Date of Graduation


Document Type


Degree Type



Chambers College of Business and Economics



Committee Chair

Ronald J. Balvers.


In this dissertation, I investigate the applicability of some asset-pricing models by exploiting the predictive power of structural variables and accommodate the characteristic of predictability of stock returns in a theoretical conditional asset-pricing framework.;In the first chapter of the dissertation, I briefly review the relevant literature on the predictability of stock returns, and discuss the motivations for my research, and finally present my contributions to the existing literature while providing some discussions for future research. Chapter 2 examines the ability of the Fama-French three-factor model and the CAPM in estimating portfolios' returns. Instead of using historical average of risk premiums as a proxy for the estimated risk premiums, I propose that the estimation of risk premiums is conditioned on structural variables. The out-of-sample results from a simple trading strategy show that at least in the short-run, the structural variables generate better estimates of risk premiums than the historical averages of risk premiums, and the Fama-French three-factor model outperforms the CAPM in estimating portfolios' returns. Chapter 3 of the dissertation investigates the common and country-specific components in national stock price indices. The proposed general dynamic state-space model not only incorporates the feature of predictability of stock returns by modeling the price index as the combination of a permanent component and a transitory component, but also allows distinguishing the common shocks shared by each country from the country-specific shocks. It is found that for the G7 countries, there exist country-specific permanent and transitory components, which implies that the price indices of these countries are not only predictable, but also they are not cointegrated around one stochastic trend. Therefore, there exist potential long-run international diversification benefits. In light of the predictability of stock returns, Chapter 4 proposes a conditional two-beta intertemporal asset pricing model to investigate how the permanent shock and the transitory shock to market return are priced. Two decomposition methods are employed: the Hodrick-Prescott approach and the unobserved component method with Markov-Switching disturbances. The traditional two-pass evaluation indicates that this two-beta model outperforms the CAPM, but it is not as good as the Fama-French three-factor model in explaining stock returns.