To test the assumption embedded in state-of-the-art, community assessment and decision-making regarding prevention of underage drinking: that there is minimal variation in the way that risk and protective factors (RPF) are associated with underage drinking across communities. Three large datasets provided the same measures of adolescent alcohol use and RPFs. Multilevel ordered-logistic regression models were carried out separately for each dataset and separately for males and females in 8th and 10th grades, testing random slopes for each RPF index. Predicted school-level coefficients were derived from these models, representing the association between RPFs and alcohol use. The variation in associations between RPFs and alcohol use across schools was greatest for antisocial peer risk and community protection; the lowest variation across schools was found for family cohesion and individual antisocial behavior. Ranges in predicted coefficients indicate large differences across schools for many RPFs. Bivariate correlations indicated that school-level associations vary across RPFs in expected directions. Policy makers should recognize that the magnitude of associations between RPFs and adolescent alcohol use vary considerably across communities, and that such variability is greater for certain RPFs than others. These findings have implications for policies regarding how prevention resources are targeted within and across communities.