City Research Online

Abstract

This thesis is composed of three essays related to empirical asset pricing. In the first essay of the thesis, we investigate recent rational explanations of the value premium using a regime-switching approach. Using data from the US stock market, we investigate the risk of value and growth in different market states and using alternative risk measures such as downside beta and higher moments. Our results provide little or no evidence that value is riskier than growth, and that evidence is specific to pre-1963 period (including the Great Depression). Within the post-1963 sample, there are periods when the value premium can be explained by the CAPM, whilst during other periods the premium is explained by the fact that the returns on value firms increase more than the returns on growth stocks in periods of strong market performance, whilst in downturns growth stocks suffer more than value, and these features are captured by different upside/downside betas or higher moments. These results are not consistent with a risk-based explanation of the value premium. The second essay of the thesis contributes to the debate about the momentum premium. We investigate the robustness of the momentum premium in the US over the period from 1927 to 2006 using a model that allows multiple structural breaks. We find that the risk-adjusted momentum premium is significantly positive only during certain periods, notably from the 1940s to the mid-1960s and from the mid-1970s to the late 1990s, and we find evidence that momentum has disappeared since the late 1990s. Our results suggest that the momentum premium has been slowly eroded away since the early 1990s, in a process which was delayed by the occurrence of the high-technology stock bubble of the 1990s. In particular, we estimate that the bubble accounts for at least 60% of momentum profits during the period from 1995 to 1999. In the final essay of this thesis, we study the question of which asset pricing factors should be included in linear factor asset pricing model. We develop a simple multivariate extension of a Bayesian variable selection procedure from the statistics literature to estimate posterior probabilities of asset pricing factors using many assets at once. Using a dataset of thousands of individual stocks in the US market, we calculate posterior probabilities of 12 factors which have been suggested in the literature. Our results indicate strong and robust evidence that a linear factor model should include the excess market return, the size and the liquidity factors, and only weak evidence that the idiosyncratic volatility and downside risk factors matter. We also apply our methodology to portfolios of stocks commonly used in the literature, and find that the famous Fama and French (1993, 1996) HML factor has high posterior probability only if portfolios formed on book-to-market ratio are used.

Publication Type:	Thesis (Doctoral)
Subjects:	H Social Sciences > HG Finance
Departments:	Bayes Business School > Finance Doctoral Theses Bayes Business School > Bayes Business School Doctoral Theses

Export

Downloads