Working Papers
In this paper, we develop a specification test for IV validity assumptions required for the identification of marginal treatment effects. The strongest testable implication for the instrument validity is characterized by a monotonicity restriction and an index sufficiency restriction on the conditional joint density of outcome and treatment on the propensity score. Our test reformulates the test for IV validity assumptions as the test for shape restrictions. Our statistics imposes the shape restriction under the null hypothesis via the least concave majorant operator. We approximate the null asymptotic distribution of the statistics using a bootstrap procedure easy to implement. The finite sample performance is examined by a Monte-Carlo experiment.
Spatial dependence among local units leads the size distortion issue in regional policy evaluation. In this paper, we analyze this kind of size distortion based on the matching estimator. Firstly, we propose an asymptotic valid variance estimator under spatial heteroskedasticity and autocorrelation consistent (SHAC) framework. Secondly, we propose two valid bootstrap procedures under the spatial autocorrelated disturbance and SHAC framework, respectively. We construct Monte Carlo experiments to compare these inference approaches in small samples. As an empirical illustration, we reevaluate one immigration policy on the unemployment rate of German local labor markets.