Things government doesn't know about racial disparities

On November 19, 2013, the House Oversight and Investigations Subcommittee held a hearing titled “A General Overview of Disparate Impact Theory” to discuss whether the Fair Housing Act (FHA) prohibits practices that, while not intended to discriminate, have a disparate impact on group’s protected by the statute. 

On January 8, 2014, the departments of Justice and the Education jointly announced release of DOE guidance on implementation of public school discipline policies.  The guidance reflects both disapproval of stringent discipline policies and concern that students from certain minority groups are disciplined several times as often as white students.  It also specifically notes that schools would be found guilty of discrimination if they failed to implement less discriminatory alternatives to practices with disparate impacts on protected groups.

ADVERTISEMENT
The content of both the subcommittee hearing and the school discipline guidance highlight a problem with the legislative and executive branches’ understanding of racial disparities, with implications far beyond fair lending and fair discipline policies. 

Executive efforts to apply disparate impact theory to the FHA began in 1994.  Concerned that blacks and Hispanics had home mortgage applications rejected several times as often as whites, federal agencies issued an Interagency Policy Statement announcing that lenders could be liable for unnecessarily stringent criteria that disproportionately disqualified minority loan applicants.    

Encouraging the relaxing of criteria accorded with federal policy in employment where lowering test cutoffs was universally regarded as reducing a test’s disparate impact because lowering cutoffs tends to reduce relative (percentage) differences in pass rates.  For example, suppose that pass rates are 80 percent for whites and 63 percent for a minority group and thus the pass rate is 27 percent higher for whites than minorities.  If the cutoff is lowered to the point where 95 percent of whites pass, assuming normal test score distributions, the minority pass rate would be about 87 percent.  With the lower cutoff, the pass rate would be only 9 percent higher for whites than minorities. 

Lending criteria operate just like test cutoffs.  By relaxing criteria, lenders tend to reduce relative differences in approval rates of white and minority mortgage loan applicants. 

But, while lowering cutoffs tends to reduce relative differences in pass rates, lowering cutoffs tends to increase relative differences in fail rates.  In the situation above, the minority fail rate was initially 1.85 times the white fail rate (37 percent/20 percent).  With the lower cutoff, the minority fair rate would be 2.6 times the white fail rate (13 percent/5 percent).

This pattern is not peculiar to test scores.  Inherent in most distributions of factors associated with experiencing favorable or adverse outcomes is a pattern whereby the rarer the outcome, the greater tends to be the relative difference in experiencing it and the smaller tends to be relative difference in avoiding it.  It can be illustrated with virtually any data showing points on a continuum of factors associated with experiencing an outcome.   Income and credit score data, for example, show how lowering an income or credit score requirement will tend to reduce relative differences in meeting it while increasing relative differences in failing to meet it.

But, unaware that reducing the frequency of an adverse outcome tends to increase relative differences in experiencing it, regulators consistently monitor the fairness of lender practices in terms of relative differences in adverse outcomes.  Thus, by responding to  encouragements to relax criteria, lenders increased  the chances that they would be sued for discrimination. 

The same holds in the school discipline context.  Reducing suspensions and expulsions as the recent DOE guidance urges schools to do, while reducing relative differences in rates at which whites and minorities avoid those outcomes, will tend to increase the relative differences in suspensions and expulsions that DOE monitors to appraise the fairness of practices at particular schools.

Despite their involvement with numerous  testing issues, neither DOE nor DOJ understands these issues.  As reflected by the November 19 hearing, neither does Congress.

The failure of understanding affects a wide range of matters in which the federal government endeavors to ameliorate differences between advantaged and disadvantaged groups.  For example, for some time close to 10 percent of the NIH budget has been devoted to the study of disparities in health and healthcare.  Practically no one conducting that research, or monitoring the value of federal dollars devoted to it, is aware that determinations of whether disparities are increasing or decreasing commonly turns on whether they are measured in terms of relative differences in favorable outcomes or relative differences in the corresponding adverse outcomes.

James P. Scanlan is a lawyer in Washington, DC, who specializes in the use of statistics in litigation.