Artificial intelligence and economic growth
An evaluator found an error in a proof supporting a central result of Aghion, Jones, and Jones; the authors confirmed a corrected proof.
The questions
We focus on research that bears on decisions people are making now, where an outside expert view is missing and the claims can be checked.
How fast
How quickly will AI change what work gets done, and what do economists, forecasters, and AI-company staff actually expect?
Where value lands
How will gains split between workers, firms, consumers, governments, and owners of capital? What happens to the labor share?
Who is left out
Do low- and middle-income countries capture the benefits, or face displacement without the infrastructure to gain from augmentation?
What to do
Which fiscal and policy responses hold up: new tax bases, income support, sovereign funds, and measurement systems that can track the shift?
Candidate research
These are drawn from our prioritization dashboard. Two are prioritized for evaluation now. The rest are candidates; the team has made no decision on them. Each links to a page with an AI-generated brief, human ratings where we have them, and room for your input.
Forecasts and scenarios
- Forecasting the Economic Effects of AI Prioritized. Karger, Kuusela, and Abaluck elicit forecasts of growth and labor outcomes from economists, AI-company staff, policy researchers, superforecasters, and the public. Already reused in fiscal and tax modeling and in commentary aimed at the Fed. Evaluation focus: sample selection, scenario definitions, and how to read disagreement between groups.
- Economic Scenarios for Transformative AI Prioritized. Korinek, Jones, and co-authors map AI capability paths onto GDP, the labor share, wages, and unemployment to 2030. Evaluation focus: model structure, calibration, labor-market adjustment, and how sensitive the scenarios are to their assumptions.
Measuring labor-market effects so far
- Labor market impacts of AI: A new measure and early evidence A framework from Anthropic's economic research team for tracking AI's labor-market effects, intended as a running baseline. Its measurement choices deserve scrutiny before it becomes a reference point.
- Still Waters, Rapid Currents Danish administrative data show precise null effects of chatbot adoption on earnings and hours, alongside task restructuring. One of the most-cited early causal studies.
- What Work Does Generative AI Do? A nationally representative survey suggesting adoption is widespread but shallow. It informs how fast AI diffuses inside jobs.
- Canaries in the Gold Mine Firms with more AI-skilled staff saw faster productivity growth recently, attributed to organization capital. The identification and the new measure need checking.
Across countries
- Disruption without dividend? Argues workers in developing countries face displacement without the digital infrastructure to gain from augmentation, and that standard exposure indices overstate effects there.
- Generative AI and the Global Redistribution of Expertise Asks whether AI erodes or extends the returns to expertise in low- and middle-income labor markets.
- Adopting Fast and Slow Firm-level data from more than 20,000 firms on how AI adoption differs by country income level and prior digitalization.
- AI Adoption in Developing Countries Which firm capabilities and complementary investments productive AI use requires, drawing on earlier digital technologies.
Policy responses
- Public Finance in the Age of AI: A Primer How tax systems should change if AI erodes labor-income tax bases: consumption taxes, taxes on compute, sovereign wealth funds, and windfall clauses.
Briefs and scores are AI-generated and checked by our team; human ratings are shown where we have them. A low priority score is not a judgment of a paper's quality. Our broader stream also covers the pace of AI progress and AI governance; see the overview.
Our past work closest to this area
Few of our 59 evaluated papers are directly about the economics of AI. Several more involve the methods this area depends on: growth models, structured forecasting, labor-market evidence, and direct cash transfers as a way to share gains.
Existential risk and growth; the returns to science
Two growth and innovation models that trade off faster technological progress against catastrophic risk, each with multiple evaluations and an author response.
Forecasting existential risks (the XPT)
Evaluations of the tournament's elicitation and aggregation methods. The same institute produced the forecasting paper we have now prioritized.
Universal basic income in Kenya; cash transfer amounts in the US
Evaluations of two large randomized trials of unconditional cash, directly relevant to proposals for distributing AI-driven gains.
We have also evaluated labor-market studies of migration and guest-worker programs, and work on public R&D and productivity. All evaluations.
How to help
- Tell us which claims you rely on. If your organization uses this research, which estimates matter, and what would change if they were wrong?
- Suggest evaluators. We need labor economists, growth and macro economists working on AI, specialists in expert elicitation and survey methods, and development economists. Join the evaluator pool or see other ways to get involved.
- Rate and nominate papers on the AI prioritization page, or submit research we have missed.
- Fund or partner. Speed and evaluator quality depend on funding. If you want independent evaluation of research your organization produces or relies on, get in touch.