HOUSEHOLD LABOR SUPPLY AND TAXES: A NONPARAMETRIC, REVEALED PREFERENCE APPROACH

This paper uses the Generalized Axiom of Revealed Preference (GARP) to test the static labor supply model. The advantages of GARP are that it tests for utility maximization without making parametric assumptions and accommodates piecewise linear and nonconvex budget sets. Further, GARP tests are an important preliminary step before estimating a parametric labor supply model because GARP helps identify which assumptions are particularly troubling. Tests of GARP are applied to the hours of work and disaggregated commodity choices of individual households' from the 1982-85 Consumer Expenditure Survey. The leisure and commodity choices of demographically similar households resoundingly reject GARP. To find the source of the violations, GARP tests on commodity choices alone are performed and the joint hypotheses of utility maximization and common preferences within demographic group are not rejected. This suggests that the source of the rejections is the labor supply decision. This paper then highlights the role that measurement error in hours, combined with calculating the hourly wage as earnings divided by hours, plays in directly leading to violations of revealed preference. Leisure and commodity choices are generally consistent with GARP when using predicted wages and the tax schedule to model the budget constraint. Therefore, as long as researchers instrument for the wage in the nonlinear budget constraint setting, the search for less restrictive empirical models of static labor supply is likely to be worthwhile.