We exploit a country-wide child labor regulation that eliminated the difference in school/work alternatives for children born at the beginning and the end of the year to identify the causal effect of education on migration at low levels of schooling. By not relying on changes to the school...