Permanent income hypothesis and the cost of adjustment

Parise, Gerald
Major Professor
Walter Enders
Committee Member
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Organizational Unit
Journal Issue

This dissertation studies the assumptions of perfect capital markets and rational expectations in modeling consumption. Analysis concentrates on reducing the level of knowledge and information imposed upon the agent while remaining within a permanent income hypothesis (PIH) framework;Imperfect capital markets are introduced into the PIH by assuming that agents may encounter two costs while formulating a consumption path. Introducing these costs permits agents to possess a degree of ignorance about the imperfections that may confront them. One cost arises from the inability to achieve desired consumption levels as inferred by the PIH, reflecting the possibility that imperfections may make formulation of a consumption plan different from its implementation. Formulation and implementation are one only if the assumptions of the PIH are fulfilled. A second cost is incurred as agents attempt to alter consumption levels because market failures may introduce rigidities into the system. Confronted with these two costs, agents will attempt to form a consumption plan;Costs are introduced into the PIH framework through a quadratic loss function. Deviations from desired consumption and the previous consumption level will generate costs that agents seek to minimize. Specifying an auxiliary system and assuming rational expectations shows that the change in consumption is a function of its lag and the expected change in permanent income. Using a complex rational expectations/VAR system, estimation suggests that these two costs of adjustment are significant. A second model, which reduces the level of information imposed upon the agent and introduces cointegration, reveals similar results. An important result of this work is the possibility of government policy affecting consumption levels;Estimation of both models relies upon the assumption of rational expectations. As an alternative, a learning model is presented that allows agents to obtain knowledge of the system as the sample progresses, permitting change to be incorporated into the system. Consumption can be determined by jointly estimating the learning rule and a time-varying auxiliary system with a Kalman filter. Estimation results suggest that standard application of the rational expectation hypothesis is incorrect because of coefficient instability. All estimation uses the GAUSS programming language.