THE CONCEPTUAL FOUNDATIONS for econometric analyses of preferencebased, intertemporal asset pricing relations can be traced back to Rubinstein (1976), Lucas (1978), Breeden (1979), and Cox et al. (1985a). They deduced general equilibrium relations among consumption decisions, asset prices, and production decisions in the context of dynamic models under uncertainty. Typically, agents had common information sets and identical preferences, access to a complete set of contingent claims markets, and equal access to all production technologies.
Grossman and Shiller (1981) were the first to study empirically the relation between consumption and asset returns implied by the representative agent models of Rubinstein (1976), Lucas (1978), and Breeden (1979). They focused on the co-movements of consumption and returns in the context of a model in which consumers were risk averse and had perfect foresight about the future. Subsequently, Hansen and Singleton (1982, 1983) developed methods for estimating the parameters of asset pricing relations implied by stochastic dynamic models that incorporate fairly general specifications of concave preference functions.
The primary objectives of this chapter are to provide an overview of the specifications of preference-based DAPMs and to summarize the large body of empirical evidence on the goodness-of-fit of these models. We begin in Section 10.1 with a presentation of the key challenges facing preferencebased models using a log-linear setting. This is followed in Section 10.2 by a discussion of alternative methodological approaches to assessing the fit of preference-based models.
The strengths and limitations of time-separable, single-good models in representing the historical co-movements of consumption and real returns on stocks and bonds are explored in Section 10.3. Among the issues addressed are the autocorrelation properties of the disturbances, the sensitivity of the results to the length of the holding period of the investments,