Algorithmic Targeting of Social Welfare Programs: Machine Learning for Prediction Model Design and Causal Effects Estimation Open Access

Abstract

Governments and aid organizations in developing countries implement algorithmic rules to identify and provide necessary aid for households in underprivileged conditions. Given demographic and background characteristics from administrative data, traditional econometric methods along with regularized linear regressions have been used for targeting social welfare programs. Non-parametric machine learning techniques, however, are less common in these contexts. In this paper, I compare non-parametric forests to parametric linear regression techniques in both prediction and causal treatment effects estimation problem settings. The standard metric of prediction accuracy suggests that random forests perform slightly better than regularized linear regressions, validated across multiple subsets of data; the estimated average treatment effects using both modeling techniques are positive, with only causal forests showing statistically significant results. There is no evidence of significant heterogeneity in individual treatment effects.

Table of Contents

Chapter 1: Introduction.........................................................................1

Chapter 2: Context.................................................................................3

Chapter 3: Methodology........................................................................7

Chapter 4: Data and Model Design...........................................................10

Chapter 5: Results..................................................................................24

Chapter 6: Discussion.............................................................................29

Chapter 7: Conclusion............................................................................34

About this Honors Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Emory College
Department	Economics
Degree	B.S.
Submission	Honors Thesis
Language	English
Research Field	Economics, Labor
Keyword	machine learning econometrics causal inference
Committee Chair / Thesis Advisor	O'Connell, Stephen, Emory University
Committee Members	Banerjee, Shomu, Emory University Rho, Seunghwa, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Algorithmic Targeting of Social Welfare Programs: Machine Learning for Prediction Model Design and Causal Effects Estimation ()	2020-04-14 00:47:44 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions