For some period of time now (it is unclear just how long), the CFPB has been using a proxy method, which combines geography and surname-based information into a simple tool that is used to determine a borrower's or an applicant's race and ethnicity for loan transactions where that demographic information would not appear in a bank's loan files (e.g., automobile loans).
The Equal Credit Opportunity Act ("ECOA") and Regulation B prohibit, with few exceptions, inquiries about race, color, religion, national origin or sex of an applicant. That information is often needed, however, in the context of a Fair Lending analysis to identify potential discrimination. The CFPB has validated a proxy process that can approximate this data to a high degree of accuracy.
To come up with the proxy, the CFPB constructs a probability that assigns race and ethnicity based on demographic information associated with a borrower's or applicant's surname and then updates that probability using demographic characteristics of the census block group associated with the borrower's or applicant's place of residence. The technique is called the Bayesian Improved Surname Geocoding method, or BISG.
Our resident expert on statistical analysis, Brandon Roberts, tells us that this proxy process is surprisingly accurate. So much so that it may be difficult to disprove the results that such a proxy process might produce. This makes it all the more likely that examiners will resort to this type of data analysis. And expanded HMDA and small business loan data reporting requirements will only make it more likely that examiners will resort to statistical analysis of your loan data when doing Fair Lending reviews. That makes it all the more important that you know well in advance of a Fair Lending exam just what your data has to say. Statistical analysis of loan data is rapidly becoming a fact of life for banks of all sizes.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
