A better way to assess disparate impact
Here’s the key issue: What if a lender applies the same credit underwriting standards to all credit applicants, but this results in different demographic groups having different credit approval-credit decline ratios? Is that necessarily a problem?
One position is that applying the same credit standards to everybody, regardless of demographic group, is exactly what every lender should be doing. Yet supporters of “disparate impact” argue that if there are different ratios for loan approvals versus loan declines among groups, it must mean there is some kind of hidden, even if entirely unintended, bias in the process.
Which side is right? There is a straightforward, data-based way to tell. It is simply to add to the report the default rates on the loans in question and compare them to the approval ratios by group.
Suppose, for example, that demographic Group A has a lower loan approval rate and therefore a higher decline rate than Group B. We must also compare their default rates. There are three possibilities: Group A either has the same, a lower or a higher default rate than Group B.
If Group A has the same default rate as Group B, then the underwriting procedure and the different approval-decline ratios were fair and appropriate, since they resulted in the same default outcome. Predicting and controlling defaults is the whole point of doing the credit analysis.
If the default rate of Group A is lower than Group B, however, that shows that it is experiencing a different credit standard, which may be a higher standard, or may be one biased one against Group A, even if it is not intended.
In the third possibility, if the default rate for Group A is higher than Group B, that shows that in spite of the fact Group A had a lower approval and higher decline ratio, it was nonetheless being given easier credit standards, or that the process was biased in its favor, even if not intended.
We need the facts of default rates to objectively and calmly address this issue. Why not simply provide them as part of the regular Home Mortgage Disclosure Act reports?
Some previous discussions of this issue have analyzed the different groups by factors such as household income or standard credit ratios. But such factors are merely attempted predictions of future default rates, not the reality of the actual default rates. It is much better to use the direct reality of defaults, since controlling defaults is the whole point of credit underwriting.
As HUD addresses the issue, a resolution based on fact should be adopted: Report the default rates on relevant loans and compare them to approval-decline rates, and then draw the logically necessary conclusions. If the question gets to the courts, judges should insist on the same fundamental logic being applied.
Image by VGstockstudio