When Can Schools Affect Dropout Behavior? A Longitudinal Multilevel Analysis

The National Educational Longitudinal Study (NELS) database was used to examine student and school factors associated with students dropping out in different grades. Specifically, a hierarchical logistic model was used to address three issues. First, are early (middle school) and late (high school) dropouts equally affected by traditionally defined risk factors? Second, do school-level factors, after controlling for differences in enrollment, account for between-school differences in school dropout rates, and can these school factors mediate individual student risk factors? Third, what impact does early predicted risk have on the likelihood of dropping out late? Results showed that the mix of student risk factors changes between early and late dropouts, while family characteristics are more important for late dropouts. Consistent with previous research, the results also indicated that being held back is the single strongest predictor of dropping out and that its effect is consistent for both early and late dropouts. School factors can account for approximately two thirds of the differences in mean school dropout rates, but they do a poor job of mediating specific student risk factors. The results indicate as well that early predicted risk, at both the student level and the school level, significantly affects the odds of a student dropping out late.