Expectations of Real Interest Rates and Aggregate Consumption: Empirical Tests

Abstract

Recently, the finance literature has included empirical analysis of consump? tion in asset pricing models based on the cross-equation restrictions implied by optimality of a representative agent's and investment plan.1 These studies have required some specification of an aggregate utility function, and power (constant relative risk aversion) utility has been predominant. The present paper extends this body of research by including models with constant absolute, as well as constant relative, risk aversion. Evidence on the empirical magnitude of aggregate risk aversion is important for several reasons. In portfolio models, the relation of asset demand to risk de? pends on the degree of risk aversion. Aggregate risk aversion is a determinant of the predicted response of to interest rate shifts and of optimal hedg? ing behavior in dynamic asset pricing models. Risk aversion, therefore, influences the structure of prices in futures markets and the term structure of interest rates, to name but two examples. Recent studies of the volatility of stock prices (e.g., [16]) also depend on the magnitude of risk aversion. This study employs quarterly data on U.S. aggregate and Treasury bill returns for 1947-1980. The results do not appear consistent with the hypothesis of aggregate risk neutrality and they suggest that consumption-based asset pricing models should emphasize nonstationarities of consumption betas over intertemporal substitution in consumption. The paper is organized as follows. Section II discusses the model relating equilibrium interest rates and anticipated real changes. This is a discrete-time model, in the spirit of Breeden's [2] intertemporal asset pricing model. It assumes time-additive, state-independent utility with linear risk toler