For those of you who are not on the Linked In Compliance Group, I wanted to share a very important observation on the article I wrote regarding proxy’s that are being used in Fair Lending reviews. I thought you would find this next level of detail preposterous yet enlightening!
This is my original article: Isn’t Life Complex Enough?
This is what was commented on the Linked In Group:
David Skanderson, PHD is a noted Fair Lending Statistician and Economist and he had this to add:
David Skanderson
Vice President – Financial Economics Practice at Charles River Associates
Tammy: It’s actually even more interesting than that. My understanding of the Bureau’s methods is that each credit applicant is assigned a probability for EACH race/ethnicity group. So, on a statistical basis, their method may indicate that Tammy is 60% African American, 10% Hispanic, 5% Asian, and 25% non-Hispanic white. If they were to conclude in an enforcement matter that there was discrimination against Hispanics and African Americans and that borrower remuneration is required, assuming that you were in the data sample, they would attribute remuneration owed to YOU in proportion to your imputed characteristics, regardless of the fact that you are actually white. Consider also the reverse scenario: Suppose a borrower who is actually African American is also in the sample, but lives in an area that is 95% non-Hispanic white. Even though the she was (theoretically) a victim of discrimination in this example, zero or nearly zero remuneration would be attributed to her. It is not at all clear, however, how they will actually pay out to borrowers, given that they don’t really know the actual races and ethnicities at the end of the day. So, even if it were the case (very hypothetically) that the proxy approach achieved the correct result on average (which is debatable), it certainly poses an interesting problem of how to remunerate those who were presumably affected by discrimination.
Another interesting question: Will they apply the proxy method routinely in examinations of mortgage lenders, such as in cases where a high percentage of the lender’s customers do not report race, ethnicity or gender?
On a positive note, I do applaud the Bureau for (finally) being transparent about its methods and expressing an openness to considering reasonable alternatives that lenders may use.