What 69 Economists and 38 Superforecasters Actually Predict About AI and Your Job

Jamie Dimon says AI will eliminate jobs. Sam Altman predicts whole job categories will disappear. Dario Amodei suggests 10-20% unemployment within five years.

But what do the people who actually study economies for a living think?

A new 224-page study from the Federal Reserve Bank of Chicago, the Forecasting Research Institute, and researchers at Yale, Stanford, and UPenn just gave us the most rigorous answer we have. The paper — "Forecasting the Economic Effects of AI" by Ezra Karger, Otto Kuusela, and 12 co-authors including legendary forecasting researcher Philip Tetlock — surveyed economists, AI industry professionals, superforecasters, and the general public about what they actually expect to happen.

The results are more nuanced than the headlines suggest. And the implications for your career depend heavily on which scenario plays out.

The Three Scenarios Economists Were Asked About

The researchers defined three levels of AI progress by 2030, each with concrete capability benchmarks:

Slow progress means AI stays at roughly PhD-student-level research assistance, handles about half of freelance coding tasks, and writes passable but not great creative content. Think: a very capable but limited assistant.

Moderate progress means AI can run semi-autonomous research labs, handle almost all freelance software engineering, produce high-quality novels, and manage complex projects with human oversight. Robo-taxis work everywhere.

Rapid progress means AI surpasses top humans in research, coding, and leadership. Autonomous researchers compress years of work into days. Robots can assist in any home or factory in the world.

Here's the critical finding: economists assigned a 61.4% probability to moderate or rapid progress by 2030. They think meaningful AI advancement is more likely than not. The debate isn't about whether AI will get more capable — it's about what happens to the economy when it does.

The Headline Numbers

Under the most likely scenario (moderate progress, 47% probability), economists forecast the economy largely stays on its current path. GDP growth continues around 2.5%, and labor force participation drifts down gradually, mostly from demographics.

Under rapid progress (14% probability but significant consequences), the forecasts shift substantially. Economists predict annualized GDP growth of about 3.5% — a level the U.S. hasn't sustained since the postwar boom of the 1950s-1960s. But this growth comes with significant labor market disruption: labor force participation falling to 55% by 2050, down from 62.6% today. The researchers estimate roughly 10 million fewer jobs attributable specifically to AI, above and beyond demographic trends.

Wealth concentration also accelerates. Under rapid progress, the top 10% of households are forecast to hold 80% of national wealth by 2050 — the highest level since the late 1930s.

The critical takeaway: even the rapid scenario doesn't produce historically unprecedented outcomes. These numbers have parallels — postwar GDP growth, pre-WWII inequality levels, the LFPR before women entered the workforce en masse. Economists are saying AI could reshape the economy significantly, but not beyond what we've seen before.

The Surprising Finding That Changes Everything

Here's what makes this study genuinely different from the usual AI predictions: a variance decomposition analysis reveals that expert disagreement about AI's economic impact is driven almost entirely by differing views on economic mechanisms — not by disagreement about whether AI will actually advance.

For 2030 GDP growth forecasts, within-scenario variance (economists disagreeing about what happens given the same level of AI progress) accounts for 94.9% of total variance. Between-scenario variance (disagreement about which AI scenario is most likely) accounts for just 5.1%.

In plain language: economists mostly agree that AI capabilities will advance significantly. Where they disagree — sharply — is on how fast the economy can absorb those capabilities. Some think adoption lags, infrastructure bottlenecks, regulatory friction, and demographic headwinds will keep growth modest even with transformative technology. Others think the gains will be faster and larger.

This matters for you personally because it means the risk to your career isn't primarily about whether AI gets better — it will. The risk is about how your specific industry, occupation, and geography adapt to those capabilities.

Which Jobs Are Most and Least Exposed

The study asked economists to rank 43 occupation groups by expected employment change from 2025 to 2030. The results largely confirm what occupation-level data from other sources shows, but with some interesting nuances.

Occupations where a majority of economists predict employment growth: personal service workers, personal care workers, health professionals, military roles, hospitality managers, protective service workers. These share common characteristics: physical presence requirements, deep human interaction, and tasks that are difficult to standardize.

Occupations where a majority of economists predict employment decline: general and keyboard clerks, clerical support workers, assemblers, machine operators, drivers. These are the classic automation-exposed roles — routine cognitive tasks or predictable physical tasks.

The interesting middle ground: business and administration professionals, IT professionals, teaching professionals, sales workers, and legal professionals all show split predictions. Some economists see growth, others see decline. This uncertainty zone is exactly where individual career decisions matter most — the same job title can be low-risk or high-risk depending on how the specific role is structured.

The Policy Divide That Should Concern Everyone

The study reveals a sharp disagreement between experts and the general public on what to do about AI-driven job displacement. Economists overwhelmingly favor targeted worker retraining (71.8% support) over broader interventions. Universal basic income gets only 37.4% economist support, and job guarantee programs just 13.7%.

The general public, by contrast, supports all these interventions — including job guarantees (57.1% support) and UBI (47.9%).

This gap matters because it suggests policy responses may be slower, more incremental, and more targeted than displaced workers might hope for. If you're waiting for a policy safety net, the data suggests it's more likely to look like retraining credits than a universal basic income.

What This Means for You

This research reinforces a point we've been making since we launched RiskQuiz: the question isn't binary. It's not "will AI replace me — yes or no." It's a probability distribution that depends on your specific situation — your role structure, your industry's adoption speed, your adaptability, and the broader economic forces shaping your occupation.

The Karger et al. study shows that even the smartest people in the room — economists who study this for a living, superforecasters with verified prediction track records — can't agree on the magnitude of the impact. What they do agree on: meaningful AI progress is more likely than not, and the distribution of outcomes is wide enough that individual preparation matters enormously.

Your risk isn't fixed. It's a function of choices you make in the next 12-18 months.

Want to know where you stand? Take the 90-second RiskQuiz assessment to get a personalized AI replacement risk score based on 9 research-backed dimensions — including the occupation-level data this study confirms. Your detailed report includes a 30-day action plan tailored to your specific work type, industry, and risk profile.

---

Source: Karger, E., Kuusela, O., Abaluck, J., Bryan, K., Halperin, B., et al. (2026). "Forecasting the Economic Effects of AI." Federal Reserve Bank of Chicago / Forecasting Research Institute. March 2026. Referenced via Marginal Revolution

---

FAQ

What is the Karger et al. AI economist survey?

A March 2026 study from the Federal Reserve Bank of Chicago and the Forecasting Research Institute that surveyed 69 economists, 52 AI industry and policy professionals, 38 superforecasters, and 401 members of the general public about their predictions for AI's impact on the U.S. economy through 2050.

What do economists predict about AI job losses?

Under the most likely scenario (moderate AI progress, 47% probability), economists predict labor force participation continues its gradual demographic decline. Under rapid AI progress (14% probability), they forecast the labor force participation rate dropping to 55% by 2050, with roughly 10 million of those lost positions attributable to AI rather than demographics.

Which jobs do economists think are safest from AI?

Occupations with the strongest positive employment growth predictions include personal service workers, personal care workers, health professionals, and roles requiring physical presence, deep human interaction, and tasks that resist standardization.

Which jobs do economists think are most at risk from AI?

Clerical roles, data entry, machine operation, assembly work, and driving occupations consistently show the highest predicted employment declines across all AI progress scenarios.

Do experts agree on AI's economic impact?

They agree that AI capabilities will advance significantly (61.4% assign moderate or rapid progress probability). They disagree sharply on economic consequences — the variance decomposition shows 94.9% of forecasting disagreement comes from differing views on economic mechanisms, not AI capabilities.