A Learning and Control Perspective for Microfinance

Xiyu Deng, Christian Kurniawan, Adhiraj Chakraborty, Assane Gueye, Niangjun Chen, Yorie Nakahira
Proceedings of The 5th Annual Learning for Dynamics and Control Conference, PMLR 211:915-927, 2023.

Abstract

While microfinance has excellent potential for poverty reduction, microfinance institutions (MFIs) are facing sustainability hardships due to high default rates. Existing methods in traditional finance are not directly applicable to microfinance due to the following unique characteristics: (a) insufficient prior loan histories to establish a credit scoring system; (b) applicants may have difficulty providing all the information required by MFIs to predict default probabilities accurately, and (c) many MFIs use group liability (instead of collateral) to secure repayment. In this paper, we present a novel control-theoretic model of microfinance that accounts for these characteristics and an algorithm to optimize the financing decision in real-time. We characterize the convergence conditions to Pareto-optimum. We demonstrate that the proposed method produces fast decisions and is robust against missing information while still accounting for financial inclusion, fairness, social welfare, sustainability, and the complexities induced by group liability. To the best of our knowledge, this paper is the first to connect microfinance and control theory.

Cite this Paper

BibTeX
@InProceedings{pmlr-v211-deng23a, title = {A Learning and Control Perspective for Microfinance}, author = {Deng, Xiyu and Kurniawan, Christian and Chakraborty, Adhiraj and Gueye, Assane and Chen, Niangjun and Nakahira, Yorie}, booktitle = {Proceedings of The 5th Annual Learning for Dynamics and Control Conference}, pages = {915--927}, year = {2023}, editor = {Matni, Nikolai and Morari, Manfred and Pappas, George J.}, volume = {211}, series = {Proceedings of Machine Learning Research}, month = {15--16 Jun}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v211/deng23a/deng23a.pdf}, url = {https://proceedings.mlr.press/v211/deng23a.html}, abstract = {While microfinance has excellent potential for poverty reduction, microfinance institutions (MFIs) are facing sustainability hardships due to high default rates. Existing methods in traditional finance are not directly applicable to microfinance due to the following unique characteristics: (a) insufficient prior loan histories to establish a credit scoring system; (b) applicants may have difficulty providing all the information required by MFIs to predict default probabilities accurately, and (c) many MFIs use group liability (instead of collateral) to secure repayment. In this paper, we present a novel control-theoretic model of microfinance that accounts for these characteristics and an algorithm to optimize the financing decision in real-time. We characterize the convergence conditions to Pareto-optimum. We demonstrate that the proposed method produces fast decisions and is robust against missing information while still accounting for financial inclusion, fairness, social welfare, sustainability, and the complexities induced by group liability. To the best of our knowledge, this paper is the first to connect microfinance and control theory.} }
Endnote
%0 Conference Paper %T A Learning and Control Perspective for Microfinance %A Xiyu Deng %A Christian Kurniawan %A Adhiraj Chakraborty %A Assane Gueye %A Niangjun Chen %A Yorie Nakahira %B Proceedings of The 5th Annual Learning for Dynamics and Control Conference %C Proceedings of Machine Learning Research %D 2023 %E Nikolai Matni %E Manfred Morari %E George J. Pappas %F pmlr-v211-deng23a %I PMLR %P 915--927 %U https://proceedings.mlr.press/v211/deng23a.html %V 211 %X While microfinance has excellent potential for poverty reduction, microfinance institutions (MFIs) are facing sustainability hardships due to high default rates. Existing methods in traditional finance are not directly applicable to microfinance due to the following unique characteristics: (a) insufficient prior loan histories to establish a credit scoring system; (b) applicants may have difficulty providing all the information required by MFIs to predict default probabilities accurately, and (c) many MFIs use group liability (instead of collateral) to secure repayment. In this paper, we present a novel control-theoretic model of microfinance that accounts for these characteristics and an algorithm to optimize the financing decision in real-time. We characterize the convergence conditions to Pareto-optimum. We demonstrate that the proposed method produces fast decisions and is robust against missing information while still accounting for financial inclusion, fairness, social welfare, sustainability, and the complexities induced by group liability. To the best of our knowledge, this paper is the first to connect microfinance and control theory.
APA
Deng, X., Kurniawan, C., Chakraborty, A., Gueye, A., Chen, N. & Nakahira, Y.. (2023). A Learning and Control Perspective for Microfinance. Proceedings of The 5th Annual Learning for Dynamics and Control Conference, in Proceedings of Machine Learning Research 211:915-927 Available from https://proceedings.mlr.press/v211/deng23a.html.

Related Material