Overestimating the Effect of Complementarity on Skill Demand

Many recent studies estimate cost function parameters to measure the influence of capital-skill complementarity on changes in skill demand. This paper argues that standard cost function estimates assuming quasi-fixed capital systematically overestimate the effect of complementarity when subject to skill-biased technological change. While previous work has considered bias due to measurement error or general endogeneity concerns, this paper shows that upward bias results directly from cost minimizing behavior. I also develop a novel instrumental variables strategy based on the tax treatment of capital to more accurately measure the effect of complementarity. Although somewhat imprecise, the IV results support the model's prediction that the standard approach overestimates the effect of complementarity.