English Abstract: Assumptions that are sufficient to identify local average treatment effects (LATEs) generate necessary conditions that allow instrument validity to be refuted. The degree to which instrument validity is violated, however, probably varies across subpopulations. In this paper, we...