Study 1
Scheuermann, Martinez–Prather, and Petrosino (2022) conducted a clustered randomized controlled trial to assess the effectiveness of a comprehensive research-based framework for implementing school-based law enforcement programs on school climate, student discipline, and students’ perceptions of police over 2 full school years. The sample included 25 schools from six separate school districts in suburban and urban areas of Texas. The treatment group included 13 schools, and the waitlisted control group included 12 schools.
In addition to the components of the framework described in the Program Description, officers in the treatment campuses all received the Texas state-mandated, school-based law enforcement training course, and school staff received training on the history of law enforcement in schools, their roles and responsibilities, how to collaborate with school law enforcement, and awareness about the school policing framework and the established campus goals and activities. Beyond implementing the framework components, treatment schools were required to meet four additional deliverables: 1) a project planning meeting with key project staff and all key campus and district stakeholders to achieve a common understanding of the research project and framework processes; 2) a goal-setting meeting with key implementation team staff and campus framework committee members in September of each school year to ensure goals were data driven, relevant to school policing, realistic, and measurable and to ensure campuses had a thorough plan for achieving their goals; 3) monthly check-in meetings from October through May with key implementation staff to provide updates on progress toward goals, review collected data, and discuss whether adjustments were needed at that time; and 4) a final debrief meeting to discuss whether goals were achieved, direction for moving into the next school year, and feedback on the framework manual and processes. Schools in the control group did not use the school-based law enforcement framework and continued to run their school-based law enforcement programs as they saw fit; however, control schools did implement state-mandated training for School Resource Officers.
School climate was measured as student’s self-reported victimization (for example, “Have you been pushed, shoved, slapped?”), delinquency (e.g., “have you been in a physical fight”), caring adult–student relationships (e.g., “There is an adult who really cares about me”), and school connectedness and safety (e.g., “I feel safe in my school”), using items from WestEd’s California School Climate, Healthy, and Learning Survey. Rule clarity items were derived from the Delaware School Surveys scale (e.g., “Students know what the rules are”) and school bonding items came from the Psychological Sense of School Membership scale (e.g., “I wish I were at a different school”). The measurement of student perceptions of police used eight items (for example, “The officer makes me feel safe”) from an internally consistent scale. To examine student disciplinary actions, student incident data were collected directly from participating school districts and contained information related to the nature of the incident, disciplinary action, and student characteristics, including race, age, grade, and disability status.
At baseline, there were 5,671 student survey responses in the treatment group and 5,036 student survey responses in the control group. There were baseline differences between treatment and control group students on demographic variables (student grade level, the number of Hispanic and white students and students with above-average grades), and the outcome variable adult–student relationships. Linear regression models controlling for difference in baseline variables were employed to estimate the differences in self-reported student victimization and delinquency, school climate, and student perceptions of police in treatment and control schools after 2 years. To compute the effects of the intervention on student discipline, a simple measure of mean differences was computed for count variables and odds ratio for prevalence variables. The magnitude of effect was then computed using Hedges’ g for the count variable and the Cox index for prevalence variables. No subgroup analysis was conducted.