What happened
Anthropic's Economic Research team, with independent researcher David Roodman and Anthropic economist Maxim Massenkoff, published a meta-analysis of 56 randomized US studies (plus European experimental evidence) on the effectiveness of worker retraining programs as a policy response to AI-driven labor disruption. The headline finding: on average, job training programs produce positive but modest effects — for each person offered a training slot, employment rises 2-3 percentage points and earnings by roughly $1,000/year, against a program cost of about $13,000, with government recovering more than half of spending through added tax revenue and reduced benefits (programs roughly break even overall). A subset of 'sector programs' that partner directly with employers in high-demand industries produce gains several times larger, but have proven difficult to replicate at scale. The authors conclude that if AI displaces workers at scale, existing retraining programs would likely fall short, and recommend investing now in demonstrating and scaling the most promising sector-based models via Anthropic's Economic Futures Research Fund.
Why it matters
This is the most rigorous evidence base yet assembled on whether retraining — the most politically popular policy response to AI job displacement — can actually work at scale, giving boards, policymakers and workforce-strategy leads a quantitative baseline for evaluating retraining investments and informing AI-labor policy positioning.
Action needed
Brief workforce-strategy and public-policy teams on the quantitative limits of retraining as a mitigation lever, and evaluate whether sector-partnership retraining models are viable additions to workforce transition planning.