How Much Redistribution Would AI-Driven Growth Require?

Summary

The article examines whether artificial intelligence will require massive redistribution if it sharply reduces labor’s share of national income. Its central argument is that labor income depends on the product of GDP and labor’s share, so a smaller share of a much larger economy can still leave workers collectively no worse off. Starting from 2% annual real GDP-per-capita growth and a 60% labor share, the author estimates that 5% annual AI-era growth could allow labor’s share to fall to about 45% over ten years while preserving the no-AI labor-income path. At 10% annual growth, the share could fall to about 28% while producing roughly the same aggregate labor income. Three scenarios illustrate the range: an econ-pessimist case with 2.1% growth and a 56.6% labor share would require transfers equal to 2.7% of GDP; an econ-optimist case with 4.1% growth and a 51.4% share would require no transfer; and a techno-optimist case with 10% growth and a 22.9% share would require a transfer of 5.3% of GDP. The author argues that this amount need not come entirely from higher taxes, because payroll taxes could be reduced and replaced with broader consumption taxation, supplemented by transfers for people without labor income. The analysis does not imply that every worker benefits: occupations could be displaced, industries could undergo severe churn, and transitions could be painful. Its broader conclusion is that labor’s share alone is a poor guide to AI’s distributional effects; the scale of productivity growth is equally important. Under very high growth, the article argues, shorter workweeks and higher hourly wages could potentially coexist with gains for both labor and capital, although this remains scenario-dependent.