Trapped Workers: Which Moves Actually Change a Worker's Options? • Bipartisan Policy Center

The Bipartisan Policy Center has published an issue brief called 'Trapped Workers,' which deserves a small round of applause for at least naming the condition out loud instead of calling it dynamic labour-market reallocation. The data inside is bracing: net inflow into more AI-exposed work peaked in 2020 at 2.2 million weighted transitions, then slid to 0.2 million by 2024. That is not a cooling-off period. That is the ladder being quietly withdrawn in public, on schedule, over four consecutive years, with nobody's permission required. Two point two million down to two hundred thousand is a trend line, and trend lines in labour economics are just distress with a regression attached.

The finer print is where it gets properly funny, in the way a slammed thumb is funny. Of the workers leaving highly AI-exposed occupations, 41% land in another job inside the same occupational cluster — same blast radius, new lanyard. That leaves 59% who do cross clusters, and their prize is that the reachable next jobs carry 6.5 percentage points less AI exposure on average. Six and a half points. That is the escape hatch: a marginally less exposable corridor leading to another marginally less exposable corridor, planned by nobody and signed off by everyone.

The title asks which moves actually change a worker's options, which presumes a menu of moves exists and somebody is at the counter choosing. What the figures describe is a workforce being invited to retrain its way out of a structural shift, using a mobility map that predominantly routes people back to the neighbourhood they were trying to leave. The honest version of this brief would be one page long: the exits are narrowing, and the people best placed to widen them are occupied publishing briefs about how narrow the exits have become.