ASA Connect

Item 1 below is a letter to the Antioch (CA) Unified School District regarding a suit recently brought against it by the NAACP where the complaint in the case reflects the view that generally reducing public school suspension rates will tend to decrease the proportion African Americans make up of students who are suspended.  The letter is similar to many I have posted here over the last year and a half ­– including letters of October 8, 2015 (item 2) and July 25, 2016 (item 3) urging ASA to explain the matter to arms of the federal government or the President – explaining that reducing any outcome will tend to increase, not decrease, the proportions more susceptible groups make up of persons experiencing the outcome.  

The Antioch situation also has the following interesting element.  The suit is based on an agreement requiring the school district to retain three experts (or expert groups) to analyze discipline issues.  The lead expert has consistently analyzed discipline disparities in terms of absolute differences between rates, which commonly decrease when discipline rates generally decline.  As discussed in Section B of the letter, that expert will commonly reach conclusions about the comparative size of the disparities (over time or from setting to setting) that are the opposite of the conclusions yielded by the approach reflected in the complaint.  As discussed in Section C, that expert will also commonly reach conclusion about the comparative size of disparities that are the opposite of the conclusions the second expert and the third expert (a group) would reach.

 1. http://www.jpscanlan.com/images/Letter_to_Antioch_Unified_School_District_Sept._12,_2016_.pdf

 2. http://jpscanlan.com/images/Letter_to_American_Statistical_Association_Oct._8,_2015_.pdf 

 3. http://www.jpscanlan.com/images/Letter_to_American_Statistical_Association_July_25,_2016_.pdf

Mr. Scanlon is obviously right, but controversies about racial disparities are driven by politics, not by logic.  

The phenomenon that James addresses is an important one for politicians and other decision makers to understand at least at the conceptual level, and all statisticians to raise when a situation calls for it. But it is a technical matter, and understanding of it will not do much, in my opinion, to affect the behavior of those who wish those who wish to use "disparity" data for political agendas.

"Disparities" are universal, and most of the time are not the product of malevolent forces or unjust discrimination, as so many individuals, bereft of good advice from statisticians are so quick to claim.

For a fairly neutral example, consider the relative economic and social positions of people of Italian origin in U.S. and Argentina half a century ago (and to a much lesser extent now). In Argentina, they have  been part of the economically successful ruling class for over a century; in the U.S. in the early part of the 20th century a larger percentage was poor and were infamous for bringing the Mafia and Cosa Nostra. The disparity, as Thomas Sowell wrote about long ago, was due to the Italian immigrants to Argentina having come predominantly from economically more prosperous northern Italy (think Milan) and Italian immigrants to the U.S. having come predominantly from the more impoverished southern Italy (think Naples, Palermo). Would it ever been reasonable to address such disparities by instituting "affirmative action" for southern Italians?

But back on track. Under ideas sometimes grouped under the rubric of "disparate impact theory," it has become fashionable to treat, at least initially, every documented disparity between the sexes or among racial groups as due to sexism or racism on the part of either individuals or institutions. The most egregious current examples would be claims of sex discrimination in pay and racial discrimination in treatment of blacks by police.  Anecdotes aside, there usually has no statistically valid back up for such claims, once all other relevant variables, are appropriately corrected for.

Yet aggressive claims to the contrary come from the president, the mainline media, scientific societies, academics, politicians, and so on. And where has ASA been on these issues? It is better positioned than any other organization to take on those who use dishonest use of raw disparity data to foster ignorance and create social division. It should be publicly blowing the whistle on those individuals, organizations, newspapers, etc.

Attached is a report providing one example of how this was done effectively on a small scale by one person but with institution wide positive  effects over at least a couple of decades. Think what ASA could do.

Item 1 below is a letter to the Antioch (CA) Unified School District regarding a suit recently brought against it by the NAACP where the complaint in the case reflects the view that generally reducing public school suspension rates will tend to decrease the proportion African Americans make up of students who are suspended.  The letter is similar to many I have posted here over the last year and a half ­– including letters of October 8, 2015 (item 2) and July 25, 2016 (item 3) urging ASA to explain the matter to arms of the federal government or the President – explaining that reducing any outcome will tend to increase, not decrease, the proportions more susceptible groups make up of persons experiencing the outcome.  

The Antioch situation also has the following interesting element.  The suit is based on an agreement requiring the school district to retain three experts (or expert groups) to analyze discipline issues.  The lead expert has consistently analyzed discipline disparities in terms of absolute differences between rates, which commonly decrease when discipline rates generally decline.  As discussed in Section B of the letter, that expert will commonly reach conclusions about the comparative size of the disparities (over time or from setting to setting) that are the opposite of the conclusions yielded by the approach reflected in the complaint.  As discussed in Section C, that expert will also commonly reach conclusion about the comparative size of disparities that are the opposite of the conclusions the second expert and the third expert (a group) would reach.

 1. http://www.jpscanlan.com/images/Letter_to_Antioch_Unified_School_District_Sept._12,_2016_.pdf

 2. http://jpscanlan.com/images/Letter_to_American_Statistical_Association_Oct._8,_2015_.pdf 

 3. http://www.jpscanlan.com/images/Letter_to_American_Statistical_Association_July_25,_2016_.pdf

Politicians may rightfully be more concerned about the number of students suspended than ratios.

Item 1 below is a letter to the Antioch (CA) Unified School District regarding a suit recently brought against it by the NAACP where the complaint in the case reflects the view that generally reducing public school suspension rates will tend to decrease the proportion African Americans make up of students who are suspended.  The letter is similar to many I have posted here over the last year and a half ­– including letters of October 8, 2015 (item 2) and July 25, 2016 (item 3) urging ASA to explain the matter to arms of the federal government or the President – explaining that reducing any outcome will tend to increase, not decrease, the proportions more susceptible groups make up of persons experiencing the outcome.  

The Antioch situation also has the following interesting element.  The suit is based on an agreement requiring the school district to retain three experts (or expert groups) to analyze discipline issues.  The lead expert has consistently analyzed discipline disparities in terms of absolute differences between rates, which commonly decrease when discipline rates generally decline.  As discussed in Section B of the letter, that expert will commonly reach conclusions about the comparative size of the disparities (over time or from setting to setting) that are the opposite of the conclusions yielded by the approach reflected in the complaint.  As discussed in Section C, that expert will also commonly reach conclusion about the comparative size of disparities that are the opposite of the conclusions the second expert and the third expert (a group) would reach.

 1. http://www.jpscanlan.com/images/Letter_to_Antioch_Unified_School_District_Sept._12,_2016_.pdf

 2. http://jpscanlan.com/images/Letter_to_American_Statistical_Association_Oct._8,_2015_.pdf 

 3. http://www.jpscanlan.com/images/Letter_to_American_Statistical_Association_July_25,_2016_.pdf

Thanks to all for these comments.  I have a few observations. 

Joel Wiesen is correct that there is much support for generally reducing the number of suspended students.  But a good deal of that support rests on the mistaken belief – promoted by Department of Education, Justice, and Health and Human Services – that generally reducing adverse outcomes like suspension will tend to decrease rather than increase  (a) relative demographic differences in adverse outcome rates and (b) the difference between the proportion disadvantaged groups comprise of students and the proportion they comprise of persons experiencing adverse outcomes.   And the government continues to monitor fairness of practices on the basis of (a) and (b), with the perverse consequence that the more an entity accedes to government pressure to reduce adverse outcomes, the more likely it is that the government (and others) will sue the entity for discrimination.  That was the subject of my December 2012 Amstat New Statistician’s View column.[1]  And the situation has only worsened since then.[2,3] 

Reference 4 is a September 20, 2016 letter to the Oklahoma City School District explaining implications of a recent agreement between the school district and the Department of Education that requires the district to substantially reduce adverse discipline outcomes while reducing (a) and (b) – as well as to distinguish among good and bad decision-makers on the basis (presumably) of the size of (a) and (b) – while all parties (and probably retained experts) mistakenly believe that reducing discipline rates will tend to reduce (a) and (b).  See reference 5 regarding the impossible situation in which the city of Ferguson, Missouri has been placed as a result of the Department of Justice’s failure to understand these issues.  Having agreed to reach a consent decree with the Department of Justice to resolve racial disparities in criminal justice outcomes, Baltimore will likely find itself in a similar situation within a few months.     

An interesting example of the Department of Education’s failure of understanding may be found in the agency’s March  2014 discipline report.[6]  In the case of physical restraint – where overall rate are below 2-tenths of one percent – the report presents the state by state disparities in terms of (b) for students with disabilities (IDEA students).   And the report (at 11) highlights three states where the proportion IDEA students comprise of restrained students is below 50% (better states in the DOE view) and three where the proportion is above 90% (extremely bad states in the DOE view).  Yet almost certainly the latter states limit the use of restraints to the most extreme cases to a greater extent than the former states.  If all states limit the use of restraints to the most unavoidable cases, the proportion IDEA students comprise of restrained students would commonly approach 100% in all states.  See reference 7 (at 5-6) regarding the way social scientists somehow think society would be better off if poverty were less restricted to comparatively inefficient economic units (and the near universal failure of persons analyzing demographic differences to recognize that the proportion a disadvantaged group comprises of persons experiencing an adverse outcome commonly varies inversely with the proportion of the group experiencing the outcome).  

In accord with what I regard to be the intent Dr. Wiesen’s comment,  I respond with regard to reducing total numbers rather than reducing the absolute difference between rates of advantaged and disadvantaged groups.  I note, however, that, given the rate ranges at issue for things like suspension from school, generally reducing adverse outcome rates will tend to reduce absolute differences between rates.  As discussed in the Antioch letter (at 10-11), an anomaly in the Antioch situation is that the lead expert the plaintiff required the school district to retain will (relying on absolute differences) tend to reach opposite conclusions about directions of changes in disparities from those the plaintiff would reach.  But increasing favorable outcomes while reducing the corresponding adverse outcomes does not always tend to reduce absolute differences.  See reference 8 (at 337-339) regarding the way reliance on absolute differences to measure healthcare disparities led Massachusetts to include a disparities element in its Medicaid pay-for-performance program that will tend to increase healthcare disparities (however measured).

Further regarding Dr. Wiesen’s comment, the support for generally reducing suspensions also lies in the belief that stringent discipline policies have a deleterious effect on the educational environment.  That belief is based on rather dubious research.[9]  The mistaken belief that generally reducing suspensions will tend to reduce (a) and (b) very likely contributes to the willingness to allow that research to go unscrutinized.

I am not sure the extent to which Stuart Hulbert’s comment on the failure of statisticians and other scientists to speak up is focused on such persons within agencies like the Departments of Education and Justice.  But it should be recognized that the persistence of a statistical understanding within government that is the opposite of reality is substantially a consequence of the fact that few members of the scientific community understand the matter any better than federal bureaucrats.  So far as the published record reveals, the overwhelming majority of persons specializing in the analyses of demographic differences regarding binary outcomes are unaware that the frequency of an outcome tend to increase (a) and (b).  And, as reflected by the ubiquitous statements that “despite” the decline in some outcome disparities in the outcome persist or have increased, probably a substantial majority of such persons believe that reducing the frequency of an outcome will tend to reduce (a) and (b) (as discussed at 36-40 of the October 2015 ASA letter).  An irony here is that use of the word “despite” is one of the few indications on the part of data analysts of an expectation that changes in the prevalence of an outcome ought to affect a measure; but the expectation as to direction of change in the measure is the opposite of the actual direction.  There is no reason to expect bureaucrats in federal law enforcement agencies to understand the matter better than experts in the analysis of demographic differences in academia and research institutions. 

The scientific community thus bears substantial responsibility for the anomaly in civil rights enforcement.  Recently, the leadership of the Population Association of America (PAA) and the Association of Populations Centers (APC), responding to a March 29, 2016 letter[10] similar to my October 2015 ASA letter, specifically declined to advise the government that reducing the frequency of an outcome like public school suspensions tends to increase, not reduce (a) and (b). Until some organization like PAA or APC – or ASA – recognizes a responsibility both to understand the issues and to correct governmental misunderstanding of the issues, the anomaly in the enforcement of civil rights laws is likely to persist.  Of course, similar responsibilities exist in the case of entities providing expert guidance to government agencies, even if (and especially if) correcting the agencies involves contradicting prior guidance given by the entities. 

Finally, I note that, while the existence of law enforcement regimes described in references 1 to 3 ought to be extremely disturbing to anyone who cares about rational government, the misunderstanding underlying those regimes is but part of the larger, and near universal, failure to recognize patterns by which measures of difference between outcome rates (or between the proportion a group comprises of persons potentially experiencing an outcome the proportion it comprises of persons experiencing the outcome) tend to be affected by the frequency of an outcome.  As a result of that failure virtually never has a researcher explored the extent to which an observed pattern is a function of the prevalence of an outcome and the extent to which it reflects something meaningful about underlying processes.  Hence, despite billions devoted to the study of demographic differences especially regarding health and healthcare outcomes, little has that has said about directions of changes in demographic difference over time or the comparative size of such differences from setting to setting – or even about whether a particular difference should be deemed large or small – has had a sound statistical basis.  See the October 2015 ASA letter and reference 8 (especially at 343-345).  See also the (more recent) reference 11 regarding facts that (a) a decade after the National Center for Health Statistics (NCHS) recognized that relative differences in favorable and adverse health and healthcare outcomes would commonly change in opposite directions as the frequency of an outcome changes no other federal agency has indicated an understanding that it is even possible for that to happen and (b) that NCHS has recently repudiated a decade of research that relied on its earlier guidance (with anyone’s noticing).

Correcting this situation is the larger purpose of my October 2015 letter to ASA.  

1. “Misunderstanding of Statistics Leads to Misguided Law Enforcement Policies,” Amstat News (Dec. 2012) http://magazine.amstat.org/blog/2012/12/01/misguided-law-enforcement/
2. “Things DOJ doesn’t know about racial disparities in Ferguson,” The Hill (Feb. 22, 2016) http://thehill.com/blogs/congress-blog/judicial/270091-things-doj-doesnt-know-about-racial-disparities-in-ferguson
3. “Things the President Doesn’t Know About Racial Disparities,” Federalist Society Blog (Aug. 5, 2016)http://www.fed-soc.org/blog/detail/things-the-president-doesnt-know-about-racial-disparities

4. http://jpscanlan.com/images/Letter_to_Oklahoma_City_School_District_Sept._20,_2016_.pdf
5. http://jpscanlan.com/images/Submission_of_James_P._Scanlan_in_U.S._v._City_of_Ferguson_Apr._11,_2016_.pdf
6. http://blogs.edweek.org/edweek/rulesforengagement/CRDC%20School%20Discipline%20Snapshot.pdf
7. “Race and Mortality,” Society (Jan./Feb. 2000) http://www.jpscanlan.com/images/Race_and_Mortality.pdf
8. “Race and Mortality Revisited,” Society (July/Aug. 2014)http://jpscanlan.com/images/Race_and_Mortality_Revisited.pdf

9. APA Zero Tolerance Study subpage of Discipline Disparities page of jpscanlan.com http://jpscanlan.com/disciplinedisparities/apazerotolerancestudy.html

10. http://jpscanlan.com/images/Letter_to_PAA_and_APC_Mar._29,_2016_.pdf
11. “The Mismeasure of Health Disparities,” Journal of Public Health Management and Practice (July/Aug. 2016)http://www.jpscanlan.com/images/The_Mismeasure_of_Health_Disparities_JPHMP_2016_.pdf

Politicians may rightfully be more concerned about the number of students suspended than ratios.

Item 1 below is a letter to the Antioch (CA) Unified School District regarding a suit recently brought against it by the NAACP where the complaint in the case reflects the view that generally reducing public school suspension rates will tend to decrease the proportion African Americans make up of students who are suspended.  The letter is similar to many I have posted here over the last year and a half ­– including letters of October 8, 2015 (item 2) and July 25, 2016 (item 3) urging ASA to explain the matter to arms of the federal government or the President – explaining that reducing any outcome will tend to increase, not decrease, the proportions more susceptible groups make up of persons experiencing the outcome.  

The Antioch situation also has the following interesting element.  The suit is based on an agreement requiring the school district to retain three experts (or expert groups) to analyze discipline issues.  The lead expert has consistently analyzed discipline disparities in terms of absolute differences between rates, which commonly decrease when discipline rates generally decline.  As discussed in Section B of the letter, that expert will commonly reach conclusions about the comparative size of the disparities (over time or from setting to setting) that are the opposite of the conclusions yielded by the approach reflected in the complaint.  As discussed in Section C, that expert will also commonly reach conclusion about the comparative size of disparities that are the opposite of the conclusions the second expert and the third expert (a group) would reach.

 1. http://www.jpscanlan.com/images/Letter_to_Antioch_Unified_School_District_Sept._12,_2016_.pdf

 2. http://jpscanlan.com/images/Letter_to_American_Statistical_Association_Oct._8,_2015_.pdf 

 3. http://www.jpscanlan.com/images/Letter_to_American_Statistical_Association_July_25,_2016_.pdf

<g class="gr_ gr_30 gr-alert gr_spell gr_run_anim ContextualSpelling ins-del multiReplace" id="30" data-gr-id="30">Scanlen</g>, and <g class="gr_ gr_62 gr-alert gr_gramm gr_run_anim Punctuation only-del replaceWithoutSep" id="62" data-gr-id="62">others,</g> seem to be implying that the government's interest in focusing on reducing DG suspensions or reducing excessive traffic fines is not a worthwhile pursuit if it doesn't "reduce other negative outcomes" among DG groups.  

This displays <g class="gr_ gr_334 gr-alert gr_gramm gr_run_anim Grammar only-ins doubleReplace replaceWithoutSep" id="334" data-gr-id="334">lack</g> of understanding of the issues by the members of this thread.  A question that should be posed in regards to suspensions or traffic fines is "are these punitive actions being consistently applied to the entire population".  That is, are DG group members being punished for actions that AG members also take but are not punished for.  The impact of these punitive actions is not inconsequential for those systemically, inequitably targeted by them.

Particularly in regards to Ferguson, excessive fines and jail for minor traffic offensives for DG groups <g class="gr_ gr_1142 gr-alert gr_gramm gr_run_anim Grammar multiReplace" id="1142" data-gr-id="1142">has</g> negative economic consequences that can further disadvantage these groups while benefitting AG groups.

Thanks to all for these comments.  I have a few observations. 

Joel Wiesen is correct that there is much support for generally reducing the number of suspended students.  But a good deal of that support rests on the mistaken belief – promoted by Department of Education, Justice, and Health and Human Services – that generally reducing adverse outcomes like suspension will tend to decrease rather than increase  (a) relative demographic differences in adverse outcome rates and (b) the difference between the proportion disadvantaged groups comprise of students and the proportion they comprise of persons experiencing adverse outcomes.   And the government continues to monitor fairness of practices on the basis of (a) and (b), with the perverse consequence that the more an entity accedes to government pressure to reduce adverse outcomes, the more likely it is that the government (and others) will sue the entity for discrimination.  That was the subject of my December 2012 Amstat New Statistician’s View column.[1]  And the situation has only worsened since then.[2,3] 

Reference 4 is a September 20, 2016 letter to the Oklahoma City School District explaining implications of a recent agreement between the school district and the Department of Education that requires the district to substantially reduce adverse discipline outcomes while reducing (a) and (b) – as well as to distinguish among good and bad decision-makers on the basis (presumably) of the size of (a) and (b) – while all parties (and probably retained experts) mistakenly believe that reducing discipline rates will tend to reduce (a) and (b).  See reference 5 regarding the impossible situation in which the city of Ferguson, Missouri has been placed as a result of the Department of Justice’s failure to understand these issues.  Having agreed to reach a consent decree with the Department of Justice to resolve racial disparities in criminal justice outcomes, Baltimore will likely find itself in a similar situation within a few months.     

An interesting example of the Department of Education’s failure of understanding may be found in the agency’s March  2014 discipline report.[6]  In the case of physical restraint – where overall rate are below 2-tenths of one percent – the report presents the state by state disparities in terms of (b) for students with disabilities (IDEA students).   And the report (at 11) highlights three states where the proportion IDEA students comprise of restrained students is below 50% (better states in the DOE view) and three where the proportion is above 90% (extremely bad states in the DOE view).  Yet almost certainly the latter states limit the use of restraints to the most extreme cases to a greater extent than the former states.  If all states limit the use of restraints to the most unavoidable cases, the proportion IDEA students comprise of restrained students would commonly approach 100% in all states.  See reference 7 (at 5-6) regarding the way social scientists somehow think society would be better off if poverty were less restricted to comparatively inefficient economic units (and the near universal failure of persons analyzing demographic differences to recognize that the proportion a disadvantaged group comprises of persons experiencing an adverse outcome commonly varies inversely with the proportion of the group experiencing the outcome).  

In accord with what I regard to be the intent Dr. Wiesen’s comment,  I respond with regard to reducing total numbers rather than reducing the absolute difference between rates of advantaged and disadvantaged groups.  I note, however, that, given the rate ranges at issue for things like suspension from school, generally reducing adverse outcome rates will tend to reduce absolute differences between rates.  As discussed in the Antioch letter (at 10-11), an anomaly in the Antioch situation is that the lead expert the plaintiff required the school district to retain will (relying on absolute differences) tend to reach opposite conclusions about directions of changes in disparities from those the plaintiff would reach.  But increasing favorable outcomes while reducing the corresponding adverse outcomes does not always tend to reduce absolute differences.  See reference 8 (at 337-339) regarding the way reliance on absolute differences to measure healthcare disparities led Massachusetts to include a disparities element in its Medicaid pay-for-performance program that will tend to increase healthcare disparities (however measured).

Further regarding Dr. Wiesen’s comment, the support for generally reducing suspensions also lies in the belief that stringent discipline policies have a deleterious effect on the educational environment.  That belief is based on rather dubious research.[9]  The mistaken belief that generally reducing suspensions will tend to reduce (a) and (b) very likely contributes to the willingness to allow that research to go unscrutinized.

I am not sure the extent to which Stuart Hulbert’s comment on the failure of statisticians and other scientists to speak up is focused on such persons within agencies like the Departments of Education and Justice.  But it should be recognized that the persistence of a statistical understanding within government that is the opposite of reality is substantially a consequence of the fact that few members of the scientific community understand the matter any better than federal bureaucrats.  So far as the published record reveals, the overwhelming majority of persons specializing in the analyses of demographic differences regarding binary outcomes are unaware that the frequency of an outcome tend to increase (a) and (b).  And, as reflected by the ubiquitous statements that “despite” the decline in some outcome disparities in the outcome persist or have increased, probably a substantial majority of such persons believe that reducing the frequency of an outcome will tend to reduce (a) and (b) (as discussed at 36-40 of the October 2015 ASA letter).  An irony here is that use of the word “despite” is one of the few indications on the part of data analysts of an expectation that changes in the prevalence of an outcome ought to affect a measure; but the expectation as to direction of change in the measure is the opposite of the actual direction.  There is no reason to expect bureaucrats in federal law enforcement agencies to understand the matter better than experts in the analysis of demographic differences in academia and research institutions. 

The scientific community thus bears substantial responsibility for the anomaly in civil rights enforcement.  Recently, the leadership of the Population Association of America (PAA) and the Association of Populations Centers (APC), responding to a March 29, 2016 letter[10] similar to my October 2015 ASA letter, specifically declined to advise the government that reducing the frequency of an outcome like public school suspensions tends to increase, not reduce (a) and (b). Until some organization like PAA or APC – or ASA – recognizes a responsibility both to understand the issues and to correct governmental misunderstanding of the issues, the anomaly in the enforcement of civil rights laws is likely to persist.  Of course, similar responsibilities exist in the case of entities providing expert guidance to government agencies, even if (and especially if) correcting the agencies involves contradicting prior guidance given by the entities. 

Finally, I note that, while the existence of law enforcement regimes described in references 1 to 3 ought to be extremely disturbing to anyone who cares about rational government, the misunderstanding underlying those regimes is but part of the larger, and near universal, failure to recognize patterns by which measures of difference between outcome rates (or between the proportion a group comprises of persons potentially experiencing an outcome the proportion it comprises of persons experiencing the outcome) tend to be affected by the frequency of an outcome.  As a result of that failure virtually never has a researcher explored the extent to which an observed pattern is a function of the prevalence of an outcome and the extent to which it reflects something meaningful about underlying processes.  Hence, despite billions devoted to the study of demographic differences especially regarding health and healthcare outcomes, little has that has said about directions of changes in demographic difference over time or the comparative size of such differences from setting to setting – or even about whether a particular difference should be deemed large or small – has had a sound statistical basis.  See the October 2015 ASA letter and reference 8 (especially at 343-345).  See also the (more recent) reference 11 regarding facts that (a) a decade after the National Center for Health Statistics (NCHS) recognized that relative differences in favorable and adverse health and healthcare outcomes would commonly change in opposite directions as the frequency of an outcome changes no other federal agency has indicated an understanding that it is even possible for that to happen and (b) that NCHS has recently repudiated a decade of research that relied on its earlier guidance (with anyone’s noticing).

Correcting this situation is the larger purpose of my October 2015 letter to ASA.  

1. “Misunderstanding of Statistics Leads to Misguided Law Enforcement Policies,” Amstat News (Dec. 2012) http://magazine.amstat.org/blog/2012/12/01/misguided-law-enforcement/
2. “Things DOJ doesn’t know about racial disparities in Ferguson,” The Hill (Feb. 22, 2016) http://thehill.com/blogs/congress-blog/judicial/270091-things-doj-doesnt-know-about-racial-disparities-in-ferguson
3. “Things the President Doesn’t Know About Racial Disparities,” Federalist Society Blog (Aug. 5, 2016)http://www.fed-soc.org/blog/detail/things-the-president-doesnt-know-about-racial-disparities

4. http://jpscanlan.com/images/Letter_to_Oklahoma_City_School_District_Sept._20,_2016_.pdf
5. http://jpscanlan.com/images/Submission_of_James_P._Scanlan_in_U.S._v._City_of_Ferguson_Apr._11,_2016_.pdf
6. http://blogs.edweek.org/edweek/rulesforengagement/CRDC%20School%20Discipline%20Snapshot.pdf
7. “Race and Mortality,” Society (Jan./Feb. 2000) http://www.jpscanlan.com/images/Race_and_Mortality.pdf
8. “Race and Mortality Revisited,” Society (July/Aug. 2014)http://jpscanlan.com/images/Race_and_Mortality_Revisited.pdf

9. APA Zero Tolerance Study subpage of Discipline Disparities page of jpscanlan.com http://jpscanlan.com/disciplinedisparities/apazerotolerancestudy.html

10. http://jpscanlan.com/images/Letter_to_PAA_and_APC_Mar._29,_2016_.pdf
11. “The Mismeasure of Health Disparities,” Journal of Public Health Management and Practice (July/Aug. 2016)http://www.jpscanlan.com/images/The_Mismeasure_of_Health_Disparities_JPHMP_2016_.pdf

Politicians may rightfully be more concerned about the number of students suspended than ratios.

Item 1 below is a letter to the Antioch (CA) Unified School District regarding a suit recently brought against it by the NAACP where the complaint in the case reflects the view that generally reducing public school suspension rates will tend to decrease the proportion African Americans make up of students who are suspended.  The letter is similar to many I have posted here over the last year and a half ­– including letters of October 8, 2015 (item 2) and July 25, 2016 (item 3) urging ASA to explain the matter to arms of the federal government or the President – explaining that reducing any outcome will tend to increase, not decrease, the proportions more susceptible groups make up of persons experiencing the outcome.  

The Antioch situation also has the following interesting element.  The suit is based on an agreement requiring the school district to retain three experts (or expert groups) to analyze discipline issues.  The lead expert has consistently analyzed discipline disparities in terms of absolute differences between rates, which commonly decrease when discipline rates generally decline.  As discussed in Section B of the letter, that expert will commonly reach conclusions about the comparative size of the disparities (over time or from setting to setting) that are the opposite of the conclusions yielded by the approach reflected in the complaint.  As discussed in Section C, that expert will also commonly reach conclusion about the comparative size of disparities that are the opposite of the conclusions the second expert and the third expert (a group) would reach.

 1. http://www.jpscanlan.com/images/Letter_to_Antioch_Unified_School_District_Sept._12,_2016_.pdf

 2. http://jpscanlan.com/images/Letter_to_American_Statistical_Association_Oct._8,_2015_.pdf 

 3. http://www.jpscanlan.com/images/Letter_to_American_Statistical_Association_July_25,_2016_.pdf