At 2:00 p.m. on July 24, 2017, with Roger Clegg moderating, I will be presenting a Federalist Society teleforum titled “Are Existing Civil Rights Policies Based on a Statistical Understanding That Is the Opposite of Reality?”  The subject concerns the fact that federal civil rights enforcement policies regarding lending, school discipline, criminal justice, voter qualifications, and employment have long been based on the belief that relaxing standards or otherwise reducing the frequency of some adverse outcome will tend to reduce (a) relative (percentage) racial and other demographic differences in rates of experiencing the outcome and (b) the proportions racial minorities and other groups more susceptible to the outcome make up of persons experiencing the outcome.  In fact, reducing the frequency of an adverse outcome tends to increase, not reduce, both (a) and (b) as to the outcome. 

In consequence of the government’s failure to understand this matter, entities covered by civil rights law that have acceded to government encouragements to relax standards have increased the chances that the government (or others) will sue them for discrimination.  In the context of consent decrees and other agreements calling for reducing adverse outcomes in order to reduce (a) and (b), the more entities and individuals attempt to comply with their obligations to reduce the outcomes, the more likely they are to be faulted for failing to reduce (or for increasing) (a) and (b). 

This issue, which was the subject of my presentation titled “The Mismeasure of Disparate Impact” at last year’s Federalist Society Fourth Annual Executive Branch Review Conference, is something I have written about here a good deal over the last 14 months.  This year, in posts of January 4 (“Will Trump Have the First Numerate Administration?”),  February 8 (“Compliance Nightmare Looms for Baltimore Police Department”), and March 20 (“Racial Impact Statement Laws in New Jersey and Elsewhere”),  I have discussed the possibility that the new administration will understand the issue better than prior administrations.  And I have noted that Department of Justice (DOJ) actions regarding police practices in Baltimore (and Chicago) may provide an opportunity for the agency to educate itself on the matter.  As of the March 20 post, the court handling the DOJ case against the Baltimore Police Department was still considering the consent decree the parties had submitted to the court in early January. 

Since that post, there have been a number of developments regarding the Baltimore Police case and other matters involving this issue.  On March 31, Attorney General Jeff Sessions issued a memorandum calling for review of consent decrees and other DOJ activities involving oversight of state and local police practices.  The memorandum was almost certainly issued without the agency’s understanding that many of the decrees are based on a fundamentally mistaken premise as to the probable effects of generally reducing adverse interactions between the police and public on the measures of racial disparity on which the agency has long relied.  That mistaken premise, however, alone provides a compelling basis for review of all decrees and other agreements aimed at reducing those measures.    

On April 6, the court entered the Baltimore Police consent decree over belated objection of DOJ attorneys who still showed no signs of understanding the decree’s mistaken premise.  Among other things, the decree calls for monitoring racial disparities in the Baltimore Police Department’s policing practices, as well as in its hiring practices.  Last month, the parties  received 26 proposals from entities maintaining they have the expertise to monitor the decree.  The public had until July 17 to submit comments regarding the suitability of the competing applications.

Meanwhile, by letter of April 13, I attempted to explain the matter to Attorney General Jeff Sessions, as I had done with letters of March 9, 2015, and April 23, 2012, to DOJ leadership in the prior administration.  On June 26, I submitted comments on proposals to monitor the Baltimore Police consent decree, explaining that it is improbable that the monitor candidates understand the statistical issues any better than the parties and stressing the importance of ensuring that the selected entity does understand the issues.  And, by letter of July 17 to the Departments of  Education, Health and Human Services, and Justice, I urged the agencies at least to explain to school administrators, that, contrary to what the agencies have previously led those administrators and the public to believe, relaxing public school discipline standards will tend to  increase relative racial and other demographic differences in discipline rates and the proportions racial minorities and other more susceptible groups make up of disciplined students. 

The three items provide short summaries of the waste of resources and other problems arising from the failure of the government and the social and medical science communities to understand how to analyze demographic differences in favorable and adverse outcome, even when actors are not proceeding on beliefs about the effects of policies that are the opposite of reality.   A longer summary may be found in my November 14, 2016 comments for the Commission on Evidence-Based Policymaking.

But the items also provide useful illustrations of the patterns to be discussed at the July 24 teleforum, and I will be using some of those illustrations at the teleforum.  Since many people, including many whose main activities involve the analysis of demographic differences, have not found these issues easy to understand, those considering joining the teleforum may find it useful to review some of those illustrations before the teleforum. 

The best collection of illustrations is found in the June 26 comments on Baltimore Police consent decree monitor proposals.  Tables 1 though 3 (which also appear in the April 13 Sessions letter and include the table used in my February 8 post on the Baltimore Police consent decree) show how lowering a test cutoff or income or credit score requirement will tend to reduce relative differences in rates of passing the test or meeting the income and credit score requirements.  This is something that most people who deal with discrimination issues probably understand. 

But the tables also show that lowering the test cutoff or income or credit score requirement will tend to increase relative differences in rates of failing the test or failing to meet the requirements.  This is something that almost no one understands.

Table 4 of the comments shows what in fact happened when, between 1998 and 2000, the U.S. Customs Service implemented a number of reforms aimed at reducing racial disproportionality in searches by generally restricting the use of searches.  As total searches decreased from about 44,000 to about 9,000, the proportion blacks made up of persons searched rose from 14 percent to 27 percent.

Table 5 of the comments presents what may be one of the easier to understand illustrations of why relaxing a standard tends to increase the proportions groups that are more susceptible to an adverse outcome make up of persons experiencing it.  Using data from a 2014 Department of Education report, the table shows that black students made up 44 percent of preschool students suspended one or more times and 48 percent of preschool students suspended multiple times.   

What, then, should we expect to happen if standards are relaxed such as to give all students a simple reprimand rather than what would otherwise be their first suspension?  It should be evident that the proportion black students make up of students suspended one or more times would increase – indeed, increase precisely from 44 percent to 48 percent.  Patterns of this nature are in fact being observed all across the country where jurisdictions have relaxed public school discipline standards in reliance on government guidance that doing so will tend to reduce such proportions (and associated relative racial differences in suspension rates).   See page 8 of the July 17 letter to the three agencies providing that guidance.

There ought to be nothing mysterious about this.   The described patterns are the natural consequences of the fact that a group that is more likely to experience some outcome than other groups will almost invariably make up a larger proportion of persons very likely to experience the outcome than it makes up of persons likely to experience the outcome.  That restricting any outcome to those most likely to experience it will tend to increase the proportions those groups make up of persons experience the outcome therefore ought to be obvious.  But, apparently, not only do the government and virtually all others discussing the matter fail to find this to be obvious, they commonly assume the opposite.        

As discussed in the January 4 post, the current administration has a much stronger interest in understanding this and related issues than the last administration.  But whether any arms of the administration will in fact come to understand the issues remains to be seen.  So far, my April 13 letter to the Attorney General has merely elicited responses from two sections of the Civil Rights Division advising that the letter did not constitute a complaint within their purview.

On June 29, 2017, the President submitted a nomination for Assistant Attorney General for the DOJ’s Civil Rights Division.  Given the stir caused by the Attorney General’s March 31 memorandum, any opposition to the appointment from members of the Senate Judiciary Committee (or the Senate itself) may focus on that memorandum and attitudes about civil rights enforcement it is deemed reflect.  It would be well if the administration, and the Senate, better understand the mistaken premises of those decrees and other civil rights enforcement activities before that happens.   

Click here to access Jim Scanlan's website, which hosts information he will reference in the Federalist Society Teleforum on July 24, 2017. For specific references to charts in the Teleforum, please visit this link.