Time-Varying Asset Pricing Models in the Context of Segmented Markets

Abstract

This paper explores and tests two multi-factor asset pricing models in an international context. One model focuses only on local risk factors and therefore assumes that the market is completely segmented. The other model focuses only on global risk factors and assumes that the market is fully integrated. The models incorporate time-variation in both the risk exposures and risk premia. The models are applied in cross-section to a range of developed and emerging markets so that varying levels of integration are examined. Expected returns are formed using time-varying estimates of risk premia that allow for out-of-sample testing. Using a range of performance metrics, the findings show that returns in developed markets are better approximated by a global pricing model, whereas returns in emerging markets are better represented by a local pricing model. These results are found to be generally robust to a range of research design issues.