Date of Submission
Spring 2024
Academic Program
Economics
Project Advisor 1
Gautam Sethi
Project Advisor 2
Sanjay DeSilva
Abstract/Artist's Statement
Addressing poverty in developing countries without relying on income data is becoming increasingly important, particularly since the Covid-19 pandemic has exacerbated the issue farther than anyone expected. This paper reviews literature on various poverty targeting models including Proxy Means Test, Community-Based Targeting, hybrid models, and machine learning-based models in hopes of finding the best method. The findings highlight the importance of model parameters, particularly in PMT, also revealing that the number of potential beneficiaries analyzed and number of indicators utilized can influence the targeting accuracy. CBT incorporates community involvement in the poverty targeting process at a lower cost than PMT, despite slightly lower accuracies. Hybrid models contain both benefits and problems with PMT and CBT, with many authors finding mixed results. The recent emergence of ML-based PMT models has shown promising results, particularly when applied to the feature selection phase. With targeting accuracies averaging 10% higher than PMT and cost efficiencies rivaling those from CBT, ML-based models might be the future of poverty targeting in developing countries.
Open Access Agreement
Open Access
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
Recommended Citation
Heidelberg, Luca Orion, "PMT, CBT, and Hybrid Models: Is Machine Learning the Future of Poverty Targeting?" (2024). Senior Projects Spring 2024. 120.
https://digitalcommons.bard.edu/senproj_s2024/120
This work is protected by a Creative Commons license. Any use not permitted under that license is prohibited.
Included in
Behavioral Economics Commons, Econometrics Commons, Growth and Development Commons, Macroeconomics Commons