The Apprentice Republic

The problem

America has about 678,000 registered apprentices. That is real growth — up from roughly 360,000 in 2015, an 88 percent increase in a decade, and one of the few genuinely bipartisan workforce successes of the period.

It is also 0.3 percent of the working-age population.

The share of working-age adults in apprenticeship is roughly five times higher in Canada, seven times higher in Germany, and eleven times higher in Switzerland. Among the young the gap is starker still: about half of Germans aged 18 to 24 are in an apprenticeship, against roughly 2 percent of young Americans. In Switzerland, Germany and Austria, something near 70 percent of young people begin their working lives as apprentices.

The usual explanation is culture, and the usual explanation is wrong, or at least lazy. America had a functioning apprenticeship system within living memory. What it has now is a structural problem with a name: the poaching problem.

An employer who trains a worker bears the full cost. A competitor who does not train can hire that worker afterwards at a premium and pay nothing toward the training. Each firm's rational move is to let someone else do it. So nobody does, everybody complains about a skills shortage, and the cost of qualification is pushed onto the eighteen-year-old, who borrows to buy it from a university that faces no obligation to make it pay off.

That arrangement was defensible when the thing worth buying was knowledge. It is becoming indefensible now, for a specific reason.

Proposal № 014 laid out the evidence that AI is levelling: it raises novice performance sharply and expert performance barely, because it lends every worker the best colleague's brain. If that holds, then the market value of taught, transferable, examinable knowledge — the thing a degree certifies — is falling. What stays scarce is judgment under real conditions, accountability, physical competence, and everything that can only be learned standing next to someone who already knows how.

That is the definition of an apprenticeship. The AI economy is, unexpectedly, an argument for the oldest training model we have.

The proposal

Solve the poaching problem with a train-or-pay levy, set national occupational standards, strip degree requirements where they are not needed, and fund an apprenticeship year at parity with a college year.

How it would work

1. Train or pay. Employers above a threshold either run qualifying apprenticeships or contribute to a national apprenticeship fund at a small percentage of payroll. Firms that train draw down more than they put in; firms that free-ride finance the ones that don't. This is not a novel American invention — the United Kingdom's apprenticeship levy and Denmark's employer-financed reimbursement scheme both work this way, with mixed but instructive results. It is the only mechanism that directly addresses why firms under-train, and every proposal that skips it is proposing exhortation.

1. National occupational standards, industry-written. Portable credentials defined per occupation by industry bodies and certified federally, so an apprenticeship completed at one employer is recognised by every other. Portability is what makes the credential worth the apprentice's time; without it, training is a firm-specific asset the worker cannot sell.

1. Parity of funding. An apprenticeship year is financed at parity with a subsidised college year — the same public money, following the student, to either path. Currently public subsidy is overwhelmingly available for one of the two and the choice young people face is not a fair comparison.

1. Strip the degree requirement where it is decorative. Federal jobs, and federal contractors, drop degree requirements for roles that do not need them, following the states that began doing this in the 2020s. A degree requirement on a job that does not require a degree is a tax on people who could not afford one.

1. Apprenticeship beyond the trades. This is the part that makes it an AI-era proposal rather than a nostalgia one. Standards should cover software, cybersecurity, health technology, laboratory work, logistics, finance operations, and public administration. Britain and Germany both apprentice bankers and IT technicians. America has largely restricted the model to construction and manufacturing, which is why it reads as a consolation prize.

1. Connect it to the covenant. The American Year of service (№ 005) carries apprenticeship tracks, and the fractional professions of № 014 open their tiered licences to apprenticeship-qualified practitioners, not only to degree holders.

The numbers

Getting from 0.3 percent to Canada's roughly 1.5 percent of working-age population means going from about 678,000 apprentices to around 3.4 million. That is a fivefold increase, and it is the modest target — Germany's share would imply nearly 5 million, Switzerland's over 7 million.

Cost depends almost entirely on who pays the wage, and in a working apprenticeship system the answer is the employer, because the apprentice is producing. The public cost is the classroom component, the administration of standards, and the funding parity. At a public subsidy of roughly $8,000 per apprentice-year — well below the public cost of a subsidised college year — reaching 3.4 million apprentices implies on the order of $27 billion a year, substantially offset by the levy itself and by reduced student lending.

Set that against what it displaces. Roughly four in ten recent American graduates work in jobs that do not require their degree (№ 014). A large share of the public money currently spent producing that outcome is money spent badly, and this proposal does not ask for it back — it asks that the alternative path be funded on the same terms.

The honest objections

"Germany's system rests on institutions America does not have." The strongest objection and largely correct. German apprenticeship depends on chambers of commerce with statutory authority, works councils, sectoral bargaining, and a century of employer coordination. You cannot legislate those into existence, and proposals that assume you can have failed before. What the levy does is address the economic obstacle — the poaching externality — without pretending to import the institutional one. That is a partial fix and we present it as one.

"Early tracking is the ugly side of the German model, and you are importing it." We are not, and the distinction matters enough to be explicit: Germany sorts children into academic and vocational streams at around age ten, which is a serious injustice and produces measurable inheritance of class. Nothing here tracks anyone. Apprenticeship is offered as a funded option at eighteen and after, entry stays open at any age, and the credential must ladder into further education rather than terminate it. If a version of this proposal creates a sorting mechanism for children, it should be defeated.

"A payroll levy is a tax on jobs, and it will cost employment at the margin." It is a payroll tax, and we should say so rather than call it a contribution. The evidence from the UK levy is genuinely mixed: it raised training spending but also produced relabelling of existing management courses as "apprenticeships." Two design responses, neither complete: national standards make relabelling harder, and the threshold keeps it off small employers. We would rather defend a real tax honestly than smuggle one in.

"Apprenticeships train people for jobs that AI will eliminate next." A serious worry, and it is why standards must be reviewed on a short cycle and why the occupations covered must extend well beyond the trades. But note the asymmetry: this is an argument against any specific training, including the four-year degree, and the degree takes longer, costs the student more, and is revised more slowly. If the future is uncertain, shorter and employer-paid beats longer and student-financed.

"This is class sorting with better branding. The rich will still go to college." Partly true and it will remain partly true. The mitigation is that parity of funding and laddering credentials make apprenticeship a genuine choice rather than a residual one, and that stripping decorative degree requirements removes the penalty for choosing it. We are not going to claim it equalises status. Status follows money and access, and the honest version of this proposal is that it improves the material outcome first and hopes status follows.

Sources

• U.S. Department of Labor, registered apprenticeship data — 678,014 active apprentices in 2025, up from 359,776 in 2015 (apprenticeship.gov)
• Third Way, America's Apprenticeship Gap and What Five Countries Can Teach America About Apprenticeships — participation five times higher in Canada, seven in Germany, eleven in Switzerland (thirdway.org)
• OECD and BIBB data on German dual vocational training participation among 18–24 year olds
• UK apprenticeship levy evaluations; Danish employers' reimbursement scheme (AUB)
• Proposals № 005 (The Service Dividend), № 014 (The Fractional Republic), № 026 (The Landing)