In this paper, I first develop a new approach to estimating the return on the aggregate wealth portfolio that accounts for human capital and financial assets other than stocks. Using the estimated return on the wealth portfolio and the quarterly U.S. aggregate data on consumption and asset returns from 1959 to 2001, I then test the asset pricing and consumption implications of the Epstein and Zin (1991) and Weil (1990) model by employing the weak-identification robust tests of Stock and Wright's (2000) in the context of continuous updating generalized method of moments. In contrast with previous studies that ignored human capital and weak identification in evaluating this model, I find that its asset pricing implications cannot be rejected at conventional significance levels for reasonable parameter values.

For example, the 95% confidence sets for unknown parameters include values of the relative risk aversion around 2 or lower, values of the elasticity of intertemporal substitution for consumption closely around 1, and the time discount factor around 0.987. Some of these parameter value combinations are able to simultaneously match the average equity premium and the average riskfree rate in the data. Furthermore, they imply that the dominant determinant of the equity premium is, surprisingly, the volatility of stock returns, the risk factor in the traditional capital asset pricing model.