Asset pricing models with and without consumption data : An empirical evaluation

This paper evaluates the ability of the empirical model of asset pricing of Campbell(1993a,b) to explain the time-series and cross-sectional variation of expected returns ofportfolios of stocks. In Campbell's model, an alternative risk-return relationship is derivedby substituting consumption out of the linearized first-order condition of the representativeagent. We compare this methodology to models that use actual consumption data, such asthe model of Epstein and Zin, 1989, 1991, and the standard consumption-based CAPM.Although we find that Campbell's model fits the data slightly better than models whichexplicitly price consumption risk, and provides reasonable estimates of the representativeagent's preference parameters, the parameter restrictions of the Campbell model, as well asits over-identifying orthogonality conditions, are generally rejected. The parameter restrictionsof the Campbell model, and the over-identifying conditions, are marginally not rejectedwhen the empirical model is augmented to account for the "size effect"