Inequitable Compliance: Implementation Failure of a Statewide Student Discipline Reform

ABSTRACT

Researchers and others in the education community are concerned about over-reliance on exclusionary discipline such as out-of-school suspension (OSS) and expulsion. Exclusionary discipline is associated with lower student achievement, higher risk of drop-out or grade retention, and involvement in the juvenile justice system. In response, many schools, districts, and states are moving toward less exclusionary consequences. In 2013, the Arkansas state legislature passed a bill prohibiting the use of OSS as a consequence for truancy. Yet, even after 3 school years, there has not been a complete elimination in the use of OSS for truancy. In this article, I use 8 years of student achievement, demographic, and disciplinary data from all Arkansas public schools to assess which school-level factors are associated with the use of OSS as a consequence for truancy in the year the law was passed, and which are associated with policy compliance in the following 3 years. I find schools that served more minority students, had higher rates of truancy, and higher rates of OSS were less likely to comply with the policy, all else equal. Combined, these results suggest that the types of schools likely targeted by this policy are the same ones not fully complying with it. As a result, the impact of policy reforms, which sound beneficial prima facie, may be limited if changes are not communicated well to schools, if there is not accountability to ensure compliance, and if there is not school capacity to handle discipline effectively.

Notes

1 For example, parents can be held civilly or criminally liable for their children's misbehavior (Geis & Binder, 1991; Siegel, 2002).

2 The primary goal was to understand the school characteristics associated with implementation of this policy. Alternatively, I could have used hierarchical linear modelling to incorporate the variables at multiple levels and predict outcomes at the student level. In the future, estimating whether individual student characteristics are associated with policy noncompliance would be a worthy topic of study as well.

3 From 2008-2009 to 2013-2014, ELA and math exams were administered in grades 3-8, and end-of-course (EOC) examinations were administered in Algebra I, Geometry, Algebra II, and 11th grade literacy as part of the Arkansas Comprehensive Testing, Assessment, and Accountability Program. In 2014-2015, the state administered the Partnership for Assessment of Readiness for College and Careers (PARCC) tests including literacy exams in grades 3-10, math exams in grades 3-8, and end of course exams for Algebra and Geometry. In 2015-2016, Arkansas administered the ACT Aspire summative assessments in math and ELA in grades 3-10.

4 Other nonspecified consequences were coded as a particular type of consequence at the school level, but when combined and reported by the state, they are grouped into an other category. This group was labeled as other when in the data provided, and is not a researcher-created group of consequences. There is a similar other category for infractions.

5 It is unknown whether days of Saturday school are included in reported attendance data.

6 About 80% of the schools reporting no truancy are elementary schools, which typically represent less than 50% of all schools. Schools reporting no truancy are proportionally distributed across the state's five regions. Schools with no truancy are slightly poorer on average (the average school with no truancy is 66-67% FRL instead of 64-65% for the average school in the state). Finally, schools with no reported truancy serve a similar share of black students (20% in the average school with no truancy, compared to 21% for the average school in the state).

7 These rates include multiple incidents per student, if applicable. Thus, given that there are some students with multiple truancy incidents per year, the percent of students written up for truancy in an average school is less than 4.1% per year.

8 I contacted an assistant superintendent/former principal, a principal, and a former teacher in three of the school districts with the largest number of instances of other consequences administered for truancy in the past 2 years (2014-2015 and 2015-2016). The superintendent/former principal stated that the response to truancy typically is ISS, but the other consequences may include loss of course credit or credit recovery (making up missed instructional time). For example, seniors may be required to come in extra days after they would usually be required to be at school. The high school principal indicated that his school primarily uses after school suspension and Saturday school as a consequence for truancy, practices which do not exclude the truant student from the regular learning environment. A former teacher from another high school indicated that some of the other consequences are likely a mix of morning detentions and students being sent home for part of the day.

9 The percent of students who are identified as limited English proficient in each school was not included as an explanatory variable due to its high correlation with percent Hispanic (r = 0.946).

10 In some models, these charter school indicators were dropped due to perfect collinearity.

11 As of 2015-16, there were 259 districts in Arkansas, 192 represented in the models restricted to schools with at least one truancy incident, and 140 represented in the models restricted to schools with at least five truancy incidents.

12 The goal was to study patterns of compliance among the schools that, up to the point of the policy change, were the ones using OSS as a consequence for truancy. Otherwise, I would be measuring patterns of compliance including some schools who had already chosen to comply pre–policy, which would confound the estimates of interest.

13 These results are available from the author by request.

14 Some coefficients lost or gained significance (school percent black, school percent other minority, high school, middle school, some region indicators, and some of the controls for prior year discipline rates). None of the significant relationships remained significant but changed direction. Tables available from author by request.

15 The correlation between school percent limited English proficient and school percent Hispanic is 0.946. Hispanic (and English language learners) are concentrated in a few large districts some of which also rely heavily on other consequences. In 2012–2013, 50% of Hispanic students and 60% of ELL students were in only five districts of the state, which are also five of the top six in terms of student enrollment (out of 258 traditional public school districts).