Homework is one of the few school policies that teachers, parents and pupils all hold a firm view on, and the research supports a more qualified position than any of them usually offers. The evidence is real but patchy, mostly correlational, and heavily weighted towards the United States. It points to a modest benefit for older pupils, little for younger ones, and a strong dependence on how the homework is designed.

This guide sets out the strongest case on each side, notes where the evidence is thin, and ends with what a policy built on it could look like.

The headline numbers

The Education Endowment Foundation's Teaching and Learning Toolkit estimates that homework adds about five months of additional progress overall: five months in secondary schools and three in primary. The EEF rates it as moderate impact for very low cost, based on very limited evidence. Evidence security is low, built on 43 studies, and many of those are not randomised controlled trials or independently evaluated. That caveat belongs next to the number every time it is quoted.

The case for setting homework

The most cited review is Cooper, Robinson and Patall (2006), which synthesised US studies from 1987 to 2003. The authors acknowledge that every study had design flaws, but found the evidence generally consistent and positive. Of 69 correlations between homework time and achievement, 50 were positive and 19 negative. For secondary-age pupils the average correlation was about .20 to .25, depending on the model used.

Earlier work by Cooper (1989), as summarised in the 2006 review, put a rough figure on what that means. The average high school student who did homework outperformed 69 per cent of students who did none. For junior high the gap was about half as large, and for elementary pupils there was none.

Fan and colleagues (2017) reached a similar conclusion in a meta-analysis of homework in maths and science from 1986 to 2015: an overall small positive relationship with achievement, stronger for elementary and high school than for middle school, and strongest in US studies.

The EEF adds two practical findings for the proponent's side. The highest-impact studies set homework about twice a week in a subject, and homework works best when it is integral to learning and followed by high-quality feedback.

The case for caution

The sceptic's argument starts with the design of the studies. Most of the evidence is correlational, and Cooper and colleagues are explicit that correlational designs cannot establish causality. Pupils who do more homework may differ in motivation, prior attainment or home support, and each of those could explain the gap on its own.

The studies that did manipulate homework are few. The 2006 review found six, all unpublished, with effect sizes from .39 to .97 and a weighted average of .60. Only one used random assignment within classes, and the average was not statistically significant under one of the review's assumptions about how pupils in the same class resemble each other. The reviewers also found no association when homework time was reported by parents (r = -.03). Their own warning about the field is worth quoting: "a few studies can always be found to buttress whatever position is desired".

International data complicates the picture further. Dettmers, Trautwein and Lüdtke (2009) analysed PISA 2003 data from 40 countries. Homework set at class or school level was positively related to maths achievement. Time spent on homework by individual students showed no consistent relationship, and was negative in many countries. The paper's own title states the conclusion: the relationship between homework time and achievement is not universal. One plausible reading is that struggling pupils take longer, which the data cannot confirm.

On quantity, the EEF reports that quality matters more than amount and that impact diminishes as time increases. Cooper's 1989 review found that gains for junior high pupils levelled off at one to two hours a night, while the benefit for high school pupils kept rising beyond two hours. That is often summarised as diminishing returns for all secondary pupils, which is a simpler claim than the original supports.

The equity question

The EEF notes that pupils eligible for free school meals typically gain additional benefit from homework, but are also less likely to have a quiet place to work, a device or parental support. That combination makes homework both potentially more valuable and harder to do well for the pupils who need it most. Homework clubs are one practical response, and the EEF is clear that homework should not be used as a punishment.

Generative AI adds a new variable. A 2026 National Literacy Trust survey reported that about one in four young people copy AI output for homework, rising to 31.6 per cent among pupils eligible for free school meals. Where completion is unsupervised, a task that can be copied tests the tool rather than the pupil.

What a policy built on this evidence could look like

Set regular, short, specific tasks. The highest-impact EEF studies set homework twice a week in a subject. That pattern is more consistent with brief, frequent tasks than with long ones.

Tie it to what was taught. The EEF finding that homework should be integral to learning points to tasks that practise or retrieve lesson content, and away from stand-alone projects that depend on home resources.

Close the loop with feedback. Homework that nobody looks at loses the feature the EEF identifies as important. Departments can decide what form of feedback is realistic, from quizzes marked in class to whole-class comments.

Provide a route for pupils without space or devices. Homework clubs and protected study time address the access gap the EEF describes.

Evaluate locally. With 43 studies and low evidence security, the honest position is that a school should trial its approach and check whether it improves outcomes for its own pupils.

Where the debate stays open

Advocates of homework can point to consistent positive associations in secondary-age pupils, to the EEF's five-month estimate, and to the extra practice time it offers. Their strongest argument is that independent practice is hard to fit into lesson time alone.

Sceptics can point to the reliance on correlational data, the near-absence of benefit in younger pupils, the lack of any relationship between individual homework time and achievement in many countries, and the risk of widening gaps for pupils without support at home. Their strongest argument is that the evidence for a causal effect is weaker than the confident claims made on either side.

Both sides are working from the same limited evidence. The more useful question for a school may be what its homework is for, and what evidence it will collect to find out whether it is working.

Sources and further reading

[1] Education Endowment Foundation. Teaching and Learning Toolkit: Homework. Review last updated August 2021.

[2] Cooper, H., Robinson, J.C. and Patall, E.A. (2006). Does homework improve academic achievement? A synthesis of research, 1987-2003. Review of Educational Research, 76(1), 1-62. Also the source for the 1989 Cooper findings quoted here.

[3] Fan, H., Xu, J., Cai, Z., He, J. and Fan, X. (2017). Homework and students' achievement in math and science: a 30-year meta-analysis, 1986-2015. Educational Research Review, 20, 35-54.

[4] Dettmers, S., Trautwein, U. and Lüdtke, O. (2009). The relationship between homework time and achievement is not universal: evidence from multilevel analyses in 40 countries. School Effectiveness and School Improvement, 20(4), 375-405.

[5] National Literacy Trust (2026). Survey findings on young people's use of generative AI, including copying AI output for homework.

Internal note: Add FAQ schema markup before publishing. Featured image alt text: 'Stack of exercise books and a planner on a teacher's desk.' Internal links to add once live: 'What the Research Says About Effective Revision Strategies,' 'Low-Effort, High-Impact Formative Assessment for Secondary Teachers,' 'How Technology Is Closing the Learning Gap (and Where It Falls Short).' Check the National Literacy Trust citation title before publishing.