The link between poverty and child labor has been regarded as a well established fact, but recent research has questioned the validity of this link. Starting from the premise that child labor is not necessarily harmful, this paper analyzes the determinants of harmful child labor, viewed as labor that directly conflicts with children's human capital accumulation, and attempts to identify the most vulnerable groups. Data were drawn from a 1997 household survey of Ghana, covering 14,514 households. Analysis focused on children aged 9-14. The economic model and econometric methodology used are discussed. Although the vast majority of children in Ghana were attending school, the incidence of harmful child labor increased with age, ranging from 1.2 percent at age 9 to 5.9 percent at age 14. Rural children were over twice as likely to engage in harmful labor than were urban children. Harmful child labor was more common among girls than boys and among children from poor households than those of nonpoor households. In addition, harmful child labor was positively related to parent being self-employed, family ownership of land and livestock, and the distances to the nearest primary school and the nearest secondary school. The findings suggest that there are structural differences in the processes underlying harmful child labor in Ghana across gender, across rural/urban location, and across poverty quintiles of households. (Contains 18 references.) (SV)