James P. Scanlan, Attorney at Law

Flawed Inferences – Discipline

(Jan. 18, 2013; rev. Jan. 22, 2014)

This page is closely related to the Offense Type Issues subpage of the Discipline Disparities page of jpscanlan.com

Eventually there likely will be on this site a page generally addressing flawed inferences based on the comparative size of some measure of disparity and discussing the failure of those drawing such inferences to understand the way the measure on which they rely is affected by the prevalence of an outcome or that, when such measure is a relative difference in experiencing an outcome, the relative difference in the opposite outcome commonly would support an opposite inference.  See discussion at pages 39-41 of Harvard University Measurement Letter.  See also the Criminal Record Effects subpage of the Scanlan’s Rule (regarding interpretations of effects of criminal records on employment opportunities of black and white job applicants) and Disparities – High Income subpage of the Lending Disparities (regarding interpretations of increasing income on borrowing prospects of black and white mortgage loan applicants).  Some of these issues are also addressed in the “The Mismeasure of Discrimination,” Faculty Workshop, University of Kansas School of Law, Sept. 20, 2013.

The instant page is devoted to addressing that issue in the context of discussion of school discipline.    

One instance where one observers drew inferences based on perceptions about the size of a disparity with respect to a discipline disparities issues is found in discussion of whether, or the extent to which, racial differences in discipline rates are results of racial differences in behavior.  In such context it is occasionally noted that that racial differences tend to be larger in circumstances where objective judgments are involved (as at page 11 of the October 2011 report of the National Education Policy Center (NEPC) of the School of Education of the University of Colorado Boulder styled “Discipline Policies, Successful Schools, and Racial Justice” that is the subject of the NEPC National Study subpage).  Those making the point typically do so to suggest that such pattern supports the view that observed disparities are in whole or in part functions of racial bias.  See the Offense Type Issues subpage.

As in other contexts where observers draw an inference on the basis of the perceived size of disparity, those making such point do so without consideration of the way the measure on which they rely (assuming they in fact know what such measure is) is affected by the prevalence of an outcome or the way other measures would tend to support different conclusions.  Whether or not, properly analyzed, one might draw an inference based on the comparative size of disparities would turn on the underlying figures.  I have so far been unable to secure the underlying data in order to determine what they might show.

I also note the following instance where in the discipline context observers discuss the comparative size of an effect with understanding the way a measure is affected by the prevalence of an outcome.  This instance may not fall squarely into the subject of this subpage, save in the sense that any discussion of the comparative size of an effect involves the drawing of an inference – as, for example, when an observation that a health disparity has changed over time is drawing an inference based on the comparative size of the difference between outcome rates at two points in time.  See page 17 of the Harvard University Measurement Letter.  But I believe it warrants mention here.

Paragraph 6 of Department of Justice complaint against Meridian, Mississippi, in discussing the detrimental consequences of contact with the criminal justice system that the complaint maintains are suggested by research,[i] notes:  “Even one court appearance during high school increases a child's likelihood of dropping out of school, and court appearances are especially detrimental to children with no or minimal previous history of delinquency.”   

But the association between a court appearance and likelihood of dropping out of school and other adverse consequences, as measured by the relative differences between the rates of experiencing such outcomes of those with such court appearances and those without such court appearance, will generally be higher among those with minimal or no previous history of delinquency simply because those with minimal or no previous history of delinquency have lower baseline rates for such outcomes.  The relative difference for rates of avoiding adverse outcomes will likely smaller for such persons.  See the Subgroup Effects and Subgroup Effects – Nonclinical subpage of the Scanlan’s Rule page. 

Flawed Inferences – Discipline

(January 18, 2013)

One instance where one sees observers drawing an inference based on perceptions about the size of a disparity with respect to a discipline disparities issues is found in discussion of whether or the extent to which racial differences in discipline rates are results of differences in behavior.  In such context it is occasionally noted that that racial differences tend to be larger in circumstances where objective judgments are involved (as at page 11 of the October 2011 the National Education Policy Center (NEPC) of the School of Education of the University of Colorado Boulder issued a report styled “Discipline Policies, Successful Schools, and Racial Justice” that is the subject of the NEPC National Study page.  Those making the point typically do so to suggest that such pattern supports the view that observed disparities are in whole or in part functions of racial bias. 

As in other contexts where observers draw an inference on the basis of the perceived size of disparity, those making such point do so without consideration of the way the measure on which they rely (assuming they in fact know what such measure is) is affected by the prevalence of an outcome or the way other measures would tend to support different conclusions.  Whether or not, properly analyzed, one might draw an inference based on the comparative size of disparities would turn on the underlying figures.  I have so far been unable to secure the underlying data in order to determine what they might show.

I also note the following instance where in the discipline context observers discuss the comparative size of an effect with understanding the way a measure is affected by the prevalence of an outcome.  This instance may not fall squarely into the subject of this sub-page, save in the sense that any discussion of the comparative size of an effect involves the drawing of an inference – as, for example, when an observation that a health disparity has changed over time is drawing an inference based on the comparative size of the difference between outcome rates at two points in time.  See page 17 of the Harvard University Measurement Letter.  But I believe it warrants mention here.

Paragraph 6 of Department of Justice complaint against Meridian, Mississippi, in discussing the detrimental consequences of contact with the criminal justice system that the complaint maintains are suggested by research,[i] notes:  “Even one court appearance during high school increases a child's likelihood of dropping out of school, and court appearances are especially detrimental to children with no or minimal previous history of delinquency.”   

But the association between a court appearance and likelihood of dropping out of school and other adverse consequences, as measured by the relative differences between the rates of experiencing such outcomes of those with such court appearances and those without such court appearance, will generally be higher among those with minimal or no previous history of delinquency simply because those with minimal or no previous history of delinquency have lower baseline rates for such outcomes.  The relative difference for rates of avoiding adverse outcomes will likely smaller for such persons.  See the Subgroup Effects sub-page of the Scanlan’s Rule page.