Comparative international assessments of academic achievement consistently indicate that US students trail behind many peers, particularly those from east Asia, in math and science. Traditional efforts to explain this finding have focused on identifying characteristics that might differentiate the United States from top-performing countries. Limitations of this descriptive, atheoretical strategy include: the ecological fallacy of inferring individual-level relations from aggregate-level data, the context specificity of predictors examined, and overreliance on simplistic methodological/analytic approaches. We advocate shifting the emphasis from cataloguing cross-national differences to: identifying core psychological achievement predictors that are stable across national settings, modeling mediators and moderators of the relations between those predictors and academic performance, and using more sophisticated analytic strategies that do justice to the complexity of the problem. We outline such an approach and then discuss how it might be used to understand the math-science achievement gap and reduce persistent ethnic disparities in educational outcomes both within and between countries.