Preferences and observed risk premia: an empirical analysis

Abstract

The fundamental prediction of the Consumption-based Capital Asset Pricing Model (CCAPM) relates asset returns to their covariance with the intertemporal marginal rate of substitution (IMRS). With utility subjected to constant relative risk aversion, the IMRS is characterized by only one economic variable namely, consumption growth. One explanation for the disappointing empirical performance of the CCAPM model may be that the constant relative risk aversion specification is too restrictive. In this article we consider alternative specifications and compare their empirical performance with the reference model using Canadian data.

Notes

¹ Data on consumption, the risk free rate and the TSE300 return is from the CANSIM database of Statistics Canada. The risky asset returns used to generate the five value-weighted industrial returns are from the TSE Western Database. There are 1384, 739, 215, 199 and 1011 securities listed during some period of time for the sample period considered in sectors 1 to 5, respectively.

² Other lists were used, but the results are very similar to the ones reported in the various tables. The discount factor is not estimated and is fixed at 0.98 per annum. This makes sure that the estimated risk aversion parameters are comparable to those calibrated in Hansen and Jagannathan (1991) and Cochrane and Hansen (1992) under the assumption of a unitary discount factor.