AN EMPIRICAL ANALYSIS OF THE PRICING OF MORTGAGE‐BACKED SECURITIES

Abstract

EQUILIBRIUM PRICES OF FIXED-RATE securities have recently been calculated for a limited set of parametric specifications of investors' preferences, production technologies, and economic uncertainty. For instance, in one example, Cox, Ingersoll and Ross (1978) solve for the equilibrium bond prices implied by a linear technology and the assumption that investors exhibit logarithmic utility. Subsequently, Dunn and McConnell (1981b) adopted these assumptions in order to calculate and study the prices of Government National Mortgage Association (GNMA) mortgage-backed securities. The distributions (and, therefore, the magnitudes and time paths) of the asset prices calculated from these models depend in an important way on the underlying specifications of the economic environment. Consequently, inferences about the time series behavior of individual and relative prices of fixed-rate securities may not be robust to alternative model specifications. Therefore, before proceeding with tests of explicit pricing equations, it seems worthwhile to investigate empirically whether the restrictions on asset prices implied by particular classes of preference functions are consistent with the data, without imposing the auxiliary restrictions associated with the specification of technology and uncertainty. The primary purpose of this paper is to test a discrete time, consumptionbased model of prices of GNMA securities and U. S. Treasury bonds in which the specification of preferences is more general than the logarithmic form and in which minimal structure is imposed on the stochastic processes generating consumption and returns. By focusing on the necessary conditions for intertemporal utility maximization, we are able to estimate the parameters characterizing preferences and test restrictions on the temporal behavior of the prices of GNMAs and U.S. Treasury bonds, without having to solve for the stochastic equilibrium. Our estimates of the preference parameters hopefully will provide guidance for structuring analyses of bond pricing models, which require that preferences be specified a priori. In addition, the issue of whether the relative prices of GNMAs