CMStatistics 2021: Start Registration
View Submission - CFE
Title: Retrieving grouped local average treatment effects via cLasso Authors:  Nicolas Apfel - University of Regensburg (Germany) [presenting]
Martin Huber - University of Fribourg (Switzerland)
Henrika Langen - University of Fribourg (Switzerland)
Helmut Farbmacher - Max Planck Society (Germany)
Abstract: In the context of an endogenous binary treatment with heterogeneous effects and multiple instruments, we propose a classifier-Lasso (C-Lasso) procedure to identify complier groups with identical local average treatment effects (LATE), in spite of relying on distinct instruments. Our procedure is based on the fact that the LATE needs to be homogeneous for any two or multiple instruments that (i) satisfy the LATE assumptions and (ii) generate identical complier groups in terms of treatment probabilities given the respective instruments. Under the assumption that a relative majority of instruments with identical complier groups satisfies the LATE assumptions, our procedure permits identifying the valid instruments (the exclusion restriction) in a data-driven way. We also provide a simulation study investigating the finite sample properties of our LATE C-Lasso approach and an empirical application investigating the effect of incarceration on recidivism in the US with judge assignments serving as instruments.