Centre for Development Economics
and
Department of Economics, Delhi School of Economics

ANNOUNCE A SEMINAR


Bayesian Learning and Statistical Discrimination in the Market for Small Business Loans.

by

Atanu Rakshit
Nazarbayev University

On

2nd August 2018 (Thursday) at 3:00 PM

Venue : Seminar Room (First Floor)
Department of Economics, Delhi School of Economics

All are cordially invited
 Abstract

We use data on small business loans in the United States to test for evidence of statistical discrimination. To distinguish between this and other causes of differential treatment, like taste-based discrimination, we model discrimination with Bayesian updating feature where banks have prior beliefs concerning each group’s repayment probability and receive noisy signals of creditworthiness (i.e credit ratings). Our model predicts that for the same credit rating, differing prior beliefs generate different loan approval rates between groups at the mediocre level of creditworthiness, but this difference all but disappears at excellent credit rating levels. We then test for this pattern in loan approval rates data constructed from the 1998 and 2003 Survey of Small Business Finances. In the 1998 survey, under a number of controls, Black, Hispanic and Asian-owned businesses are more likely to be denied credit compared to their White-owned counterpart, (only Black-owned businesses experience this discrimination in the 2003 survey). More importantly, these higher denial ratings are observed only at mediocre credit ratings and not at excellent (low risk) credit rating. We contrast these findings against alternative, taste-based types of discrimination where a higher standard of creditworthiness is required by all disadvantaged minorities. We conclude that the data is consistent with statistical discrimination and find no evidence of taste-based racial discrimination.