| Abstract: | Abstract: Studies of child labor in developing countries have largely relied on data collected through national household surveys. But government‐sponsored household surveys are arguably prone to both sample selection bias and reporting bias. In this paper, we compare the demographic and labor market characteristics of working street children obtained from the 2000 Tanzanian Integrated Labor Force Survey, a government‐sponsored household survey, with the results obtained in a 2004 survey of working street children in the city of Mwanza in northern Tanzania. Our comparisons help illumine the potential biases introduced when child labor studies rely on household samples; specifically, we show that children in the non‐household based survey work longer hours, have lower educational attainment and suffer worse health outcomes than their counterparts in the government‐sponsored household survey. This suggests that studies based on national household surveys may significantly underestimate both the extent and the consequences of child labor. |
