Underwriting
Relative to mortgage lending, underwriting in the indirect automotive space is often less automated (since there is no industry
standard automated underwriting model similar to those used in
the mortgage lending business) and, therefore, more judgmental.
Each indirect automotive finance company develops customized
scorecards and underwriting criteria. Understanding these scorecards, underwriting policies, and areas of discretion is a necessary
starting point for conducting a fair lending analysis.
Also common in the automotive finance industry are counter
offers, which may involve reducing the loan-to-value (LTV) or
payment-to-income (PTI) ratios, shortening the term of the
loan, or requiring a larger down payment. The dealership has a
number of alternatives for adjusting the deal to meet the counter
offer. Those may include offering a higher trade-in value, reducing
the price of the vehicle, or changing the composition of add-on
products. The treatment of these counter offers in fair lending
underwriting models is often driven by limitations in the storage
of relevant electronic data in the underwriting system.
Conceptually, counter offers can be thought of as two applica-
tions. The first, as submitted by the dealership, may be considered
a declined application. The second may be considered as the terms
upon which the bank or finance company was willing to purchase
the contract. How these are analyzed depends on which data fields
the underwriting system stores. Optimally, an underwriting system
stores all of the relevant information submitted by the dealership.
Once the appropriate treatment of counter offers has been de-
termined, a fair lending analysis of underwriting can be conducted
to identify meaningful differences correlated with race, ethnicity,
gender, or age. If regression analysis is used, the factors included
in a regression model should be based on the bank’s underwriting
policies and procedures. Common controls include deal specific
attributes (e.g., age of vehicle, LTV, PTI, rebate), applicant credit
worthiness and stability (e.g., FICO/customer credit score, pay-
ment history, time in job, own versus rent), and dealership specific
attributes (e.g., recourse versus non-recourse, 6 relationship with
bank, performance of contracts previously assigned to the bank).
Where statistically significant differences exist, manual file
reviews of “matched pairs” are often used to further analyze the
circumstances surrounding the applications in question. 7 Often,
the manual file review identifies relevant information that was
not available electronically and could not easily be incorporated
in the regression model.
Underwriting analysis should also consider the risks associated
with custom scorecards. The highly competitive auto market has
led to the desire to obtain better credit predictions. Numerous third
parties are marketing sophisticated, predicative tools that may include
non-traditional credit factors. When these are used as a “black box,”
with the bank having little, if any, information on how the models
were developed and which factors they include, fair lending risks
increase when expressly prohibited attributes or factors highly
correlated with prohibited basis are factors within the “black box.”
Pricing
The CFPB’s March 21, 2013 bulletin has focused attention on the
pricing of indirect automotive contracts. The use of judgmental
or discretionary pricing by the dealership or the finance company
presents heightened fair lending risk, with many issues unique to
the pricing of indirect automotive contracts. The contract rate paid
by the buyer is a combination of two components: the wholesale
rate and the dealer reserve. The CFPB’s analytical framework
examines these two components separately.
The wholesale rate is commonly called the “buy rate” and
represents the minimum rate at which the finance company is
willing to purchase the contract. It is generally agreed that the
buy rate is a risk-based price, 8 meaning that fair lending analyses
will include controls for the credit worthiness of the buyer. In
order to apply the correct analytical framework, one must first
understand the process by which the bank established the buy
rate. In the indirect automotive market, there are a number of
ways to establish buy rates. Many finance companies publish rate
sheets for the dealerships from whom they purchase contracts.
These rate sheets outline buy rates available to the dealership for
different types of contracts and often vary by credit score ranges,
age of vehicle, length of contract, LTV, and loan amount, as well
as dealership specific factors. For example, dealerships that have
a floor plan line or other commercial loan arrangement with the
bank or finance company may receive lower buy rates.
In some cases, the rate sheet reflects set rates, but in others
they represent the starting point for a negotiation between the
dealership and the finance company. Once this process is understood, a regression analysis may be completed to look for any
meaningful differences in buy rate correlated with a prohibited
basis. The controls included in a buy rate regression should reflect
the key factors used to set the buy rate as reflected in the bank’s
rate sheets and other documentation. Statistically significant differences may be further analyzed using a manual file review of
matched pairs. This may be particularly relevant when the bank
used a relatively judgmental process to establish the buy rate. For
example, additional analyses may examine the extent to which
buy rate exceptions were granted by race or ethnicity.
Unexplained differences in buy rates correlated with a prohibited
basis may be interpreted by regulators as disparate impact created
by pricing models and/or the unfair application of discretion.
While the CFPB may be focused on dealership discretion, it is
shut
te
r
stO
ck
