Grade retention: what the research shows

How many students repeat a year, and how the practice is evolving

In France the annual rates carry the story. At the start of the 2025 school year, 3 percent of first-year primary pupils in public schools repeated the year, up from 2.7 percent a year earlier; at the other primary levels, rates held steady after the 2024 rise (DEPP, briefing note 25.58, October 2025). The swing is recent: rates were at a historic low in 2017, rose in 2018 after the February 2018 decree, stayed flat until 2023, then climbed at every primary level in 2024, from 2.7 percent in the first year (up 0.8 points) to 0.7 percent in the last (up 0.2 points) (DEPP, briefing note 24.41, October 2024).

The long trend is clearer than these jitters. The share of 15-year-olds who had repeated at least one year fell from roughly 40 percent in 2003 to 22 percent in 2015 and 11 percent in 2022, one of the steepest declines the OECD has measured (OECD, PISA 2022, Volume II). In 2009, France had the third-highest rate in the OECD; in 2022 it sits just above the OECD average of 9 percent (OECD, PISA 2022, Volume II).

The profile of repeaters is stable. On average across OECD countries, a disadvantaged student is more than three times as likely to have repeated a grade as an advantaged one; boys repeat more often than girls, and students with an immigrant background more often than others (OECD, PISA 2022, Volume II). Even after test performance is accounted for, repetition keeps tracking social background (OECD, PISA 2022, Volume II).

What the research says

The benchmark meta-analysis frames the debate. It pools 22 studies published between 1990 and 2007 and 207 effect sizes: in studies with weak controls, retention is tied to clearly lower achievement (minus 0.30 standard deviations); in the best-designed studies, which compare students of equivalent prior attainment, the average effect falls to plus 0.04 standard deviations, indistinguishable from zero (Allen et al., 2009). The gap widens over time: the estimated effect erodes by about 0.11 standard deviations per additional year of follow-up in grade-based comparisons (Allen et al., 2009).

Quasi-experimental designs sharpen the picture. In Chicago, a programme paired summer school with retention for students below a promotion cut-off: judged around that cut-off, it lifted third graders' achievement by roughly 33 to 41 percent of an average year of learning in year one, before weakening in year two (Jacob and Lefgren, 2004). For older students, the sixth-grade equivalent, the effect is nil after one year; after two, the authors find a reading shortfall of about a quarter of a year's learning (Jacob and Lefgren, 2004).

Florida's long-run evidence confirms the pattern. Students retained at the end of third grade under the state reading test gained at first, but after six years the advantage vanishes as soon as they are compared with peers of the same age; it survives only against same-grade peers; the probability of earning a high-school diploma is unchanged, while final grades rise and remedial courses drop (Schwerdt, West and Winters, 2017). The OECD draws a cautious reading: short-term gains that are real but fragile, uncertain long-term effects, and a higher dropout risk among repeaters (OECD, PISA 2022, Volume II).

What international comparisons show

Grade repetition remains deeply uneven. In 2022, 9 percent of 15-year-olds reported having repeated a grade at least once on average across the OECD, against 11 percent in France; 36 systems stay at or below 5 percent, and 14 exceed 20 percent, including Colombia (about 40 percent) and Morocco (about 46 percent) (OECD, PISA 2022, Volume II). France, third-highest in the OECD in 2009, now retains less than Austria, Belgium, Germany, Spain, Portugal or Switzerland (OECD, PISA 2022, Volume II).

Rules explain part of the spread. Automatic promotion applies in about 28 percent of systems at primary level and 24 percent at lower secondary, including Japan, Korea, Iceland, Israel, Mexico, Norway and the United Kingdom, systems where repetition is excluded in principle (OECD, PISA 2022, Volume II). Where repetition is restricted, the share of repeaters in primary education reaches 6.7 percent, against 3.6 percent in automatic-promotion systems and 7.6 percent where no limits apply (OECD, PISA 2022, Volume II).

These numbers come with limits the OECD documents itself: they rest on student self-reports, and in Japan and Norway the question was not asked, with the rate set to zero by agreement (OECD, PISA 2022, Volume II). One French finding also deserves attention: falling repetition was accompanied neither by a decline in mathematics results before the pandemic nor by gains in social fairness, a shift the OECD ties mainly to the rule changes of 2013 and 2018 (OECD, PISA 2022, Volume II).

Documented alternatives: targeted support, summer programmes, tutoring

Programmes that raise learning without an extra year are documented. In India, schemes that group children by actual learning level rather than age produced gains of 0.15 to 0.70 standard deviations across delivery models, including programmes run inside public schools (Banerjee et al., 2016). In France, the OECD lists the levers already in place in primary and lower secondary schools: remise à niveau support weeks, complementary teaching activities, and the Devoirs faits homework programme, to be used before retention is considered (OECD, PISA 2022, Volume II).

Tutoring shows the steadiest results. The meta-analysis of experimental studies by Nickow, Oreopoulos and Quan (2020) finds an average effect of 0.37 standard deviations on learning, stronger when tutors are teachers or trained staff, when sessions run during the school day rather than after hours, and in the early grades (Nickow, Oreopoulos and Quan, 2020).

Summer stays useful time. In Chicago, mandatory summer school raised reading and mathematics achievement, with gains that did not fade immediately (Jacob and Lefgren, 2004). In France, remise à niveau courses run over three holiday weeks and target French and mathematics in small groups (OECD, PISA 2022, Volume II). The OECD's recommendation points the same way: use such support before a student is held back, not after (OECD, PISA 2022, Volume II).

Five anchors for a school leader

In France, retention remains a school-level decision, framed by the 2013 law and the 2018 decree: it must stay exceptional and come with an individual success programme, decided by the teachers' council in primary schools and by the principal in lower secondary (OECD, PISA 2022, Volume II, France box). Five anchors, each tied to the evidence cited above.

• Compare before deciding: at equivalent prior attainment, the best-designed studies find no durable benefit from retention (Allen et al., 2009).
• Document the exception: retention must stay rare and carry an individual success programme, decided by the teachers' council or the principal (OECD, PISA 2022, Volume II).
• Fund tutoring and targeted support: 0.37 standard deviations on average for tutoring (Nickow et al., 2020), 0.15 to 0.70 for learning-level regrouping (Banerjee et al., 2016).
• Use the holidays: summer remediation produced measured gains in Chicago (Jacob and Lefgren, 2004); France already runs three-week remise à niveau courses (OECD, PISA 2022, Volume II).
• Track effects and equity: short-term gains fade within a few years and the diploma odds in the Florida scheme did not change (Schwerdt, West and Winters, 2017), while a disadvantaged student is more than three times as likely to repeat a year (OECD, PISA 2022, Volume II).