Working Paper: NBER ID: w31194
Authors: Sren Blomquist; Anil Kumar; Cheyuan Liang; Whitney Newey
Abstract: This paper is about the nonparametric regression of a choice variable on a nonlinear budget set under utility maximization with general heterogeneity, i.e. in the random utility model (RUM). We show that utility maximization and convex budget sets make this regression three dimensional with a more parsimonious specification than previously derived. We show that nonconvexities in the budget set will have little effect on these results in important cases. We characterize all the restrictions of utility maximization on the budget set regression and show how to check these restrictions in applications. We formulate budget set effects that can be identified by this regression and give automatic debiased machine learners of these effects. We consider use of control functions to allow for endogeneity. Throughout we take as the main example the effect of taxes on taxable income including accounting for productivity growth. In an application to Swedish data we find the taxable income elasticity of a change in the slope of each segment to be .52, that the regression satisfies the restrictions of utility maximization at the values chosen for over 95% of observations, and that a productivity growth rate we estimate is close to other estimates.
Keywords: Nonparametric Regression; Random Utility Model; Taxable Income; Budget Set; Elasticity
JEL Codes: C14; C24; H31; J22
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| taxes (H29) | taxable income (H24) |
| slope of budget segments (H60) | taxable income (H24) |
| budget set variables (H61) | taxable income (H24) |