Automation, labor share, and productivity: plant-level evidence from U.S. manufacturing

This paper provides direct evidence of automation's role in production using establishment-level data from the U.S. Census Bureau's Survey of Manufacturing Technology. The data indicate that more automated plants have lower production labor share and higher capital share, higher labor productivity, and a smaller fraction of workers in production who receive higher wages. To understand the connection between automation and total factor productivity, we estimate a CES model of production where a plant chooses the degree of automation by adjusting the relative weight of capital and production labor given their relative price. The results indicate that, in the presence of heterogeneity in the extent of automation, productivity estimates are more dispersed and skewed relative to Cobb-Douglas residuals, and that the choice of functional form affects the level of productivity estimates. Overall, broad and deep automation is concentrated in larger plants with higher total factor productivity and lower labor share, consistent with a role for automation in contributing to dispersion in input utilization and market share.