Essays in Microeconometrics

Abstract

This dissertation consists of three essays in microeconometrics.The first essay establishes the first set of inference results for using marginal treatment effects (MTEs) to extrapolate local average treatment effects (LATEs) that are robust to limited instrument variation. These results apply not only to inference on the MTE itself but also to other causal parameters, such as policy-relevant treatment effects, which are of particular interest to policymakers.The second essay, co-authored with Federico Bugni and Jackson Bunting, studies hypothesis testing for the marginal homogeneity assumption, an assumption where time-specific marginal distributions of the panel data remain homogeneous or time-invariant. We develop three inference methods to test this hypothesis based on asymptotic approximation, bootstrap, and permutations.The third essay, co-authored with Matthew Masten and Alexandre Poirier, introduces a general class of relaxations of the unconfoundedness assumption, encompassing several existing approaches as special cases. We use this class to derive a variety of new identification results which can be used to assess sensitivity to unconfoundedness.