Cross-country evidence on student achievement might be hampered by omitted country characteristics such as language or legal differences. This paper uses cross-state variation in Germany, whose sixteen states share the same language and legal system, but pursue different education policies. Education production function models are estimated using state-level PISA-E data, where possible pooling three subjects and three waves to obtain up to 138 test-score observations. The same results found previously across countries hold within Germany:Higher mean student performance is associated with central exams, private school operation, and socio-economic background, but not with spending, while higher equality of opportunity is associated with reduced tracking. In models that pool German states with OECD countries and combine up to 54 state and country observations, these institutional determinants do not differ significantly between the sample of German states and the sample of OECD countries, indicating that the existing cross-country evidence is not substantially biased by unobserved country-specific factors.