While the family is a critical determinant of educational achievement, methodological difficulties and the availability of data limit estimation of the family contribution in school effectiveness models. This study uses multilevel modeling to estimate the proportion of variation in student educational achievement between families, family-level intraclass correlation coefficients, and specific family structure effects (family size, birth order, birth spacing, sibling sex ratio). We use cross-classified random effects to account for school and neighborhood variation. We analyze Swedish administrative education records linked with birth records for four academic cohorts of students, with siblings identified from a wider pool of 21 cohorts. We show that almost half of the variation in student achievement described as "between students" in traditional school effectiveness studies would be better described as variation "between families," suggesting effectiveness research might give greater consideration to family-based interventions in tandem with existing student- and school-based approaches to raising low achievement.