Essays in applied microeconometrics

Abstract

In the Regression Discontinuity (RD) design,discontinuities in the density function of the running variable may harm identification and bias estimations as in the manipulation and heaping cases. This paper proposes a new robust approach that consistently estimates the Average Treatment . Effect near the cutoff under discontinuities in the distribution function of the running variable. The approach consists of sample-reweighting the outcome prior to the estimation of the causal effect.This paper also discusses the limitation of existing indirect tests about the validity of the RD estimation results and presents sufficient conditions for identification that are directly linked to those tests.Simulated examples are presented to assess the finite simple performance of the proposed approach, while non-simulated examples use real data and compare to existing correction methods that partially identify the effect. The Reweighted-RD design applies to any setting with or without discontinuities in the conditional or marginal distributions while maintaining all the distinctive features of th econventional RD design.