Do School Phone Bans Improve Learning? The Evidence After the First Randomized Trials

The question

Do enforceable restrictions on student smartphones during the school day improve academic achievement or adolescent well-being in secondary schools? If they do, which channel carries the effect: fewer interruptions during instruction, changes in peer interaction, less total digital exposure, or a change in school discipline?

The question needs narrowing: a school-day rule changes access for part of the day, not home use or platform design. And "ban" covers different treatments: a teacher asking for silence, collection at the classroom door, a pouch sealed until dismissal, or phones permitted for teacher-directed instruction. These differ in enforcement, in how much use they remove, and in what students do instead.

The stakes are concrete. By 2026, many U.S. states and countries had laws limiting student phone access, and Brazil's federal Law 15.100/2025 required schools to restrict personal devices. Schools and governments are making durable organizational choices on the strength of a literature that only recently acquired randomized evidence.

Best current answer

The first link in the causal chain is solid. Stringent enforcement reduces phone access during school. The 2026 national U.S. study of lockable pouches found that in-class personal phone use fell from 61% to 13%, confirmed by GPS activity and teacher reports. Rio de Janeiro's policy also produced a substantial fall in in-school use. My confidence that pouches, lockers, or collection systems reduce in-school phone access is about 90%.

Academic effects are real but modest and context-dependent. The strongest studies span a precise average null in U.S. pouch schools, a 0.06 SD relative gain in Rio, second-year gains in one Florida district, about 0.07 SD in England, and zero in Sweden. Norway finds no average effect but gains for girls; Chile finds no test-score gain. A 2026 randomized trial in India finds that collection raises higher-education grades, with the largest gains among lower performers. Two Chinese trials find unstructured use worse than a ban, but teacher-guided use better than either.

My synthesis assigns about 70% probability that the average effect of a well-enforced secondary-school restriction falls between -0.02 and +0.07 SD over its first three years. I assign about 65% probability to an effect above zero. Schools with high baseline phone use or weak classroom control have more room to benefit.

Population mental-health benefits have weaker support. Some studies report less distress, fewer mental-health visits among girls, or less bullying; others find no difference in screen time, well-being, bullying, or mental-health outcomes. A 2026 English economic evaluation finds minimal differences in quality-adjusted life years or well-being, alongside a modest staff-time saving. I assign about 80% probability that the average well-being effect is negligible over one to three years, and about 85% confidence that current evidence cannot justify school bans as a population mental-health intervention.

Schools also pay an implementation cost. U.S. pouch schools and the Florida district recorded more discipline in year one; Florida's increase fell most on Black students. The effects faded, but a temporary rise in exclusion still counts as harm.

Phone-free instructional time is a reasonable management default when a school can enforce it with little punishment. Claims about repairing adolescent mental health, ending bullying, or reversing cognitive decline exceed the evidence. The newest trials qualify this: where teachers can integrate devices, a blanket ban may not be the best rule.

What the studies actually test

The literature is often read as one debate about whether phones are good or bad. It is better read as comparisons between different treatments, populations, outcomes, and follow-up periods.

• Classroom-only restrictions remove recreational use during lessons but may leave breaks unchanged. Chile's teacher-level variation is the clearest example.
• Bell-to-bell bans prohibit use throughout the day, often with storage in bags, lockers, or pouches. The U.S. pouch study and Rio's second phase are the strongest examples.
• Guided-use policies permit phones but require teachers to direct when and how they are used. The Chinese trials test this.
• Enforcement intensity varies within each category. A written rule that teachers rarely apply is a different treatment from a sealed pouch.

Outcomes differ too. Standardized scores are not teacher-awarded grades, which can respond to calmer classrooms without matching a learning gain. Attendance, suspensions, attention, bullying, and well-being scales capture different things. A policy can improve one and worsen another. The strongest studies measure a first stage — actual phone use — because a null is uninformative if the policy never changed behavior.

Academic evidence

The large natural experiments

Allcott and colleagues' 2026 U.S. study combined pouch-provider records with GPS pings, teacher and student surveys, standardized scores, and administrative records. Their staggered difference-in-differences compared schools before and after adoption with non-adopters. Pouches cut measured phone activity, yet pooled test-score effects stayed near zero over three years, precise enough to reject an average gain above about 0.01 student-level standard deviations. High-school mathematics rose slightly; middle-school scores fell slightly. Attendance, attention, and online bullying changed little. Discipline rose and well-being fell in year one, then both faded; well-being was slightly positive by year two. A strong first stage and precise null make weak enforcement an unlikely explanation. Adoption timing, vendor capture of one policy type, and GPS measurement remain limits.

Lichand and colleagues compared Rio de Janeiro's municipal middle schools with weak prior rules to matched schools that already enforced strict ones, before and after the 2023 classroom policy and its 2024 full-day expansion. In-school use fell more in the first group. Mathematics and Portuguese scores rose by 0.06 SD relative to the already-strict schools, with similar pre-trends; the main comparison covered 154 schools and about 97,000 students. This is a relative effect, not a clean ban-versus-none contrast: both groups faced the city policy. Their 0.1–0.2 SD estimate against no restriction is an extrapolation, not the headline.

Figlio and Özek's Florida study used pre-policy differences in building-level phone activity in one large district. Scores improved in year two and unexcused absences fell; suspensions rose in year one, then subsided. The pattern fits an adjustment path, but one district and an exposure-intensity design limit inference.

Older replications disagree. Beland and Murphy linked policy timing in four English cities to national exams and estimated about 0.07 SD, concentrated among low achievers. Kessel, Hardardottir, and Tyrefors found no gain in Swedish secondary schools, precise enough to reject small positive effects; Swedish teachers may have restricted phones before formal adoption. Abrahamsson found no average Norwegian effect, but girls' GPAs and academic-track placement improved, especially among disadvantaged girls, with fewer psychological health-care visits and less bullying. Adoption timing looked unrelated to observed trends, but many outcomes and subgroups raise the risk of unstable estimates. Shi and Villarroel used Chilean teacher and cohort variation: recreational use fell for eighth graders but not tenth, scores did not improve, and eighth graders reported lower academic capability. This identifies uneven decentralized enforcement, not a storage regime.

The randomized evidence

Sungu, Choudhury, and Bjerre-Nielsen randomized cohorts covering 16,683 students at 10 higher-education institutions in Odisha, India to mandatory phone collection before lectures or usual practice. Collection raised course grades, with the largest gains among lower-ranked students; attendance showed no consistent effect. Support for the ban rose, fear of missing out rose mildly, and total phone-use time, phones brought to school, and online harassment did not change. Spot checks found less use and disruption. This is a causal estimate for Indian higher education; it makes attention more credible for adolescents but cannot size a K–12 effect.

Deng, Cheng, Ferreira, and Pavlou ran two large randomized trials in Chinese vocational classrooms with four arms: ban; unguided free use; teacher-guided instructional use; and ban plus paper-based aids. Unguided use reduced performance relative to the ban. Guided use significantly improved learning relative to the ban. Paper aids added nothing over the ban. Video showed similar time on task, but teacher-directed phone use produced large marginal learning gains that outweighed distraction. Guided use narrowed gender and performance gaps while risking a digital divide by major and region.

This reframes the debate: a ban and guided use are different instructional designs, not two intensities of one treatment. A ban beats unstructured use but may lose to purposeful use when teachers can direct it. The result comes from vocational classrooms in one country and needs replication, but it is the strongest evidence that what replaces the phone matters.

A preregistered 2024 scoping review found 22 studies through May 2023, no randomized trial, mixed definitions, and weak comparability. Coverage improved in 2026, but most K–12 studies still rely on natural experiments.

Well-being, bullying, and total exposure

The mental-health hypothesis assumes that school restrictions reduce total digital dose. Goodyear and colleagues' SMART Schools study challenges that premise. Across 1,227 students in 30 English schools, restrictive policies correlated with 0.67 fewer hours of phone use and 0.54 fewer hours of social-media use during school, but weekday and weekend totals did not differ. Well-being scores also did not differ. The cross-sectional design cannot establish effects, but the time-use pattern is direct evidence for substitution.

A 2026 economic evaluation adds cost. Perry and colleagues found minimal differences in quality-adjusted life years (0.009, 95% CI -0.014 to 0.032) and mental well-being adjusted life years (-0.004, 95% CI -0.044 to 0.036). The point estimate for per-pupil cost was £94 lower in restrictive schools, but the confidence interval crossed zero: restrictions may save staff time without measurable health gains. A 2026 U.S. state-policy study by Saffer also found no clear screen-time or well-being improvement, though follow-up was short.

The U.S. pouch study found a more complex path: well-being fell by about 0.20 SD in the adoption year and rose by about 0.16 SD by year two, while discipline rose and faded. A short staggered panel cannot separate adaptation from cohort differences, so the estimates fit an implementation shock without proving a better equilibrium.

Norway offers positive counterevidence: Abrahamsson found fewer mental-health consultations among girls and less bullying among both genders. The consultation result came from the intensive margin — girls who used care made fewer visits — not from fewer diagnoses. An emulated South Australian trial found a small positive result. Baggio and colleagues compared 1,062 students in schools that had or had not yet implemented the state policy, with weighting for measured confounders. The ban correlated with about a 0.9-point reduction on the K10 distress scale (95% CI 0.03 to 1.85) and less negative affect. The lower bound sits near zero and timing was nonrandom.

A Dutch comparison points the other way. Vanluydt and colleagues surveyed 1,398 students in 24 schools with partial or full bans. Full bans showed no well-being or bullying advantage, correlated with lower student-teacher connectedness, and were associated with less school belonging among girls. Selection and school culture remain competing explanations.

School restrictions can change life inside school without changing total screen exposure. Benefits tied to classroom interruption remain plausible; benefits that require a large reduction in daily social-media dose face a substitution problem.

Mechanisms and competing explanations

Attention and instructional time. Phones create interruptions, and a shared rule can reduce teachers' enforcement burden. The Indian trial shows that collection can turn this into grade gains; England, Rio, and Florida fit the same account. But removing a distraction does not specify the counterfactual. Students can shift to laptops, peers, or daydreaming, and teachers may spend saved time on administration. Allcott found less phone activity without better attention or average scores. The Chinese trials make the point explicit: the same device can distract or teach, depending on who directs it.

Peer coordination. An individual student pays a social cost for disconnecting while peers remain online. A common rule removes part of that cost and may ease face-to-face interaction. Rio's school-wide phase produced clearer gains than its classroom phase, and strict Norwegian policies produced larger estimates than lenient ones. Neither study assigned strictness at random, though; school leadership can cause both enforcement and better outcomes.

Exposure substitution. Mental-health effects linked to sleep loss, late-night use, or platform content require less exposure outside school. SMART Schools found no difference in total weekday or weekend use. A policy may simply move use into the afternoon, preserving dose-related harms while classes improve.

Enforcement and adaptation. Confiscation can provoke conflict, and schools can respond with suspension, which removes instruction and can widen racial disparities. The U.S. pouch and Florida results show this cost during adoption; the later fade could be adaptation, weaker enforcement, or both.

Instructional and access costs. Teachers sometimes use phones for research, authentication, photography, or accessibility. Chile's decentralized setting may capture a tradeoff that pouch studies suppress: a restriction removes an instructional tool along with recreation. Medical and disability exceptions reduce harm but complicate enforcement, and no study measures this cost well enough to estimate a net effect.

Five threats complicate the quasi-experiments: adopting schools may change discipline, curriculum, or staffing concurrently; researchers often code the written rule rather than the enforced one; short panels mix adoption shocks with stable effects; teacher grades can improve without matching learning gains; and subgroup searches can produce unstable winners.

Strongest counterevidence and the live dispute

The strongest counterevidence comes from the 2026 U.S. pouch study. It captured a policy that changed phone behavior, covered many schools, preregistered its analysis, and estimated an average near zero, precise enough to reject a modest gain. Any theory predicting a general academic benefit must explain it.

The best positive challenge comes from Rio: a 0.06 SD standardized-score gain with similar pre-trends that survived alternative matching. England, Florida, and the Indian trial point in the same direction. A blanket null also fails to fit the record.

The Chinese trials cut across both camps. They suggest that the relevant comparison is not "ban versus no ban" but "ban versus structured use versus unstructured use." A school that bans phones and replaces them with nothing may do better than a school that permits unstructured use, and worse than a school that teaches with devices purposefully. That is a design claim, not a verdict on phones.

Context remains the leading explanation for the disagreement, but "context" needs measurable content. Baseline phone use, prior teacher enforcement, instructional quality, student age, and use of recovered time could moderate effects; researchers have not tested a common set of moderators across sites. The grade split in the U.S. study also clashes with claims that younger students or weaker self-control should gain most.

Well-being carries more uncertainty. Norway reports gender-specific benefits, while U.S., English, Dutch, and other Australian evidence ranges from null to small harm or benefit. Outcome definitions differ: health-care visits, short mood scales, broad well-being, belonging, and bullying are not interchangeable. Researchers also test many time points and subgroups. Current evidence cannot identify a stable mental-health mechanism.

Evidence quality and coverage limits

The natural experiments are credible but not interchangeable. Staggered adoption can be biased when timing responds to local trends, and the U.S. pouch authors note that adopters differ systematically from non-adopters. Rio's estimate is relative to schools already under a city policy, not a clean ban-versus-none effect. Florida relies on within-district exposure intensity, which may proxy for unmeasured family and neighborhood differences. Event studies can be confounded by concurrent reforms.

The randomized evidence is stronger on identification but narrower on population. The Indian trial is higher education in one state; the Chinese trials are vocational classrooms in one country. Neither is a multi-year K–12 trial in a high-income country, and neither measures long-run well-being. The 2024 scoping review found no K–12 randomized trial; coverage improved in 2026 but remains thin.

Measurement is a further limit. Test scores are comparable only within their own scale and population; teacher grades are not. Phone use is measured by GPS, teacher reports, self-reports, and spot checks, each with different error. Well-being scales are not interchangeable, and a null on a short scale does not rule out effects on clinical outcomes or sleep.

This is a documented synthesis, not a formal systematic review. I searched primary studies, reviews, and publisher records through 22 October 2026, checked citations against Crossref and abstracts, and prioritized studies with a measured first stage and a credible comparison. I did not run dual screening, a registered protocol, or a meta-analysis, and I did not reanalyze any data. English-language coverage is better than Portuguese, and unpublished nulls may be missing. For Brazil, the 2025 federal law created a common minimum, but implementation varied and the strongest causal evidence remains Rio's municipal policy; a national evaluation using census and assessment data is feasible but unpublished.

Calibrated confidence and update conditions

• 90% confidence: physical collection, lockers, or pouches reduce in-school phone access.
• 70% confidence: the average academic effect in secondary schools lies between -0.02 and +0.07 SD over three years. Site effects can fall outside that range.
• 65% confidence: the average academic effect is positive. The precise U.S. null keeps this near equipoise.
• 80% confidence: the average well-being effect is negligible over one to three years. Subgroup effects, especially for girls, remain plausible.
• 85% confidence: evidence does not support marketing bans as a population mental-health intervention.
• 75% confidence: punitive enforcement creates a short-run discipline cost that schools can influence.
• 70% confidence: rule design and the replacement activity matter more than the presence of a ban.
• 60% confidence: teacher-guided use can outperform a blanket ban with strong instructional support; two trials in one country need replication.

A multi-year cluster-randomized K–12 trial would change these estimates most. It should assign classroom-only restriction, bell-to-bell storage, guided use, and usual practice; collect objective device activity and time use; preregister standardized achievement, well-being, discipline, and subgroup outcomes; and test non-punitive enforcement. Two years of follow-up would separate adoption conflict from a stable effect.

Short of a trial, a coordinated prospective study across jurisdictions could use a common policy taxonomy and outcome battery. I would raise confidence in broad academic benefit if two independent sites replicated Rio's gain with a strong first stage and no concurrent reforms. I would lower it if longer U.S. follow-up retained the precise null after the adjustment period. I would revise the mental-health conclusion if total-use reductions mediated durable symptom or health-record changes in more than one country, and the guided-use conclusion if the Chinese result failed to replicate elsewhere or with younger students.

What follows for schools and policy

Separate three decisions that are often bundled.

Phone-free instruction. A school can justify a simple, consistently enforced rule that keeps phones out of lessons. The expected academic gain is modest. The case is strongest where students use phones during class, teachers spend time negotiating use, and devices can be stored without punitive escalation.

Bell-to-bell storage. A full-day ban makes a larger claim. It may change lunch and break interaction, but current studies do not establish a consistent social or well-being benefit over classroom-only restrictions. Schools choosing that model should state the social goal and measure whether students talk, move, bully, or withdraw.

Guided use. The newest trials suggest that purposeful, teacher-directed use can beat both a ban and unstructured use. That option requires trained teachers, reliable devices, and instructional materials. It is not a justification for free phone use; unstructured use was the worst tested condition.

Implementation should treat first-year discipline as an outcome. A school can set a simple storage routine, allow defined medical and accessibility exceptions, and avoid suspensions for ordinary noncompliance. It should track scores, classroom observation, attendance, referrals, and belonging by subgroup; the same rule can have a favorable average and an unacceptable cost for one group.

Parents should expect a ban to protect instructional time, not to reduce home screen use, repair sleep, or treat anxiety without other changes. Governments should fund measurement alongside mandates: a national ban without implementation data is a natural experiment run blind. Brazil's 2025 law and its uneven enforcement create an opportunity to learn which version helps, for whom, and at what cost.

References

1. Abrahamsson, S. (2026). Smartphone Bans, Student Outcomes and Mental Health. Journal of Human Resources, 61(5).
2. Allcott, H., et al. (2026). The Effects of School Phone Bans: National Evidence from Lockable Pouches. NBER Working Paper 35132, revised September 2026.
3. Baggio, S., et al. (2025). Psychological consequences of school mobile phone bans: Emulated trial of a natural experiment in South Australia. Computers in Human Behavior, 172, 108767.
4. Beland, L.-P., & Murphy, R. (2016). Ill Communication: Technology, Distraction & Student Performance. Labour Economics, 41, 61–76.
5. Campbell, M., et al. (2024). Evidence for and against banning mobile phones in schools: A scoping review. Journal of Psychologists and Counsellors in Schools, 34(3), 242–265.
6. Deng, Z., Cheng, Z., Ferreira, P., & Pavlou, P. A. (2026). From Smartphones to Smart Students: Learning vs. Distraction Using Smartphones in the Classroom. Information Systems Research, 37(1), 276–293.
7. Figlio, D. N., & Özek, U. (2025). The Impact of Cellphone Bans in Schools on Student Outcomes: Evidence from Florida. NBER Working Paper 34388.
8. Goodyear, V. A., et al. (2025). School phone policies and their association with mental wellbeing, phone use, and social media use (SMART Schools): a cross-sectional observational study. The Lancet Regional Health – Europe, 51, 101211.
9. Kessel, D., Hardardottir, H. L., & Tyrefors, B. (2020). The Impact of Banning Mobile Phones in Swedish Secondary Schools. Economics of Education Review, 77, 102009.
10. Lichand, G., Moreno-Louzada, L., da Costa, T., & Gentzkow, M. (2026). The Educational Impacts of School Phone Bans: Evidence from Brazil. NBER Working Paper 35233.
11. Perry, S. J., et al. (2026). Health economics analysis of restrictive school smartphone policies in secondary schools in England (SMART Schools). BMJ Mental Health, 29(1), e301892.
12. Saffer, H. (2026). Youth Mental Health and School Smartphone Bans: Early Evidence. NBER Working Paper 35181.
13. Shi, Y., & Villarroel, F. (2026). The Consequences of Cellphone Restrictions in Classrooms. IZA Discussion Paper 18426.
14. Sungu, A., Choudhury, P. K., & Bjerre-Nielsen, A. (2026). Phones impede learning: Evidence from a massive phone-free classroom experiment. Science, 394(6820), 178–181.
15. Vanluydt, E., et al. (2026). Disconnect To Reconnect: How Variations between Types of Smartphone Bans Influence Students' Well-being and Social Connectedness in Dutch Secondary Education. Journal of Youth and Adolescence, 55, 551–569.
16. Brazil. (2025). Lei nº 15.100, de 13 de janeiro de 2025. Diário Oficial da União.