Essays in labor and development econometrics

This thesis consists of three essays that develop nonparametric data-based smoothing methods for estimands popular in labor and development economics.;In the first essay, I investigate the problem of optimal selection of smoothing parameter or bandwidth for finite-dimensional smooth functionals of a regression function. Examples of such functionals include average treatment effects, nonparametric versions of the Blinder-Oaxaca decompositions, nonparametric policy effects and average derivatives. I show that this problem is in general characterized by a form of ill-posedness: small changes in the underlying joint distribution of the data can cause large changes in chosen bandwidth. I propose a simple solution and prove its optimality properties. As an example I focus on how to choose bandwidth for average treatment effects.;The second essay outlines an easily implemented rule of thumb bandwidth selection algorithm for the Average Treatment Effect (ATE) I also analyze the algorithm using two real datasets as bases for Monte Carlo simulations. In the first dataset, the key functional is an average treatment effect that quantifies the impact of a training and job search/placement program on earnings. In the second, the estimand is the effect of unearned lottery winnings on labor supply in the form of post-win earnings.;The third essay investigates the problem of optimal selection of bandwidth for the popular regression discontinuity estimator, a very different sort of estimand than the one considered in the first chapter. The focus is on estimation by local linear regressions. The paper propose a simple reference rule that estimates an optimal bandwidth as a function of the data and the chosen kernel. We establish some optimality properties of the rule.