Linked administrative data on education, health, social services, and crime from British Columbia, Canada, are used to document the relationship between measures of secondary educational attainment and indicators of poor outcomes later in life. Poor outcomes are seen to manifest primarily among high school dropouts. Next, we document the ability of characteristics observed in administrative data in grade 4 to predict high school graduation using a very simple model. It is straightforward to identify more than one fifth of future dropouts reasonably accurately. Non-cognitive measures (esp. social and emotional characteristics) are better predictors of educational attainment than cognitive ones. We discuss the implications of these findings for a scientific approach for developing interventions to prevent poor outcomes later in life.