Individual, Academic, and School-Related Predictors of Offending: Early Versus Late Onset
Methods: The current study used a merged state-level administrative data set containing educational, juvenile justice, and adult correctional information about 407,800 children enrolled in grades 7-12 in public schools in a state in the Deep South. The overall sample (N=31,454) was drawn from a 10-year cohort (1996-2008) of children who were identified as either early starters (n=14,349; 45.6%) or late starters (n=17,105; 54.4%). Binary logistic regression (LR) was used to determine the combination of variables that best predicted the two distinct patterns of offending: early starters having a juvenile offense record and late starters having an adult offense record. Using the statistical software, STATA, two models were tested and predictors were entered in four blocks. Three demographic variables were entered (gender, race, poverty status) in the first block. The second block included four school discipline variables (in-school suspension, out-of-school suspension, in-school expulsion, out-of-school expulsion). The third block included four school-related predictors (attendance, truancy status, dropout status, unplanned school transitions). The final block included the two academic performance predictors (highest grade completed, ELA standardized test score).
Results: The majority of early starters were male (77.3%) and African American (62.8%), whereas a larger proportion of males (84.1%) and a similar proportion of African Americans (61.3%) comprised the late starters. With regard to early starters, LR results showed that the final model was statistically reliable in distinguishing between having and not having a juvenile offense record (-2 Log Likelihood =-50059.92;X2(=24122.22;df=15;p<.001). LR results indicated that the model predicting late onset was statistically reliable in distinguishing between having and not having an adult offense record (-2 Log Likelihood=60672.374; X2=20631.34;df=15;p<.001). Wald statistics showed that all predictors, except one academic predictor (ELA standardized test score), significantly predicted both patterns of offending. Odds ratios indicated that males were nearly three times as likely as females to have a juvenile record, at Exp(B)=2.89; and nearly five times as likely to have an adult record, at Exp(B)=4.78. Odds ratios for expulsion, however, indicated large changes in the likelihood of having a juvenile record and an adult record (Exp(B) = 112.65 and 22.33), respectively. Students who were expelled were over 100 times as likely to be early starters as those who were not, and were 22 times as likely be late starters as students who were never expelled.
Conclusions: Results underscore the importance of implementing educational and psychosocial interventions to minimize school expulsion. Additional longitudinal research is needed to understand how school discipline practices affect children with academic and behavior problems and how expulsion, in particular, influences the trajectories of boys and girls.