The Landing

The problem

When a machine takes an American's job, the standard promise is retraining. It has been the standard promise since the Trade Act of 1974. It is bipartisan, it is reassuring, and the evidence that it works is poor.

Trade Adjustment Assistance is the closest thing we have to a controlled experiment in helping displaced workers. It ran for nearly fifty years, spent billions, and was evaluated repeatedly. The results are not ambiguous in the direction people expect. Four quasi-experimental impact evaluations found TAA ineffective at raising participants' wages; one found participants ending up about 10 percent below their comparison group. The Department of Labor's own national evaluation found that four years in, the net effect on participants was negative by roughly $26,800 — income from all sources came in below the matched comparison group, because the earnings loss during and after training was not offset by the income support received during it.

Read that carefully, because it is the opposite of the intuition. People sent to retraining did worse than statistically similar people who were not. The training years were years out of the labor market, and the skills acquired did not command enough of a premium to repay them.

TAA's authorization lapsed in 2022. Almost nobody noticed, which is its own verdict.

But buried inside the program was a second instrument that worked, and it worked best for exactly the people the main program failed. Wage insurance — Alternative and later Reemployment TAA — did not send anyone to a classroom. It paid a displaced worker a portion of the gap between their old wage and their new, lower one, for a limited period, on condition that they took the job. The evaluations found it halved earnings losses for workers without a high school diploma, and reduced them by 10 to 15 percent for everyone else.

The least-credentialed workers, the ones retraining helped least, were helped most by simply being paid to land.

So America ran the experiment, got the answer, and kept funding the arm that failed while the arm that worked stayed small, capped, age-restricted, and eventually lapsed with the rest of it.

The proposal

Make wage insurance the primary federal response to displacement. Pay half the wage gap for two years, to anyone displaced by automation, trade, or plant closure, on the condition that they are working.

Not instead of training. Training remains available and is genuinely right for some people. But it stops being the default answer, and it stops being the thing a worker must do in order to receive help.

How it would work

1. The benefit. A worker displaced from a job held at least two years, who takes new work at a lower wage, receives 50 percent of the difference between the old and new wage, for 24 months, capped at $15,000 total. Paid through the existing unemployment insurance infrastructure, which already knows the prior wage.

1. Conditioned on working, not on training. The payment starts when the new job starts. This is the design's whole logic: it rewards re-entry rather than subsidising time out of the market, and it removes the perverse incentive in the current system to hold out for a job that matches the old wage while skills and attachment decay.

1. Cause-neutral eligibility. Trade, automation, closure, relocation — the worker's mortgage does not care which. Restricting help by cause was TAA's original sin: it produced a decade of litigation about whether a given layoff was "trade-related," and it created the political spectacle of two workers in the same town with the same loss receiving different treatment.

1. Training stays, unbundled. Anyone eligible for wage insurance may instead take a training entitlement, and the two may be combined part-time. What changes is that nobody is required to enrol in anything to be caught.

1. Publish the outcomes. Reemployment rate, wage-recovery rate, and cost per dollar of earnings preserved, released annually. If the numbers turn against wage insurance the way they turned against retraining, we should be the first to say so.

The numbers

Roughly 1 to 2 million American workers a year experience a long-tenure displacement. Assume 600,000 take up wage insurance — a high estimate relative to TAA's historical caseload of tens of thousands, because eligibility here is far broader.

The median displaced worker who is re-employed takes a wage cut in the neighbourhood of 15 to 20 percent. On a $55,000 prior salary, that is a gap of about $9,000; half of it is $4,500 a year, or roughly $7,500 over two years against the $15,000 cap.

600,000 × $7,500 = about $4.5 billion a year.

That is the whole program. It is roughly one-fifteenth of the Caregiver's Wage (№ 021), it is smaller than what data centers added to PJM electricity bills in four auctions (№ 024), and it is a fraction of what TAA spent per participant to produce a negative result. Wage insurance is cheap precisely because it pays a fraction of a gap rather than a full income, and because it ends when the worker recovers.

The honest caveat on that arithmetic: take-up is the number we are least sure of, and the range that matters is wide. If uptake doubled, so does the cost. We would rather publish a defensible mid-case than a flattering low one.

The honest objections

"This subsidises employers to pay less. They will cut wages knowing the government tops them up." The strongest objection, and it has a name — incidence. Some of the benefit will be captured by employers. The limits are that the payment attaches to the worker's history rather than the job, so an employer cannot see or price it reliably; it is temporary, so it cannot become a permanent wage floor; and it applies only to workers displaced from a prior long-tenure job, who are a small share of any employer's applicant pool. That reduces the problem. It does not eliminate it.

"Paying people to accept worse jobs entrenches downward mobility." This is the objection we take most seriously, because it is partly true: wage insurance is a shock absorber, not an escalator. Our answer is that the alternative on offer is not an escalator either. It is retraining that the evaluations say leaves people worse off, or nothing. A policy that halves the earnings loss of a worker without a diploma is worth having even though it does not restore them. And the escalator, in this catalog, is elsewhere: apprenticeship (№ 028), the fractional professions (№ 014), and ownership (№ 001).

"You are giving up on retraining because it is hard." We are reporting what the evaluations found, including the ones that embarrass a position we would prefer to hold. Training that is employer-linked and paid — apprenticeship — has much better evidence than classroom retraining after displacement, which is why № 028 exists. The failure is specific: mid-career, post-layoff, classroom-based retraining, undertaken while out of work. That is the thing the data condemns, and that is the thing that has been America's default answer.

"Automation displacement is mostly quiet attrition, and those workers were never 'displaced' in the legal sense." Correct, and it is a genuine hole. A worker whose role is never eliminated but whose employer stops hiring at their level does not qualify for anything here. That is exactly why № 025 exists — you cannot design eligibility for a phenomenon you cannot measure. This proposal catches the visible cases and we say plainly that the invisible ones remain uncaught.

"Why should a displaced worker get this when a worker who was never displaced gets nothing?" Because the covenant is about protection against shocks the individual did not cause and could not insure against privately. If the objection is really that everyone deserves a floor, the answer is № 001, and we would rather build that than means-test misfortune more finely.

Sources

• Mathematica Policy Research / U.S. Department of Labor, The Evaluation of the Trade Adjustment Assistance Program — net participant impact of approximately −$26,800 (dol.gov)
• U.S. International Trade Commission, review of TAA evaluations — four quasi-experimental studies finding no wage gains (usitc.gov)
• Evaluations of Alternative / Reemployment Trade Adjustment Assistance wage insurance — earnings losses halved for workers without a high school diploma, reduced 10.5–15.3% for other groups
• Trade Act of 1974 and the 2022 lapse of TAA authorization
• Proposals № 014 (The Fractional Republic), № 021 (The Caregiver's Wage), № 025 (The Automation Ledger), № 028 (The Apprentice Republic)