Late again, whithout Monotonicity
暂无分享,去创建一个
Monotonicity is not necessary for the Wald ratio to identify a Local Average Treatment Effect. Under random assignment and exclusion restriction, if for every value of potential outcomes there are more compliers than defiers, the Wald ratio identifies the average treatment effect within a subpopulation of compliers. I use a simple Roy selection model to show that this "less defiers than compliers" condition is substantially weaker than monotonicity. It has two implications which are testable from the data, and it is closely related to those testable implications. Similarly, the local monotonicity condition in Huber & Mellace (2012) is not necessary for their identification results to hold and can also be replaced by a substantially weaker condition
[1] Xavier d'Haultfoeuille,et al. Late Again with Defiers , 2012 .
[2] M. Huber,et al. Relaxing monotonicity in the identification of local average treatment effects , 2012 .
[3] Kevin Lang,et al. School Entry, Educational Attainment and Quarter of Birth: A Cautionary Tale of LATE , 2009 .
[4] Alberto Abadie. Semiparametric instrumental variable estimation of treatment response models , 2003 .