This article examines the relationship between youthful drinking and educational attainment using data on same-sex siblings pairs from the 1979–90 panels of the National Longitudinal Survey of Youth. We consider different estimators that can be constructed using siblings data, including estimators that adopt key restrictions of the standard regression, family fixed effect, and instrumental variable approaches. We also consider the properties of these estimators under more general conditions and show that under very plausible assumptions the effect of drinking on schooling can be bounded. The study finds that estimates of the schooling consequences of youthful drinking are very sensitive to specification issues. The research concludes that the actual effects of youthful drinking on education are likely to be small.