On April 2, 2012, the Justice Department’s Civil Rights Division (“Justice”) filed suit against GFI Mortgage Bankers, Inc. (“GFI”) alleging that GFI charged African American and Hispanic borrowers higher rates and fees solely based on their race. Previously, in January 2010, the Department of Housing and Urban Development found a “pattern or practice” of discrimination at GFI and referred GFI to the Department of Justice.
GFI used a web-based mortgage pricing software called Optimal Blue, which had been “customized” for GFI’s products and pricing strategy. The individual lender would select loan products, price loans, and set fees based on Optimal Blue. From there, the lender was allowed to “manipulate the search parameters and criteria” to generate a wide range of pricing options. Optimal Blue generated a range for loan price for each product type. Thus, the individual lender had three points of discretion: (1) the lender could vary the price within the range set by Optimal Blue, (2) the lender could “mark up” the interest rates generated by Optimal Blue and (3) the lender could impose additional fees after the loan product and interest rate were chosen.
Apparently, GFI did not provide any criteria or factors its lenders were supposed to use in exercising this discretion. The Complaint makes repeated reference to “subjective and unguided” decisions.
African Americans were, on average, charged interest rates 19 to 41 basis points higher, and fees 73 to 105 points higher, than similarly situated white borrowers for the years in question, and Hispanics were charged rates 20 to 23 points higher, and fees 24 to 51 points higher, than similarly situated white borrowers for the years examined.
Justice apparently employed matched pair analyses and regression analyses. The pricing disparities noted above compared African American/Hispanics against “similarly-situated” white borrowers and “account[ed] for all factors related to the borrowers’ credit risk and loan characteristics.” Justice did not describe its precise methodology for its regression analyses and the credit factors it “regressed” out of the model to produce the conclusion that race was “statistically significant” in the pricing disparity. In Justice’s November 2011 Complaint against Countrywide, Justice noted in a footnote that:
“Statistical significance is a measure of probability that an observed outcome would not have occurred by chance. As used in this Complaint, an outcome is statistically significant if the probability that it could have occurred by chance is less than 5%.” (emphasis added)
Presumably Justice is using the same 5% benchmark in the GFI Complaint, however, there is no footnote describing that benchmark as there was in Countrywide.
It would be helpful to know the parameters around the selection of similar-situated white borrowers. Only with that information could one get a complete picture of the strength of Justice’s case.
Justice alleges disparate impact in the Complaint, using the familiar language:
GFI’s pricing policies and practices . . . had a disparate impact . . . and are not justified by business necessity or legitimate business interests.
Much of the Complaint, however, alleges specific conduct of lenders based on discriminatory treatment based on race: lenders charged African Americans and Hispanics more because they were African American or Hispanic. In a direct discriminatory treatment case, disparate impact is irrelevant. And, in a disparate impact case, evidence of discriminatory intent is irrelevant.
GFI appears to be primarily an “upcharging” case – charging African Americans and Hispanics more than the lender’s “normal” pricing. The December 2011 Justice Department case against Countrywide includes the flip side of this coin – “downcharging” white borrowers or giving white borrowers preferential rates versus “normal” pricing. White borrowerswere placed into prime loans, Hispanics and African Americans were left with subprime, higher priced, loans:
By early 2007, Countrywide originated as many as half of certain loan products as exceptions to its underwriting policies. Countrywide made tens of thousands of non-subprime loans . . . based on criteria other than strict adherence to its published underwriting guidelines. Countrywide did not grant these exceptions to Hispanic and African-American borrowers on a basis equal to that for non-Hispanic white borrowers. (emphasis added)
Justice apparently performed a “matched pair” type analysis to “factor out” credit factors (again it would be helpful to see their actual model and its factors to make a judgment on its validity and the inclusion of all relevant factors):
African Americans had odds . . 2.1 and 2.7 higher than similarly situated [white borrowers] of receiving a subprime loan instead of a prime loan after accounting for objective credit qualifications” (emphasis added)
Revised Truth in Lending removes lender incentive to upcharge, the gravamen of the GFI complaint. So that will reduce the odds for this kind of case.
Lenders should also consider:
- Eliminate all lender discretion. This means discretion to charge a higher rate/fee, and discretion to charge a lower rate/fee.
- Another way of saying it: one rate, one set of fees, no exceptions. What about lowering the rate to meet the competition? What about lowering the rate for a customer with a large deposit with the bank? A large commercial loan customer? What about lowering the rate for a customer who is paying you off to take a better rate from another lender? These are certainly justified from a legitimate business purpose perspective, but if they are received disproportionately by white borrowers, disparate impact and the regression models are in play.
- Perform your own regression analysis on matched pairs. Of course, you may be hoisting yourself on your own petard.
- Train, train and train again.
- Provide financial planning/management classes/assistance for borrowers/community.
- Make sure every customer gets the best rate for which she/he qualifies.
- Mystery shop with matched pair borrowers (remember the 60 minutes show of years ago!).
- Effectively monitor individual and aggregated lender activity.
- Validate your credit approval/pricing model, making sure that every factor can meet the “business necessity” and “legitimate business interest” disparate impact litmus tests. Get an opinion from an expert.