during the same period. After replicating their results, we update the study to the most current data. We find that the t-statistics for size and book-to-market equity are more significant during the 1963-1998 period than they were for the 1963-1990 period. We also confirm that [beta] is statistically insignificant during the 1963-1998 period.