Grade retention

Why does making a struggling child repeat a year so often leave them further behind than before?

The logic of holding a child back is close to unarguable. She has not mastered the year's work; the next year builds on it; giving her the year again supplies the missing foundation before it is needed. Every alternative moves her on with a known gap.

And yet "retention leaves children further behind" is repeated as settled fact. Both cannot be straightforwardly true. So before asking whether retention works, ask a duller question: further behind than what? I suspect the whole difficulty lives in the answer.

Take that seriously. A retained child now sits among children a year younger. Compare her with them and she is older, has seen the material before, and will very likely be near the top — the retention looks like a success. Compare her with the children she started school with, now a grade ahead, and she is a full year of curriculum behind them, permanently. Same child, same year, opposite verdicts.

So "further behind" is not a fact about the child; it is a fact about the yardstick. And notice which yardstick each party reaches for. The school sees the same-grade comparison daily, and it looks encouraging. Researchers reach for the same-age comparison, because that asks the counterfactual question — what would have happened had she been promoted. Much of the disagreement between teachers and researchers here is not about evidence at all. It is two different denominators.

Is the same-age comparison sound? Not on its own, and this is where the older literature went wrong. Retained children were selected precisely because they were struggling, so comparing them with promoted classmates compares weaker children to stronger ones and finds — inevitably — that the retained do worse. That cannot separate the effect of retention from the reason for it, and for decades it was the basis of the confident claim that retention is harmful.

What breaks the deadlock is a policy that assigns retention by a rule rather than a judgement. Where a jurisdiction retains everyone below a fixed test score, children just below the line and just above it are near-identical in everything except the treatment they received. Comparing across that line is a genuine experiment that nobody had to run. Studies of that kind — the Florida third-grade reading promotion rule is the most examined, and Chicago's promotion policy was analysed similarly by Jacob and Lefgren — produce a more awkward picture than either camp expected: measurable gains in the years immediately after retention, which shrink as time passes, with the results depending heavily on the age at which it happens.

Why would a gain appear and then fade? Two things are bundled into the retained year that have nothing to do with remedy. She is a year older when she sits the same assessment, and maturity alone moves scores. And she is re-sitting material she has already met, so the test measures partly familiarity. Neither is a durable repair. If the year is otherwise the same year — same curriculum, same pace, no diagnosis of what actually went wrong — there is no reason to expect the second pass to succeed where the first failed. Under those conditions the fade is exactly what the mechanism predicts.

Meanwhile the costs do not fade, because they are structural rather than academic. Being a year older than the cohort is permanent, and it accumulates against a fixed date: the age at which a young person may lawfully leave school. A retained pupil arrives at that date one year further from a qualification, with one more year of schooling required to obtain it. Of everything reported about retention, raised likelihood of leaving without a qualification is the most consistent finding — and it is the one that needs no psychology to explain.

Three refinements.

First, the claim is not that a repeated year cannot help, but that repetition as such supplies no new mechanism. What could supply one is what is done differently during the year — diagnosis of the specific failure, changed instruction, intensive reading support. Where retention policies have shown the clearest gains, remediation was attached: the promotion rule triggered extra teaching rather than merely a delay. That reading is uncomfortable for both sides, since the effective ingredient may be the remediation, which does not require retention to deliver.

Second, this is a path-dependence story more than a learning story: the lasting consequences follow from her position relative to two clocks — her cohort and her legal school career — not from what she knows.

Third, on the evidence: the widely-taught claim that retention harms children rests substantially on the confounded comparisons above, and became consensus before the design problem was taken seriously. The cutoff-based studies are far better identified, and they do not say retention is harmless — they say the short-run effect is often positive, decays, and that long-run completion is the worry. I decline to quote magnitudes: they vary with grade, jurisdiction, accompanying remediation, and how long afterwards the outcome was measured.

What would refute this account? Its load-bearing claim is that the early gain is maturity plus familiarity, not repair — so it should fade wherever the repeated year is instructionally unchanged, and persist where it is not. A well-identified study finding durable gains from a bare repeat with no changed teaching would break it, and repetition would have to be doing something I have not identified.

• What a promotion rule that triggers remediation without repetition would have to look like.
• Why test-score cutoffs make such good natural experiments, and where that design misleads.

Every term the collection defines is gathered in the glossary.