Every creditor with sub-prime customers has experienced—on one occasion or another, we expect—someone judging it unfavorably for allegedly taking advantage of consumers by charging higher interest rates. That “tsk-tsk” reaction might come from a consumer advocate who is unschooled in the realities of credit losses, servicing costs, and other factors that increase the cost of extending credit to higher-risk borrowers. It might come from members of Congress who believe that a certain APR should be enough for even the greediest creditor. Sometimes it comes from a government regulator who should really know better. So, what does the Consumer Financial Protection Bureau think? A recent report gives a clue.
Most creditors price credit using models that predict default risk based on experience data and facially objective creditworthiness criteria. The more creditworthy the applicant, the better the terms of credit that applicant can typically expect. And, of course, the reverse is true—the riskier the applicant’s credit profile, the more the creditor will price that risk into the offer of credit, and the more expensive the credit will be. Creditors use complicated formulas based on data they have collected about default rates to try to predict how likely consumers with certain credit profiles (e.g., credit score, income, homeownership, etc.) are to default on a vehicle-secured credit obligation, and they use those formulas to price credit for consumers according to the risk of default.
In September, the CFPB issued an interesting Data Point, “Sub-prime Auto Loan Outcomes by Lender Type” (note that the CFPB uses the term “loan” loosely in the Data Point to refer to any vehicle finance credit transaction, whether a direct loan or a dealer-originated retail installment sale contract). In the Data Point, the CFPB noted that sub-prime credit pricing varies across creditor types, and it wanted to review how much the variation in credit pricing was attributable to relative default risk. The CFPB acknowledged that risk-based pricing models are built to recognize some likelihood of default. The study looked at how impactful the rate of default experienced by a creditor was on the finance charge rates for the accounts on which the CFPB had data. In other words, does the increased risk fully justify the higher cost? If not, does this fact confirm the cynical view that sub-prime creditors are exploiting the vulnerabilities of less creditworthy consumers? And beneath the surface lies the question the CFPB may yet try to answer: does more expensive credit for sub-prime consumers lead to higher default rates, and might those consumers default with less frequency if they paid lower finance charge or interest rates?
The short answer we get from the study is that, based on the information in the CFPB’s model, the increased risk of default does not, on its own, fully explain why some sub-prime customers pay higher rates than others. But that apparent result was qualified in the report, as it should be: first, the model the CFPB used to conduct the study was missing some information critical in credit underwriting; second, there are differences in the business models of various creditors that can further explain the differences in credit pricing in ways the statistical model cannot capture. So, readers should resist the temptation to focus on the notion that credit risk does not fully explain credit pricing, without taking into account the limitations in the model that the Bureau acknowledged.
The Data Point also looked at pricing across creditor type, identifying some of the apparent differences among creditors and what we generally know motivates their business models. For example, the CFPB noted that captive non-bank vehicle finance companies that have corporate relationships with vehicle manufacturers tend to price credit rather favorably, even at a 0% finance charge rate, because their primary motivation is to facilitate the sale of new cars. Other creditors do not have an incentive to finance car sales at extremely low rates, so their rates tend to be higher. These business model differences are important because unless a statistical study can control for those differences, the differences may actually inform results in ways that the study does not capture.
The CFPB noted in the Data Point that while default risks and default rates might be similar among creditor types based on objective criteria, like FICO score, the cost of credit for sub-prime customers can differ across creditor types, such that even though the credit default risk is the same for two sub-prime customers with identical credit profiles, those customers could pay finance charge rates that differ by several percentage points depending on their sub-prime financing source. For the sake of consistency in the Data Point, the CFPB examined the likelihood of a consumer defaulting within the first three years after consummation.
Let’s review some of the details of the Data Point more closely.
What the CFPB studied. The CFPB used a robust data set of de-identified account-level information for six million accounts originated between 2014 and 2016 (to avoid life-of-loan data being complicated by the impact of the COVID-19 pandemic) obtained from the national consumer reporting agencies. Those consumer records included important information about the consumers—whether they had mortgages, credit cards, student loans, etc., where they were located, their credit score, and their birth year.
Section 3 of the Data Point provides intricate details of the data set and its limitations. One limitation worth noting, though, was the absence of information in the data set about the vehicle the consumer bought with the credit being studied. The CFPB correctly identified that vehicle information, along with information about down payments and origination fees that can impact the cost of credit, could have further explained pricing discrepancies across creditor types such that apparent differences in pricing across creditor types may have diminished or gone away. The CFPB also noted that not all vehicle finance creditors furnish account data to the consumer reporting agencies, so there was less information available from small buy-here-pay-here dealers than there was for banks and credit unions. But data set limitations are inherent in statistical work, and the CFPB was forthcoming about those limitations in this study.
What the CFPB found. The CFPB identified things we may have already known—like banks, which do not rely primarily on income from the vehicle finance business and which tend to extend credit to less risky customers, charge finance charge or interest rates that are lower on average than buy-here-pay-here dealerships and certain finance companies by 25-50%.
Therefore, the cost differences are not really apples-to-apples comparisons. The nonprime customers of banks and credit unions are often limited to “shallow” sub-primes with credit scores that approach the higher end of typical sub-prime customer categories.
Finance companies’ customers often include more “deep” sub-prime customers with lower credit scores.
The CFPB found that “the likelihood of a sub-prime auto loan becoming at least 60 days delinquent within three years is approximately 15 percent for bank borrowers and between 25 percent and 40 percent for finance company and buy-here-pay-here borrowers.” The CFPB also found that the difference in default risk across creditor types was not able to fully explain the difference in pricing across creditor types, based on its model and the information it had. And, as it expected, the CFPB found that customer default rates were higher with sub-prime creditors that charged higher rates.
The Data Point highlighted that creditors focus on differing customer bases: banks and credit unions are focused on their customers to whom they can offer multiple financial services, captive finance companies operate largely in the new car buyer market, and buy-here-pay-here dealers tend to focus on credit-challenged consumers in sometimes rural or otherwise underserved communities. In the Data Point, the CFPB identified some overlap in consumer characteristics across creditors that typically extend credit to sub-prime consumers (NOTE: for purposes of the Data Point, “sub-prime” means a credit score of 620 or less), noting that some sub-prime consumers who might have qualified for credit with a bank or credit union instead financed their cars at a buy-here-pay-here dealer and paid more in finance charges but had the same default rate. In fact, the CFPB observed that consumers whose relatively high (but still sub-prime) credit scores put them in the 90th percentile of buy-here-pay-here dealers in terms of creditworthiness had credit scores that were about 60 points higher than the credit scores of customers in the bottom 10th percentile of bank customers.
The observations about the differences among customer bases, and the corresponding relationships different kinds of creditors have with their customers, were interesting to note, and the CFPB acknowledged in the Data Point that those differences may or may not inform credit pricing in ways the study could not measure. But, based on the finding that consumers with similar credit profiles have similar likelihoods of defaulting within the first three years without regard to their creditor type, the CFPB observed that the difference in the risk of default alone did not explain the differences observed in average pricing of credit across creditor types.
What else was missing, and why it matters. Ultimately the CFPB identified several other factors that may help better explain the differences in credit pricing in addition to those noted above—specifically, differences among consumers, such as their income, their access to information (like bank financing rates), and their financial sophistication (e.g., understanding that the local dealer is not the only place where the consumer can buy and finance a car and that some aspects of vehicle financing are negotiable), and creditor-related differences, like underwriting capacity and sophistication, relative costs of repossession and collection, and relative ability to absorb credit losses. In the Data Point, the CFPB did not fault more expensive creditors for the differences in pricing; it just noted the differences and suggested future research that might help it better understand how to educate consumers about options in vehicle financing.
The absence of certain important information is critical to the reliability of a statistical study. It is possible that when you control for data points not included in the CFPB’s analysis, the apparent differences in pricing go away. For example, customer income is an important driver of credit pricing because it informs the likelihood that the customer will be able to repay the debt for which she is applying. While the CFPB’s analysis was not able to include that data, further study may reveal that the income variable helps more fully explain the differences in pricing across creditor types because consumers whose profiles look similar based on the data available may not be as similar when you account for their income and other important factors. A model that includes more consumer-level and/or transaction-level (e.g., vehicle) data might not yield the apparent pricing differences the CFPB identified in this study that did not include such data (like income).
In the Data Point’s conclusion, the CFPB suggested further and more expansive research to better understand consumer behavior when it comes to financing vehicle purchases. That research would presumably pick up some of the factors that go into consumer and creditor decision-making that did not make it into the model used in this Data Point. It is important to note that the CFPB did not draw any conclusions from this research, except that more research was necessary to reach meaningful conclusions.
A careful reader of this study will appreciate its limitations. It is likely, however, that many readers will draw the unfounded conclusion that certain sub-prime auto creditors charge customers more than these consumers’ risk warrants and thus “gouge” them unfairly. We will watch for and report on any follow-ups to this Data Point in future issues of Spot Delivery.
Charles F. Dodge, Jr., is a partner in the Maine office of Hudson Cook, LLP. Chuck can be reached at 207.210.6825 or by email at [email protected]
L. Jean Noonan is a partner in the Washington, D.C., office of Hudson Cook, LLP. Jean can be reached at 202.327.9700 or by email at [email protected]
Copyright © 2021 CounselorLibrary.com LLC. All rights reserved. This article appeared in Spot Delivery®. Reprinted with express permission from CounselorLibrary.com.