Date of Award

June 2019

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Economics

Advisor(s)

Alfonso Flores-Lagunes

Second Advisor

Leonard M. Lopoo

Keywords

causal inference, heterogeneous treatment effect, job training program, korea, random forest, selection bias

Subject Categories

Social and Behavioral Sciences

Abstract

This dissertation comprises two chapters that are related to the research topics in Korean Job Training Programs using administrative data. The first chapter examines the effectiveness of the programs on the probability of re-employment and on wages. This observational study tackles multiple biases by adopting: (1) a propensity score matching approach for selection into treatment, and (2) a principal stratification framework for selection into employment when looking at wages. This chapter finds positive effects on employability (0.026), but negative effects on wages (-8.4%), both of which are statistically significant. We conduct sensitivity analysis to violations of our main identifying assumption that confirms our results are robust to unobserved confounding. The second chapter analyzes the heterogeneous treatment effects of the programs on employability using a recent causal forest estimator, which is a machine learning technique. This chapter finds that almost a third of trainees (31.0%) is likely to experience negative effects, in spite of a positive and significant average treatment effect on the treated (0.029). We illustrate distinctive characteristics of the two groups that are affected positively and negatively. Based on these characteristics, we suggest some alternative assignment rules and find that several of the suggested assignment rules outperform the current one.

Access

Open Access

Available for download on Sunday, July 05, 2020

Share

COinS