OBJECTIVE: To determine the effect of employment in childhood on self-reported health in adulthood. METHOD: A cross-sectional household survey, with households selected through two-stage sampling, in urban and rural areas in the northeast and southeast of Brazil. A total of 4940 individuals, aged between 18 and 65 years, were included. The main outcome measure was self-reported health. RESULTS: There has been a marked reduction in the proportion of people starting work during childhood although, even in the youngest age group, nearly 20% of males began work when under 10. Early entrance into the labour market is strongly associated with low levels of both education and income, with income differentials remaining at later ages. Age starting work is also linked to current household income, with approximately 35% of those starting work when 15 or over currently in the top quartile of household income, compared with 12% of those starting work when under 10. Males, those living in rural areas, and non-whites are most likely to start work early. In univariate analyses, the younger a person started working, the greater the probability of reporting less than good health status as an adult. This persists through all ages, although the difference attenuates with increasing age. In multivariate analyses, adjustment for education or household income substantially reduces the effect but fails to eliminate it in several age bands up to the age of 48, indicating that age starting work has an independent effect on self-reported health in adulthood. CONCLUSIONS: The debate about the appropriate policy response to child labour is complex, requiring a balance between protecting the health of the child and safeguarding the income of the family. These findings indicate the need for more research on the long-term sequelae of beginning work at an early age.