This website uses cookies.

By clicking the "Accept" button or continuing to browse our site, you agree to first-party and session-only cookies being stored on your device to enhance site navigation and analyze site performance and traffic. For more information on our use of cookies, please see our Privacy Policy.

« Back to Results

New Approaches to Measuring Technology and Innovation

Paper Session

Friday, Jan. 3, 2020 2:30 PM - 4:30 PM (PDT)

Marriott Marquis, Presidio 1 - 2
Hosted By: American Economic Association & Committee on Economic Statistics
  • Chair: Ellen Hughes-Cromwick, Third Way

Measuring Technology Adoption in Enterprise-Level Surveys: The Annual Business Survey

David Beede
,
U.S. Census Bureau
Erik Brynjolfsson
,
Massachusetts Institute of Technology
Cathy Buffington
,
U.S. Census Bureau
Emin Dinlersoz
,
U.S. Census Bureau
Lucia Foster
,
U.S. Census Bureau
Nathan Goldschlag
,
U.S. Census Bureau
Kristina McElheran
,
University of Toronto
Nikolas Zolas
,
U.S. Census Bureau

Abstract

The Annual Business Survey (ABS) started in 2017 and, in its first collection year, will provide an economy-wide view of technology from its sample of almost 850,000 firms across almost all sectors of the economy. The ABS is conducted by the Census Bureau in partnership with National Center for Science Engineering Statistics (NCSES) and replaces the Survey of Business Owners (SBO), Annual Survey of Entrepreneurs (ASE) and the Business R&D and Innovation Survey for Microbusinesses (BRDI-M). In addition to regularly occurring questions, the 2017 ABS includes a module of questions on the adoption and use of new technologies. The technology module was developed by the Census Bureau and a group of external researchers who specialize in technology adoption by firms and its effects. This paper describes the development of this module and some of the challenges faced by Census, including which technologies to collect data on, how to define certain technologies, what types of measures
(extensive versus intensive) and what time frame to use for the adoption and use of the technologies. We describe how the cognitive testing of the survey was performed and how companies interpreted questions posed by Census and the external collaborators. We also discuss how the module fits in with current Census data collection efforts to better measure technology in the Annual Survey of Manufactures (ASM), as well as Annual Capital Expenditures Survey (ACES). Future versions of the paper will provide results from this survey.

Quantifying the Impact of AI on Productivity and Labor Demand: Evidence from United States Census Microdata

Dean Alderucci
,
Carnegie Mellon University
Lee Branstetter
,
Carnegie Mellon University
Ed Hovy
,
Carnegie Mellon University
Andrew Runge
,
Carnegie Mellon University
Nikolas Zolas
,
U.S. Census Bureau

Abstract

Artificial intelligence has entered a new era of rapidly advancing capabilities that may increasingly permit human labor to be supplemented by – and sometimes replaced by –algorithms and machines. Like past periods of rapid and substantive technological progress, these developments are likely to bring both opportunities and challenges. This paper directly addresses that challenge by creating a new, detailed, granular measure of AI-related innovation utilizing a suite of machine learning algorithms to parse the full text of U.S. patent grant documents and identify those related to AI. This process has created a dynamic, detailed map of AI invention that identifies far more patenting in this space than prior efforts. Using the U.S. Census Bureau’s well developed mapping from patent assignees to firm enterprises, we have linked AI-inventing firms with detailed data on employment, output, productivity, and investment. We have also linked these firms to data on the full distribution of wages as recorded in the LEHD. Using an event study methodology, we find that AI invention is associated with greater employment growth, higher levels of output per worker, and increases in within-firm earnings inequality.

Data Development and Measurement of the Economic Geography of Robotics

Nancy Green Leigh
,
Georgia Institute of Technology
Ben Kraft
,
Georgia Institute of Technology
Heon Yeong Lee
,
Georgia Institute of Technology

Abstract

The lack of public use data on the robotics industry or employment is a fundamental barrier to measuring the economic geography of robotics. There is neither a North American Industrial Classification System (NAICS) code to single out robotics manufacturers or consulting firms, or a Standard Occupational Code (SOC) to identify employees who work directly with robots. This paper first describes three datasets the authors developed in order to analyze the evolving economic geography of robotics in the U.S. manufacturing industry. These include: a robotics “census” created by mining proprietary business databases for robotics firms; analysis of proprietary real time labor market data and use of machine learning to identify three robotics occupations; and a national survey of manufacturers’ use of advanced technology and robotics. Combining these three datasets with socio-economic data below the national level, we identify robotic regions and metro areas and model impacts on skill, employment and income levels.
Discussant(s)
Pascual Restrepo
,
Boston University
Enghin Atalay
,
University of Wisconsin
Susan R. Helper
,
Case Western Reserve University
JEL Classifications
  • O3 - Innovation; Research and Development; Technological Change; Intellectual Property Rights
  • D2 - Production and Organizations