A revisit to low volatility anomaly

Abstract

The relation between risk and return has been debated for a long period. Apart from the debates on the relation being negative or positive, academics have reached the consensus that the security market line is much flatter than as implied by CAPM. Recently an increasing amount of empirical studies have provided evidences to support the existence of low volatility anomaly, arguing that bearing additional risk would give rise to lower expected returns. Some even move one step further, documenting that volatility earns significant risk premium therefore volatility should be incorporated into the asset pricing models to explain the cross section of stock returns. Different opinions strike back by concluding that after controlling for well-known anomalous factors, volatility factor becomes insignificant. In this thesis, total volatility and idiosyncratic volatility are examined separately. Evidences show that on the basis of risk-adjusted returns, low volatility anomaly strongly persists. Different weighting approaches on portfolio formation should also be taken into consideration since value-weighted approach tends to generate better results supporting low volatility anomaly. Volatility factors are also constructed based on the 2x3 portfolios after the double sorting. Results from Fama-MacBeth two-step regression presents that over the period from February 1900 to December 2015 (311 months), neither total volatility factor nor idiosyncratic volatility factor earns significant risk premium. Regression results from Carhart four-factor model prove that large, value and momentum lie behind low volatility anomaly. Robust profitability and conservative investment are also captured by Fama French five-factor model. But persistent variables independent of either Carhart or FF framework maintain explanatory power on low volatility anomaly.