This page can be compared to the Oakland (CA) Disparities page in that both pages involve situations where, apparently as a result of the failure to recognize that absolute differences between rates and relative differences between rates can (or typically do in the case of discipline disparities) change in opposite directions as discipline rates decline generally. Very likely, all the situations discussed in the prior paragraph where general reductions in discipline rates were accompanied by increased relative differences in discipline rates, the absolute difference declined.
In August 2018, the University of Pittsburgh released a study titled “Just Discipline and the School-to-Prison Pipeline in Greater Pittsburgh: Local Challenges and Solutions.” Broadly, like many documents discussing ways to reduce racial disparities in discipline, the study reflected the view that generally reducing discipline rates would tend to reduce relative racial differences in discipline rates. But the study also revealed that general reductions in discipline rates had been accompanied by an increase in the relative difference in discipline rates, although that could be divined only from a careful reading of the report (aided by the understanding that typically general reductions in discipline rates tend to be accompanied by increased relative differences in discipline rates).
The study summarizes the situation at page 3 in the following terms. of less repeats at page 14):
Our findings on racial disparities in local suspension use were startling. Overall, in 2015-2016, Black students across Allegheny County were suspended at a rate of approximately 41.0 students per 100, as compared to only 5.6 suspensions for every 100 non-Black students. This difference equates to Black students in Allegheny County being subjected to suspension rates that are 7.3 times [as high as (see the Times Higher subpage of the Vignettes page of jpscanlan.com] the rate of non-Black suspensions, a disparity rate that is above the statewide level of 5.5 to 1. In terms of individual districts, 73% (37 of 51) of Allegheny County districts had suspension rates for Black students that were at least double the rate of their non-Black counterparts.
The study thereafter makes many references to the size of the ratio of the black suspension rate to the white rate, restating (at page 14) essentially verbatim the description of the 2015-16 disparity on page 3.
On page 18, in Table 3, the study shows larger percentage point reductions in suspensions for black than non-black students in particular districts, which is what typically occurs when suspension rates generally decline, and which may in some cases (though usually does not) also involves larger percentage reductions for blacks than whites. One would need the actual rates to know how the black and non-black percentage changes.
On page 19, in Table 4, the study shows percentage point reductions for blacks in certain districts where the white rates did not decline, which would necessarily mean larger percentage reductions for blacks than whites.
But the report then summarizes the situation in the following terms (at 19):
In addition to overall suspension rates, we also examined the degree to which race-specific trends were visible, and results show that county-wide the suspension rates for both Black and non-Black students decreased over this time span. The greater absolute decrease happened among Black students, with the overall suspension rate for this group decreasing by 7.4 percentage points, from 47.4 per 100 in 2012-2013 to 41.0 per 100 in 2015-2016 (a 14% rate reduction). The decrease for non-Black students was much more modest in absolute terms, but still notable—from 7.0 in 2012-2013 to 5.6 per 100 in 2015- 2016.
But the percentage reduction for non-black students was 20% compared with the 14% reduction for black students, which results in an increase, not a decrease, in the ratio of the black rate to the non-black rate. As reflected by the 2012-13 figures, in that school year, the black rate was 6.8 times the non-black rate. The 7.3 black/non-black ratio highlighted in the report is thus an increase, not a decrease, from the earlier period when suspensions were more widely used.
Thus, while there may have been some variation in certain districts, the overall result was what usually occurs in the circumstances, a larger percentage point reduction for the blacks than non-blacks but a larger percentage reduction for non-blacks than blacks, with a corresponding decrease in the absolute difference between rates but an increase in the relative difference between rates that is commonly highlighted as a reason to generally reduce discipline rates.
The study also discusses that the overall disparity is a function of high suspension rates in districts where black students are concentrated and exceptionally high racial disparities in suburban districts. The exceptionally high racial disparities (in relative terms) are functions of the fact that rates are generally low in such districts., as discussed on the Suburban Disparities page. But the suburban districts will also have comparatively small absolute differences between rates. For an example, see Table 7 of the Commission on Civil Rights testimony, regarding Loudoun County, Virginia, which shows comparatively small absolute differences between rates. In that case, however, the district also shows a smaller relative racial difference than if found nationally. That is because all measures of difference are functions of both (a) the prevalence of an outcome, (b) the strength of the forces causing black and white outcome rates to differ. That latter is comparatively small is Loudoun County. Nevertheless, that district believes it has comparatively large disparities because it measures them in terms of the relative difference between the proportion blacks make up of students and the proportion they make up of suspended students. That approach tends to cause areas where blacks make up comparatively small proportions of students to seem to have comparatively large disparities. See the Loudoun County (VA) Disparities page and the Letter to Loudoun County Public Schools (Sept, 5, 2017).
It is important to distinguish between the (a) the relative and absolute differences between rates and (b) the relative and absolute differences between the proportion a group makes up of students and the proportion it makes up it makes up of suspended students. When discipline rate are generally reduced, the relative difference between black and white rates tend to increase while the absolute difference between those rates tends to decrease; but when rates are generally reduced, both the relative and absolute differences between the proportion blacks make up of students and the proportion they make up of suspended students tend to increase.
The relative and absolute differences between rates are unsound measure of association because they tend to be affected by the prevalence of an outcome, as discussed, for example, in “Race and Mortality Revisited,” Society (July/Aug. 2014) and score of other places. Relative and absolute difference between the proportion a group makes up of persons potentially experiencing an outcome (the pool) the proportion it makes up of persons actually experiencing an outcome are unsound measures of association for the same reason, but they are additionally unsound because of the varied effects of the proportion a group make up of the pool on the those two differences. One aspect of the latter problem is summarized fairly succinctly in Letter to the Comptroller General of the United States (Apr. 12, 2018). More complex issues are addressed in the IDEA Data Center Disproportionality Guide page slides 98 to 108 of the 2014 University of Maryland workshop.
See the Rhode Island Disparities page regarding a situation where a study confused the proportions blacks and whites made up of suspended students with the proportion of each group that were suspended. Due to that confusion, the study reported that the black rate had reached a high point over a particular period peak while the white rate had reached a low point, in fact both rates were at or about low points for the period.