A new study commissioned by the American Financial Services Association found significant bias and high error rates in the proxy methodology used by the CFPB to determine discrimination in the indirect auto finance market. Central to the study was an examination of the Bayesian Improved Surname Geocoding (BISG) proxy methodology used by the CFPB to determine a disparate impact to legally protected groups. BISG estimates race and ethnicity based on an applicant’s name and census data. AFSA’s study calculated BISG probabilities against a test population of mortgage data, where race and ethnicity are known. One of the primary findings was that when the proxy uses an 80 percent probability that a person belongs to an African American group, the proxy ...