Study 1
Cook, Murphy, and Hunt (2000) conducted a randomized controlled trial to examine the effect of Comer’s School Development Program on school climate and student behaviors in 19 inner-city Chicago, Ill., middle schools.
The program was implemented over 4 years with a pilot year and two phases. The pilot year (the 1991–92 school year) included four treatment schools where the program was implemented. Phase 1 (1992–93) included eight middle schools, which were matched for 2 prior years on academic achievement and racial composition, before being randomly assigned to treatment or control status. Phase 2 (1993–94) included 12 middle schools from less economically affected parts of the city compared with Phase 1 schools, using the same matching technique. For various reasons, 5 of the 24 schools dropped out of the study, leaving a final sample of 19 schools. The pilot schools were added to the Phase 1 schools, thus creating 8 Phase 1 schools (4 treatment and 4 control), and 11 Phase 2 schools (6 treatment and 5 control). Control schools had the option to enter the Comer School Development Program 4 years later.
There were 10,306 male and female students in grades 5 through 8 who were surveyed on at least one measurement occasion (referred to as the cross-sectional sample). Of these, 1,685 were first eligible to be in the study in grades 5 or 6 in the fall of 1993 or 1994 and remained in the same school for 3 or 4 years between grades 5 and 8. The longitudinal sample comprised these students. Students in both treatment and control schools were mostly minority, including Black (81 percent of the treatment schools, and 74 percent of control schools), Hispanic (13 percent of treatment schools, and 18 percent of control schools), or Asian (4 percent of treatment schools, and 5 percent of control schools). Most students also came from a low-income background (91 percent of treatment schools, and 94 percent of control schools) and lived with at least one biological parent (45 percent of treatment schools, and 46 percent of control schools). There were no statistically significant differences between treatment and control schools on school demographics and student reports of family demographics and processes, but treatment students scored below control school students on achievement tests (math and reading scores).
The Chicago Public School System provided annual data on each school’s enrollment, attendance, student mobility, percentage of students classified as having limited English proficiency, and percentage of low-income students. In the first study year (or the second year for pilot schools), questionnaires were administered to students in late fall to provide a baseline for the written outcome measures. Subsequent outcome testing took place in the late spring of each school year. Only students from grades 5 through 8 were sampled because of the need to complete written questionnaires on outcome measures and school climate. Student outcome measures were organized into four domains: 1) mental health (including anger), 2) negative social behaviors (consisting of an “acting out” scale that covered normative beliefs about misbehavior, questions about substance use, etc.), 3) positive social behaviors, and 4) academic achievement. The student questionnaires also included reports of the student’s sex, race, parents’ or guardians’ education and work status, and household composition, and family process variables such as measures of home academic support, use of reasoning in parenting practices, parental rule setting, and the quality of parent–child affective relations.
To examine student outcomes measuring lack of anger, acting out, substance use, and disapproval of misbehavior at the 3-year follow-up, analyses were conducted at both the school-level and individual-level. The CrimeSolutions review of this study concentrated on student-level results. Hierarchical linear modeling analyses that used occasions of measurement within individuals as the first level, individuals within schools as the second, and schools as the third were tested. The last observation (eighth grade) was chosen as the zero-reference point for the longitudinal student sample. At the individual student-within-schools level, race, gender, family composition, parental employment status, and third-grade combined Iowa Test of Basic Skills test scores in math and reading were used as covariates. The school-level covariates were Comer, phase, and the Comer-phase interaction. No subgroup analysis was conducted.