Karger et al. — What Economists and Superforecasters Predict About AI and Jobs

Predicting AI's economic impact requires navigating deep uncertainty, which is exactly the territory where professional forecasters are supposed to outperform casual opinion. This study surveyed two distinct expert groups — academic economists who study technology and labor, and superforecasters with documented track records of accurate predictions — about AI's likely effects on GDP, employment, and labor force participation under different AI progress scenarios. The disagreements between the two groups are as informative as the consensus.

Key Findings

• Under rapid AI progress, GDP grows but jobs shrink. Both groups agree that fast AI advancement would boost GDP by approximately 3.5% over the study's forecast horizon. However, they also predict roughly 10 million fewer jobs in the economy — growth that benefits capital and productivity metrics but not necessarily workers.
• 41% of economists predict declining labor force participation. Nearly half of the economists surveyed believe that AI will cause a measurable decline in labor force participation — meaning not just unemployment but people leaving the workforce entirely, whether through early retirement, disability, or discouragement.
• Superforecasters are more pessimistic about job displacement than economists. On most measures related to job losses and worker displacement, the superforecasters gave darker predictions than the economists. The forecasters with the best track records of predicting complex outcomes see more downside risk for workers.
• Both groups see AI as different from previous technological shifts. A strong majority in both groups rejected the analogy that "AI will be like the internet or industrial revolution — disruptive but ultimately job-creating." They see AI as potentially breaking the historical pattern where new technology creates more jobs than it destroys.
• Wage effects are highly uncertain. Both groups expressed low confidence in predictions about wage impacts. Some expect AI to raise wages for workers who complement AI capabilities. Others expect downward wage pressure for the majority, with gains concentrated at the top. The range of estimates was unusually wide even by forecasting standards.

What This Means for Your Career

The GDP-up-but-jobs-down finding is the headline to internalize. Economic growth does not automatically translate into job growth when the growth is driven by automation. If AI boosts productivity by replacing human labor rather than enhancing it, the economy can expand while employment contracts. This has happened before in specific sectors (agriculture, manufacturing), but economists and forecasters are now saying it could happen economy-wide.

The divergence between economists and superforecasters deserves attention. Superforecasters are selected specifically for their ability to make accurate predictions about complex, uncertain events. When they are more pessimistic than domain experts, it is worth asking why. One interpretation: economists may anchor on historical patterns (technology has always created more jobs), while superforecasters weight the specific characteristics of this technology more heavily.

For career planning, the 41% labor-force-participation finding is a warning. It does not mean 41% of people will lose their jobs. It means a significant fraction of experts believe the workforce itself will shrink — that some people will simply not find a viable role in an AI-augmented economy. The best defense against this scenario is to be in the group that complements AI rather than competes with it: workers who use AI tools to amplify their output, not workers whose output AI can replicate independently.

Data Highlights

• 69 economists and 38 superforecasters surveyed
• +3.5% GDP expected under rapid AI progress scenario
• 10 million fewer jobs predicted alongside that GDP growth
• 41% of economists predict declining labor force participation
• Superforecasters consistently more pessimistic than economists on job displacement

Methodology

The study surveyed 69 economists specializing in technology, labor markets, and macroeconomics, along with 38 superforecasters drawn from established forecasting tournaments (individuals with documented above-average prediction accuracy). Each participant was presented with defined AI progress scenarios — baseline, moderate, and rapid — and asked to provide calibrated probability estimates for outcomes including GDP growth, employment levels, labor force participation, and wage distributions. The survey used structured elicitation techniques to reduce anchoring bias and captured confidence intervals alongside point estimates. Results were analyzed by expert group to identify systematic differences in outlook between economists (domain experts) and superforecasters (calibration experts).