ASSET PRICES, FINITE HORIZONS, AND PROXIES FOR THE INTERTEMPORAL MARGINAL RATE OF SUBSTITUTION

The Intertemporal Marginal Rate of Substitution (IMRS) plays a central role in the Intertemporal Capital Asset Pricing Model (ICAPM). This dissertation is concerned with theoretical and empirical aspects of modeling the IMRS.;The IMRS is unobservable. In the empirical section of this dissertation, we present a methodology for generating proxies for the IMRS from observable data. This methodology includes specifing and to estimating an exact model of conditional first and second moments of the realized risk premia on a set of traded assets. To guarantee that the generated conditional variance-covariance matrices are positive definite, we represent the conditional variance-covariance for each observation in Cholesky decomposion form and let each non-zero element of the square root of the matrix be modeled as a linear combination of observable data.;We incorporate our methodology for proxying the IMRS into a testable asset pricing model. By imposing certain constraints across the parameters of the first and second conditional moments, we are able to produce a single model of both risk premia and the IMRS. Using monthly asset market data, we test several versions of the model. The model is strongly rejected in all cases. However, there is some indication that the realized risk premia on treasury bills are much more strongly related to the IMRS then the realized risk premia on an aggregate stock portfolio.;A standard assumption of the ICAPM is that consumers have infinite planning horizons. This usually implies that the Ricardian Equivalence Theorem holds. In the theoretical model of this dissertation, we assume that consumers have uncertain life spans and no bequest motive. Hence, they have finite planning horizons and the Ricardian Equivalence Theorem does not hold. Thus, changing the policy mix of debt and lump-sum taxes does affect asset prices. The model is a pure exchange general equilibrium model with complete markets for contingent claims over the indefinite future. New consumers are born at the same rate that old ones die. We show that in the context of this model, the degree to which the Ricardian Theorem fails depends both on the death/birth rate and on the parameters of the utility function.