Fair-housing regulators could benefit from a lesson in statistics. Lenders that follow federal guidance increase chances the government will sue them for discrimination.
In recent years, the Department of Justice settled claims of racial and ethnic discrimination in lending with recoveries totaling more than half a billion dollars. The largest occurred in United States v. Countrywide Financial Corporation ($335 million) and United States u. Wells Fargo Bank ($175 million). [paragraph] The complaints in both cases fault lenders for failing to implement less-discriminatory alternatives to practices believed to cause minorities to receive subprime rather than prime loans at higher rates than whites. The complaints also fault lenders for various practices that led generally to greater frequency of subprime loans.
That emphasis reflects an aspect of federal fair-lending enforcement that has long been based on a perception about racial and ethnic differences in outcome rates that is the exact opposite of reality.
But that perception is not the only problem with fair-lending enforcement.
To put things in context, one must look back to the 1990s. At that time there was great concern raised over the fact that minorities had their home mortgage loan applications rejected several times as often as whites.
In 1994, belief that a substantial part of the rejection-rate differences resulted from the greater difficulty minorities had in meeting standard lending criteria prompted federal agencies monitoring fair-lending laws to issue an Interagency Policy Statement on Fair Lending. That policy statement announced that regardless of the absence of the intent to discriminate, lenders could be held liable for unnecessarily stringent criteria that disqualified minorities at higher rates than whites.
The policy statement's encouragement to relax lending criteria accorded with federal policy in the fair-employment context. Lowering cutoffs on hiring or promotional tests was universally regarded as reducing a test's disparate impact on minority job applicants because lowering cutoffs tends to reduce relative (i.e., percentage) differences in pass rates.
For example, suppose that at a particular cutoff, pass rates are 80 percent for whites and 63 percent for minorities. At this cutoff, the white pass rate is 1.27 times the minority pass rate.
If the cutoff is lowered to the point where the white pass rate is 95 percent, assuming normal (bell-shaped) test score distributions, the minority pass rate would be about 87 percent. With the lower cutoff, the white pass rate would only be 1.09 times the minority pass rate.
These numbers are shown in the three data columns on the left side of Figure 1.
But while lowering a cutoff tends to reduce relative differences in pass rates, it also tends to increase relative differences in failure rates.
As shown in the three columns on the right side of Figure 1, in the aforementioned situation, the minority failure rate was initially 1.85 times the white failure rate. With the lower cutoff, the minority failure rate would be 2.6 times the white failure rate.
The pattern by which relative differences in a favorable outcome and relative differences in the corresponding adverse outcome tend to change in opposite direction as the frequency of an outcome changes is close to universal. And it is evident in all sorts or data.
For example, income and credit score data show that lowering an income or credit score requirement will tend to reduce relative differences in meeting the requirement while increasing relative differences...