A scathing article this week in Forbes offers a startling look at some dubious research practices used by the Consumer Financial Protection Bureau when investigating alleged discriminatory lending in the auto loan industry. These practices were brought to light by a recent study of CFPB enforcement actions. The study was commissioned by the American Financial Services Association, which is the national trade association for the consumer credit industry. The regulatory actions of the CFPB naturally come under its scrutiny, and the study found “troubling questions” about CFPB research and stated that the Bureau “may have taken a wrong turn.”
The CFPB is tasked with identifying violations of the Equal Credit Opportunity Act, which prohibits discrimination in lending based upon race. Violations in mortgage loans can easily be found by researching racial data that borrowers self-report on their applications. However, federal law prohibits lenders from inquiring into the race of borrowers when offering auto and other consumer loans. This presents a challenge to anyone hoping to enforce the Equal Credit Opportunity Act.
In order to research the race of auto loan borrowers, the CFPB utilized an analytical technique known as Bayesian Improved Surname Geocoding (BISG). BISG relies on Census Bureau data relating to zip codes and surnames to predict race. It uses a list of 151,671 surnames gleaned from the 2010 Census along with each individual’s reported race. Thus, for a name such as Smith, the Census Bureau list indicates what percentage of the overall population with the last name Smith are white, African American, Asian, etc. The Census Bureau also indicates the racial breakdown of each ZIP code. The name and ZIP code data is then combined to provide a probability as to the applicant’s race.
The AFSA study tested the accuracy of the CFPB’s methodology by applying it to a pool of residential mortgage loans where the race of the applicants was known. The results were unsettling. In a group that the CFPB held to be 80 percent likely to be African American, they correctly identified only 24.2 percent of the actual African American consumers. Also, the CFPB’s method produced 22.4 percent false positives, meaning that it identified as African American significant numbers of consumers who actually were white, Hispanic or Asian.
The AFSA study concluded that the CFPB’s research “overstated disparities” and presented “overstatements of alleged consumer harm.” Such discrepancies are no doubt troubling to an auto industry subject to substantial penalties at the hands of the CFPB for alleged racial discrimination.