Work

A Machine Learning Approach to Credit Allocation

Year: 2020
Type: article
Abstract: This dissertation seeks to understand the shortcomings of contemporaneous credit allocation, with a specific focus on exploring how an improved statistical technology impacts the credit access of societally important groups. First, this dissertation investigates a variety of limitations of conventional credit scoring models, specifically their tendency to misclassify borrowers by default risk, especially for relatively risky, young, and low income borrowers. Second, this dissertation shows that an improved statistical technology need not to lead to worse outcomes for disadvantaged groups. In fact, the credit access for borrowers belonging to such groups can be improved, while providing more accurate credit risk assessment. Last, this dissertation documents modern-day disparities in debt collection judgments across white and black neighborhoods. Taken together, this dissertation provides valuable insights for the design of policies targeted at reducing consumer default and alleviating its burden on borrowers and lenders and across societally important groups, as well as macroprudential regulation.
Language: English
Subfield: Accounting
Open Access status: green
OpenAlex ID: W3145793988
FWCI:
Cites: 76
Cited by:
Related to: 10