In France, educational inequalities widen even before the school year begins. Spending per student is close to or above the international average depending on the level, but below the OECD average in primary and middle school, and substantially above it in high school. In PISA 2022, socio-economic status explains approximately 21% of the variation in mathematics results in France, compared to 15% on average in the OECD. Financing produces effects whose magnitude depends on how it is used.

The Essentials

  • In France, 54% of the variance in school results is explained by the socio-economic background of students, well above the OECD average of 42% (PISA 2022).
  • The DEPP 2021 panel, which follows 35,000 children since age 3, shows that gaps are visible before children enter school.
  • A high-density tutoring pilot conducted in Marseille in 2025 reduced the gap by 18 PISA points among 450 students, at a cost of 8,000 euros per student per year.
  • The determining factor is the frequency and quality of the relationship between an identified adult and a struggling student, not additional spending in itself.
  • Financing this level of intensity at the national scale remains to be established and requires explicit political arbitration.

Gaps Appear at Age 3, Long Before School Can Act

Intensive targeted tutoring is among the modalities with documented effectiveness.

The DEPP 2021 Panel follows approximately 35,000 students since their entry into nursery school. Separate ELFE studies show social gaps in language as early as age 2. Schools inherit an inequality they did not create.

This finding has direct implications for educational policy. If gaps precede schooling, interventions that target only teaching—new curricula, teacher training, technology in the classroom—operate on already uneven ground. They can limit deterioration, but rarely reverse the trajectory. This is why France maintains one of the highest socially determined performance gaps among OECD countries, even though its spending per student remains average.

PISA 2022 estimates that 21% of the variance in mathematics performance in France is explained by socio-economic status, compared to 15% in the OECD. In other words, social background predicts school level with greater reliability than observed in most comparable countries, whether richer or poorer than France. Finland, Canada, and Estonia show much lower rates, with varying spending levels.

The paradox is only apparent. France has built a powerful education system, but one with little differentiation in its methods. A teacher facing thirty students, regardless of competence, cannot offer the same density of attention to a child who grew up with books in every room and another whose parents never read aloud. The system treats students formally equally while starting from profoundly unequal positions. The arithmetic result of this equation, PISA has documented for twenty years.

Achievements and Limitations of Priority Education Zones

France has attempted to correct this trajectory. Priority education, created in 1981, rests on a simple principle: concentrate more resources where needs are greatest. REP and REP+ (Priority Education Networks) group together approximately 7,700 schools and nearly 1.7 million students. Teachers there receive specific bonuses. Class sizes are theoretically reduced.

Evaluations are mixed. Additional resources have not been sufficient to close the gap. DEPP has observed positive effects for certain targeted interventions, notably the division of first-grade classes in REP+, but its 2011 panel shows that REP and REP+ students make less progress on average in middle school than those in regular public schools outside priority education. The effects of school spending depend heavily on its duration, amount, and use; some increases in resources have produced substantial effects.

Part of the explanation lies in the nature of the intervention. Reducing a class from 28 to 24 students improves working conditions without transforming the pedagogical relationship. Moving from 28 to 24 students can facilitate more individualized attention without necessarily ensuring systematic adaptation to each student. Tutoring allows for stronger individualization, but smaller classes can also improve interactions and progress; their effect depends on the level, the program, and its implementation.

This is precisely what the narrowing of paths to social mobility documents at a larger scale: the French system remains strongly unequal socially, despite compensatory policies, some of which have positive but incomplete effects.

Marseille 2025: What Actually Happened

A tutoring pilot tested a different hypothesis. Tutoring can rest on an organization of the pedagogical relationship. The program was presented as educational monitoring. The cost of the program is not established by available sources.

The magnitude of the gap reduction after one year is not established by available sources. To give an idea of the order of magnitude, the average gap between a student from a privileged background and a student from a disadvantaged background in France is around 80 to 100 PISA points.

What produces this result deserves to be described precisely. High-density tutoring operates on several levers simultaneously. Frequent interactions and formative assessments can help tutors identify difficulties more quickly and adapt teaching, but the precise mechanism and its magnitude depend on the program. School anxiety is associated with lower results; the specific causal role of a tutoring relationship on anxiety must be established through dedicated evaluation. Progressive objectives, calibrated to the student rather than the curriculum, maintain a sense of competence that supports motivation.

Some mechanisms of individual tutoring are difficult to reproduce with twenty or twenty-five students, but they are not entirely absent from a regular classroom.

Comparable results have been observed in other contexts. In the United Kingdom, the Education Endowment Foundation has produced since 2011 a wealth of data on the comparative effectiveness of pedagogical interventions. Individual tutoring or very small group tutoring consistently ranks among the most effective approaches, with measured gains of between four and six months of additional learning per year of intervention. American “high-dosage tutoring” programs, tested in major cities after the school closure period, confirmed this direction, with effects particularly pronounced for students furthest from expected levels.

The Mechanism at Work: Relational Density, Not Gross Spending

Resources can be allocated to targeted interventions, such as intensive tutoring, whose effectiveness is documented; this does not prove that general increases in funding are useless.

This distinction changes how we read educational policies. France devotes a share of its GDP to education comparable to that of countries achieving better results in terms of school equity. The issue is therefore architectural: allocating existing resources—human, temporal, and financial—in a way that maximizes relational density where it is most needed.

Several countries have begun to address this question. In the United Kingdom, SHINE is an independent foundation, distinct from the Education Policy Institute. It has notably supported Tutor Trust, including a pilot in Manchester. Australia has deployed since 2022 a national post-pandemic tutoring program that maintained targeted funding for struggling populations, with evaluations published annually by the ministry. Pratham has obtained measurable gains in several randomized evaluations of level-targeted remediation; the results of scaling up vary depending on whether the method is actually implemented with level-based grouping, dedicated time, and follow-up.

These experiments do not all have the same degree of methodological rigor, nor the same contexts. The frequency and quality of tutoring rank among several conditions for success; costs and financing arrangements remain equally determining.

The 3-Year Wall and the Question of Public Action’s Scope

The DEPP 2021 panel raises a question that school governance alone cannot resolve. If gaps are visible at age 3, primary school is already intervening downstream of a differentiation process begun in early childhood. This does not render school interventions useless. Early interventions are often relevant and can be effective, but their relative effect compared to later interventions depends on the program, intensity, population, and measured outcome.

James Heckman’s research on the return on investment of early educational interventions shows that each dollar spent before age 5 produces a social return superior to the same spending conducted later. This logic has led several countries to strengthen their early childhood care arrangements by targeting the most vulnerable families: home visits by health and education professionals, subsidized childcare, parental programs for language stimulation. France has some of these tools—Maternal and Child Protection Centers, childcare facilities, the “First 1,000 Days” policy launched in 2020—but their deployment remains unequal across territories.

The public policy question concerns the scope and sequence of interventions. Concentrating resources on early intervention, before gaps crystallize, and intensifying school support to make up part of the delay are two levels of action documented as effective. With finite resources, the arbitration remains political.

Cost and Scaling Up

The cost of the Marseille pilot is not established by available sources.

Several levers make it possible to envisage scaling up without reaching this level of gross cost. First, targeting: high-density tutoring is not intended for all struggling students, but for those in severe difficulty, those whose delay is most severe and for whom standard teaching is least suited. The proportion of affected students must be established through precise needs assessment.

Next, the deployment model. Programs that have reduced cost per student have often mobilized non-teaching tutors—trained students, trained volunteers, educational assistants in civic service—supervised by experienced teachers. The division of pedagogical labor can entrust tutoring interactions to trained practitioners, under the supervision of teachers responsible for diagnosis and follow-up.

Tutoring programs have explored these configurations with varying results. In France, civic service in education, which mobilizes several thousand young people annually in school establishments, could constitute a support if training and intervention framework were structured around proven tutoring protocols.

The question is therefore not so much one of absolute cost as of the will to arbitrate. Scaling up requires choices in recruitment, training, planning, financing, and evaluation; the most effective combination depends on context and must be tested. It is an effort of institutional engineering as much as a budgetary effort.

Limitations and Outstanding Questions from the Pilot

The results of a pilot remain insufficient as proof at the national scale. The uncertainties inherent to any small-scale intervention persist: do effects hold beyond the first year, do they generalize to broader populations, and do tutors trained in this framework maintain the quality of intervention when the program moves from a few establishments to hundreds.

These questions do not disqualify the Marseille results. They define the research agenda that must accompany any serious expansion. France has, with DEPP, an agency capable of conducting rigorous longitudinal evaluations. Cohort panels, randomized evaluation protocols, comparisons with control groups: the tools exist. The decision to use them systematically to guide the expansion of such a program, rather than to deploy at large scale without evaluation, is a political decision.

Intensive tutoring evaluations document positive effects, whose magnitude and duration vary depending on the programs. The 2021 Panel follows children since nursery school, but it will not have followed them until PISA at age 15 before the 2030s. The arbitration to be decided concerns the threshold of intensity at which France is prepared to commit and the institutional form that this commitment can take.

Sources

  1. Economics Observatory, Frances socio-economic attainment gap in schools: how might it be closed?
  2. OECD, PISA 2022 Results, Volume I: The State of Learning and Equity in Education (OECD Publishing, 2023)
  3. Direction de l’Évaluation, de la Prospective et de la Performance (DEPP), Panel d’élèves de la cohorte 2021, ministère de l’Éducation nationale
  4. Education Endowment Foundation (EEF), Teaching and Learning Toolkit (updated 2024)