The Estimation of Labor Supply Models Using Experimental Data

For many years there has been interest in replacing the existing complex transfer system in the United States with a nationwide negative income tax (NIT) program.' The feasibility and desirability of an NIT, however, depend on its effects on aggregate labor supply (and its cost). Interest in predicting these aggregate effects has motivated considerable empirical research on labor supply. The first studies used existing data, usually cross-sectional, to estimate the parameters of labor supply functions.2 Unfortunately, the range of estimates in these studies is disturbingly large and of limited usefulness to policymakers.3 Consequently, a new approach to labor supply research has been followed social experimentation.4 Several experiments have been funded by the federal government to test the effects of alternative NIT programs on labor supply. The first experiment, the New Jersey Experiment, was conducted in New Jersey and Pennsylvania from 1968 to 1972.5 Other experiments have taken place in Gary, Indiana from 1970 to 1974, and in rural areas of Iowa and North Carolina from 1969 to 1973. The largest and most comprehensive of these experiments began in 1971 in Seattle, Washington and Denver, Colorado and is still taking place. In principle, a controlled experiment affords the opportunity to overcome most of the problems inherent in nonexperimental research, because in an experiment, the budget constraints of individuals are exogenously shifted in a measurable way. In practice, however, the experiments have been beset with their own unique set of econometric problems. These problems include the nonrandom assignment of experimental treatment, small samples, truncation of response, limited duration, participation in other welfare programs both before and during the experiment by sample members, and the selection of nonrepresentative samples.6 In this paper, a methodology is presented that attempts to deal with these problems. Experimental data from the Seattle and *Economists, SRI International. The research reported in this paper was performed under contracts with the states of Washington and Colorado, prime contractors for the Department of Health, Education, and Welfare, under contract numbers SRS-70-53 and SRS-71-18, respectively. The opinions expressed in the paper are our own and should not be construed as representing the opinions or policies of the states of Washington or Colorado, or any agency of the U.S. government. An earlier version of this paper was presented at the Summer 1976 meetings of the Econometric Society and in seminars at the National Bureau of Economic Research and Mathematica Policy Research. Jodie Allen, Yoram Barzel, David Betson, Michael Boskin, Glen Cain, Joseph Corbett, Irwin Garfinkel, David Greenberg, Terry Johnson, Richard Kaluzny, Richard Kasten, Robert Lerman, Stanley Masters, Myles Maxfield, Robert Moffit, Larry Orr, Harold Watts, and Robert Willis provided valuable comments on various drafts of this paper. We are, of course, solely responsible for the views presented and for any remaining errors. Helen Cohn, Diane Hollenbeck, Paul McElherne, Gary Stieger, and Steven Spickard provided expert programming assistance. I Milton Friedman is usually credited with developing the concept of a negative income tax. Robert Lampman and James Tobin (1965) among others also made early contributions to the concept. 2An excellent collection of such studies is presented in Glen Cain and Harold Watts. 3See Keeley for a survey of these studies and a discussion of some of the econometric difficulties that lead to such a wide range of estimates. 4Heather Ross (1966) is credited with first conceiving the idea of an NIT experiment. Guy Orcutt and Alice Orcutt (1968) first published a paper outlining an experimental design. 5The New Jersey Experiment is described in David Kershaw and Jerilyn Fair. Watts and Albert Rees (1977a, b) and Joseph Pechman and P. Michael Timpane present the results from this experiment. 6See Henry Aaron, Keeley, and Keeley and Robins for a critical discussion of many of these problems.