This paper investigates whether there are simple versions of the permanent income hypothesis which are consistent with the aggregate U.S. consumption and output data. Our analysis is conducted within the confines of a simple dynamic general equilibrium model of aggregate real output, investment, hours of work and consumption. We study the quantitative importance of two perturbations to the version of our model which predicts that observed consumption follows a random walk: (i) changing the production technology specification which rationalizes the random walk result, and (ii) replacing the assumption that agents' decision intervals coincide with the data sampling interval with the assumption that agents make decisions on a continuous time basis. We find substantially less evidence against the continuous time models than against their discrete time counterparts. In fact neither of the two continuous time models can be rejected at conventional significance levels. The continuous time models outperform their discrete time counterparts primarily because they explicitly account for the fact that the data used to test the models are time averaged measures of the underlying unobserved point-in-time variables. The net result is that they are better able to accommodate the degree of serial correlation present in the first difference of observed per capita U.S. consumption.

This paper investigates whether there are simple versions of the permanent income hypothesis which are consistent with the aggregate U.S. consumption and output data. Our analysis is conducted within the confines of a simple dynamic general equilibrium model of aggregate real output, investment, hours of work and consumption. We study the quantitative importance of two perturbations to the version of our model which predicts that observed consumption follows a random walk: (i) changing the production technology specification which rationalizes the random walk result, and (ii) replacing the assumption that agents' decision intervals coincide with the data sampling interval with the assumption that agents make decisions on a continuous time basis. We find substantially less evidence against the continuous time models than against their discrete time counterparts. In fact neither of the two continuous time models can be rejected at conventional significance levels. The continuous time models outperform their discrete time counterparts primarily because they explicitly account for the fact that the data used to test the models are time averaged measures of the underlying unobserved point-in-time variables. The net result is that they are better able to accommodate the degree of serial correlation present in the first difference of observed per capita U.S. consumption.

The money demand literature presents much conflicting evidence on this question. For example, Lucas (1988) reports unrestricted money demand regressions which seem to imply that long-run money demand elasticities are highly unstable across subsamples. At the same time, he also presents evidence from money demand regressions with the income elasticity restricted to unity which seem to suggest stability. We conduct a formal analysis which weighs these apparently conflicting facts to determine which hypothesis is more plausible; the hypothesis that money demand is stable, or the hypothesis that money demand is unstable. We find that the stability hypothesis is the more plausible one. Thus, according to our data set, the answer to the question in the title is "yes".

This technical appendix supports "Liquidity Effects, Monetary Policy, and the Business Cycle" in Journal of Money, Credit and Banking (November 1995, Vol. 27, No. 4, Pt. 1, pp. 1113-1136), https://doi.org/10.2307/2077793.

This article studies the accuracy of two versions of Kydland and Prescott's (1980, 1982) procedure for approximating optimal decision rules in problems in which the objective fails to be quadratic and the constraints fail to be linear. The analysis is carried out using a version of the Brock-Mirman (1972) model of optimal economic growth. Although the model is not linear quadratic, its solution can nevertheless be computed with arbitrary accuracy using a variant of existing value-function iteration procedures. I find the Kydland-Prescott approximate decision rules are very similar to those implied by value-function iteration.

This paper shows how to derive the family of models in which Cagan’s model of hyperinflation is a rational expectations model. The slope parameter in Cagan’s portfolio balance equation is identified in some of these models and in others it is not—a fact which clarifies results obtained in several recent papers.