For more than 25 years, a vigorous debate has existed over whether non-White motor vehicle buyers are illegally discriminated against in pricing credit. Both the facts and the law are complicated.
When we consider the separate question of whether finance sources that buy dealers’ retail installment sales contracts should be liable for discrimination—if indeed discrimination exists—the issues get really complicated.
The focus has been on the amount the customer pays above the finance source’s wholesale buy rate, called “rate spread,” “dealer participation,” or “markup”—different terms for the same thing.
We will use the term rate spread. (But language matters. Consumer advocates prefer to use markup, arguing that the buy rate is the rate the consumer “qualifies for” and implying that any higher amount is price gouging.)
An economist formerly at the Consumer Financial Protection Bureau (CFPB), Jonathan A. Lanning, began researching the question of whether data on millions of transactions supported a claim of discrimination by dealers and, if so, what motivated it.
Spoiler alert: Lanning concluded illegal discrimination by dealers exists and is caused by dealer employees’ prejudice against Black and non-White buyers.
Lanning first published his research online in 2021, after he left the CFPB for a position at the Federal Reserve Bank of Chicago. In January, he published a summary of his research in an FRB-Chicago online publication, Profitwise News and Views. Here is an overview of his conclusions: Black, Hispanic and Asian buyers pay substantially higher rate spreads than non-Hispanic Whites, causing these minority consumers to pay hundreds or thousands of dollars more in interest rates than their White counterparts. The cause of these higher rates is likely to be auto dealers’ racial animus or, in other words, pure prejudice.
Wow.
Is this good research? I expect this question to be debated for some time. We like to think the answers should be clear. This reminds me of an old joke about the economist who was asked, “How much is 2 + 2?” The answer: “It depends on your assumptions.”
That punch line, on this question, has quite a bit of truth to it.
My thoughts on the Lanning research fall into two categories:
- Positions that seem odd.
- Positions that seem interesting.
Let’s look at both.
Some points seem odd or simply wrong. For instance, Lanning rejected the possibility dealers may vary their compensation based on their workload to get a customer approved for credit. Some customers are a shoo-in for approval—those with good credit, low debts and high, verifiable income. Others will be approved only with stipulations requiring a lot of work and may ultimately not be satisfied, costing the dealer time and perhaps no sale.
Lanning says “there is little evidence” rate spreads vary in proportion to the work required. Why? Because finance sources “often disallow or substantially limit markups on subprime” contracts.
This fact does not “contradict this justification,” as Lanning says. Though some finance sources do not allow the maximum rate spread on subprime contracts, dealers are still motivated to get the greatest compensation available for customers, which will require a lot of work.
Only some finance sources limit rate spread based on credit tier; others do not. Creditors limiting rate spread do so to help customers who are already paying the highest rates. This means dealers are also doing more work for often the most time-consuming applications but are being compensated at a lower rate.
Just because dealers accept this reality does not negate the extra work required nor prevent them—when they can—from charging higher rate spreads to receive fair compensation.
Here is another head-scratcher. Lanning (correctly) says illegal discrimination occurs when the rate spread “paid by two otherwise identical buyers differs based on a characteristic such as race.” He makes clear, though, he did not control for all the factors that distinguish buyers. Lanning concedes he controlled only for prohibited characteristics and not for the many other important ways in which buyers differed. This is the biggest weakness of his research.
Think about it. He is essentially arguing all consumers of the same race or ethnicity are “identical.” Nothing except their race or ethnicity matters.
Lanning discusses how Bayesian Improved Surname Geocoding (BISG) proxies’ accuracy depends on how similar buyers using auto credit must be to general population-level data. He notes the financing rate appears similar across racial and ethnic groups and says “there is little room for selection bias.” However, “similar” does not mean “the same.” If White and non-White consumers seek credit at different rates, shouldn’t the proxy methodology take the difference into account? With disparities of only a few basis points, slight differences in proxy accuracy can matter.
Other points seem interesting and even support claims industry members have made to the CFPB for years without gaining much traction. The first is a table showing wildly different average disparities by race in various parts of the country. The swing can be half a percentage point or more across the nine Census divisions.
If accurate, what accounts for this enormous variation, and how we should be controlling for geography? Would looking at smaller geographic divisions provide even more insight?
On this issue, Lanning presents data showing most rate spreads in his dataset of over 7 million contracts are either at the maximum or at zero, and Blacks are more likely than Whites to pay the most. This does not necessarily prove an individual dealer’s Black customers are more likely than Whites to pay the most.
I often see dealers who consistently charge all, or almost all, their customers the same rate spread, regardless of race. This fact suggests we should be using a dealer pricing practice control in our analyses.
In short, these findings indicate the importance of controlling for factors in which buyers are different in ways that can matter. This is consistent with Lanning’s definition of discrimination, but his former colleagues at the CFPB do not seem to agree.
Another interesting point is whether considering competitive offers can justify rate spread differences by race or ethnicity. The Department of Justice has sensibly held that lowering a rate spread to meet a competitor’s offer is legitimate. The NADA Fair Credit Compliance Program reflects this position. Some have challenged this view, suggesting it legitimizes discrimination because some groups may be inherently better than others at comparison shopping or negotiating better rates.
I have seen no evidence to support this concern. Lanning has not seen it either; he says his research shows Black buyers are just as good as Whites at negotiating the price they will pay, and that fact indicates Blacks and Whites must have similar negotiation abilities.
Lanning’s work illustrates we have come a long way from believing that simply comparing average rate spreads by population group can prove illegal discrimination. For all the sophistication in some parts of Lanning’s work, he inexplicably fails to adopt other methodologies shown to largely explain observed racial and ethnic variations in average rate spreads.
Identifying illegal discrimination is far too important to rely on flawed methods. This research demonstrates the value of some of Lanning’s insights, but also—and perhaps unintentionally—shows us ways to investigate discrimination better.
Jean Noonan is a partner in Hudson Cook’s Washington, DC office. She advises clients on consumer financial services, fair lending, marketing, financial privacy, and consumer protection matters. She counsels financial institutions and others in complying with laws related to consumer credit, privacy, telemarketing, and unfair trade practices. Noonan is a former executive with the Federal Trade Commission.
