France devotes approximately 32 billion euros per year to continuous vocational training — roughly 1.4% of GDP, one of the highest ratios in Europe. Yet six months after training financed by Pôle emploi, fewer than one in two job seekers has found a stable contract. In Germany, this rate exceeds 60%. In Sweden and Denmark, it hovers around 70%. The gap is not marginal. And it cannot be explained by the amount of investment.

This anomaly reveals something precise about how France conceives vocational training: as a public policy to be managed from Paris, with national specifications, centralized certifications, and budgets negotiated sector by sector within tripartite bodies far removed from local territories. Better-functioning systems work differently. They build a direct link between the needs of local employers and the training pathways offered to workers. This is not a nuance of design. It is a fundamental divergence in governance.

The Essentials

  • France spends approximately 1.4% of GDP on continuous vocational training, versus 0.8% for Germany — yet the rate of return to stable employment six months after training is 15 to 20 percentage points lower according to DARES data and the OECD Employment Outlook.
  • Germany, Denmark, and Sweden organize training at the regional and sectoral level, in direct connection with employers — sectors concretely define expected competencies, and regions adapt content in real time.
  • In France, the 2018 reform simplified access to training through the Personal Training Account (Compte Personnel de Formation, CPF), but generated an explosion of short, poorly certifying training programs with no demonstrated link to employment — a structural problem identified by the Court of Auditors in 2022 and 2024.
  • The CPF reform introduced in 2023 (100 euros then 50 euros co-payment) reduced registrations by 30% but did not resolve the issue of training program relevance.
  • The territorial governance overhaul, announced multiple times, remains politically stalled.

Spending More to Train Less Well: Proof by the Numbers

International comparisons are often misleading because they measure different scopes. On continuous vocational training, the OECD has constructed a harmonized indicator that allows comparison of comparable things: public and parafiscal spending devoted to training working adults or job seekers.

According to the OECD Employment Outlook, France regularly ranks among Europe’s top performers on this spending indicator. Germany devotes a significantly smaller share of its GDP to this effort — yet outcome indicators there are consistently superior. This apparent contradiction lies at the heart of the literature on vocational training effectiveness for at least twenty years.

DARES data allows for refinement. In France, among job seekers who completed training prescribed by Pôle emploi (now France Travail), the rate of access to stable employment at six months hovers around 45 to 48% depending on cohorts. It is an average result — it masks significant gaps depending on occupations and employment basins. But it suffices to raise the question: why do countries that spend less per trained worker obtain substantially higher rates of return to employment?

The answer is not in the amount. It is in the chain between training and need.

The German Model Is Not a Miracle: It Is an Institutional Architecture

Bertrand Martinot, an economist specializing in labor market analysis who spent part of his career on comparative analysis of training systems, identified the central mechanism in several works published notably at the Institut Montaigne. His thesis, summarized schematically: effective training systems are not managed from a capital. They are organized around a direct link between the real needs of local employers and the design of training pathways offered to workers.

In Germany, this link is called Verbundausbildung for apprenticeship and its equivalent for continuous training: an architecture in which chambers of commerce, Länder, and sectoral employer associations co-construct competency frameworks and continuously adapt content to local market needs. German Chambers of Commerce do not validate abstract national diplomas. They certify competencies whose demand they know because they experience it from within.

Denmark and Sweden operate on comparable logic, with an additional dimension: training there is conceived as collective insurance against the risk of skills degradation, not as a curative tool for inserting job seekers most distant from the market. The Danish worker accesses training before losing employment, in anticipation of retraining. The French worker accesses it afterward, in response to already-existing unemployment. This temporal shift changes everything about the device’s effectiveness.

This logic recalls a broader observation that institutionalist economists have formulated in other contexts: institutions are not neutral; their architecture determines behavior. As a recent analysis on Mexican nearshoring showed, raw competitive advantage amounts to nothing without institutions capable of transforming it into sustainable economic results. The volume of training spending, without institutions to direct it, produces the same phenomenon.

What the 2018 Reform Really Changed — and What It Failed to Resolve

The 2018 vocational training reform, carried by the “Professional Future” law, is often presented as a major advance: it created the Personal Training Account (Compte Personnel de Formation, CPF) in euros, simplified access to training, and eliminated the monopoly of Competency Operators (OPCO) on prescribing training pathways. Individualization of training was the stated objective. It was partially achieved.

But the Court of Auditors, in two separate reports published in 2022 then in 2024, documented the negative effects of the device. The CPF in euros triggered an explosion in the supply of short training programs, often unrelated to labor market needs. Between 2019 and 2022, CPF registrations tripled. But the bulk of this growth benefited so-called “certified leisure training” — driver’s licenses, skills assessment training, language courses — rather than targeted retraining in tight labor market sectors. The Court of Auditors estimated that fraud and training without added value represented several hundred million euros in annual spending.

The government responded in 2023 with a 100-euro co-payment per training program, later reduced to 50 euros. The effect on registrations was immediate: a 30% decline according to data from the Caisse des dépôts managing the system. But the real problem was not too-easy access to training. It was the absence of a link between training supply and employers’ real demand. A 50-euro co-payment does not build this link.

This is where Martinot’s critique touches on something structural. The 2018 reform gave individuals freedom to choose their training. But it did not give them the tools to make this choice usefully. And it did not create the institutional architecture that would allow directing this freedom toward competencies the economy needs.

Why National Management Asphyxiates Local Responsiveness

French vocational training suffers from a problem of temporality as much as governance. The training certification cycle is long — several years elapse between identifying a need, designing the competency framework, certifying the qualification, and deploying training across the territory. In an economy where skills needs can transform within twelve to eighteen months due to technological change, this delay is not an operational detail. It is a structural factor in mismatch.

French employers experience this concretely. In tight labor market sectors — energy transition, digital, logistics, personal services — companies struggle to find workers trained in the precise skills they need, while certifying training in less-demanded occupations continues to fill because it exists in the catalog and Pôle emploi advisors prescribe it for lack of known alternatives.

Germany handles this problem differently. The Länder have competency over continuous training and dispose of their own budgetary levers to respond to their industrial basins’ needs. When the automotive transition requires high-voltage electricians or battery management technicians, Bavaria can adapt its programs in a few months. France centralizes this decision in national tripartite bodies where negotiations between sectors take years.

This tension between centralization and subsidiarity is not unique to training. It runs through the entire organization of the French employment system. But it is particularly costly here because training is precisely the tool supposed to ensure labor market fluidity. If this tool arrives two years late compared to real needs, it does not ensure fluidity at all.

The Competing Reading: Unequal Access as the Blind Spot of Governance

Martinot’s institutionalist thesis is solidly grounded. But it has a blind spot that must be named.

Decentralization toward employment basins and sectors presupposes that all workers have equivalent access to information about available training, tight labor market occupations, and local employment prospects. This is not the case. Economists who work on the distribution of labor market gains, like Dani Rodrik on “good jobs,” have documented how reforms that improve a system’s overall efficiency can simultaneously widen access inequality to the disadvantage of the less qualified.

In France, DARES data show concerning concentration in access to training. The most educated workers access qualifying training two to three times more often than workers without qualifications. Yet the CPF, by giving individuals freedom of choice, mechanically reproduced this inequality of use: those who know how to identify the right training, mobilize their account, and orient themselves toward the right providers are precisely those who need it least.

Decentralization toward employment basins can worsen this bias if it is not accompanied by strengthened effort in guidance and support for workers most distant from the market. The Danish model knows this and responds with massive investment in individual counseling. It is not simply a decentralization of training: it is decentralization coupled with a service of personalized support, publicly funded, capable of compensating for information inequalities.

The issue therefore is not to choose between efficiency and equity — as if one must be sacrificed to the other. Nordic experiences show that the two are compatible when institutional architecture is coherent. The question is whether France is ready to build this architecture, which requires simultaneously solving the problem of territorial governance and that of support for the most vulnerable populations.

It is a political question as much as a technical one. The same tension between institutional design and concrete results traverses the debate on digital education: governance choices determine effects on the ground, often more than amounts invested.

Paths That Advance, Slowly

It would be inaccurate to present the French system as static. Several experiments deserve mention because they test precisely the hypotheses at stake.

The “Regional Competency Investment Pacts” (PRIC), launched as part of the Skills Investment Plan (PIC) between 2018 and 2022, gave regional councils the capacity to co-finance and co-manage training with the State. Evaluations conducted by DARES and France Stratégie show contrasting but encouraging results in certain regions: rates of access to employment for beneficiaries were above the national average in regions that had built strong connections with local industrial basins. The case of Pays de la Loire, often cited, illustrates what effective coordination between regional council, OPCO, and manufacturing sector employers can produce.

France Travail, the operator resulting from the 2023 Pôle emploi reform, carries an ambition of better coordination between employment and training actors at the local level. The law of December 18, 2023, explicitly creates local employment committees, associating territorial authorities, employers, and training organizations. This is an architecture that aligns with what international comparisons suggest. But the device is new, and first feedback on experience will be available at best in 2025-2026.

The question therefore is not whether the direction is right. It probably is. It is whether institutional actors — State, regions, professional sectors, employers — are capable of building sufficient operational trust together so that local devices actually function, without being recentralized at the first change of government. Economists who have studied how the same data can produce opposite prescriptions know that the problem is never merely technical: it is also a problem of the political coalition capable of sustaining a reform over time.

Thirty-two billion spent each year to train less well than a country that spends less — this is not a problem of will or budget. It is a problem of architecture. And the architects are already at work, in a few regions, in a few basins. The question is whether what works locally can become the rule, rather than the exception that reports cite to show that the system can, sometimes, produce what it was designed to produce.

Sources

  1. DARES — Data on vocational training and return to employment: https://dares.travail-emploi.gouv.fr/
  2. OECD — Employment Outlook, compared indicators of vocational training and rates of access to employment (OECD, annual editions)
  3. Court of Auditors — Reports on the Personal Training Account, 2022 and 2024 (Court of Auditors, Paris)
  4. Caisse des dépôts — CPF usage data, 2023 annual report
  5. France Stratégie — Evaluation of the Skills Investment Plan (PIC), reports 2021-2023
  6. Bertrand Martinot — Works on vocational training and labor market, Institut Montaigne: https://www.institutmontaigne.org
  7. Dani Rodrik — Good Jobs, Harvard Kennedy School (works on distribution of labor market gains)