# Why Do Economists Keep Getting AI Wrong?

According to Alex Imas, a behavioral economist at the University of Chicago Booth School of Business, the core problem is that economists are applying historical frameworks to a technology that is fundamentally different. In a paper published in the American Economic Review in April 2026, Imas argues that previous automation technologies replaced routine, codifiable tasks. AI, by contrast, can replicate human judgment and decision-making – the very skills economists assumed would remain safe.

"The historical record shows that technological unemployment is rare," Imas said in the Bloomberg interview. "But that record is based on technologies that did one thing: automate physical or simple cognitive tasks. AI is doing something else entirely." This distinction, he argues, is why the consensus view that AI will primarily augment rather than replace workers may be dangerously wrong.

# What Evidence Supports Imas's Challenge to the Consensus?

Imas's argument is not merely theoretical. He points to a series of experimental studies where AI systems demonstrated superior performance in tasks requiring nuanced judgment – from medical diagnosis to legal contract analysis to hiring decisions. In each case, the AI outperformed human experts, not just on speed but on accuracy and consistency.

The National Bureau of Economic Research (NBER) published a working paper in 2025 that corroborates Imas's thesis. The NBER study found that AI adoption in professional services firms led to a 15-20% reduction in junior-level hiring within two years – a displacement effect that standard economic models had not predicted. "The models assumed these roles would be augmented, not replaced," the NBER paper noted. "The data suggests otherwise."

# Who Benefits and Who Loses If Imas Is Right?

If Imas's analysis is correct, the winners and losers are sharply defined. Winners: Companies that can rapidly integrate AI into judgment-intensive roles will gain significant productivity advantages. Tech firms like OpenAI, Google DeepMind, and Anthropic, which build the underlying models, will see demand surge. Consulting firms that pivot to AI-driven advisory services will also benefit.

Losers: Traditional professional services firms – law, accounting, consulting – that rely on a pyramid model of junior talent will face disruption. Workers in roles that involve judgment but not direct client relationships are most vulnerable. Governments that have not updated their social safety nets or retraining programs will face political and economic strain.

# How Should Policymakers Respond to This New Evidence?

Imas's research suggests that current policy approaches – which largely focus on upskilling and tax incentives for AI adoption – may be insufficient. "If AI is replacing judgment, not just tasks, then retraining for higher-level judgment work may not be a viable solution," Imas said. "We need to think about broader social insurance and perhaps a different relationship between work and income."

This is a sharp departure from the OECD and World Bank policy frameworks, which assume that AI will create as many jobs as it displaces. Imas's evidence challenges that assumption directly. Policymakers in the EU, which is drafting its AI liability directive, and in the US, where the White House is preparing a labor strategy for AI, should take note.

# What Remains Uncertain About AI's Labor Impact?

Not all economists agree with Imas. Critics argue that his experimental settings may not translate to real-world complexity. "AI can diagnose a skin lesion from an image, but it cannot run a clinic," said Dr. Sarah Chen, a labor economist at MIT, in a response published alongside Imas's paper. "The complementarity between AI and human workers may be stronger than Imas assumes."

Imas acknowledges this uncertainty. "We don't know the speed or the magnitude yet," he said. "But we know the direction. And the direction is more displacement, not less. Economists who assume otherwise are making a bet with little supporting evidence."

# Predictions

1. By Q4 2027, the OECD will revise its estimate of AI-driven job displacement upward by at least 50%, citing Imas's research and subsequent NBER studies.
2. Within 12 months, at least two major US consulting firms (McKinsey, BCG, or Deloitte) will announce significant restructuring of their junior talent models in response to AI displacement evidence.
3. The EU AI Office will incorporate labor displacement risk assessment into its high-risk AI classification framework by Q2 2027, directly referencing Imas's findings.

# Timeline

• April 2026: Alex Imas publishes paper in American Economic Review challenging consensus on AI and jobs.
• 2025: NBER working paper finds 15-20% reduction in junior hiring in professional services due to AI.
• 2024-2025: Multiple experimental studies show AI outperforming humans in judgment tasks.
• 2023-2024: Prevailing economic consensus holds that AI will augment, not replace, workers.

# Article Summary

• Economists have been using flawed historical analogies to predict AI's labor impact.
• AI's ability to replicate judgment, not just tasks, makes it fundamentally different from past automation.
• Imas's research provides a credible, evidence-based challenge to the dominant consensus.
• The burden of proof has shifted: those who claim AI will be different must now provide evidence.
• Policymakers and business leaders should prepare for faster and more profound displacement than current forecasts suggest.