We saw before that unemployment has a floor made of friction and skills mismatch that printing money can't touch. This essay is about the only thing that can move it: reform and retraining. It's a single living system — change any underlined number or drag any chart and the whole argument, including these sentences, re-reasons around you.
Start with the force that keeps re-opening the gap. Automation and trade displace workers from declining roles at roughly of the workforce each year. That doesn't have to mean fewer jobs overall — new roles appear elsewhere — but they appear in different places and need different skills. At a year,
Displaced people don't slot straight into the new openings. Some are between jobs and find one quickly; many face a skills gap (their trade is shrinking, the growing one needs training) or a geography gap (the jobs are in another city they can't easily move to). Two channels pull them back into work: formal retraining, currently absorbing of the workforce a year, and natural reabsorption (ordinary hiring, helped by mobility) at . Together that's flowing back in against flowing out — so reabsorption displacement, leaving .
This is why displacement is so costly to people, not just statistics: workers who lose a job to structural change suffer lasting earnings losses. In your scenario the model puts the long-run wage scar at . Displaced-worker studies (Jacobson, LaLonde & Sullivan 1993; Davis & von Wachter 2011) find earnings still 15–20% below trend many years later.
Two things make this genuinely hard. First, the treadmill: the faster automation moves, the faster skills go stale, so retraining loses traction exactly when it's needed most. At displacement, automation is moving . Second, time: retraining costs money now and pays off later. Over a horizon, structural unemployment is its floor, ending near .
There is no single fix; there's a mix, and you hold the dials. Spend of GDP on active labour-market programs, at effectiveness, with of possible mobility support (help to move, or to bring work to people). Meta-analysis of 200+ programs (Card, Kluve & Weber 2018) finds job-search help works fast and cheap, training pays off over 2–3 years, and pure public job-creation rarely does.
So is it worth it? On these settings the public investment . And the wider scoreboard — the part that matters changes with whose eyes you read through:
The model above isn't invented from nothing; it's shaped to echo a few robust findings. Flexicurity — Denmark spends close to 2% of GDP on active programs and pairs easy hiring/firing with strong retraining, keeping structural unemployment low. Dual apprenticeships (Germany, Switzerland) smooth the school-to-work and retrain-to-work transitions so the mismatch never grows as wide. Short-time work (Germany's Kurzarbeit) held workers in jobs through 2008 and 2020 instead of scarring them. And the sobering counterweight: mobility is falling (Molloy, Smith & Wozniak 2011) — people move to opportunity far less than they used to, which is why the dial matters as much as the training one.
A deliberately simplified, stylised model — internally consistent and shaped to match the direction of the cited research, but the numbers are illustrative, not a forecast or a fit to any one economy. The studies referenced are real; the dials are a teaching device. Not policy advice. Everything recomputes live from what you change.