GGrantIndex
← Search

Tests of Independence Conditions Using a Discontinuously Distributed Covariate

$97,758FY2019SBENSF

University Of Rochester, Rochester NY

Investigators

Abstract

This research will develop statistical tests for selection on observables and validity of instrumental variables assumption necessary for identifying causal effects under mild restrictions on structural functions. Existing methods impose other stringent restrictions to test identifying assumptions. This test will avoid such stringent restrictions by exploiting discontinuities that arise in the distribution of unobservable in the model conditional on the treatment variable as well as the existence of another covariate. The testing procedure will also exploit continuity of the structural functions in this special covariate and monotonicity of the structural function in one of the unobservable variables. The test is applicable to continuous or discrete treatment without instrumental variables in the context of non-parametric, non-separable, single equation model. This research will also develop a procedure for testing validity of instrumental variable for continuously distributed endogenous treatment variable in the context of a nonparametric triangular model. For both problems, the models can be nonparametric and non-separable in the unobserved heterogeneity, as well as settings with more stringent parametric or semi-parametric restrictions. The research results will encourage other researchers to identify such settings, and thus extend the range of empirical problems in which identification conditions can be tested. The results of this research will improve the ability to establish causal relationships, hence improve economic policy making, policy evaluation, and economic efficiency generally. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

View original record on NSF Award Search →