"Competing with Robots: Firm-Level Evidence from France"

by Acemoglu, Daron; Lelarge, Claire; Restrepo, Pascual (2020)

Abstract

Using several sources, we construct a data set of robot purchases by French manufacturing firms and study the firm-level implications of robot adoption. Out of 55,390 firms in our sample, 598 have adopted robots between 2010 and 2015, but these firms account for 20% of manufacturing employment and value added. Consistent with theory, robot adopters experience significant declines in labor share and the share of production workers in employment, and increases in value added and productivity. They expand their overall employment as well. However, this expansion comes at the expense of their competitors (as automation reduces their relative costs). We show that the overall impact of robot adoption on industry employment is negative. We further document that the impact of robots on overall labor share is greater than their firm-level effects because robot adopters are larger and grow faster than their competitors.

Key Passage

Automation substitutes capital for tasks previously performed by labor, reducing the labor share of value added and increasing value added per worker in the process. While the higher productivity from automation tends to increase labor demand, its displacement effect may outweigh this positive impact and may lead to an overall decline in employment and wages. Acemoglu and Restrepo estimate negative effects from the introduction of one of the leading examples of automation technology, industrial robots, across US local labor markets, suggesting that the displacement effects could be significantly larger than the productivity effect. Firm-level evidence is useful as well for understanding how automation is impacting the production process and productivity. But its interpretation is made complicated by the fact that firms adopting automation technologies reduce their costs and may expand at the expense of their competitors. In this paper, we study firm-level changes associated with robot adoption using data from France between 2010 and 2015. Consistent with our theoretical expectations (which are developed further in the Appendix), we find that firm-level adoption of robots coincides with declines in labor shares, increases in value added and productivity, and declines in the share of production workers. In contrast to their market-level effects, however, overall employment increases faster in firms adopting robots. This positive employment effect may be because firms with greater growth potential are more likely to adopt robots, generating a classic omitted variable bias. Equally important, this positive effect may be a consequence of reallocation of output and labor towards firms that reduce their costs relative to their competitors. We show that such reallocation accounts for the positive firm-level impact of robots. Firms whose competitors adopt robots experience significant declines in value added and employment. In fact, the overall impact of robot adoption (combining own and spillover effects) is negative and implies that a 20 percentage point increase in robot adoption (as in our sample) is associated with a 3.2% decline in industry employment. (p.2)