This site is a work in progress and has not been widely shared. Content may contain errors. Feedback is welcome.
This site is undergoing review. Some annotations were human-generated, some AI-generated — all are being verified.
Back to papers

Automation and Jobs: When Technology Boosts Employment

Bessen

2018The Economics of Artificial Intelligence: An Agenda (NBER book chapter)130 citations
Theoretical / conceptualTheoretical model
Automation / RobotsAI (General)General automationAugmentation vs. substitution
DOI: 10.2139/ssrn.2935003
Summary

Bessen develops a theoretical model showing how demand elasticity mediates technology's employment effects, validates it using historical data from US textiles, steel, and automotive industries (1800s-2000s), and discusses implications for future AI adoption.

Main Finding

Technology's impact on employment depends critically on demand elasticity: when demand is elastic (>1), automation increases employment; when inelastic (<1), jobs decline. Historical manufacturing industries exhibited inverted-U employment patterns as demand elasticity fell from highly elastic (2.13-6.77) to highly inelastic (0.02-0.16) over 100+ years.

Primary Datasets

US Census Bureau Historical Statistics; BLS Current Employment Statistics; Historical production and employment data for textiles (SIC 2211, 2221), steel (SIC 3312), and automotive manufacturing

Secondary Datasets

Parker and Klein (1966) agricultural productivity data; 1950 census occupation codes

Key Methods
Develops theoretical model of demand elasticity based on distribution of consumer preferences; estimates lognormal demand functions for historical industries; simulates employment trajectories
Sample Period
1810-2011
Geographic Coverage
United States
Sample Size
Three manufacturing industries tracked over 100+ years; cotton textiles 1810-1995, steel 1860-1982, automotive 1910-2007
Level of Analysis
Industry, Occupation
Occupation Classification
1950 census occupation codes mentioned
Industry Classification
SIC codes (2211, 2221, 3312 mentioned)
Notes
Economic Policy, vol. 34, no. 100, pp. 589-626 [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions. [Claude classification]: This is a book chapter published in NBER's 'The Economics of Artificial Intelligence: An Agenda' (2019). The paper is primarily theoretical with historical validation using manufacturing data. AI discussion is prospective/conceptual rather than empirical. The model assumes competitive markets and labor-augmenting technical change. Estimates assume lognormal preference distributions.