We test whether work in childhood impacts on health. We focus on agricultural work, the dominant form of child work worldwide. Data are from the Vietnam Living Standards Survey, 1992-93 and 1997-98. We correct for both unobservable heterogeneity and simultaneity biases. Instruments include small area labour market and education conditions obtained from community level surveys. We use three indicators of health: body mass index; reported illness; and, height growth. There is clear evidence of a healthy worker selection effect. We find little evidence of a contemporaneous impact of child work on health but work undertaken during childhood raises the risk of illness up to five years later and the risk is increasing with the duration of work. There is no evidence that work impedes the growth of the child.