AI Expected to Boost Annual Labor Productivity by 1.8% Over the Next Decade... "White-Collar Jobs to Be Replaced by AI" (Comprehensive) [American Economic Association 2026]

2026 Annual Meeting of the American Economic Association (ASSA)
Anthropic: "Labor Productivity to Rise by 1.8%P Annually"
OECD Forecast: Total Factor Productivity Up 0.9%P Per Year
Productivity Falls Initially, Then 'J-Curve' Rebound
Potential

"With the capabilities of current artificial intelligence (AI) models alone, labor productivity could be increased by about 1.8 percentage points per year over the next decade," said Peter McCauley, Chief Economist at Antropic.

"Companies with a high level of AI utilization see their short-term productivity fall by about 1.3%, but after passing through the transition period, sales, employment, and productivity all improve," said Christina McElheran, Professor at the University of Toronto.

On January 3 (local time), at the opening of the '2026 Annual Meeting of the American Economic Association (AEA)' in Philadelphia, Pennsylvania, USA, there was a heated debate on the impact of AI on the economy, including productivity and employment. Unlike last year, when tariffs, inflation, and monetary policy were the main topics, this year saw a significant increase in AI-related sessions, reflecting the recent AI boom. Participants generally agreed that while short-term transition costs are inevitable during AI adoption, in the medium to long term, AI is likely to become a key growth driver for the overall economy.

"Current AI Alone Can Raise Labor Productivity by 1.8% Points Annually"

Peter McCauley, Chief Economist at Antropic, is speaking at a panel discussion at the Annual Meeting of the American Economic Association (ASSA) 2026 held on the 3rd (local time) in Philadelphia, Pennsylvania, USA.

Peter McCauley, Chief Economist at Antropic, a generative AI developer, stated at the session "The Economics of AI and Decision-Making: Who Holds the Initiative?" that "if current AI models are applied to jobs across the economy, labor productivity could be increased by an average of about 1.8 percentage points per year over the next decade."

He noted, "There is a possibility of overestimation when considering bottlenecks in the application process, but since the analysis is based on the capabilities of current models, if AI develops rapidly in the future, it could actually be an underestimation. The current AI is already ready to transform the economy."

A more conservative analysis was also presented. Matthias Schiff, an economist in the Structural Policy Studies Branch at the Organisation for Economic Co-operation and Development (OECD), estimated that "AI could raise the annual total factor productivity (TFP) growth rate in the United States by about 0.3 to 0.9 percentage points over the next decade." TFP is a broader measure of productivity that excludes the direct contributions of labor and capital from total output, serving as a key indicator for assessing growth potential.

However, he predicted that the productivity effects of AI would vary significantly by industry. "The impact will be substantial in knowledge-intensive service sectors such as professional and technical services, law, and finance and insurance, but may be limited in field-based industries such as construction and accommodation," he said. "However, if labor mobility is smooth, the widening gap between industries will not have a major negative impact on overall growth," he added.

Matthias Schiff, an economist at the Organisation for Economic Co-operation and Development (OECD), spoke at a lecture during the American Economic Association Annual Meeting (ASSA) 2026 held on the 3rd (local time) in Philadelphia, Pennsylvania, USA. Photo by Haeyoung Kwon

Initial AI Adoption May Lower Productivity... "J-Curve Rebound After the Transition"

Professor Christina McElheran of the University of Toronto spoke at the 'Annual Meeting of the American Economic Association (ASSA) 2026' held on the 3rd (local time) in Philadelphia, Pennsylvania, USA. Photo by Haeyoung Kwon, Philadelphia Correspondent

Research findings were also presented suggesting that corporate productivity may actually decline in the initial phase of AI adoption. Christina McElheran, Professor at the University of Toronto, introduced research analyzing data from the U.S. manufacturing sector, stating, "Companies with a high level of AI utilization experience a short-term productivity decline of about 1.3%."

She explained that this short-term performance drop "is not due to the technology itself, but rather to the costs incurred during the internal organizational transition process." In fact, during the early stages of AI adoption, shocks from adjustments such as workforce restructuring, inventory increases, and the introduction of industrial robots can worsen short-term performance.

However, she made it clear that this adjustment phase is temporary. Professor McElheran emphasized, "Companies that endure this transition period see improvements in sales, employment, and productivity. The effect of AI is initially negative, but over time it follows a J-curve pattern of improvement." Once companies absorb the initial costs and reach a threshold where organizational transformation and AI utilization stabilize, productivity and performance are expected to improve rapidly.

Will AI Replace White-Collar Workers... Potential for Worsening Income Inequality

The impact on the labor market also emerged as a major issue. Monetary authorities agreed that AI could be a growth engine, but remained cautious about its employment effects. Anna Paulson, President of the Federal Reserve Bank of Philadelphia, said, "We are entering an initial phase where productivity growth rates are stepping up due to AI and regulatory easing," but pointed out that "initial AI investments are concentrated in areas such as data centers, which have limited potential to create jobs." She added, "Once AI is fully integrated into the economy, strong growth phases may occur despite relatively little job creation."

In particular, the potential for changes in white-collar jobs drew attention. Lawrence Schmidt, Professor at the Massachusetts Institute of Technology (MIT), analyzed that AI could weaken the demand for professional jobs requiring higher education or skills in the medium term. He predicted, "AI will favor jobs that are low-education, low-wage, and male-dominated," and that the impact of AI advancement on labor demand will move in the opposite direction from previous technological progress.

Laura Veldkamp, Professor at Columbia University, pointed out that productivity improvements could worsen income inequality, emphasizing, "We need to find ways for the benefits of AI-driven productivity improvements to be shared across society."