Losing track of the asset markets : the case of housing and stock

This paper revisits the relationships among macroeconomic variables and asset returns. Based on recent developments in econometrics, we categorize competing models of asset returns into different "Equivalence Predictive Power Classes" (EPPC). During the pre-crisis period (1975-2005), some models that emphasize imperfect capital markets outperform an AR(1) for the forecast of housing returns. After 2006, a model that includes both an external finance premium (EFP) and the TED spread "learns and adjusts" faster than competing models. Models that encompass GDP experience a significant decay in predictive power. We also demonstrate that a simulation-based approach is complementary to the EPPC methodology.