An empirical test for the effectiveness of alternative CAPMs in Korean stock returns

Abstract

The failure of standard CAPM to explain the cross-sectional variation in asset returns has led to the development of many alternative empirical asset pricing models. One of recent approach is higher moment CAPM in which researchers assume investors have preferences for higher moments in return distribution beyond mean and variance. Arditti (1971), Fielitz (1976), and Gibbons et al.(1989) show that skewness and kurtosis cannot be diversified by increasing portfolio size. While most former empirical studies focus on testing whether a particular asset pricing model is literally true or not, the performance comparison of different asset pricing models is recently in the spotlight of CAPM field.

Since stock returns show different skewness and kurtosis from normal distribution, we cannot assume normal distribution. In this case, the GMM can be used to test model specifications because GMM does not assume any special distribution. I analysed Korean stock market data to compare the performance of several CAPMs. The two four-moment CAPM models showed better performances than any other CAPM models according to the research results. The J-tests find that only the two models are correctly specified to capture the movement of the Korean industry portfolios. In addition, the estimated coefficients of the two models are statistically significant in the GMM system estimation.