In the large empirical literature that investigates the causal effects of education on outcomes such as health, wages and crime, it is customary to measure education with years of schooling, and to identify these effects using the exogenous variation provided by school reforms increasing compulsory education and minimum school leaving age. If these reforms are correlated to changes in school quality, and school quality is an omitted variable, this identification strategy may fail. We test whether this is the case by using the information provided by two distinct test scores on mathematics and reading and find that we cannot reject the internal validity of this popular identification strategy.