Member login
American Financial Services Association

Disparities in Indirect Auto Loans

Disparities in Indirect Auto Loans

Recent research by economist Jonathan Lanning, formerly of the Consumer Financial Protection Bureau (CFPB) and currently with the Federal Reserve Bank of Chicago, documents racial and ethnic disparities in indirect auto loans.

Using CFPB data on more than 7.5 million loans from 2008 to 2013, Lanning estimates Black borrowers pay an average of 2.1 percentage points more on a dealer reserve than non-Hispanic White borrowers, while Hispanic and Asian and Pacific Islander borrowers pay 1.8 and 1.1 more respectively than non-Hispanic White borrowers. Further statistical analysis leads Lanning to attribute the estimated disparity in dealer reserves between Black and White borrowers to racial prejudice.

The author’s “threshold” result of the presence and extent of racial and ethnic disparities is at odds with earlier research conducted by Charles River Associates (CRA) on behalf of AFSA. CRA’s work highlights the complexities of the indirect auto lending market and identifies several important methodological considerations for estimating such disparities, that were they incorporated by the newer study might have led to more accurate insights.

First, since dealers and lenders are prohibited by law from collecting racial and ethnic information, studies of discrimination must necessarily find other ways to accurately account for race and ethnicity of the borrower. To do this, Lanning employs the BISG procedure, which uses data on surname and geographic location. However, CRA notes that BISG and other proxy methods can be flawed. “So, while BISG may be relatively less inaccurate than proxies based on geography or surname alone, BISG remains subject to significant biases.”

Second, Lanning’s threshold result comes from a simple model relating dealer reserve to race and ethnicity. (A companion set of results also identifies the presence of racial and ethnic disparities in most, but not all, broad geographic regions.) Controlling for geographic differences in greater detail, ideally at the metro area, might account for differences in local market characteristics. Other factors such as loans for a new or used vehicle, loan terms, and credit characteristics of the borrower should also be controlled. Although the author examines many of these factors at later stages of his research, they are not considered in finding that disparities exist in the first place.

Third, CRA recommends excluding loans with certain characteristics from the estimation sample, including loans with zero dealer reserve or those from dealers with zero variance in reserve across borrowers.

Incorporating these recommendations leads to a rather different conclusion. Indeed, CRA estimates a model using data on nearly 4.3 million loans from 2012 and 2013 and finds that “…appropriately considering the relevant market complexities and adjusting for [BISG] proxy bias and error, the observed [racial and ethnic] variations in dealer reserve are largely explained.”

Finally, the author’s sample of loans is from the years 2008 to 2013. A timeframe that includes a period of significant disruptions in credit markets and a substantial downturn in the broad economy and the auto market. The sample window closes a decade ago, during which time the structure of the indirect auto lending market has likely evolved.

Monitoring for and identifying illegal discrimination in indirect auto lending is of critical importance. Such efforts would benefit from addressing the issues identified in the CRA study, biases or inaccuracies due to sample selection, and from an up-to-date analysis using more recent data.

March 8th, 2023 by

Recent Posts