Returning to the issue of mortgages, attempts to prove discrimination by looking at loan approval rates, as the Markup / AP team attempted to do, are wrong. External analysts don’t have the full set of information about loan applicants, including credit scores, that financial institutions use to guide their decisions.

A more revealing approach is to look not at the “front end” of the process, so to speak, but the “back end”. How often do borrowers default on their mortgages? If lenders make their decisions without prejudice, then the default rates for, say, white and Hispanic customers should be about the same. That is, default rates would show that lenders are just as willing to take a risk with a white candidate as they are with an otherwise comparable Hispanic candidate.

If, on the other hand, a lender takes ethnicity into account and assumes that whites are more trustworthy – more likely to repay their loans – than Hispanics, then the lender will give more loans to white applicants than are Hispanics. ‘to Hispanic applicants with the same income. , debts, credit scores and financial history. This, in turn, means that the aggrieved lender will experience a lower average default rate for Hispanic customers than for white customers.

The statistical analysis involved is a bit more complicated than any brief explanation can express, I admit, and over time economists have argued convincingly that default rate data must themselves be risk-adjusted. Yet if someone claims to assess the extent of loan discrimination without paying close attention to default rates, that person is either misinformed or is actively trying to misinform.