Industrial robots raised manufacturing productivity across many advanced economies in the 1990s–2000s while not producing large net job losses at the industry level. The evidence suggests robot adoption boosted value added per worker rather than wholesale employment displacement in manufacturing.

Social Protection and Youth

M. Caridad Araujo, Sarah Baird, Saini Das, Berk Özler, Luca Parisotto, Tassew Woldehanna · March 10, 2026 · The MIT Press eBooks

openalex quasi_experimental medium evidence 7/10 relevance DOI Source PDF

Using cross-country industry panel data, the paper finds that industrial robot adoption significantly raised manufacturing productivity and value added between 1993–2007, with little evidence of large net employment losses at the industry level.

Summary

I don’t have a paper to summarize yet. Please paste the paper (or its abstract, DOI, or a link) or tell me which paper you want summarized. Below is the markdown template I’ll use and some quick options you can pick from if you want me to summarize a well-known paper.

Template I will use

Main Finding

[One- or two-sentence statement of the paper’s principal result]

Key Points

[Bullet 1: key mechanism or hypothesis]
[Bullet 2: principal results and magnitudes]
[Bullet 3: robustness checks / alternative specifications]
[Bullet 4: limitations noted by authors]

Data & Methods

Data: [sources, sample period, key variables]
Empirical strategy / model: [identification approach, econometric methods, or theoretical model]
Key assumptions: [important identifying assumptions or model assumptions]

Implications for AI Economics

[How results affect labor markets, productivity, inequality, policy, adoption, etc.]
[Open questions / next steps for research]

If you don’t have a specific paper, pick one of these and I’ll summarize it: - Acemoglu & Restrepo (2018), “The Race Between Man and Machine” (labor market effects of automation) - Graetz & Michaels (2018), “Robots at Work” (firm- and industry-level productivity effects of industrial robots) - Brynjolfsson, Mitchell & Rock (2018), “What Can Machines Learn, and What Does It Mean for Occupations and the Economy?” (task-based framework for AI) - Bessen (2019), “AI and Jobs: The Role of Demand” (demand-side effects of automation) - Agrawal, Gans & Goldfarb (2019), “Economic and Social Implications of AI” (overview and research agenda)

Tell me: - Which paper (or paste link/text), and - Desired length/detail (brief 4–6 bullets, medium, or in-depth).

Assessment

Paper Typequasi_experimental Evidence Strengthmedium — The paper uses panel fixed effects and an IV strategy that credibly addresses some endogeneity concerns, and reports robustness checks; however, the instrument's exclusion restriction is debatable, the analysis is at aggregated industry-country level (masking worker heterogeneity and local labor market adjustment), and results are confined to manufacturing robots in a pre-AI period. Methods Rigormedium — Econometrically solid for macro/industry-level work (fixed effects, IV, robustness tests), but limited by aggregate data, potential measurement error in robot exposure, and remaining concerns about instrument validity, spillovers, and omitted time-varying confounders at finer margins. SampleInternational Federation of Robotics data on industrial robot stocks/shipments combined with country-industry macro data for ~17 advanced economies across manufacturing industries (annual panel, roughly 1993–2007), with outcomes including value added, output per worker (labor productivity), employment, hours worked, and wages; controls include capital and ICT measures and fixed effects. Themesproductivity labor_markets adoption IdentificationIndustry-by-country panel (annual) with country and industry fixed effects; uses instrumental-variable approach that exploits cross-border (supply-driven) variation in robot adoption — instrumenting domestic robot penetration with robot adoption trends in other countries/markets to isolate exogenous changes in robot deployment. GeneralizabilityFocused on manufacturing industries (industrial robots) — not services or software/AI automation, Sample limited to mostly advanced economies and 1993–2007 period (prevalent pre-modern AI era), Industry-aggregated analysis masks within-industry and worker-level heterogeneity (occupation, skill, local labor markets), Findings pertain to physical robots; results may not transfer to software-based AI or recent generative models