Two stages least squares with two stages heteroskedasticity

Seminario de investigación del IIEP

Two stages least squares with two stages heteroskedasticity

Expositor | Gabriel Montes Rojas | IIEP UBA-CONICET

Abstract

Linear regressions with endogeneity are widely used to solve endogeneity and estimate causal effects. We show that the presence of conditional heteroskedasticity in the structural and first-stage regressions render the two-stage least squares (2SLS) estimator to be inconsistent. Thus, in empirical work it is paramount to evaluate the presence of heteroskedasticity when employing the 2SLS estimator. We propose an augmented control function estimator that solves this issue. We establish the statistical properties of the estimator, say consistency and asymptotic normality, and propose valid inference procedures. Monte Carlo simulations provide evidence of the finite sample performance of the proposed methods. We consider an empirical application and evaluate different implementing procedures.

Lugar | FCE-UBA | Aula 235