Investigating the robustness of the skill-biased technical change hypothesis, this analysis incorporates two novel features. First, effective labor is modeled as the product of a quantity measure - number of employees with a given level of education - and a quality index, depending on, i.a., demographic characteristics and fields-of-study. Second, low-skilled labor is more disaggregated than in earlier studies. A fully specified structural model is used, containing demand equations for four categories of labor, two types of capital and intermediate goods. The empirical application covers 24 industries in the Swedish manufacturing sector 1985 - 1995. The skill-bias is further corroborated: it is confirmed although the specification of effective labor is supported. Substantial differences are, however, found among the low-skilled.