by. Henry Meier
Yesterday, the CFPB, which prides itself on beinga statistics-driven, cutting edge agency of the 21st Century, announced a new rating system for its employees which deemphasizes statistics. For several months now, the CFPB has been dogged by increasingly strident accusations that its managers engaged in discriminatory practices. These accusations were bolstered by an internal reporthighlighted in yesterday’s CU Times showing statistical disparities based on race in the performance review process. For example, 20.3 percent of white employees received the highest rating (a 5 on a 1-5 scale), while only 10.5% of African-American employees received this rating. The CFPB is responding to this “proof” of racial disparity by implementing a pass-fail system of employee evaluations, doing away with those troublesome numbers. Instead, employees will retroactively be classified as either solid performers or unacceptable ones.
CFPB’s retreatspeaksvolumes about statistics and theirlimits. Disparity impact analysis, where regulators and litigators argue that a facially neutral lending policycan be proven to discriminate against individuals based on statistical analysis, is predicated on the assumption that statistics don’t lie.Advocates of this approach argue that at some point statistical disparities demonstrate that even facially neutral policies reflect discriminatory undertones and/or practices.
On the other end of the spectrum, on which I would place myself, are those who take a jaundiced view of disparate impact analysis. Statistics only tell a fractionof the story. For instance, the CFPB’s statistical chart can’t tell you about how often an employee had to be pushed to get his work done. Similarly, statistics alone can’t capture thefull extent of negotiations that went on between a mortgageoriginator and a consumer who happened to be African-American. Nevertheless, the explosion of data makes it more, not less, likely that statistics will be used to judge the effectiveness of anti-discrimination laws. This is why I find theCFPB’s response so telling. Rather than defend itsevaluations, it implicitly assumes that its managers must be racially biased. Remember, these are the same peoplewho will ultimately be reviewing lending trends andusing increased HMDA data to spotdiscrimination.
The pre-eminence of disparate analysis is going to have real life consequences. For instance, the reality is that as lenders heighten their underwriting standards to make sure that they can document why a borrower can repay a mortgage loan ordecide to only make so-calledqualified mortgages, these decisions will have a disproportionately negative impact on minority groups that,in the aggregate, have less income.
