Panel Paper: Regional Economic Polarization and the AI-Economy

Friday, November 9, 2018
8229 - Lobby Level (Marriott Wardman Park)

*Names in bold indicate Presenter

Henry Renski, University of Massachusetts


There has been growing concern that artificial intelligence will soon advance to the point where machines and algorithms could displace a considerable portion of the labor force. Some predict that as many as 40 to 50 percent of present-day jobs are at a serious risk of automation in the foreseeable future (Frey and Osborne 2017, Manyika, Lund et al. 2017). Others offer a more cautious interpretation, emphasizing the limitations of computers in solving unstructured problems, working with new information, and in performing non-routine manual tasks (Levy and Murnane 2013).

There has been little discussion, thus far, of whether and how the anticipated automation of work will have differential impacts in different types of cities. Technological change has long had a variable impact on cities, largely conditioned by regional industrial specializations. However, the impacts of automation and computerization are anticipated to be far more widespread—impacting entire job categories across a wide spectrum of skill sets and industries, as opposed to being contained within a limited number of product markets.

This paper examines the possible spatial ramifications of the anticipated computerization of work among U.S. cities and regions. It matches occupations that are most and least “susceptible” to automation with regional data on occupational employment to measure the possible impact across different cities. I find that the jobs that are most at risk tend to be relatively spatially ubiquitous. That is, they are common to many metropolitan areas where they exist in roughly the same proportion. By contrast, the types of jobs that are least vulnerable to replacement by automation are heavily concentrated. These “safe” jobs tend to locate in places with a strong University presence or in large MSAs that are already leaders in the knowledge economy, such as Boston, Silicon Valley, and Austin.

If current predictions are accurate, the concentration of “AI-safe” jobs coupled with widespread decline in “AI-threatened” jobs will likely lead to the further spatial polarization of the U.S. economy between leading and lagging regions. The paper concludes by discussing the prospects for a renewed role for the federal government in addressing the anticipated growth in regional economic disparities.