Bootstrap Tests for Distributional Treatment Effects in Instrumental Variable Models

摘要: This article considers the problem of assessing distributional consequences a treatment on some outcome variable interest when intake is (possibly) nonrandomized, but there binaryinstrument available for researcher. Such scenario common in observational studies and randomized experiments with imperfect compliance. One possible approach to this compare counterfactual cumulative distribution functions without treatment. shows how estimate these distributions using instrumental methods simple bootstrap procedure proposed test hypotheses, such as equality distributions, first-order second-order stochastic dominance. These tests estimators are applied study effects veteran status civilian earnings. The results show negative effect military service during Vietnam era that appears be concentrated lower tail ...