Does arts education raise academic achievement?

Verdict

No effect. This is the second clean no-effect verdict in the archive and it lands the same way as the first: teaching an art form teaches the art form, and the spillover to other subjects is the part that does not survive a randomised test. The parallel with motor competence and transfer is exact, down to the political economy — physical education and arts education are both defended in budget meetings on the strength of effects they do not have.

Be precise about what is being denied. Nobody here is claiming arts education is worthless, harmful, or a waste of a timetable slot. The trials say the opposite about harm: arts schools tested the same as control schools. What fails is one specific claim, the one that gets written into grant applications: that adding arts provision raises reading and mathematics attainment.

What the evidence shows

Source	Design	Grade	Key effect
EEF Learning About Culture	five cluster RCTs, 400 schools / ~8,500 pupils	A	Every primary attainment estimate null; pooled KS2 writing −0.04 [−0.14, 0.07], n = 5,255
Bowen & Kisida 2023 (Houston)	school-randomised, 42 schools	B	Reading −0.019, maths −0.001; writing +0.128 (n = 4,842); discipline p = 0.14 under randomisation inference
Ludwig 2017 (AIR/ESSA)	ESSA-tiered review + meta, 27 studies / 122 effects	C	g = 0.11 [0.07, 0.16]; maths 0.06 [−0.17, 0.30] null; 1 of 44 interventions reaches Tier I
Hetland & Winner 2001 (REAP)	ten meta-analyses, 188 reports / 275 effects	C	Arts-rich education → achievement: correlation yes, causation no; 7 of 10 links unsupported

The EEF programme is the strongest evidence in either subject and it is worth reading closely. Five separate school-randomised trials, four different art forms, two key stages, harmonised standardised outcome measures chosen before the trials began, independent evaluators, and 400 schools. Every primary attainment estimate is indistinguishable from zero, and so are creativity, ideation and social skills. The exploratory pooled Key Stage 2 estimate — the best-powered single number anyone has on arts education and academic attainment — is −0.04.

Houston is the one trial that could have shown a positive and mostly did not. It randomised schools rather than pupils, doubled the arts provision in treatment schools at trivial cost, and used administrative test data with essentially no attrition. Reading and mathematics, the two outcomes with more than 11,000 students behind them, are both null and both point-negative. The writing effect (+0.128 SD) is real and worth having, and it is also the outcome closest to the treatment, measured on the smallest sample, on the one Texas test with an open-response composition — the archive's measure-proximity rule applies to it directly.

The advocacy literature knew this. REAP looked at everything published between 1950 and 1999, found the correlation between arts-rich schooling and achievement, and refused to call it causal: "no evidence was found that studying the arts causes academic indicators to improve." Its authors' proposed explanation was institutional selection — schools that take the arts seriously tend to be schools that take everything seriously. Their warning is the best sentence in the literature: "Those who live by instrumental claims risk dying by such claims."

The field's own evidence standard tells the story. When AIR applied the ESSA tiers with WWC- certified reviewers, 34 of 44 catalogued arts-integration interventions had Tier IV evidence only — a rationale and nothing measured. One reached Tier I.

Hereditarian-lens assessment

Risk: low, unusually for this archive, because the two load-bearing sources randomise at the school level. Genotype cannot differ systematically between arms, and the effect is absent anyway.

The lens still does one job here, and it is the important one: it explains the correlation that everyone starts from. Arts participation is chosen — by families, by schools, by children — and the choosing is not random. The music side of this archive has the number: instrument engagement is 78% heritable with no detectable shared environment (Gustavson 2021), and it is genetically correlated with verbal IQ. Arts participation has no comparable twin study, but there is no reason to think it behaves differently. Every observational arts-and-achievement finding is estimating that selection.

Note what the randomised nulls therefore establish: it is not that the correlation is small, it is that the correlation is entirely attributable to who takes part.

Boundaries & what critics say

• This is not a case for cutting the arts. Every trial here also found no harm to attainment. The honest summary is "costs nothing academically, buys nothing academically." What the arts buy is art — see does teaching art teach art — and cultural exposure, where the evidence is better (museum and theatre trips).
• Writing is a genuine exception and may not be transfer at all. Houston's +0.128 SD on writing, and the fact that three of the five EEF programmes were writing-focused, point the same way: arts activity that involves composing, narrating or illustrating a text sits adjacent to writing rather than distant from it. That is near transfer, and near transfer is common. The EEF writing trials were still null, so even this reading is not secure.
• Drama is the strongest counter-case and is filed separately. REAP's drama meta-analysis found medium effects on verbal outcomes that generalised to non-enacted texts. See drama and verbal skills; the EEF's own drama trial was null.
• Dose could always be higher. Houston delivered about ten arts experiences per school year; the EEF programmes ran a school year of weekly sessions. Nobody has tested a genuinely intensive arts curriculum against a randomised control, and nobody is likely to.
• Non-academic outcomes are a different question and are weaker than advertised. Houston's headline social results — engagement, compassion, aspirations — are an elementary-school subgroup, are negative for middle schoolers, and lose significance under Lee bounds for differential survey response. "Emotional empathy" in the peer-reviewed version is a single item retained after its scale failed reliability.
• Teachers see effects that the tests do not. In the EEF trials, 85–97% of teaching staff perceived improved creativity and 88–98% improved engagement, against measured effects of essentially zero. Both facts can be true — but a school board deciding on evidence has to know which one is the measurement.

Practical guidance

• Defend the arts on the arts. It is the only defence the evidence supports, and it is the one that cannot be taken away by the next trial.
• Refuse to buy an "arts integration" package sold on achievement gains. The field average is g = 0.11 with maths null, one intervention in forty-four clears the top evidence tier, and the two randomised programmes covering 440+ schools found nothing.
• If writing is the goal, teach writing. The archive's writing evidence supports direct instruction at effect sizes several times anything here; the arts route to writing is the long way round.
• Do not accept teacher perception as an outcome measure. The EEF contrast — near-universal perceived creativity gains against null measured ones — is the single most useful line in the report for anyone evaluating a programme in their own school.
• Cost is the real argument for keeping it. Houston's provision cost $14.67 per pupil per year, 0.08% of per-pupil spending. A cheap thing worth having in itself does not need an achievement story.

Open questions

• Whether arts effects compound over multiple years is untested — every estimate here is end-of-year, and the Houston authors say explicitly that they do not know whether their effects compound, persist or fade.
• Whether an intensive, multi-year arts curriculum (as opposed to added arts experiences) would move achievement has never been randomised.
• Whether the writing result reflects genuine near transfer from composition-adjacent arts activity or is a chance finding on the smallest sample in the trial is unresolved.
• No genetically informative study of arts participation exists; the selection story is inferred from the music literature.