![Sample our Education journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5932/TOC-Subject-Sample-2021-Education.jpg"})
ABSTRACT
To address disproportionalities in suspensions for students of color, many districts have prohibited schools from suspending students for willful defiance of school authorities and implemented restorative justice programs (RJP) that address student misconduct using alternative conflict resolution practices. However, there is limited evidence on the efficacy of these new policies. This article examines how the Los Angeles Unified School District's (LAUSD) suspension ban and RJP are associated with student suspensions over time, as well as suspensions across different groups of students and schools targeted by these policies. We employ an interrupted time series design using 12 years of studentlevel administrative data from the 2003–2004 to 2014–2015 school years, which allows us to examine trends in student suspensions in LAUSD before and after its suspension ban in 2011–2012 and rollout of restorative justice practices and training to schools in 2014–2015. We find large rates of decline in suspensions in the years following LAUSD's suspension ban relative to the years leading up to the ban, as well as evidence of reduced suspension gaps between frequently disciplined students and their less-disciplined peers. Additionally, we find the district's public identification of schools most in need of continued reform, and provision of restorative justice training to these schools, is associated with further reductions in suspensions. Despite this progress, suspension gaps between black and non-black students, and between special education (SPED) and non-SPED students, still persist in our data, suggesting that districts may need more time and comprehensive strategies to fully resolve these inequities.
Acknowledgments
We gratefully acknowledge support from the United States Department of Education Investing in Innovation Grant. We also greatly appreciate assistance from Matt Hill, Cynthia Lim, Donna Muncey and many other administrators within the Los Angeles Unified School District. In addition, we have benefited greatly from helpful feedback from the editors and a set of anonymous reviewers. The views expressed in this article do not necessarily reflect those of the University of Southern California, Michigan State University, Los Angeles Unified School District, or the study's sponsors. Any remaining errors are our own.
Notes
1 As we discuss herein, LAUSD released public documents prioritizing schools with these demographics and suspension trends for intervention in its suspension ban and for receiving restorative justice training.
2 Losen and colleagues (Citation2014) compared suspension rates in California as a whole and in LAUSD from 2011–2012 to 2012–2013. Both years fall within LAUSD's suspension ban, again making it difficult to assess if suspension rates fell in this period to a greater extent than in years prior to the ban. Nevertheless, these results are still insightful in that they show a reduction in suspension gaps between students of color and white students.
3 These intervention strategies included Tier 1 practices for building schoolwide norms for positive student behavior and reinforcing and monitoring such behavior (e.g., teaching and modeling schoolwide behavior expectations, positively reinforcing good behavior); Tier 2 practices that utilize school supports to meet the needs of students who are unresponsive to Tier 1 practices (e.g., targeted social skills instruction, academic support); and intensive Tier 3 practices that provide specialized supports for students who are unresponsive to Tier 1 and 2 practices (e.g., individualized behavioral intervention plans, intensive academic support and social skills counseling).
4 The district defined disproportionality in suspensions as black students being over-represented in the population of suspended students relative to the district's total student enrollment (LAUSD, Citation2012).
5 In this publicly released list, the district counted magnet centers operating on traditional school campuses as separate schools. However, in our data, these magnet centers have the same school code as their colocated traditional schools, which is consistent with how LAUSD reports its school-level data to the California Department of Education. As such, the number of schools that we report as receiving training in our article is lower than the total number reported by the district. During the time frame of our analysis, LAUSD had not yet developed procedures for requiring approximately ∼220 independent charter schools operating in the district to adhere to its suspension ban and RJP. As such, we do not include these schools in our analysis.
6 Given this phased training schedule, LAUSD would have provided complete restorative justice training to 19% of its schools by 2015-2016 (Cohort 1), 37% of schools by 2016-2017 (Cohorts 1 and 2), 62% of schools by 2017-2018 (Cohorts 1-3), 82% of schools by 2018-2019 (Cohorts 1-4), and 100% of schools by 2019-2020 (Cohorts 1-5).
7 For example, schools that received restorative justice training received in-depth professional development (developed by the district's restorative justice specialist and counselors) on restorative justice principles, empathy/team building, defusing disruptive behavior in the classroom, and procedures for training school staff members on restorative justice practices. Schools that were not yet scheduled to receive restorative justice training participated in a Positive School Climate Awareness Kick-Off meeting, received a principal toolkit with guidelines on how to initiate restorative justice practices and communicate the district's new philosophy on student discipline at their school sites, and could receive ongoing support from their ESCs.
8 The restorative justice teacher adviser was provided as an additional full-time staff member for schools. LAUSD hired from across the district to fill these roles.
9 Our original dataset includes 7.32 million student-year observations and 1.51 million individual student observations of students enrolled in K-12th grade in 804 LAUSD schools, excluding students in independent charter schools. From this sample, we drop ∼62,000 individual student observations of students who do not demonstrate normal grade progression (4.1% of students in our original sample). We further restrict our sample to students in schools that continued to be operational until the 2014–2015 school year (the year in which LAUSD publicly identified cohorts of schools for restorative justice training), dropping 18 schools that closed before the 2014–2015 school year.
10 We create separate flags for structural student mobility (defined as students switching schools to enroll in a school that serves their grade-level) and nonstructural student mobility (defined as students who leave their school for other reasons). We do not count matriculating students, students who leave the district under extenuating circumstances, or students who moved between PSCI relief schools and the neighboring feeder schools.
11 We cannot track in-school suspensions in our data which could lead us to under-report the number of student suspended. However, drawing on publicly available data on student suspensions from the California Department of Education (CDE) which tracks both in- and out-of-school suspensions from 2011–2012 to 2016–2017, we can confirm that the unduplicated counts of suspended in our administrative data are similar to those reported by LAUSD to the CDE. This suggests that we are still capturing most instances of student suspensions in our data.
12 For example, the Great Recession may have resulted in increased student-teacher ratios, which could have lowered the quality of classroom management and increased student suspension rates during the time frame of our analysis. To that end, our data show that student-teacher ratios increased from 19.2 in the 2008-2009 school year to 21.0 in the 2009-2010 school year and stayed at this level until the 2013-2014 school year. This may lead us to underestimate the true effects of the suspension ban. In addition, the Great Recession may have increased incidences of student misbehavior due to extenuating circumstances at home (e.g., job loss, economic hardship).
13 We use a logistic regression model because it respects the boundary of our dependent variable (which ranges from 0 to 1) and allows for differential rates of change at the low and high ends of our independent variables. This latter affordance is important because we might expect to see slower rates of decline in the probability of suspensions over time as LAUSD gets closer to its goal of reducing student suspensions. We obtain similar results when we run probit models (available upon request).
14 Although we would have liked to model suspension trends separately for other race students who are Native American, Alaskan Native, Pacific Islanders, and Filipino, LAUSD enrolls small numbers of these students, precluding us from observing precise effects. We group Asian students with white students since both student groups are relatively advantaged in terms of achievement and exposure to suspensions.
15 LAUSD has several schools (around ∼100,000 students) that do not follow traditional grade-level configurations. For these schools, we identify students’ school level by their enrolled grade (e.g., grade 1 is coded as elementary).
16 Our indicator for students who moved to a new school from the previous school year includes both structural movers who transitioned out from a school that does not serve their grade level and nonstructural movers who transitioned out from a school for other reasons.
17 We do not control for student or school-level achievement in our models because our policy overlaps with California's transition from the California Standards Tests to the Smarter Balanced Consortium Assessments for Common Core. Because of this transition, we do not have student assessment data in the 2013–2014 school year, which is right in the middle of the suspension ban. That said, our indicators for low-performing schools allow us to control for some of the differences in school achievement and performance in our data.
18 We do not control for the percent of under-represented minority students in schools because of its high correlation with the percent of FRL eligible students. Results are consistent when we use percent minority.
19 Ideally, we would have interacted our continuous year variable (Yeart) with covariates for frequently disciplined students and school cohorts to control for differences in pre-treatment trends (Somers, Zhu, Jacob, & Bloom, Citation2013). Results are consistent when we do so, and most of the pretreatment trend coefficients in our models are not statistically significant (available upon request). However, these models suffer from multicollinearity issues that do not allow us to predict marginal effects from our logit models, so we report marginal effects from non-interacted models.
20 It is important to emphasize that we can only comment on differential trends for student groups and cannot comment on disproportionalities in suspensions as per LAUSD's definition. For example, if we see a negative and significant effect for β6Blacki*YR1t, this would suggest that black students experience a larger drop in the probability of suspension relative to white and Asian students in year 1, but does not tell us if black students are more or less represented among suspended students relative to their enrollment in LAUSD schools.
21 We confirm that Cohort 1 and 2 schools meet the district's criteria for prioritizing restorative justice training by comparing these schools to later cohorts in terms of the percent of enrolled students who are black, FRL and SPED eligible, as well as school suspension rates and school disproportionalities in suspension for black students (measured as the difference in the percent enrollment of black students and the percent of suspended students who are black). We make these comparisons in the 2010-2011 school year (before LAUSD adopted its suspension ban) and in 2012-2013 (before LAUSD released the list of schools to be prioritized for restorative justice training in 2013-2014). In both years, we find that Cohort 1 schools enrolled a significantly higher proportion of black students and had significantly higher suspension rates and disproportionalities in suspensions for black students than Cohort 2-5 schools. We also find similar differences between Cohort 2 schools and Cohort 3-5 schools, although the magnitude of these differences is smaller. We find that Cohort 1 schools are comparable to all other cohorts of schools in terms of SPED and FRL enrollments. However, Cohort 2 schools enroll significantly higher numbers of FRL students than later cohorts.
22 These predicted suspension rates adjust for student and school-level covariates specified in models 1-2 that might influence the probability of suspension (e.g., race/ethnicity, ELL status, pupil-teacher ratios).
23 Because we only have two years of pretreatment data prior to LAUSD's adoption of the SWPBIS program, we cannot verify these descriptive trends in our ITS models.