Essays on Distributional Treatment Effects with Panel Data

Abstract

In my dissertation, I develop new methods to understand the distributional effect of participating in a program or experiencing a treatment. This goal is different from most research in economics which either (i) restricts the effect of participating in a treatment to be the same across all individuals or (ii) allows for heterogeneous treatment effects but estimates the average effect of participating in a treatment. I consider the case where a researcher has access to panel data and wants to exploit having access to panel data by allowing for time invariant unobserved heterogeneity. Chapters 1 and 3 introduce new methods for comparing the observed distribution of treated potential outcomes for the treated group to their counterfactual distribution of untreated potential outcomes. Chapter 2 considers a more general class of distributional treatment effect parameters that depend on the joint distribution of treated and untreated potential outcomes; these are parameters that would not be identified even if a researcher had access to experimental data. Methodologically, the main innovation of my dissertation is to replace the unknown dependence, or copula, between potential outcomes with observed dependence in the past or observed dependence for the untreated group. I provide details on estimation and consider applications on (i) the distributional effects of participating in a job training program and (ii) the distributional effects of job displacement of older workers during the Great Recession.