Fair lending compliance in indirect auto lending continues to grab the attention of both regulators and industry leaders. Here are 9 steps to take for effective indirect auto data analysis.
In the United States, automobiles are vitally important to consumers, which means the regulators are paying close attention to automobile financing.
Auto loans are the third-largest source of outstanding household debt after mortgages and student loans, according to the CFPB. The Federal Reserve estimates that 86 percent of US families have an automobile. There were approximately 58 million cars sold in the United States (~16m new car sales and ~42m used car sales), according to the 2014 analysis by National Independent Automobile Dealers Association (NIADA). 85 percent of new autos and 54 percent of used autos are financed, according to Experian's Q2 2014 State of the Auto Finance Market report.
For example, the CFPB has said that they are conducting “ECOA Targeted Reviews” of financial institutions that practice indirect auto lending. In those reviews, the Bureau says that they may conduct fair lending statistical modeling and analysis using the institution’s indirect auto lending data.
According to the CFPB, the targeted reviews assess “underwriting, buy rates, dealer markup and compensation policies and practices for ECOA compliance.” These reviews evaluate policies and procedures as well as various aspects of the Compliance Management System, including internal controls and monitoring.
The CFPB team considers the following in their targeted reviews:
“While fair lending analyses of mortgage lending are simplified by the availability of lender data reported under HMDA, this is not the case with indirect auto lending. Information on race, ethnicity, and gender is typically not permitted to be collected as part of an auto lending transaction," according to the CFPB's Fair Lending Report. "Therefore, the Bureau [and other regulators] use proxy methodologies to differentiate among consumers based upon these characteristics."
The CFPB's proxy analysis for indirect auto lending uses publicly available Social Security and Census data.
Regardless of your regulator, financial institutions should be paying close attention to fair lending compliance for indirect auto lending, especially when the size and importance of the auto finance market is taken into account.
Banks were responsible for 35 percent of new and used originations in Q2 2014, captive auto 27 percent, credit unions 17 percent, finance companies 14 percent, and Buy Here Pay Here dealerships 8 percent, according to Experian.
Here are the 9 steps that all financial institutions should employ when it comes to analyzing their auto data:
It all starts with performance context. You need to understand how your institution operates. Without this fundamental understanding of the business, it's very difficult to analyze any type of data.
Questions to Consider: How are the indirect relationships set up? Many financial institutions have different structures and approaches. Does it make sense to aggregate data? Should you analyze the data in unique groups based on business model? Without this fundamental understanding of the business, it is very difficult to draw meaningful conclusions about your data.
Based on the business model(s), determine what risks you're looking to analyze. With fair lending reviews, two of the most common analyses include a review of pricing risk (e.g. disparities in price and denial rates, and underwriting risk). Product mix can also be an important part of your analysis, (e.g. steering risk).
Questions to Consider: What layers of pricing are currently used (e.g. buy rate, mark-up and/or contract rate)? Do you truly understand the underwriting process and how the systems and workflow operate? Do you understand how the system data maps to the initial buy rate, dealer mark-up and contract rate?
After you've decided what risks you'll be analyzing, you'll need to determine the key variables needed to understand the risk. Underwriting and pricing are the two most common areas to analyze, so start by identifying the key variables your organization uses to underwrite and price. Some of the more common variables, or data fields, used include: Credit Score or Credit Quality Factor, Down Payment or Loan-to-Value, Payment-to-Income, and Product Type (New vs. Used).
Questions to Consider: What decisioning factors does your underwriting process include? Do you employ loan level pricing adjustments? What other variables are used in pricing and understanding?
This is one of the more passionately discussed topics in compliance today. Some analyses employ the Federal Reserve's process of reviewing gender and ethnicity based on look-up tables. Others rely on the CFPB's process that assigns race and ethnicity based on probabilities, using demographic characteristics of the census block group associated with place of residence (this method is referred to as Baynesian Improved Surname Geocoding, or BISG).
Questions to Consider: Do you know which of these methods makes the most sense for you? Is there a different approach that works for your specific market areas? Can you support the chosen proxy methodologies with logic?
Conduct a cursory review of your data and look for anomalies and missing data fields. Unlike HMDA data, indirect auto data has not been scrubbed and typically requires a thorough review. This step is critical. The law of GIGO (garbage in, garbage out) will prevail; invalid data on the front end will result in an invalid output.
Questions to Consider: Are there any 999 credit scores? What percentage of the fields are completed? In short, is the data strong enough to analyze? If not, what can you do to make it stronger?
We recommend that financial institutions look to understand both the aggregated-level data as well as the individual dealer-level data. The aggregated view will give you a macro perspective to determine if there are concerns within your organization. The dealer-level analysis will help you determine if those concerns are systemic, or related to a specific dealer or group of dealers.
Questions to Consider: What types of relationships exist with your dealers? Are they all the same?
Once you have confidence in the input data, the real analysis can begin. To identify potential disparities, compare prohibited basis groups to the control group. Your control group will be those individuals who are least likely to be discriminated against. Some examples include: Female vs. Male (Control); Hispanic vs. Non-Hispanic (Control); and Black vs. Non-Black (Control).
Start with simple comparisons, like comparing average pricing (e.g. buy-rate, mark-up, and/or contract rate) by credit score band. If your data is large enough, you may need to conduct more sophisticated analysis, like regression. With regression, you can include individual credit characteristics in order to compare similarly situated individuals.
Questions to Consider: Do you have disparities between groups? Is the data statistically significant?
When disparities exist, it is important to determine whether the disparities are statistically significant.
Questions to Consider: Are the disparities statistically significant? Are there logical reasons for disparities? Is additional analysis needed (e.g. regression)? Is comparative file review in order?
It is very likely that you will find disparities in the indirect auto data that you examine. A fair lending compliance program does not rely on a one-time analysis. By regularly conducting data analysis, you'll get snapshots showing how your data changes over time. Those snapshots will help determine if there is a pattern or practice that is worthy of additional attention.
Questions to Consider: Do you know if identified disparities are anomalies? Do some disparities appear consistently? Do you have a process for comparing historical to current data?
Every organization is unique, therefore your analysis strategy will have unique focal points. The steps you take to analyze the data will be based on your business model and institution structure, which will help determine focal points. In addition, the intensity of your analysis will be determined by some of the following factors:
The auto finance market is a big part of a consumer's life. At Ncontracts, we believe in the importance of auto finance and the essential role financial institutions play in providing access to consumers. We know that most organizations don't intend to discriminate, and want to create a fair marketplace. Knowing what your data says about you can help protect your institution, and relieve some of the stress and pressure associated with compliance.
Ncontracts helps financial institutions choose the analysis path that makes the most sense for each unique organization. If you have questions, just give us a call. We'll be happy to help you find the path that makes the most sense for you.