Objectives: This study sought to understand (1) whether parents and teachers reports of childhood behavioral problems could predict smoking and alcohol consumption 35 years later, and (2) whether propensity for smoking and alcohol consumption differed on the basis of informant agreement in their classification of behavioral syndromes.

Methods: Participants included those from the Aberdeen Children of the 1950s study (ACONF) with full childhood information as well as self-reports in adulthood (n = 1342). Latent Class Analysis was conducted to identify patterns of childhood problem behaviors as rated by teachers and parents. Regression models were then conducted predicting adulthood smoking and alcohol consumption. Informant agreement categories were constructed and differences across categories in both smoking and alcohol consumption were tested.

Results: Three subtypes of childhood behavior problems were identified by both teachers and parents: “Normative,” “Externalizing,” and “Internalizing,”. Parents also identified a distinct fourth group “Mixed”. Teacher's classification of the child as externalizing significantly predicted greater likelihood of being a current smoker, and parents' classification as internalizing predicted lower likelihood of being an ex-smoker. Parents' ratings as externalizing and mixed also predicted lower levels of alcohol consumption, which was opposite to the predicted effect. Additionally, informant agreement of externalizing indicated a greater propensity of smoking in adulthood, but did not indicate differences in alcohol consumption.

Conclusions: This study suggests that it is important to consider additive information from multiple informants when examining the life-course effects of childhood behavioral problems on risky health behaviors in adulthood.