Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/249459 
Authors: 
Year of Publication: 
2021
Series/Report no.: 
ADBI Working Paper No. 1280
Publisher: 
Asian Development Bank Institute (ADBI), Tokyo
Abstract: 
Lending institutions' reluctance to lend to MSMEs or to offer them competitive interest rates stems from the relatively costly information acquisition for small loans. The central idea is to bridge the information gap between the demand and the supply side by creating a credit analytics sharing infrastructure through federated learning, which completely respects data privacy. Pooling credit information across multiple lending institutions, particularly rare default events, enables the construction of a more informative credit model for MSMEs, which can then serve as a common good among lenders. The technology also allows for lender-specific models, which in essence share the model's parameters on the common prediction variables while differing in their respective alternative data fields. The lenders in the MSME space can work like a coopetition and continue to compete with their varying risk appetites, loan rates, and banking services. We use real MSME credit data to demonstrate the feasibility of the sharing technology and to study the impact of the COVID-19 pandemic via a portfolio that we assembled from four hypothetical banks operating in six ASEAN countries.
Subjects: 
COVID-19
coopetition
alternative data
federated learning
default
JEL: 
C1
C8
G21
Creative Commons License: 
cc-by-nc-nd Logo
Document Type: 
Working Paper

Files in This Item:
File
Size





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.