[This is the 3rd and final installment of a multi-part series on voluntary protection products. You can see previous posts Blog 1, Blog 2 here]
Introduction
If auto lenders weren’t concerned about fair lending risk in their portfolio before now, President Biden’s recent Executive Order on the AI and other Federal Agency announcements have put discrimination right at the top of the agenda.
So how do auto lenders manage this risk? One area of particular complexity relates to the sale of voluntary protection products (VPPs). VPPs provide additional coverage and protection for certain vehicle components or services beyond the manufacturer’s original warranty.
VPPs are complex because lenders don’t sell these products; dealers do. So in order to manage bias, lenders need to put in guidelines on what products are permissible.
To better understand what is going in the marketplace, Informed looked at the demographics of how these products are sold.
Here is what we found: The study found that VPP purchases were relatively constant among racial and ethnic groups, but we did see variations in purchase practices related to age and gender.
THE STUDY:
Informed looked at an anonymized and deidentified data set of more than 300,000 auto purchases involving loans from the first nine months of 2023 (January 1st to September 30th) from a cross-section of auto lenders of varying size and customer profile.
The voluntary protection products we examined include:
- Anti-Theft and Theft Protection Services
- Appearance Protection Plans
- Bundled Products
- GAP Insurance
- Maintenance Plans
- Paintless Dent Repair Contracts
- Key Replacement Contracts
- Tire and Wheel Plans
- Windshield Protection Contracts
It should come as no surprise that different groups have different purchasing needs and preferences when visiting a dealership. Our key findings are below.
Retirees Purchases VPPs at Almost 3x Than Others
Retired individuals purchase more Voluntary Products than employed Individuals. At an aggregate level, retired individuals are three times more likely to purchase more than one of the Voluntary Protection Products described above. When broken down by specific product categories, retired people purchase products focused on safeguarding their vehicles.
Women are 24% more likely to purchase a VPP than Men
For every product category in our data, females were more likely to purchase the product than males.
VPP Purchase Practices Were Relatively Constant Across Racial and Ethnic Groups
When looking at VPP purchasing practices across racial and ethnic groups, we did not see much variation in purchasing practices. Non-Hispanic whites purchased VPPs 10.11% of the time. Hispanics and African Americans purchased multiple products slightly less frequently, and Asian Pacific Islanders purchased these products slightly more often.
GAP Insurance Products are the Most Commonly Purchased Voluntary Product
We found GAP Insurance was by far the most common VPP that consumers purchased. 33% of all auto loan applications in our analysis contained a GAP Insurance Contract.
Conclusion
The data sheds light on consumer preferences and priorities when it comes to safeguarding and maintaining vehicles. We found that consumers purchase multiple voluntary products in 8.4% of vehicle deals.
Consumers purchased GAP products (including GAP waivers) more often than any other product – between 25% and 46% of the time. GAP Waivers and Insurance are products that protect consumers from being accountable for the balance of their loan if their vehicle is totaled or stolen. These differences may reflect varying levels of access, opportunity, or experience in the lending process.
While we didn’t see industry-wide problems of bias in VPP products, results can vary company by company. And with AI, companies now have the capacity to intelligently monitor their portfolios. Companies who wish to avoid regulatory scrutiny would be well advised to monitor the bias trends in their VPP product sales and prices. We believe strongly that companies leveraging digitization and AI technologies in lending can be more effective at reducing bias than those using older technologies.
Methodology
To predict race and gender for fair lending analysis, we employed open-source algorithms such as the Zest Race Predictor, Bayesian Improved Surname Geocoding (BISG), which is used by the CFPB for assigning race, and the Social Security Administration’s Name Gender Lookups. We also leveraged our internal datasets. To reduce noise in the analysis, we narrowed down the dataset to focus on applications primarily intended for auto loan verifications, thereby minimizing extraneous information. These internal datasets helped us gain insights into the age and employment types of the applicant pool.
In our analysis, we established target groups and defined comparison groups as follows:
- Race and Ethnicity: Non-Hispanic White applicants serve as the comparison group for African American, Hispanic, Asian, and American Indian/Alaskan Native applicants.
- Gender: Male applicants serve as the comparison group for female applicants.
- Age: Non-Senior applicants serve as the comparison group for senior applicants.
- Employment Status: Employed applicants serve as the comparison group for all Employment Types.
By establishing these comparison groups, we were able to conduct a comprehensive analysis of how various factors, such as race, gender, age, and employment status, impacted VPP purchases.
Our analysis uses the metric of odds ratios. The odds ratio compares the likelihood of a test group purchasing a product versus the control group. An odds ratio greater than 1 indicates that the test group is more likely to purchase the product, while a ratio less than 1 suggests that control groups are more likely.
