Evidence Desk · Beland & Murphy 2016
School phone bans lifted GCSE scores in four English cities. Read that sentence to the end before you legislate.
A small average gain, a bigger one for struggling pupils, nothing for the top of the class, and a Swedish sequel that didn't get the memo.
vizmaya desk · 3 November 2026
Exhibit A
91 schools, eleven years of GCSEs, one questionnaire
Economists Louis-Philippe Beland and Richard Murphy asked secondary schools in Birmingham, Leicester, London and Manchester when they banned phones, then lined that up against national exam records.
Only 21% of the schools they asked replied, a hit rate familiar to any teacher who has ever said "phones away, please".
- Study
- Beland & Murphy, "Ill communication: Technology, distraction & student performance"
- Venue
- Labour Economics 41, 61–76 (2016), peer reviewed
- Design
- Natural experiment: difference-in-differences on the year each school banned
- Sample
- England (United Kingdom): 91 schools, 130,482 pupil observations, GCSEs 2001 to 2011
- Outcome
- Total GCSE score, standardised nationally
- Paper
- doi.org/10.1016/j.labeco.2016.04.004
The finding in a number
The average effect: real, small, and wearing a 10% significance badge
Average rise in GCSE scores after a school banned phones: 6.4% of a standard deviation.
Its 95% interval runs from about −0.01 to 0.14 SD, so it clears the 10% significance bar but not the 5% one.
Benchmark: fake bans dated one or two years early "raise" scores by 0.02 to 0.03 SD, not significantly. The real thing is about twice that. Exciting, in the way a modest pay rise is exciting.
Source: Beland & Murphy 2016, Table 4 col. 5 (SE 0.0373, clustered by school) and Table 7 col. 5. n = 130,482.
Who gained
The gain went to the bottom of the class. The top fifth got a participation certificate.
Pupils grouped by their test scores at age 11. Source: Beland & Murphy 2016, Table 6 col. 4; middle fifth 0.065 and fourth fifth 0.023, neither significant.
In everyday terms
In school-calendar money: five extra days, for the price of a sign on the door
For the lowest-attaining fifth, the authors call the gain "equivalent to an additional hour a week in school", or adding five days to the school year.
That conversion is theirs, borrowed from earlier studies, and it applies to that group only. Somewhere a finance ministry just sat up straight.
Chart 1
The bottom two fifths gained. The top fifth didn't notice. The fake bans did nothing, as fake bans should.
Effect of a phone ban on GCSE scores, in standard deviations, with 95% intervals. Only the two lowest-attaining fifths clear zero.
Source: Beland & Murphy 2016, Table 4 col. 5, Table 6 col. 4, Table 7 col. 5. n = 130,482 pupil observations. Interval = estimate ± 1.96 × clustered SE. No comparable attainment estimate exists from the other studies on this page, so none is plotted.
What the design can show
Why "caused" is allowed here (terms: England, 2001 to 2011)
Schools banned phones in different years, so early banners are compared with schools that hadn't banned yet. That works only if their results were on the same path beforehand.
The checks pass: no rising trend before bans, nothing from fake ban dates, no change in the pupils arriving. Still possible: some other school improvement that showed up with the ban. Causal for those schools, that decade. Not for your nephew's school in 2027.
Schematic, not data. Checks: Beland & Murphy 2016, Fig. 1 (event study), Table 7 (placebo dates), Table A.2 (intake balance).
Enforcement
Stricter bans did not beat lax ones. Somewhere, a confiscation drawer weeps.
Every interval here crosses zero, and the lax bans post the biggest point estimate. For the lowest fifth the gap is wider still: 0.305 SD for less strict bans against 0.109 for strict ones.
Schools chose their own strictness, so this is not a randomised comparison and not a licence to go soft. The authors say the reason is unclear.
Source: Beland & Murphy 2016, Table A.1 col. 5 and footnote 19. SD, 95% intervals from clustered SEs.
The lesson is who might gain, and under what conditions. It is not a promise that a ban lifts grades everywhere.
The other side, part one
Sweden filmed the sequel. Plot twist: nothing happens.
Sweden · Kessel and colleagues, 2020 · peer reviewed
A partial replication of the English design in Swedish schools reports no improvement in attainment.
The authors point to bans being defined differently and schools being different. Translation: two countries can use the word "ban" for two different policies.
It does not show the English result was wrong, or that every school rule fails. It shows the result doesn't travel automatically. Unlike the policy speeches.
Source: Kessel et al. 2020, pp. 1–2. No effect size is quoted here because the desk did not extract a comparable estimate.
The other side, part two
England, 2025: strict schools cut phone time. Eight other measures shrug.
SMART Schools compared 20 restrictive and 10 permissive English schools (1,227 pupils). Restrictive schools logged 0.67 hours less phone use in school, but no detected difference in total daily use. The phones, it seems, simply went home.
A snapshot, not an experiment, so it can't prove rules do nothing. But the wellbeing interval excludes the 3-point gap the study was planned to detect.
Source: Goodyear et al. 2025, Table 3. Adjusted differences; each row on its own scale. Cross-sectional: associations, not effects. n = 1,227 pupils, 30 schools.
Chart 2: the lens
Six policies on the table. These studies back one of them, with an asterisk.
What Beland & Murphy, SMART Schools and the Swedish replication can say about each. Everything else on the list is being argued from somewhere else.
Desk's reading of the three studies above, not a ranking of all evidence on each policy.
Lens: what policy should take
What a minister should take from this (besides the credit)
Treat a school phone rule as a learning intervention, not a moral statement. Pick the outcome and the smallest gain worth the hassle before it starts, record a baseline, find a fair comparison, and check who gains. The best estimate here lands on lower-attaining pupils, so judge it there.
For lawmakers, the paper picks no platform feature to ban and justifies no age limit. For parents, it says nothing about screens at home. For anyone selling wellbeing, fewer phones at school is a means, not a health outcome.
Interpretation by the desk, not a tested result. Laminating "phones away" is not an evaluation strategy.
Where it stands
236 citations and one stubborn Swedish footnote
The collection that selected this paper rates its theme's evidence "mostly correlational; few RCTs". This is one of the sturdier designs in that pile, and it is still one country, one decade, one exam.
What to watch: replications that report results by prior attainment and record how strictly rules were enforced. Fame is not proof; its field-weighted impact is 100 times the field average.
Source: OpenAlex W2128474361 (CC0), read 3 November 2026. 2026 is partial. One 2013 citation predates the journal article and is not shown. Citations measure attention, not validity.
Go further
Read the originals. They have more footnotes than we do.
- Beland & Murphy 2016, Labour EconomicsThe study. Author copy (PDF)
- Goodyear et al. 2025, SMART SchoolsCross-sectional, England. Open copy (PDF)
- Kessel et al. 2020, Swedish replicationPartial replication, no attainment gain
- OpenAlex recordCitation counts, CC0. No public preregistration or replication data found.
- More vizmaya data stories and evidence explainersvizmaya.fyi/stories
Study: Beland, L.-P., & Murphy, R. (2016). Ill communication: Technology, distraction & student performance. Labour Economics, 41, 61–76. https://doi.org/10.1016/j.labeco.2016.04.004. Selected from: Social Media and Its Effects on People and Society: An Annotated Research Collection (2026 edition). Citation counts: OpenAlex (CC0).
Sources and claims log
The receipts
Every number comes from the papers' own tables: Beland & Murphy 2016 (Tables 4, 6, 7, A.1, A.2; Fig. 1; fn. 19), Goodyear et al. 2025 (Summary, Table 3, Statistical analysis), Kessel et al. 2020 (pp. 1–2), OpenAlex (read 3 Nov 2026). The collection's summary was used as a lead only; its "6.4%" is 0.064 of a standard deviation, not 6.4% higher scores, and it omits that the average is significant only at the 10% level.
Not found: a minimum GCSE gain worth a ban's costs; causal estimates for age limits, design rules or home rules; a preregistration or public replication data for the study.
Photos: davidhc, bartosz.maciejewski and Citizen 4474 on Flickr (CC BY 2.0, CC BY-SA 2.0, CC BY 2.0), via Openverse; full credits on each slide.
No phones were confiscated in the making of this deck.