Are skilled and unskilled labour complements or substitutes?

Using theoretical and empirical approaches, this thesis asks whether skilled and unskilled labour complement or substitute one another in production. We primarily investigate whether an increase in the proportion of workers with skills would raise or lower demand for those who remain unskilled. A secondary issue is the role of factor prices in labour demand. To study the role of factor prices, we estimate labour demand elasticities and Alien elasticities of substitution between capital and up to five occupations in South Africa. We supplement firmlevel data with household survey information and confirm theoretically that the elasticities can be estimated from a cost function under non-constant returns to scale. We show that separable disaggregated inputs can be used to find aggregate elasticities: more skilled and less skilled aggregates are p-complements, so a fall in skilled wages would lead to a rise in demand for less skilled labour. Disaggregated estimates suggest unskilled workers are p-complements with semi-skilled workers but p-substitutes with skilled/artisanal labour. We investigate the effects of a rise in skill supply on the relatively unskilled by estimating Hicks elasticities of complementarity and factor price. Aggregated estimates suggest more skilled and less skilled labour are q-complements, so an exogenous rise in the supply of skilled labour would raise demand for less skilled labour. Disaggregated estimates suggest skilled/artisanal and unskilled labour are q-complements while semi-skilled and unskilled labour are q-substitutes. The results allow for imperfectly elastic product demand and rigid wages. Using an endogenous growth model, we show technological progress is skill-biased in the South if it is in the North, resulting in rising wage inequality in developing countries. Assuming skilled and unskilled labour are perfect substitutes, we model expanded educational access as it adds relatively educated cohorts to the labour market. A rising skill composition causes accelerated skill-biased technological change and wage inequality. Relaxing the assumption of perfect substitutability, a one-off rise in skill supply only raises wage inequality if the elasticity of substitution is high, higher than existing empirical estimates.