American Psychological Association Zero Tolerance Report
(May 27, 2012)
In August 2006, the American Psychological Association (APA) Zero Tolerance Task Force issued a report styled “Are Zero Tolerance Policies Effective in Schools?: An Evidentiary Review and Recommendation” summarizing its review of ten years of research into the effects of zero tolerance public school discipline policies. An August 9, 2006 Press Release described its findings to the effect that “mandatory discipline can actually increase bad behavior and drop out rates.” An abbreviated version of the report was later published as an article in December 2008 issue of American Psychologist.
The report strongly recommends substantially modifying zero tolerance policies. But its main arguments are unpersuasive and commonly fail to respond that to the points to which they purport to respond. The report’s approach appears to be colored by its view (at 64) “that the use, and especially the overuse, of disciplinary removal carries with it an inherent risk of disparity for students of color and possibly for students with disabilities.” But given APA’s involvement with testing issues, one would think that, even if no one else understands that lower test cutoffs lead to large relative differences in test failure rates, APA would understand that pattern. And it should be evident to anyone who understands that pattern that overuse of suspensions will tend to cause smaller, not larger, differences in suspension rates.
In any case, given that the report’s finding have already been relied upon in further studies of the utility and harms of zero tolerance policies, and that the report’s finding may importantly influence legislative or administrative treatments of school discipline issues, many of its analytical weaknesses may eventually be treated here in some detail. Just a few aspects of the report, mainly as described in the American Psychologist article, are addressed below.
A. Consistency of Discipline
Responding to an argument that zero tolerance policies increase the clarity of disciplinary message to students, the article (at 854-55) states that discipline policies vary widely across schools and school districts.
The variation across schools and school districts, however, has no bearing on the clarity of the message to students, whose focus is solely on what to expect within his or her own school, and the article says nothing about within-school consistency. Moreover, variation from school to school – which means the some schools are not committed to a rigorous, inflexible zero tolerance policies – does not appear to be something the report is concerned about. The article’s recommendations call for greater flexibility.
B. Effects on School Climate
Responding to arguments that removal of disruptive students will improve school climate, the article (at 854) points to research showing a negative relationship between school suspensions/expulsions and school climate as well as academic achievement “even when controlling for demographics such socioeconomic status.” The article also states (id.): “Although such findings do not demonstrate causality, it becomes difficult to argue that zero tolerance creates more positive school climates when its use is associated with more negative achievement outcomes.”
As discussed on the Underadjustment Issues sub-page of the Lending Disparities page and the Disparate Treatment sub-page of the Discipline Disparities page, studies almost never adequately adjust for differences in socioeconomic status and other outcome related characteristics. More important, whereas it may too often be observed that correlation does not prove causation when there is reason to expect causation, here there is absolutely no reason to expect causation. Even where two schools have exactly the same socioeconomic situation, one would expect schools with the less satisfactory environment to have higher discipline rates because of those schools’ efforts to deal with the unsatisfactory environment. Similarly, because lower achieving students tend to be more disruptive than higher achieving students, one would expect low achievement invariably to be correlated with higher discipline rates. The findings thus provide no information whatever about whether the removal of students is making the environments better than they otherwise would be.
The report itself even suggests that school security measures make schools less safe, noting that (at 74) “[o]f schools with no reported crime, only 5% of principals reported moderate or stringent security measures; in contrast, 39% of schools with serious violent crimes reported using moderate to stringent security.” The report then sensibly notes (id):
“From one perspective, correlations showing a relationship between levels of school violence and increased use of security measures are unsurprising, and possibly influenced by the safety of the surrounding community. That is, unsafe schools might well be expected to employ more extreme measures.”
But the report then goes on to say that the data still do not support a claim that security measures make a positive contribution to school environment. The data may well not demonstrate that security measures (or stringent discipline policies) improve the school environment. But given what are likely to be strong correlations between such measures/policies and the unsatisfactory environments that cause administrators to impose them, it may be impossible to show that such measures improve school environments.
C. Deterrent Effect of Stringent Discipline
Responding to arguments that stronger discipline policies will “have a deterrent effect upon students,” the article states (at 854) that “[r]ather than reducing the likelihood of disruption, however, school suspension in general appears to predict higher future rates of misbehavior and suspension among those students who are suspended.”
1. Effects on Observers of Discipline
Putatively predictive correlations regarding disciplined students and subsequent adverse outcomes experienced by those student tells nothing about whether strong discipline policies are deterring students from engaging in conduct that will cause them to get suspended at all, which is obviously a key, if not the principal, deterrence consideration. The report itself at one point describes the deterrence issue entirely in terms of the effects on observers of the punishment, stating (at 21):
“Deter potential troublemakers (“send a message”) through the use of more severe consequences.
“Historically, the purpose of severe punishment has always been the deterrent effect that witnessing punishment might have upon others who may witness that punishment (Noguera, 1995). Ewing (2000) argues that zero tolerance ‘appropriately denounces violent student behavior in no uncertain terms and serves as a deterrent to such behavior in the future by sending a clear message that acts which physically harm or endanger others will not be permitted at school under any circumstances.’ Advocates of a zero tolerance approach may also suggest that the converse applies: that failure to punish misbehavior sufficiently will “send a message” that a school is not serious enough about safety (Larson & Ovando, 2001). “
But in its subsequent treatment of the issue, the report ignores the matter. It even casts the issue in terms solely related to the individuals who are punished, using the following heading (at 48, emphasis added):
“4. To what extent do data suggest the application of out-of-school suspension and expulsion result in improved student behavior for students who were so disciplined?” See also page 5, where in summarizing the issue the report, like the article, ambiguously refers to a “deterrent effect upon students.”
Then, after stating nothing whatever about deterrent effects on non-punished students in its treatment of the issue (at 48-50), in summarizing said treatment, the report again states (at 50, emphasis added): “Zero tolerance philosophy and practice relies to some extent on an assumption that disciplinary removal can lead to improved student behavior, either by experiencing that removal, or through the deterrent effect of observing others being removed for disciplinary infractions.”
It thus appears that an earlier draft intended in some manner to deal with the crucial question of the deterrent effects on non-punished students, but a decision simply to avoid addressing the issue was not fully implemented in the revision.
The research recommendation (item A.3.1) in the report (at 12) and the article (at 858) is explicitly limited to effects on students “who are disciplined or expelled for school due to zero tolerance policies.” Thus, the deterrent effect on non-punished student, highlighted at one point in the report, is deemed not to warrant study at all.
2. Effects on Disciplined Students
The correlations cited in the article concerning suspended students do not even suggest that the suspensions fail to have a deterrent effect among students who are suspended. To state the obvious, students who are suspended at least once are a subset of students defined by the fact that they have behavior problems of some sort. For that reason, students who are suspended commonly are more likely to be suspended a second time than other students are to be suspended a first time. First suspensions could materially deter students from engaging in conduct causing second suspensions and the group of students experiencing a second suspension easily would remain more likely to receive another suspension than students never previously suspended. The same holds for the types of measures of the report recommends.
Suppose that suspension was entirely eliminated for sixth graders and that sixth graders who would previously have been suspended are instead provided with intensive counseling. Suppose also that the counseling should prove extremely effective in reducing disruptive behavior of students diverted to it – which is to say that it would reduce that behavior in a large proportion of cases. One still would likely find that students who were referred to counseling in the sixth grade are more likely to be suspended in the seventh grade than students who were not referred to counseling. That would hardly suggest that counseling in fact is causing suspension rates to be higher than they otherwise would be.
The report itself (at 46) discusses a negative relationship between the number of suspensions for a 6th grader and his/her achievement in math and reading in the 7th and 8th grades. But it hard to imagine that there would not be such a correlation. Students who do well in school find it a rewarding experience, which tends to favorably influence their conduct; students who do poorly find it a frustrating experience, which tends to adversely influence their conduct. Even if suspensions had a positive influence on student achievement in subsequent years, there still would be reason to expect a negative correlation between suspensions and achievement in subsequent years.
Assuming that there are further efforts to study the effects of discipline on disciplined students, one hopes it will be carried out with a better recognition of the extent to which correlations between punishment and future adverse outcomes (which is to say between behavior deemed by some to warrant discipline and future such behavior and other adverse outcomes) are inevitable than is reflected in the report. One also should recognize that the more discipline and expulsion are restricted to increasingly severe forms of misconduct, the stronger will be the correlation between receipt of such punishment (as correlation is typically measured)[i] and future adverse outcomes.
D. Racial Disparities
Responding to arguments that zero tolerance policies, by reducing discretion, may reduce racial disparities, the article (at 854) states that the evidence does not support such claim, observing that “the disproportionate discipline of students of color continues to be a concern” and that “overrepresentation in suspension and expulsion has been found consistently for African Americans.”
But that statement says nothing whatever about whether the limitation of discretion has made disparities smaller than they otherwise would be. Assuming that racially disparate treatment of students who engage in like conduct plays some role in racial difference in discipline rates, the notion that limitation on discretion would reduce that role is certainly plausible enough. It may well be that it is difficult to study whether zero tolerance reduces the possible role of disparate treatment in racial differences in discipline rates, or that no effort has been made to study it. But the absence of study does nothing to undermine the plausibility of the claim.
Whether reducing discretion reduces disparate treatment is another issue that is difficult to study, given that (as discussed in the Disparate Treatment sub-page) it is difficult to determine what role of disparate treatment plays in difference in discipline rates. Any effort to address that question must be informed by recognition of the flaws in standard measures of differences in outcome rates during a period of changing overall discipline rates. That is, suppose that under more stringent policies, the white suspension rate increased from 5% to 10% (a 100% increase) while the black rate increased from 12.7% to 21.8% (a 71% increase). Correspondingly, the ratio of the black rate to the white rate would decline from 2.5 to 2.2. While most researchers would regard these changes to reflect a reduction in the disparity, in fact (as reflect in Table 1 of BSPS 2006) the changes were entirely consistent with no change at all in the strength of whatever force is driving the differences. See the Relative Versus Absolute sub-page of the Measuring Health Disparities page.
With regard to whether bias is playing any role in the observed differences in rates, the article observes (at 854, bracketed numbers added):
“[1] The evidence shows that such disproportionality is not due entirely to economic disadvantage (Skiba et al.,2002; Wu et al., 1982), [2] nor are there any data supporting the assumption that African-American students exhibit higher rates of disruption or violence that would warrant higher rates of discipline.”
Yet the implication in [1] that economic disadvantage (and behavior associated with it) play some role in the disparity contradicts the claim in [2] that there is no evidence of difference in behavior. Whether differences in treatment for similar conduct play a role, and the size of that role, has never been properly studied – i.e., with a recognition of the inevitability of underadjustment of differences in outcome-related characteristics and the ways standard measures of differences are affected by the overall prevalence of an outcome (as discussed on the Disparate Treatment sub-page and two paragraphs above).
E. Disparities by Gender and Disability
The article and report discuss disparities in suspension by gender and disability. While the report leave leaves open the possibility that the disparities are results of differences in behavior, the report (at 13), as Recommendations A.3.4 and A.3.5, calls for further research on the matter. The article’s listing of recommendation does not include such recommendations. Item A.3.5 has been eliminated and A.3.4 has been changed to a recommendation for studying the economic consequences of removing students from school.
A general shortcoming of the report is the failure to recognize the compelling reasons to expect differences by gender and by disability (when behavioral disorders are included among disabilities). When things are observed that it would be absurd not to expect, a question arises as to whether they even warrant discussion much less whether they warrant further research. Thus, the article’s elimination of the recommendation for study of such issues was certainly a sensible course. Ideally, however, the decision to eliminate that recommendation from the article would have been explained. For the report’s recommendation may yet cause resources to be devoted to highly questionable research.
See the Disabilities – PL 108-446 sub-page of the Discipline Disparities page, which sub-page addresses provisions of the Disabilities Education Improvement Act that require responses to observed disability-related differences in discipline rates that would likely increase those differences.