Seen at CBS News:
Bailey’s superiors encouraged her to apply, she says — but turned her down after discovering her bad credit history.
Bailey, with her lawyer, has lodged a complaint against Harvard charging racial discrimination. The reason: Studies show that minorities are more likely to have bad credit, but credit problems have not been shown to negatively affect job performance.
Setting aside my concern with blanket statements like “credit problems have not been shown to negatively affect job performance”, since I can easily imagine a scoring model being built to predict, say, risk of employee fraud or early termination, using credit bureau data, which might not closely correlate with credit scores for lending, or insurance scores…. I suppose that it would be wrong to note that in the U.S., modelers are prohibited from using data that would enable them to adjust the models for any alleged differences in norms in particular groups.
Overseas, where race and gender differentiation are less likely to provoke a civil rights backlash, it’s not unheard of for modelers to consider the backgrounds of different groups of folks who may be measured by their models.
For example, in Asia, there is a tendency for ethnic Chinese to try to wrap up debts, or at least pre-pay bills ahead of the Chinese New Year celebrations. A financial institution would perhaps like to consider this when assessing the potential profitability of a client, as both the default and prepayment rates would be affected by this tendency. Perhaps a late payment or loan default during Chinese New Year is a more significant piece of information depending on the consumer’s background.
I wouldn’t be comfortable with loosening controls in the U.S. to permit the use of ethnic data, due to the potential for public backlash — the likelihood that modelers would be accused (correctly or incorrectly) of inappropriate discrimination.
I just find the irony oddly amusing.