Every year, millions of American families buy a car – and it will be one of the most significant purchases they make.
One key priority for us is protecting consumers from the silent pickpocket of discrimination. Discriminatory markups in auto lending may result in tens of millions of dollars in consumer harm each year. The average loan for a new car is up to $26,691, so a higher interest rate can make the total cost of the car much higher.
In March, we released a bulletin to help lenders that offer auto loans through dealerships make sure they are following the law. That bulletin explained that so-called “dealer markup” policies that give dealerships discretion in what interest rates to charge consumers and that create incentives for charging higher interest rates may be implemented in a way that violates the law. Research indicates that lenders’ markup policies may lead to minorities being charged higher markups than other, similarly situated, white consumers.
The auto bulletin indicated that we are engaged in the same type of fair lending analysis and scrutiny that our fellow regulators and the Department of Justice have engaged in for many years. In addition, responsible auto lenders have regularly engaged in similar analyses to monitor their own lending practices for compliance with the law.
We know that many lenders are committed to fighting unlawful, discriminatory practices and creating a fair marketplace for all consumers. In the mortgage market, laws and regulations require most lenders to collect and report demographic information about their borrowers so that they and their regulators can analyze which mortgage loans are made or denied and how they are priced, for potentially discriminatory patterns.
However, auto lenders and other non-mortgage lenders are not generally allowed to collect demographic information. Since they don’t collect this data, they use various approaches to make sure they are being fair to their customers.
Let’s say a responsible auto lender wanted to make sure that their female customers are not paying more for a loan than similarly-situated men. Before analyzing the pricing patterns, the lender needs to calculate the likelihood that a borrower is male or female. Without actually recording the gender of each borrower, to substitute, or “proxy,” for gender, responsible lenders often rely on a from the Social Security Administration. The public database contains counts of individuals by gender and birth year for first names occurring at least five times for a particular gender in a birth year. Using statistics, they can determine a probability that a particular applicant is male or female based on the distribution of the population across gender categories for the applicant’s first name.
There are a greater variety of methods to proxy for race and national origin. One method used by lenders to check the probability that an applicant is Hispanic or Asian is to use the published by the Census Bureau, in which the Census Bureau reports, by race and national origin, the percentage of individuals with a given surname. Another method to proxy for race and national origin uses the demographics of the census geography (e.g., census tract or block group) in which an individual’s residence is located, and assigns probabilities about the individual’s race or national origin based on the demographics of that area as reported by Census. This method is also used to proxy the probability that an applicant is African American, and it can be used to proxy for other racial and ethnic groups as well.
The role of regulators
The CFPB and other agencies are charged with making sure lenders are following fair lending laws, whether those lenders are engaged in mortgage lending or other types of consumer lending. For auto and other types of non-mortgage lending, the federal regulatory and enforcement agencies typically engage in similar analyses that use a variety of proxy methods, often drawing from the same public databases used by responsible lenders. Our method integrates two common approaches by combining the respective probabilities generated by the last name and geographical proxies. has found that this approach produces proxies that correlate highly with self-reported race and national origin and is more accurate than relying only on the borrower’s last name or geographic location.
Statistical methods are often refined over time. We are committed to staying in dialogue with our sister agencies, lenders and researchers to refine our proxy methods over time, so that we can stop the silent pickpocket of discrimination in various consumer finance markets.