Racial disparities in the small business loan market

ABSTRACT

We investigate patterns of racial bias in small business loans denial rates in the US across different credit risk scores. Using data constructed from the 1998 Survey of Small Business Finances and Kauffman Firm Survey, we find disparities in loan approval ratings between Black and White entrepreneurs in intermediate risk categories, but not for the best and worst categories. We explain these findings with a simple and generalizable statistical discrimination model where banks hold prior beliefs of repayment probability based on the applicant’s group and observe noisy signals of creditworthiness. Our model predicts that differences in loan denial rates across groups are more pronounced at middle range values and disappear at very high and very low credit scores. Identifying the likely cause of differential treatment in the market is an important first step in improving representation of minority groups as entrepreneurs. Our findings contribute to this critical effort by delineating the nature of racial bias in the credit market and informing remediation policies.

KEYWORDS:

• Small business loans
• statistical discrimination
• race
• entrepreneurs
• access to credit

JEL CODES:

• D83
• G21
• J71
• M13

Acknowledgments

We thank Gerald Oettinger, Otis Gilley, and Patrick Scott for their helpful comments on the theory. We also thank David Robinson, Jennifer Francis, Giuseppe Lopomo, Jeremy Petrankas, and seminar participants at Duke University. We thank Rohini Somanathan, Ashwini Deshpande, Mausumi Das, and Shreekant Gupta, and various seminar participants at Delhi School of Economics for helpful comments.

Code availability

The statistical analysis to derive the results in this study was conducted using Stata 16. The codes are fully available from the authors on request.

Authors’ contributions

Both authors contributed equally to the study conception and design, as well as analysis, drafting and revisions. Both authors read and approved the final manuscript.

Disclosure statement

All authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest or nonfinancial interest in the subject matter or materials discussed in this manuscript.

Data availability statement

The empirical analysis in this study uses data from US Federal Reserve’s Survey of Small Business Finances, which is publicly available from https://www.federalreserve.gov/pubs/oss/oss3/ssbf98/ssbf98home.html.

Moreover, this study also uses confidential Kauffman Firm Survey microdata, which requires paid subscription to National Opinion Research Center (NORC) Data Enclave at the University of Chicago, and is available from https://www.norc.org/Research/Projects/Pages/kauffman-firm- survey-scholars-program.aspx.

Notes

¹ There is also some mixed evidence of discrimination in mortgage markets (see, Berkovec et al., 1998; Day & Liebowitz, 1998; Munnell et al., 1996).

² Using data from other countries, a number of other studies find little to no evidence of discrimination by banks for small business loans (see, Bruder et al., 2011; Howell, 2019; Cowling et al., 2021). A further discussion of this and its implications are located in Discussion and limitations.

³ These attitudes could be held by any stakeholder of a bank, such as loan officers, bank managers, or even other clients.

⁴ Guryan and Kofi Charles (2013) discuss the empirical challenges with distinguishing between different discrimination types.

⁵ We also expect this information based differential treatment among exceedingly young entrepreneurs with absolutely no credit history at all.

⁶ Wozniak (2015) finds that more accurate information about drug use improves Black employment as better information on drug use reduces such uncertainties for this group.

⁷ Naturally, any discrimination, statistical or animus will change the banks decision when the probability is sufficiently close to the threshold level.

⁸ Lang (1986) argued, for example, that the speech patterns of some groups are less understood by their employers giving nosier information about individual workers thus contributing to discrimination.

⁹ There is some compelling justification for this. Lang (1986) suggests that communication is more reliable when there is a shared culture. If loan officers are more often White, information about Black-owned businesses is weakly noisier than information about White-owned businesses.

¹⁰ In extreme cases the expected loan value of the low type is zero and the high type is the present value of future payment obligations. However, by using L and H we allow more for more flexibility.

¹¹ Although there is evidence in the literature that loan repayment rates may differ across races (Fairlie & Robb, 2007).

¹² Neumark (2012) illustrates another type of statistical discrimination involving risk aversion and differences in ability variance across groups. We avoid this mild complication because quality and variance of quality are both solely determined by probability of high-type firm, Λv,k. Also, the risk neutrality of bank officers in our model is a simplifying rather than a crucial assumption.

¹³ The difference in expected payment is maximized at the credit score equal to the geometric mean of the two groups’ odds of being low-type.

¹⁴ Dougal et al. (2019) find that despite both being equally low-risk, historically Black colleges and universities (HCBUs) pay a higher cost of financing compared to non-HCBUs, particularly in the south.

¹⁵ A small business is defined as a for-profit nonfinancial, nonfarm, nonsubsidiary business enterprise that has fewer than 500 employees.

¹⁶ Some of the studies that have used the SSBF data to look at questions of race and access to capital are (Asiedu et al., 2012; Blanchard et al., 2008; Blanchflower et al., 2003; Cavalluzzo et al., 2002; K. Cavalluzzo & Wolken, 2005).

¹⁷ For use of KFS data, see Ballou et al. (2008), Bates and Robb (2014), Cole and Sokolyk (2018) and Fairlie et al. (2022).

¹⁸ Entrepreneurial skills tend to be passed onto members through working in family members’ business. Therefore, low rates of entrepreneurial success are passed on from one generation to another.

¹⁹ One possible solution could be to selectively limit the discretion of loan officers at certain credit levels. Another possible solution is internal or external monitoring of decisions by loan officers where they justify each decision. Both of these may be costly if there is value in giving the loan officer discretion.