A class of utility functions is proposed which relaxes some of the restrictive assumptions constrains current models. Specifically, time and state separability are no longer required. With respect to time separability, a recursive (due to Koopmans) specification permits the consumption-leisure tradeoff in any period to depend on consumption and leisure levels in subsequent times. This dependence weakens the a priori constraints imposed in the standard models on the relative responses of consumption and leisure to price and income shocks and this permits one to model more flexibly the intertermporal substitutability of each of consumption and leisure. With respect to state separability, this relaxed assumption has the flexibility to disentangle intertemporal substitutability from risk aversion. A further contribution of this work is that it allows for the estimation of alternative utility specifications. The predictions of the model developed in the theoretical section differ significantly from both the expected utility model and the separable model. These alternatives can be tested wither separately or jointly. Expected utility is a nested special case and can be tested as a simple parameter restriction. Time- separability utility can be tested using non-nested tests. This project will provide the first such empirical testing of a model that embodies both time and state nonseparabilities. The approach taken here will enable a very flexible modeling of intertemporal substitution elasticities, allowing some separation between risk preferences and substitution preferences and also allowing a greater separation between wealth and substitution effects. The empirical testing will permit a quantitative analysis of the importance of both state and time nonseparabilities for dynamic economic analysis.
