« Back to Results

Empirical Research on Automation and “Smart” Technologies

Paper Session

Friday, Jan. 3, 2020 8:00 AM - 10:00 AM (PST)

Marriott Marquis, Grand Ballroom 10
Hosted By: American Economic Association
  • Chair: James Bessen, Boston University

Competing with Robots: Micro Evidence from France

Daron Acemoglu
,
Massachusetts Institute of Technology
Claire Lelarge
,
Paris-Sud University
Pascual Restrepo
,
Boston University

Abstract

We use firm-level data for firms in France to study the implications of industrial automation. We start by comparing firms that are purchasers of robots---or robotized firms---to other firms in the same detailed industries that do not invest in robots over our period of study (2010/15). Robotized firms are larger than other firms and concentrate at the top of the sales distribution of each industry. Purchases of robots coincide with declining labor shares, rising value added and sales, higher TFP, and lower export prices. In OLS models, we find that robotization correlates with rising sales and employment at the firm-level. However, the expansion of robotized firms comes mostly at the expense of competitors using more labor-intensive technologies. This reallocation of economic activity from labor intensive to capital intensive firms might result in a decline in aggregate labor demand. We validate these findings using an instrumental variables strategy that exploits adoption patterns by firm size across industries with different opportunities for automation.

Automatic Reaction – What Happens to Workers at Firms that Automate?

James Bessen
,
Boston University
Maarten Goos
,
University of Utrecht
Anna Salomons
,
University of Utrecht
Wiljan van den Berge
,
Bureau for Economic Policy Analysis Netherlands (CPB)

Abstract

We provide the first estimate of the impacts of automation on individual workers by combining Dutch micro-data with a direct measure of automation expenditures covering firms in all private non-financial industries over 2000-2016. Using an event study differences-in-differences design, we find that automation at the firm increases the probability of workers separating from their employers and decreases days worked, leading to a 5-year cumulative wage income loss of about 8% of one year's earnings for incumbent workers. We find little change in wage rates. Further, lost wage earnings are only partially offset by various benefits systems and are disproportionately borne by older workers and workers with longer firm tenure. Compared to findings from a literature on mass layoffs, the effects of automation are more gradual and automation displaces far fewer workers, both at the individual firms and in the workforce overall.

Technological Change in Occupational Attribute Space

Richard Freeman
,
Harvard University
Ina Ganguli
,
University of Massachusetts-Amherst
Michael Handel
,
Northeastern University

Abstract

Analysis of technological change from robotization to AI and digitalization of work has focused on what job tasks and activities are likely to be impacted by new software and hardware. Most analyses believe the new technologies will target routine cognitive and non-cognitive activities so that jobs with a lot of routine work will see declines in demand for human labor, with resultant loss of employment or falls in wages. Our study uses the U.S. Department of Labor's Occupational Information Network (O*NET) database to examine the stability and change in the attributes of occupations defined by the skills, knowledge, abilities, education, work context, activities, and style. We develop a new landscape of occupations based on O*NET's over 300 measures of occupational attributes from the early 2000s to the 2010s linked to the Current Population Survey to examine the relationship between changes in these occupational attributes, wages and employment.

Machine Learning in Healthcare? Evidence from online job postings

Avi Goldfarb
,
University of Toronto
Florenta Teodoridis
,
University of Southern California
Bledi Taska
,
Burning Glass Technologies

Abstract

This paper documents a puzzle. Despite the numerous popular press discussions of machine learning and artificial intelligence in healthcare, there has been relatively little adoption. Using data from Burning Glass Technologies on the skills listed in millions of online job postings over ten years, we find that AI adoption in healthcare remains substantially lower than in most other industries. Roughly 1 in 1,250 hospital jobs required AI-related skills in 2015-2018 compared to approximately 1 in 174 in finance & insurance, 1 in 88 in professional, scientific, and technical services, and 1 in 72 in information. Combining the job posting data with data on US hospitals, we document that under 3% of the hospitals in our data had posted any jobs requiring AI skills from 2015-2018. The low adoption rates mean any statistical analysis is limited. Nevertheless, the adoption we do see in the data shows that larger hospitals, larger counties, and integrated salary model hospitals are more likely to adopt.
Discussant(s)
Susan R. Helper
,
Case Western Reserve University
Robert Seamans
,
New York University
David Deming
,
Harvard University
Daniel Rock
,
Massachusetts Institute of Technology
JEL Classifications
  • O3 - Innovation; Research and Development; Technological Change; Intellectual Property Rights
  • J2 - Demand and Supply of Labor