We suggest new evidence on minimum age regulations using a natural experiment. In 1998, a constitutional reform has changed the minimum working age from 14 to 16 in Brazil. The reform was the legislative counterpart of a broad set of measures taken by a government strongly committed to fighting child labor. We document the fact that enforcement and compliance may have been heterogeneous across regions and job types. The setting allows improving upon past approaches based on the comparison of employment rates of children below and above the minimum age. Precisely, we observe 14-year old children the year after the reform and exploit discontinuous treatment depending on their birthdate (only those who turned 14 after mid-December 1998 are banned). Regression discontinuity and difference-in-discontinuity designs show no effect of the ban overall, nor a reallocation towards less visible activities, or a substitution of labor within families. Importantly, however, we …find a significant drop in child labor among those with highest chances of compliance, namely children in visible activities and in regions characterized with an above-average intensity of labor inspections. We provide power calculation and extensive sensitivity checks.