Joseba Martinez, London Business School

"Automation Potential and Diffusion"

Abstract

We introduce a novel methodology to measure the invention and diffusion of labor-replacing technology in the US economy. First, we measure the relevance of US patents introduced from 1900-2020 to work tasks performed by human workers using a natural language processing algorithm. After controlling for the confounding effects of the evolution of language, we obtain a measure that we call the automation potential of newly introduced technology: the potential for that technology to eventually replace human workers in the performance of tasks. In a local projections framework, we project long-horizon changes in task demand onto automation potential and find large negative effects, suggesting i) that our measure captures the development of automation technology, and ii) that such technology diffuses slowly into the economy.

Contact person: Katja Mann