B1844
Title: Detecting grouped local average treatment effects and selecting true instruments
Authors: Nicolas Apfel - University of Regensburg (Germany) [presenting]
Abstract: Under an endogenous binary treatment with heterogeneous effects and multiple instruments, a two-step procedure is proposed for identifying complier groups with identical local average treatment effects (LATE) despite relying on distinct instruments, even if several instruments violate the identifying assumptions. The fact that the LATE is homogeneous is used for instruments which (i) satisfy the LATE assumptions (instrument validity and treatment monotonicity in the instrument) and (ii) generate identical complier groups in terms of treatment propensities given the respective instruments. A two-step procedure is proposed, where the propensity scores are clustered in the first step and groups of IVs are found with the same reduced form parameters in the second step. Under the plurality assumption that within each set of instruments with identical treatment propensities, instruments truly satisfying the LATE assumptions are the largest group, the procedure permits identifying these true instruments in a data-driven way. It is shown that the procedure is consistent and provides consistent and asymptotically normal estimators of underlying LATEs. A simulation study is also provided, investigating the finite sample properties of the approach and an empirical application investigating the effect of incarceration on recidivism in the US with judge assignments serving as instruments.
