Early warning systems use school record data--such as attendance rate, behavior records, and course performance--to identify students at risk of dropping out. These are useful predictors of graduation-related outcomes, in large part because they indicate a student's level of engagement with school. However, these data do not indicate how invested students are in education--information that could help school counselors and other staff understand and intervene when students are falling off the path to graduation. To examine whether student engagement surveys have additional predictive value beyond data readily available in school databases, we followed a cohort of students, who completed a survey of cognitive/affective engagement as ninth graders, to one year beyond their expected high school graduation. Some engagement factors measured by the survey met rigorous tests of predictive value in terms of identifying which students were falling off the graduation path, even when controlling for other powerful predictors of the outcome.