NCRC, in collaboration with our academic partners, conducted pre-application multi-layered matched testing for race and national origin in the Los Angeles MSA. Multi-layered matched testing consists of a control and members of two different protected classes, contacting a bank branch to determine if there is a difference in treatment. The multi-layered matched testing only occurs in the pre-application phase with the tester requesting information about loan products. For this round of testing, we conducted 30 male and 30 female multi-layered matched tests consisting of White, Black and Hispanic testers. We had a total of 180 interactions with the bank branches.
In these tests, we contacted 60 bank branches representing 47 different financial institutions. The financial institutions were randomly selected from all small business lending institutions, both national and regional banks, and comprise over 50% of the institutions operating in the market. The sample of banks in the Los Angeles market were randomly selected to represent a broad cross-section of the small business lending marketplace. The banks selected ranged from lenders with assets over $10 billion to community banks. The purpose of the research is to determine the baseline customer service level that male and female testers of different racial and national origin backgrounds receive when seeking information about small business loans to help keep their businesses open during the COVID-19 crisis. NCRC started testing on July 27, 2020, and completed the testing on August 7, 2020, the day before PPP lending ended.
Testing is a critical tool for fair lending enforcement used to assess equal access to credit. Federal agencies also conduct similar testing to investigate when they have been alerted to suspicious behavior. Matched pair testing, such as this, is the only method to measure pre-application discrimination and discouragement since banks do not collect data on loan inquires prior to an application being submitted. The Department of Housing and Urban Development (HUD) partners with fair housing groups to do testing under their Fair Housing Initiatives Program (FHIP). The Department of Justice (DOJ) Civil Rights Division has taken on many cases developed from its testing program. Additionally, the Consumer Financial Protection Bureau (CFPB) has taken enforcement action against Bancorpsouth after implementing testing that revealed multiple threats to consumers’ fair and equal access to mortgages. The U.S. Supreme Court has upheld testing as a crucial private enforcement tool in fighting against civil rights violations.
NCRC and our academic partners conducted these tests as phone inquiries since many bank branches remain closed to in-person services due to the pandemic. Tester profiles were controlled with racially identifiable names and each tester was required to pass a voice panel test to determine whether or not their perceived race could be determined over the phone. Testers were only used if their voice was perceived to be racially identifiable by a blind panel. Research shows that the race of an individual can often be determined by their names alone. We selected profile names after researching names most often perceived with a particular race, and like testers’ voices, we presented those names to a national survey panel to identify names perceived to be highly correlated with gender (male and female) and race/national origin (Black, Hispanic, White). Research also reveals that linguistic profiling occurs over the phone, and that many Americans are able to accurately guess social demographics such as race over the phone after a few sentences.
As with our previous work, this study was designed to answer the following research questions: Are minority and non-minority small business owners with similar economic and business profiles:
- Presented with the same information?
- Required to provide the same information?
- Given the same level of service quality and encouragement?
The matched-pair mystery shoppers had nearly identical business profiles and strong credit histories to inquire about small business products to maintain their business during the COVID-19 pandemic. The profiles of all testers were sufficiently strong that, on paper, they would all qualify for a loan. Furthermore, the protected Black and Hispanic tester profiles were intentionally designed to be slightly better than their White counterparts in terms of income, assets and credit scores. This was done to make it a more conservative test of any differential treatment. Immediately following the interaction, testers were asked to answer either yes or no about whether specific behaviors, queries and comments were made by bank employees. They were also required to provide a narrative of the interaction.
Analysis of the data was done under two different reviews: chi-square statistical difference tests and a fair lending review. For the chi-square difference, we applied statistical analysis in evaluating whether differences in the interactions between White testers, Black testers and Hispanic testers, were significant across all of the banks visited. The chi-square test for independence, is a particularly robust way for social scientists to evaluate whether or not there are substantial differences in outcomes between groups. Testers were asked simple “yes or no” questions in a survey about their interaction with bank personnel and the number of interactions observed ensures a high level of validity. To make certain we were analyzing and comparing cases that reached a similar point-of-contact with the bank, we only included in these analyses those cases where the tester was able to speak to a bank representative regarding their small business loan request. In total, there were 22 White male, 27 Black male, 27 Hispanic male, 30 White female, 22 Black female and 28 Hispanic female tests that met those requirements and were analyzed using Chi-square difference tests.
For the fair lending review, each of the narratives were analyzed independently by two fair lending experts as a matched-pair set (White to Black and White to Hispanic) to determine a difference in treatment under fair lending standards. We used four categories to determine a difference in treatment: lack of encouragement, difference in products, difference in information provided and difference in information requested. For each case where one of these differences was identified, we added a tally to report a total number of discrimination counts. The testing once again revealed major differences in treatment.
Any differences in treatment between White and Hispanic/Black testers are particularly troubling because the combined effect of these various differential treatments may lead to feelings of discouragement and despondency among minority entrepreneurs in the financial marketplace. There is some evidence that this may already be happening. The Federal Reserve Small Business Credit Survey 2019 Report on Nonemployer Firms states that 13% of all small business owners do not apply for credit because they are discouraged. However, in the same report, the rate of discouragement among minority entrepreneurs is markedly higher at 27% for Black entrepreneurs and 21% for Hispanic entrepreneurs.
For our fair lending review, we looked to the Equal Credit Opportunity Act (ECOA), the fair lending law that applies to small business loans. ECOA makes it illegal for a member of a protected class to be discriminated against in any aspect of a credit transaction, which includes the pre-application arena. Protected classes under ECOA include: race, color, religion, national origin, sex, marital status or age (provided the applicant has the capacity to contract); to the fact that all or part of the applicant’s income derives from a public assistance program; or to the fact that the applicant has in good faith exercised any right under the Consumer Credit Protection Act. Discrimination occurs when a protected applicant is offered different products, provided different information or experienced discouragement to apply or pursue a loan compared to the non-protected applicant.
Discrimination can be found through either overt statements, disparate impact or disparate treatment. Disparate impact is a neutral policy that has an adverse disproportionate effect against a protected class. This can be shown through the use of statistical tools. Disparate treatment compares the treatment between two individuals with one of the individuals being a member of a protected class. Disparate treatment can range from subtle differences in treatment to more overt cases. The chi-square analysis revealed statistical significance and supported the analysis that disparate impact was seen across the marketplace, which will require large-scale reform. The fair lending analysis of the individual matched-pair tests revealed disparate treatment between classes, which can be addressed individually through cases filed under ECOA.