An improved bootstrap test for restricted stochastic dominance

Bootstrap Testing for restricted stochastic dominance of a pre-specified order between two distributions is interest in many areas economics. This paper develops new method improving the performance such tests that employ moment selection procedure: tilting empirical distribution procedure. We propose amount be chosen to maximize likelihood subject restrictions null hypothesis, which are continuum unconditional inequality conditions. characterize sets population on modified test (i) asymptotically equivalent its non-modified version first-order, and (ii) superior according local power when sample size large enough. report simulation results show versions leading noticeably less conservative than their counterparts have improved power. Finally, an example discussed illustrate proposed method.