Just Transition Policies for AI Job Displacement

The phrase 'just transition' comes from the labor and environmental movements. It was first used about workers in fossil fuel and chemical industries facing closures. Applied to AI, it asks who pays when automation raises overall productivity but erases particular jobs. The evidence from past waves is sobering. Studies of workers laid off in mass layoffs, including a well-known 1993 study by Jacobson, LaLonde and Sullivan, found that long-tenured workers suffered large earnings losses that persisted for years. Research on the 'China shock' by David Autor, David Dorn and Gordon Hanson found that trade-exposed US communities saw lower employment and wages for well over a decade, not a quick adjustment. Studies of industrial robots have found similar local employment effects. The lesson is that aggregate gains can coexist with concentrated, lasting harm. The main policy tools each have trade-offs. Retraining is popular but uneven. Evaluations of Trade Adjustment Assistance found mixed results, partly because training takes time and does not always lead to available jobs. Sector-based programs tied to employer demand, such as Project QUEST and Year Up, have shown strong earnings gains in randomized trials. Wage insurance pays part of the difference when a worker takes a lower-paying job, which encourages quick reemployment. The US offered a version to older trade-displaced workers. Unemployment insurance reform could widen eligibility to gig and part-time workers and add training allowances. Short-time work programs keep workers attached to employers during temporary shocks, but they are less suited to permanent job loss. A common misconception is that retraining alone solves displacement. Older workers, those in weak local labor markets, and those who need income while they train often need several tools combined.