Artificial Intelligence and Aggregate Labor Productivity: Evidence from Patent Data

This paper investigates the aggregate impact of Artificial Intelligence (AI) technology on labor productivity, exploiting patent data in OECD countries during 2000-2017. We first document that after accelerating in 2000, the issuance of AI technology related patents has more than tripled by 2017 and OECD countries held about 89 percent of all these patents across the world during the same period. Second, using a production function approach, we find that considering the pace of AI patent applications over the 2000-2017, labor productivity, measured as output per worker, has consequently increased by between 0.8 percent to 1.2 percent. The estimates also imply that innovation in AI technology could raise aggregate labor productivity by up to 3.8 percent in the long-run. These suggest that productivity gains could be larger in the future as the pace of innovation in AI gains new speed and spread to more sectors of the economy. These results are robust to accounting for a proxy of AI technology spillovers from other countries.  Further we find that countries with a high share of employees in professional and managerial roles benefit from larger labor productivity gains  and AI technology. Furthermore, these gains from task complementarity may take time to materialize as workers learn to fully appropriate the technology. Finally, labor market flexibility could also play a role in shaping labor productivity gains by enabling workers mobility.