The Duncan-Ali Letter
(April 28, 2012, rev. May 18, 2012)
Note: This page concerns how the Department of Education may act when advised, by my letter of April 18, 2012, to Secretary of Education Arne Duncan and Assistant Secretary for Civil Rights Russlynn H. Ali, or otherwise, that its perception that stringent discipline policies lead to large racial differences in discipline rates is the exact opposite of reality. The Department replied by letter dated May 4, 2012, from the Customer Service Team of the Office of Civil Rights, which apparently received the letter on May 1, 2012. Referring to my letter of April 18, 2012, as one of March 9, 2012, and attempting unsuccessfully to summarize the letter, the response provided a page and a half of boilerplate language mainly describing the agency’s functions and explaining how I might file a complaint concerning any violation of the laws enforced by the agency. The failure to get the date of my letter correct is a mistake that anyone might make, and even the failure to accurately summarize my letter may be understandable. But, based on my own substantial tenure in the federal government, I find the evident absence of any effort to comprehend the letter surprising. In any case, one may assume that as yet the Department remains unaware that its perception about the consequences of stringent discipline policies is incorrect.
The main Discipline Disparities page on this site discusses the fact that both the Department of Education (DOE) and the Department of Justice (DOE) regard large racial differences in public school discipline rates to result from stringent discipline policies when the exact opposite is the case. The discussion on that page as to how remarkable is this particular misunderstanding of correlations between relative differences and prevalence of an outcome might be seen as questioning the competence of these agencies. The failings of these agencies in the interpretation of data on group differences, however, must be appraised in light of the fact that, as reflected by the references on the Measuring Health Disparities page, including articles such as “Can We Actually Measure Health Disparities?” (Chance, Spring 2006), “Race and Mortality” (Society, Jan/Feb 2000), “Divining Difference” (Chance, Fall 1994), “The Perils of Provocative Statistics” (Public Interest, Winter 1991), and the approximately 130 journal comments collected in Section D of that page, as well as the discussion on Subgroup Effects sub-page of the Scanlan’s Rule page, one finds the same misunderstanding in virtually all work on groups differences of the statisticians and epidemiologists of the world’s leading universities. Thus, the failure of these agencies to understand certain fundamental statistical principles relating to their enforcement roles would seem somewhat excusable at least until the agencies have been squarely confronted with the fact of their misinterpretation. At that point, the agencies’ competence, and their character, must be appraised with regard less to their statistical expertise than their ability, on receipt of information from the public pertinent to their missions, to dispassionately consider that information and resist the forces that sometimes cause institutions to subordinate their missions to concerns about how they are perceived by the public.
On April 18, 2012, I sent a letter to Secretary of Education Arne Duncan and Assistant Secretary for Civil Rights Russlynn H. Ali explaining clearly enough that the Department of Education’s assumption that stringent discipline standards are leading to large racial differences in discipline rates is incorrect. Over the ensuing weeks and months we shall learn whether that agency can understand the relevant concepts and act responsibly in light of such understanding.
Similar issues exist regarding the Department of Justice to which I sent a letter on April 23, 2012, addressing that agency’s misperceptions of the correlations between the stringency of discipline standards, as well its view that reducing the frequency of adverse lending outcomes will reduce the relative difference in experiencing those outcomes, which view is also the exact opposite of the truth, as discussed on the Lending Disparities page and in, among many published articles, most recently the April 2, 2012 National Law Journal article styled “The Lending Industry’s Conundrum.” The Holder/Perez Letter sub-page to the Lending Disparities will discuss the Department of Justice’s situation more fully.
Like issues apply with respect to a number of institutions (or individuals) I have contacted formally or informally over the years, some of which are among the recipients of letters collected on the Institutional Correspondence sub-page of MHD. The manner in which those institutions (or individuals) have acted in response to such contacts may receive substantial attention on this site. For the manner in which institutions and individuals act when confronted with information indicating that premises underlying much of what they do are unfounded raise important issues going well beyond the measurement issues addressed on the various pages of this site. At this time, I note just a few matters.
In 1998, on learning of the Race and Health Initiative, I first contacted the Department of Health and Human Services seeking to explain the way that relative differences in favorable outcomes and relative differences in adverse outcomes tended to change systematically in opposite directions as the prevalence of an outcome changes. Toward the end of the 2000 Society article “Race and Mortality,” I allude to a response from the Director of the National Center for Health Statistics (NCHS). The letter itself reflected a failure to understand of the point, but it did indicate that the materials I sent would be circulated among NCHS staff. When I later brought “Race and Mortality” and other items to the attention of NCHS statisticians directly involved with health disparities measurement issues, it appeared that any materials circulated by the director had failed to secure their attention. But “Race and Mortality” and the 2004 Chance article “Divining Difference” did secure their attention.
Unfortunately, however, rather than causing NCHS to address the difficulties in the measurement of health and healthcare disparities arising from the way standard measures of differences between outcome rates tend to be affected by the overall prevalence of an outcome, these contacts simply led NCHS to adopt the position whereby it would measure all health and health care disparities in terms of relative differences in adverse outcomes. A key implication of that course was that in many circumstances where NCHS would have found improvements in healthcare to lead to decreasing racial disparities in healthcare it would now find increasing racial disparities in healthcare. For example, in an October 26, 1998 Progress Review: Black Americans, the Department of Health and Human Services found a decrease in the racial difference in pneumococcal immunization rates among persons over 65 between 1989 and 1995 on the basis of the following figures. The black and white rates were 6% and 15% in 1989 and 23% and 34% in 1995. Thus, whereas the black rate was 60% lower than the white rates in 1989, it was only 32% lower in 1995. Put another way, the difference was cut almost in half. Under the approach NCHS adopted in response to the referenced articles, what was close to a 50% decrease in the disparity (measured in term of vaccination rates) became a 55% increase in disparity (measured in terms of relative differences in rates of failing to be vaccinated). See also Table D of Methodological Issues in Measuring Health Disparities, where NCHS shows how instead of finding a 22% decrease in the Hispanic-white disparity in mammography between 1990 and 1998 (as it would have done in 1998), NCHS was finding a 43% increase in the disparity in rates of failure to receive mammography. For more on the unfortunate consequences of the NCHS failure to act responsibly in the matter, see Section A.6 of the Scanlan’s Rule page, Section E.7(a) of the Measuring Health Disparities page, and section [2] of the Mortality and Survival page.
Contrasting situations in the United Kingdom also warrant mention. The Subgroup Effects sub-page of the Scanlan’s Rule discusses the mistaken (indeed illogical) assumption that a rate ratio observed in a clinical trial can be applied across a range of baseline rates to estimate the clinically relevant absolute difference and corresponding number needed to treat pertinent to each baseline rate. Patient.co.uk, a non-academic entity involved in providing guidance to physicians, agreed to change its guidance on calculating number needed to treat within nine days of my initially contacting it. On the other hand, Oxford’s Centre for Evidence Based Medicine took only a few hours to respond that it would not consider revising its guidance until the points I made appeared in a peer-reviewed journal.
But the issues confronting NCHS and the two UK entities just mentioned are considerably more complex than those facing the Department of Education. It is doubtful that the Department can find a statistician who, having carefully thought the matter through, will fail to advise the agency that lenient policies are likely to yield larger disparities in discipline rates than stringent policies (just as it is doubtful that the agency can find a statistician, who, carefully thinking the matter through, will fail to advise it that lowering test cutoffs will tend to increase relative difference in failure rates). Once recognizing such fact, the only responsible course for the agency is to advise the public and the public schools whose policies it monitors that, contrary to what the agency previously has said, stringent discipline policies do not tend to yield larger racial disparities in discipline rates than lenient policies but in fact tend to yield smaller ones.
|