Custodes Futurifor the advancement of humanity

Automation and job displacement

Automation and job displacement is a barrier to human progress within the technological factors. The evidence for it, and the methods proposed to address it, follow.

Acemoglu and Restrepo (2020) studied industrial robots in U.S. labor markets.

Evidence

  • Employment and wages. One more robot per thousand workers reduced the employment-to-population ratio by about 0.2 percentage points and wages by about 0.42%.
  • Who is affected. Effects were concentrated among workers in routine manual jobs.
  • Wage inequality. In a separate study, Acemoglu and Restrepo looked at changes in the US wage structure from 1980 to 2016. They estimated that 50% to 70% of these changes are accounted for by relative wage declines of worker groups specialized in routine tasks in industries undergoing rapid automation. The link held after controlling for market power, deunionization, and other forms of capital investment. They also found large rises in inequality alongside modest productivity gains. The automation measure may miss some technologies. The period largely predates modern artificial intelligence (Acemoglu and Restrepo, 2022).
  • Offsetting in Germany. A study used detailed administrative data on German local labor markets. It found robot exposure was associated with job losses in manufacturing that were fully offset by new service jobs. Fewer young people entered manufacturing. Incumbent workers had more stable employment, often by taking on new tasks in their original plants. Young people shifted from vocational training toward college. Germany is a large manufacturing exporter, so results may not carry over to other economies (Dauth et al., 2021).
  • Exposure to artificial intelligence. An IMF staff analysis covered 142 countries. It estimated that almost 40% of global employment is in occupations exposed to AI. The share is about 60% in advanced economies, 40% in emerging market economies, and 26% in low income countries. In advanced economies, the analysis says about half of exposed jobs may be negatively affected and the rest may gain from higher productivity. These are modelled exposure estimates, not observed job losses. Exposure measures vary widely (IMF, 2024).

Why it matters: Technology can raise overall productivity while pushing specific workers out of their jobs. Without retraining and support, these workers face the economic scarcity described in the economic factor. Newer research on artificial intelligence is still emerging, and its effects are not yet established.

What can be done

Actions rated on the strength-rating scale (A strong and replicated, B solid but limited, C weak or debated, D contested or failed), applied to the specific claim made.

  • Context: In U.S. local labor markets, one more industrial robot per thousand workers reduced the employment-to-population ratio by 0.2 percentage points. It also reduced wages by 0.42% (Acemoglu and Restrepo, 2020).
  • Active labor market programs that build skills, judged over two to three years (A for the pattern over time, B for size). A meta-analysis of over 200 evaluations looked at active labor market programs (training, job search help, subsidized jobs). It found that their average effects are close to zero in the short run. But the effects become more positive 2 to 3 years after the program ends. The meta-analysis found larger gains for programs that build skills, for women, for the long-term unemployed, and in recessions (Card, Kluve, and Weber, 2018). A critical review of evaluations in developing countries found that many of these programs had no significant effect on employment or earnings (McKenzie, 2017).
  • Vocational training combined with on-the-job internships (B). A randomized trial looked at Colombia's Jóvenes en Acción program for disadvantaged youth. Women offered training earned 19.6% more. They were also 6.8 percentage points more likely to have paid work, mainly in formal jobs (Attanasio, Kugler, and Meghir, 2011). Long-term administrative data showed that the effects persisted. The data showed more formal work and more work in larger firms, with similar gains for men and women (Attanasio et al., 2017). A reanalysis estimated that between 10.2% and 13.4% of always-employed women moved into formal work because of the program. But the lower bound could not be statistically distinguished from zero (Possebom and Riva, 2025).
  • Community college retraining in technical fields (B). Among displaced workers in Washington State, an academic year of community college raised long-term earnings by about 9% for men and 13% for women. For technically oriented vocational and academic math and science courses, the gains were about 14% for men and 29% for women. Less technical courses had very low and possibly zero returns (Jacobson, LaLonde, and Sullivan, 2005). A replication confirmed that the study's control for each person's earnings trend matters (Dynarski, Jacob, and Kreisman, 2018). In California, enrolling in a community college nursing degree program raised earnings by 44%, estimated using admissions lotteries (Grosz, 2020). However, these students were not all displaced workers.
  • Registered apprenticeship for displaced workers (C). In North Carolina, 219 workers were laid off in 2008 to 2010 and enrolled in a registered apprenticeship within a year. They were compared with 219 matched laid-off workers who did not. Nine years after layoff, the apprentices were 16.9 percentage points more likely to be employed. They earned $9,691 more in that year, counting all members of each group (Berger-Gross, 2021). The study is not randomized, and the author states that selection bias cannot be fully ruled out.
  • Job search assistance through public employment offices (B, with displacement limits). A national randomized evaluation looked at the U.S. Workforce Investment Act Adult and Dislocated Worker programs. It found that people offered intensive services such as counseling and assessment earned about $7,100 (or 20%) more over 30 months (Fortson et al., 2017). A clustered randomized trial covered 235 French labor markets. In that trial, job placement help for young educated job seekers raised stable employment after eight months. But the gains were temporary and came partly at the expense of other job seekers, leaving very little net benefit (Crépon et al., 2013).
  • Mixed: U.S. Trade Adjustment Assistance retraining with extended benefits (C). A national evaluation used a matched comparison group of unemployment insurance claimants. It found that participants earned about $3,300 less than comparisons in the fourth year after job loss, while working about the same number of weeks (33 versus 35). It also found that effects for older workers stayed negative throughout (Schochet et al., 2012). A later study used the quasi-random assignment of cases to more or less lenient investigators. It found that approved workers had $50,000 greater cumulative earnings ten years later, although annual gains faded (Hyman, 2022). Wage insurance within the same program is in the proposed methods below.
  • Not shown to work: generic publicly funded training in the short run (B). The same national U.S. evaluation found the evidence on Workforce Investment Act training "inconclusive but suggests it was not effective" over 30 months (Fortson et al., 2017). For training that did raise earnings, see the sector-focused training entry in the proposed methods below.
  • Not shown to work: hiring subsidies for older unemployed workers (B). In 2002, Germany extended hiring subsidies to all unemployed people aged 50 and over. A study compared people just over and just under 50. It found that exits into subsidized jobs rose by about two percentage points. But the effect on leaving unemployment overall was not significant, except for women in eastern Germany. The study found this was because subsidized hires largely replaced unsubsidized ones (Boockmann et al., 2012, summarized in Boockmann, 2015). The study covers older unemployed people in general, not only workers displaced by technology.

Proposed and experimental methods

Methods that are proposed, under trial, approved in some places, or tried and then failed. Each shows a stage label and an evidence rating. A stage label shows how far a method has progressed, not whether it works. The stage labels are explained on the technological factor page.

  • Sector-focused job training (Large trial, A). Programs screen applicants. Then they combine occupational and soft-skills training with support services for jobs in specific higher-paying industries. A review of randomized evaluations found earnings gains of 12% to 34% that lasted after training. The review found these gains came mainly from placing workers in higher-paying industries and occupations rather than from raising employment rates (Katz et al., 2022). The programs studied served low-wage workers in general, not specifically workers displaced by automation.
  • Using AI to augment workers (Early trial, B for productivity, C for jobs). One study looked at 5,172 customer support agents. Access to a generative AI assistant raised issues resolved per hour by 15% on average. Less experienced and lower-skilled workers improved in both speed and quality. The most experienced and highest-skilled workers saw small gains in speed and small declines in quality (Brynjolfsson, Li, and Raymond, 2025). In an online experiment with 453 professionals, ChatGPT cut writing time by 40% and raised quality by 18% (Noy and Zhang, 2023). But linked Danish survey and administrative data showed precise null effects of chatbot adoption on earnings and hours. These data ruled out effects larger than 2% two years after ChatGPT's launch (Humlum and Vestergaard, 2025).
  • Wage insurance for displaced workers (Approved but not scaled, B). Workers who take a lower-paying new job receive part of the pay difference. One study used an age-based eligibility cutoff in the U.S. Trade Adjustment Assistance program. It found that eligibility raised short-run employment and long-run cumulative earnings, mostly through shorter periods without work. The authors conclude the program pays for itself under conservative assumptions (Hyman, Kovak, and Leive, 2024).
  • Job guarantee for the long-term unemployed (Early trial, B). This builds on the guaranteed income entry in "Learned helplessness and low sense of control" in the psychological factors. From October 2020 to March 2024, every long-term unemployed resident of Gramatneusiedl, Austria (the town of the 1930s Marienthal unemployment study) was offered a job. Long-term unemployment nearly disappeared. Participants gained income, economic security, and social recognition. Well-being rose, but by an amount that became marginally insignificant after correction for multiple outcomes. There were no effects on physical health, and the net public cost was about €237 per participant per month (Kasy and Lehner, 2026). The randomized part included only 62 people in one town.
  • Taxing robots during the transition (Theoretical, C). The authors build a model with automation in which workers choose skills. They find that robots should be taxed while current routine workers who cannot easily switch to non-routine jobs remain in the workforce. They also find that the optimal robot tax falls to zero once those workers retire (Guerreiro, Rebelo, and Teles, 2022). No country has adopted such a tax.

Sources cited on this page

  1. Acemoglu, D., & Restrepo, P. (2020). Robots and jobs: evidence from US labor markets. Journal of Political Economy, 128(6), 2188-2244. DOI B Moderate
  2. Acemoglu, D., & Restrepo, P. (2022). Tasks, automation, and the rise in U.S. wage inequality. Econometrica, 90(5), 1973-2016. DOI B Moderate
  3. Dauth, W., Findeisen, S., Suedekum, J., & Woessner, N. (2021). The adjustment of labor markets to robots. Journal of the European Economic Association, 19(6), 3104-3153. DOI B Moderate
  4. Brynjolfsson, E., Li, D., & Raymond, L. R. (2025). Generative AI at work. The Quarterly Journal of Economics, 140(2), 889-942. link B Moderate
  5. Guerreiro, J., Rebelo, S., & Teles, P. (2022). Should robots be taxed? The Review of Economic Studies, 89(1), 279-311. link C Limited
  6. Humlum, A., & Vestergaard, E. (2025). Still waters, rapid currents: Early labor market transformation under generative AI (NBER Working Paper No. 33777). National Bureau of Economic Research. link B Moderate
  7. Hyman, B. G., Kovak, B. K., & Leive, A. (2024). Wage insurance for displaced workers (NBER Working Paper No. 32464). National Bureau of Economic Research. link B Moderate
  8. Kasy, M., & Lehner, L. (2026). Employing the unemployed of Marienthal: Evaluation of a guaranteed job program. American Economic Journal: Economic Policy (forthcoming). link B Moderate
  9. Katz, L. F., Roth, J., Hendra, R., & Schaberg, K. (2022). Why do sectoral employment programs work? Lessons from WorkAdvance. Journal of Labor Economics, 40(S1), S249-S291. link A Strong
  10. Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187-192. link B Moderate: C for job outcomes
  11. Ma, Y., Fairlie, R., Loyalka, P., & Rozelle, S. (2024). Isolating the "tech" from EdTech: Experimental evidence on computer-assisted learning in China. Economic Development and Cultural Change, 72(4), 1923-1962. link B Moderate
  12. Berger, L. M., Kerkhof, A., Mindl, F., & Münster, J. (2025). Debunking "fake news" on social media: Immediate and short-term effects of fact-checking and media literacy interventions. Journal of Public Economics, 245, 105345. link B Moderate
  13. Attanasio, O., Guarín, A., Medina, C., & Meghir, C. (2017). Vocational training for disadvantaged youth in Colombia: A long-term follow-up. American Economic Journal: Applied Economics, 9(2), 131-143. link B Moderate
  14. Attanasio, O., Kugler, A., & Meghir, C. (2011). Subsidizing vocational training for disadvantaged youth in Colombia: Evidence from a randomized trial. American Economic Journal: Applied Economics, 3(3), 188-220. link B Moderate
  15. Card, D., Kluve, J., & Weber, A. (2018). What works? A meta analysis of recent active labor market program evaluations. Journal of the European Economic Association, 16(3), 894-931. link A Strong
  16. Crépon, B., Duflo, E., Gurgand, M., Rathelot, R., & Zamora, P. (2013). Do labor market policies have displacement effects? Evidence from a clustered randomized experiment. Quarterly Journal of Economics, 128(2), 531-580. link B Moderate
  17. Dynarski, S., Jacob, B., & Kreisman, D. (2018). How important are fixed effects and time trends in estimating returns to schooling? Evidence from a replication of Jacobson, Lalonde, and Sullivan, 2005. Journal of Applied Econometrics, 33(7), 1098-1108. link B Moderate
  18. Fortson, K., Rotz, D., Burkander, P., Mastri, A., Schochet, P., Rosenberg, L., et al. (2017). National experimental evaluation of Adult and Dislocated Worker programs: Study highlights. Mathematica Policy Research. link B Moderate
  19. Grosz, M. (2020). The returns to a large community college program: Evidence from admissions lotteries. American Economic Journal: Economic Policy, 12(1), 226-253. link B Moderate
  20. Hyman, B. G. (2022). Can displaced labor be retrained? Evidence from quasi-random assignment to Trade Adjustment Assistance (Working Paper CES-22-05). U.S. Census Bureau, Center for Economic Studies. link B Moderate
  21. Jacobson, L., LaLonde, R., & Sullivan, D. G. (2005). Estimating the returns to community college schooling for displaced workers. Journal of Econometrics, 125(1-2), 271-304. link B Moderate
  22. McKenzie, D. (2017). How effective are active labor market policies in developing countries? A critical review of recent evidence. World Bank Research Observer, 32(2), 127-154. link B Moderate
  23. Possebom, V., & Riva, F. (2025). Probability of causation with sample selection: A reanalysis of the impacts of Jóvenes en Acción on formality. Journal of Business & Economic Statistics, 43(2), 391-400. link B Moderate
  24. Schochet, P. Z., D'Amico, R., Berk, J., Dolfin, S., & Wozny, N. (2012). Estimated impacts for participants in the Trade Adjustment Assistance (TAA) program under the 2002 amendments. Mathematica Policy Research and Social Policy Research Associates for the U.S. Department of Labor. link B Moderate
  25. Boockmann, B. (2015). The effects of wage subsidies for older workers. IZA World of Labor, 189. link B Moderate
  26. Boockmann, B., Zwick, T., Ammermüller, A., & Maier, M. (2012). Do hiring subsidies reduce unemployment among older workers? Evidence from natural experiments. Journal of the European Economic Association, 10(4), 735-764. link B Moderate
  27. Berger-Gross, A. (2021). The long-term impact of apprenticeship on the employment outcomes of displaced workers (LEAD Technical Paper No. 1). North Carolina Department of Commerce, Labor & Economic Analysis Division. link C Limited

Every source for this factor is listed on the technological factor page.