This paper explores the relationship between capital composition and productivity using a unique and highly detailed data set on firm-level investment in the U.S. We develop a succinct methodology for modeling the separate effects of a large number of capital types in a production function framework. We then use this methodology, combined with recently developed techniques for accounting for unobserved productivity, to identify these effects and back out the implied marginal products of each capital type. The results indicate that information and communications technology (ICT) capital–specifically, computers, software, and communications equipment–are positively and statistically significantly associated with output, even after conditioning on total capital, labor, and various proxies for unobserved productivity. We compare the implied marginal products for different capital types to official data on rental prices and find that the marginal products of the ICT capital types are substantially above their measured rental prices. Lastly, we provide evidence of complementarities and substitutabilities, both among capital types–a rejection of the common assumption of perfect substitutability–and between certain capital types and labor.
About the Author
Daniel Wilson is a vice president in the Economic Research Department of the Federal Reserve Bank of San Francisco. Learn more about Daniel Wilson