Making Inferences from Controlled Income Maintenance Experiments

There are a wide variety of issues related to how controlled income maintenance experiments should be interpreted. This paper addresses a single question: If an experiment of limited duration is conducted under "ideal" conditions, what can be inferred from the experimental results about individual behavior in a world where a negative income tax is adopted permanently? Part I of the paper summarizes a conventional analysis of the effects of a negative income tax, given a one period, static model of behavior. Part II i~troduces a simplified multiperiod model to compare the effects of a temporarily and a permanently adopted negative income tax. Part III considers how one might make inferences about permanent behavior based upon the results of a temporary experiment and summarizes how the results derived from the presented two good model can be extended to apply to the multigood case. Part IV briefly explores the implications of the results outside'the context of income maintenance experimentation and outlines some mitigating complications which would alter the results presented'in this paper. MAKING INFERENCES FROM CONTROLLED INCOME MAINTENANCE EXPERIMENTS Introduction Since 1968 substantial resources have been devoted to the design and execution of controlled income maintenance experiments. l In the typical experiment, a sample of low income households is placed on a variety of negative income tax plans for a temporary period of time (usually three years). The primary objective of these experiments has been to determine the impact of a negative income tax on work incentives and the effect of this impact on the monetary cost of adopting such a welfare scheme nationally. While there are a wide variety of issues related to how the results of such experiments are to be interpreted, this paper addresses a single question. If an experiment of limited duration is conducted under "ideal" conditions, what can be inferred from the experimental results about individuai behavior in a world where a negative income tax is adopted permanently? By assuming "ideal" conditions, we mean (a) that individuals (or . 2 households) in the experiment are utility maximizers in the conventional sense, given the structure of the experiment, and (b) that the experiment is sufficiently well designed to measure the differential response of individuals to the experiment. 3 The discussion is limited to a partial equilibrium analysis of an individual's supply of labor, or, equivalently, his demand for leisure time. Part I of the paper summarizes a conventional analysis of the effects of a negative income tax given a one period, static model of behavior.