TESTING FOR UNOBSERVED HETEROGENEOUS TREATMENT EFFECTS WITH OBSERVATIONAL DATA

Unobserved heterogeneous treatment effects have been emphasized in the recent policy evaluation literature (see, e.g., Heckman and Vytlacil (2005, Econometrica 73, 669–738)). This paper proposes a nonparametric test for unobserved effect model with binary assignment, allowing individuals’ self-selection to treatment. Under standard local average assumptions, i.e., no defiers condition, we derive testable restrictions hypothesis of effects. Furthermore, show that if outcomes satisfy monotonicity assumption, these are also sufficient. Then, propose modified Kolmogorov–Smirnov-type which is consistent simple implement. Monte Carlo simulations our performs well finite samples. For illustration, apply study Job Training Partnership Act on earnings impacts fertility family income, where null homogeneous gets rejected second case but fails be first application.