WORKING PAPERS
Testing Conditional Stochastic Dominance at Target Points with Ivan Canay and Federico Bugni
- Link: paper
This paper introduces a novel test for conditional stochastic dominance (CSD) at specific values of the conditioning covariates, referred to as target points. The test is relevant for analyzing income inequality, evaluating treatment effects, and studying discrimination. We propose a Kolmogorov--Smirnov-type test statistic that utilizes induced order statistics from independent samples. Notably, the test features a data-independent critical value, eliminating the need for resampling techniques such as the bootstrap. Our approach avoids kernel smoothing and parametric assumptions, instead relying on a tuning parameter to select relevant observations. We establish the asymptotic properties of our test, showing that the induced order statistics converge to independent draws from the true conditional distributions and that the test is asymptotically of level alpha under weak regularity conditions.
- Link: JMP version (2023)
This article considers the problem of testing sign agreement of a finite number of means. Examples of this problem include detecting heterogeneous treatment effects with opposite signs, refuting the assumptions of local average treatment effect, and testing political affiliation alignment. For the null hypothesis that the means are all non-negative or all non-positive, I propose two novel statistical tests: the Least Favorable test and the Hybrid. The main result is that both tests control their sizes uniformly over a large class of distributions for the observed data in large samples. Compared to popular multiple testing procedures, the Least Favorable test has superior power. In the existing literature on tests of sign agreement, both tests are the first to accommodate arbitrary dependence among estimators for any finite number of means. Results from simulation studies indicate that, with finite samples, the rejection probabilities of both tests reach the nominal level under the null hypothesis. These studies further suggest that, overall, the Hybrid test exhibits higher power than the Least Favorable test when there are more than three means; this relationship reverses when considering only two means. I demonstrate the utility of both tests in an application inspired by Angelucci et al. (2015), in which I study heterogeneous impacts of microloans on various groups and outcomes.