Principal Investigators: Eric Schaps, Daniel Soloman, Victor Battistich
Level of Intervention: Universal
Target Population: elementary-aged children
References: Battistich, Schaps, Watson & Solomon (1996); Solomon, Watson, Battistich, Schaps & Delucchi (1996); Watson, Battistich, & Solomon (1997); Solomon, Watson, Delucchi, Schaps, & Battistich (1988)
Theory (Risk & Protective Factors Targeted):
Derived from socialization, learning and motivation, and prosocial development theories, the Child Development Project is a comprehensive educational "reform" model, intended to transform schools into "caring communities of learners". Its focus is on enhancing protective factors, including school bonding, and recognizing the role of social context in fostering adaptive developmental outcomes. CDP is built upon a theoretical model that hypothesizes that satisfying students basic needs will lead to greater attachment or bonding to the school community, which in turn will promote commitment to the school communitys norms and values, as reflected in behavior consistent with those norms and values.
Description of Intervention:
CDP is a comprehensive school-based model that focuses on creating a cooperative and supportive school environment. Components include school staff training in the use of cooperative learning and a language arts model that fosters cooperative learning, cross-grade "buddying" activities, as well as a developmental approach to discipline that fosters self-control by engaging students in classroom norm-setting and providing them with opportunities to actively participate in classroom decision-making. School-wide community-building activities are used to promote school bonding, and parent involvement activities such as interactive homework assignments reinforce the family-school partnership.
The study consisted of approximately 4,500 third- through sixth-grade students in 24 schools (12 treatment schools, 12 control schools) from 6 diverse districts throughout the United States. Approximately 1/3 of the total sample was used for dependent measures of delinquent behavior. The schools in the total sample represented geographically diversity, including the west coast, south, southeast and northeast United States, and included large and small urban, suburban and rural schools. Intervention and control samples were generally well matched demographically each year. In year one, the intervention group was 54% white, 17% African American, 21% Hispanic, 7% Asian, and 3% other, while the comparison group was 47% white, 22% African American, 21% Hispanic, 9% Asian, and 2% other. The proportion of non-white students increased in both groups across time. The overall sample included slightly fewer boys than girls (48.2% and 51.8%, respectively). Most students were between 11 and 12 years old at time of assessment.
A quasi-experimental, cohort sequential design was used. Data were collected after 1 year and 2 years of intervention. District administrators selected treatment schools, with comparison schools matched for student characteristics (SES, ethnicity, English proficiency, and academic achievement).
To measure fidelity of implementation, teachers were assessed through four, 90-minute observations each year and annual teacher questionnaires. Student assessment was conducted through self-report surveys conducted in the top three grades in each school (either 3-5 or 4-6). Self-reports of drug use and delinquent behavior were limited to the top grade (5 or 6) in each school. In addition to questions about cigarettes, alcohol and marijuana, delinquent behavior was assessed based on the frequency (on a 5-point scale) of involvement in 10 specific behaviors during past year. The students who completed the survey were well-matched to the total population of upper-grade students with respect to gender, ethnicity and SES.
Multivariate ANCOVAs indicated overall program status x year interaction significant at p<.04. Univariate analysis showed significant differences in alcohol use (p<.02), though changes in marijuana use did not achieve statistical significance and tobacco use declined at both E and C schools. Among the delinquency measures, only vehicle theft saw a statistically significant (p<.01) effect in univariate analysis.
Owing to the difficulty in implementing with fidelity a complex intervention in the short period of time covered in the study, a second set of analyses were conducted which accounted for degree of implementation. Using a classification of high, moderate, and low implementation (based on changes from baseline in six observational measures of program implementation and practice) multivariate ANCOVAs showed a program status x year interaction which was significant for high-implementation (p<.008) and moderate implementation (p<.03) schools, but was not significant for low-implementation schools.
Using this implementation-level classification scheme, statistically significant effects were found in the high implementation schools for marijuana use, vehicle theft, and carrying a weapon. By the second year of implementation, students in high implementation schools showed significantly lower rates of skipping school, carrying a weapon, and vehicle theft (ps<.01).
Strengths & Limitations:
The Child Development Project targets multiple risk and protective factors across both individual and ecological contexts, though with a stronger focus on school ecology. The selection of program schools by administrator was based on perceived willingness to implement the program, with comparison schools matched on demographics, SES, and student achievement scores. There was a strong correlation between fidelity of implementation and program effects, though the measures used to identify the degree of program implementation are limited in that they do not account for (a) school and family involvement activities which are an integral part of the intervention, or (b) within-school variation or variation across years in program fidelity. The study also failed to account for differential program effects for different groups of students and schools (though the study did statistically control for gender, ethnicity, and age).
Because of extreme skewing of both the drug use and delinquency responses, data were recoded to dichotomous variables. Though effect sizes are relatively small, the authors note an a priori expectation that the complexity of the contextual change involved with the program would require as much as 3 years to fully implement. The extreme geographic and demographic diversity of the study lend to its generalizability. There has been no independent replication of the program.