Case Study: Fair Lending Analysis of Indirect Auto Lending

This report analyzed indirect auto lending practices for fair lending compliance. Regression analysis was used to assess pricing disparities between protected and non-protected groups (Black, Hispanic, Asian, Male/Female vs. Non-Hispanic White/Female) using proxy methods like Bayesian Improved Surname Geocoding (BISG) for race/ethnicity. The analysis considered loan data from over the period of a year, incorporating rate sheet criteria as controls. Results indicate some statistically significant pricing differences, prompting further review of specific loan cases.

The information in this case study is based on the provided report and is intended to illustrate Premier Insights, Inc.'s analytical capabilities. The names of the financial institution and market areas have been changed for confidentiality purposes.

Introduction

Premier Insights, Inc. conducted a comprehensive analysis of indirect auto lending practices at First Horizon Credit Union to evaluate potential disparities in loan pricing between protected and non-protected groups. This study aimed to assess the credit union’s lending performance in the context of fair lending regulations. The analysis was a follow-up to a similar analysis conducted earlier in the year, focusing on whether pricing differences exist for loans processed through the indirect lending channel. The analysis used the contract note rate as the measure of pricing.

Data and Methodology

The data analyzed included loan applications processed over a year period, totaling 6,650 loans. Because government monitoring information is not gathered for consumer lending, proxy methods were used to designate target and control groups. The study examined four target-control group combinations:

• Black – Non-Hispanic White
• Hispanic – Non-Hispanic White
• Asian – Non-Hispanic White
• Male – Female

Proxy Methods:

• Race and Ethnicity: The Bayesian Improved Surname Geocoding (BISG) method was used to identify race and ethnicity. This method used a joint probability based on the customer's last name and the racial/ethnic makeup of their residential area. BISG is a recommended approach by the CFPB. A description of the methodology as provided by the CFPB was included as an Appendix.
• Gender: Gender assignments were based on first names using name databases.

Descriptive Statistics:

The loan distribution by race and ethnicity was:

• White: 81.84%
• Black: 0.87%
• Hispanic: 15.24%
• Asian: 2.04%
• Other or Unclassified: 19.10%

The loan distribution by gender was:

• Male: 59.92%
• Female: 40.08%
• Unclassified: 7.10%

Regression Analysis:

Regression models were developed based on rate sheet criteria, including credit score, vehicle model year, and loan term. Two sets of regression estimates were used:

1. Rate sheet variables entered independently.
2. Rate sheet variables “interacted,” which effectively controlled for each cell on the rate sheet independently.

The regression analysis was conducted for all data. The regression tables showed the differences in pricing for each group, the direction of the difference (positive or negative), and whether the differences were statistically significant. A negative number indicated lower pricing for the target group, while a positive number indicated higher pricing. Statistical significance was determined with a p-value < 0.05.

Key Findings

The analysis revealed several statistically significant disparities in pricing:

• Black Borrowers: Black borrowers experienced significantly higher interest rates compared to non-Hispanic White borrowers. The difference in rates was statistically significant both before and after controlling for rate sheet criteria. For the overall data, the raw difference in average rate was 3.629, and after controlling for pricing factors the difference was 0.644. The p values were .000 and .012 respectively, indicating statistical significance.
• Hispanic Borrowers: Hispanic borrowers also experienced significantly higher interest rates compared to non-Hispanic White borrowers. For all the data, the raw difference in average rate was 1.886, and after controlling for pricing factors the difference was 0.305. Both values were statistically significant as indicated by the p values of 0.000.
• Asian Borrowers: The results were mixed for Asian borrowers, with some statistically significant differences when controlling for rate sheet criteria but not in all iterations. For all the data, the raw difference in average rate was -0.107 (not statistically significant). After adding controls, the difference was 0.348 (statistically significant).
• Male Borrowers: There were no statistically significant differences in pricing between male and female borrowers in any iteration.

Interacted Rate Sheet Criteria

• When the rate sheet criteria were interacted, the results remained significant for Black borrowers and Hispanic borrowers.
• For Asian borrowers, some statistically significant results were found when the rate sheet criteria were interacted, but not in all iterations. When all the data was analyzed, the raw difference in average rate was -0.107 (not statistically significant), while the difference after adding controls was 0.517 (statistically significant).
• There were no statistically significant differences in pricing between male and female borrowers when the rate sheet criteria were interacted.

Follow-Up Review Samples of loans where disparities remained after controlling for rate sheet criteria were provided for review. These samples consisted of target and control group observations which were not explained well by the model estimates.

Conclusion

The analysis revealed statistically significant disparities in pricing for Black and Hispanic borrowers in First Horizon Credit Union’s indirect auto lending program. These findings suggested a potential need for the credit union to review its lending practices to ensure fair lending compliance. The application of regression analysis and the use of proxy methods allowed for a detailed and nuanced understanding of the potential disparities in pricing that were present in the dataset. These findings provided critical insights for First Horizon Credit Union to address and remediate their lending practices.

Premier Insights, Inc. Expertise

This case study demonstrates Premier Insights, Inc.’s expertise in conducting comprehensive fair lending analyses. Our approach provides financial institutions with the necessary insights to understand their lending practices and ensure compliance. With our expertise, we can help financial institutions identify areas for improvement, mitigate risk, and promote fair lending practices.