Home · C36

C36 alive

School phone bans raise test results more for lower-achieving pupils than for higher-achieving pupils.

Status

alive

Last change

2026-11-13

Serves

Q2 (How does that work in practice: in school, with screens, and with AI as one of the tools?)

Ages

grundskola-hog, gymnasium (school evidence: England, Sweden grade 9, Norway lower secondary, US middle and high schools, Chile grades 8 and 10, Rio de Janeiro grades 6–9); vuxen-livslangt for a large Indian higher-education RCT (Sungu et al. 2026; mean age 19.4, range 16–22)

Area

Screens, devices and attention

Relates to

C36 relates-to C6 (C6 is whether an average gain from banning or restricting pupils' own phones during lessons is established; C36 is whether any gain is larger for lower-achieving pupils. They share sources and settings but ask different questions. Thematic, not a dependency). C36 relates-to C9 (C9 is about school-provided one-device-per-pupil programmes, not rules for pupils' own phones. Thematic, not a dependency).

Bounds

Scope: school bans or restrictions of pupils' own phones during lessons, compared with freer access, measured on test results or grades, split by prior achievement. Not about calm, wellbeing or reported distraction alone. School-provided 1:1 programmes are C9. C6 covers whether an average gain is established; this claim is about who gains more. Evidence that shows the pattern, from studies that split by prior achievement: Beland & Murphy 2016 (Labour Economics 41, 61–76, DOI 10.1016/j.labeco.2016.04.004; full text): England, difference-in-differences; high-stakes age-16 scores rose by 6.41% of a standard deviation on average (0.0641 in the most flexible specification); by prior achievement quintile at age 11, the lowest quintile gained 0.1423, the second 0.0986, the third 0.0654, the fourth 0.0229, and the top quintile −0.0254 (not significant). The authors report that the top quintile was neither helped nor harmed. Limits of this study (full text, corrected 2026-11-13): the average rise is significant only at the 10% level (0.0641, SE 0.0373; Table 4); by subject (Table 10), the lowest quintile's gain came from subjects other than English and maths (+0.1652, SE 0.0451), while in English and maths together it was +0.0273 (not significant), smaller than the second to fourth quintiles (+0.0635, +0.0538, +0.0388), and in maths alone the lowest quintile fell (−0.0726, SE 0.0238, p < 0.01; mean maths effect 0.0035); looser rules (phones off or on silent) helped the lowest quintile more than strict bans (+0.3052 vs +0.1085; Table A.1); the age-14 outcome rose 0.0245 (not significant). So the pattern holds on total exam results, not shown for English and maths, and reversed in maths, in this study. Lichand, Moreno-Louzada, da Costa & Gentzkow 2026 (NBER WP 35233, preprint, not peer reviewed when added; full text): Rio de Janeiro city-wide ban; school-wide phase +0.061 SD overall in the matched sample. Table D3, full sample, school-wide ban by baseline grade quartile: bottom 0.080 (SE 0.046, p < .10), then 0.040, 0.010, 0.003; classroom phase, bottom 0.062 (SE 0.025, p < .05). In the matched sample the same school-wide figures were 0.099, 0.078, 0.044, 0.040, none significant. The authors say effects are concentrated at the bottom of the pre-ban distribution. Contrast is exposure intensity under one ban, not ban vs no ban. Sungu, Choudhury & Bjerre-Nielsen 2026 (Science 394(6820), 178–181, DOI 10.1126/science.aef7249; peer reviewed; Science page read): preregistered RCT in Indian higher education, end-of-semester GPA +0.073 SD overall; larger and significant among students below median prior achievement, near zero and not significant above. Higher education (mean age 19.4), not school pupils; lecture-only ban. Subgroup point estimates are not quoted from the earlier working paper, whose overall estimate (0.086) already differs from the Science article. Related but not a prior-achievement split: Abrahamsson 2024/2026 (Journal of Human Resources early view, DOI 10.3368/jhr.0224-13403r2; NHH SAM 01/24 full text): Norway lower secondary, average effect null; girls' GPA +0.08 SD, teacher grades +0.09, externally marked maths +0.22; larger for girls whose fathers had vocational education (GPA +0.09 to 0.14 SD three and four years after). No effect on boys, but the author “cannot reject the null hypothesis that the coefficients between girls and boys are equal, in any of these results” (discussion paper p. 22); girls' GPA +0.08 is p = 0.064 and teacher grades +0.09 p = 0.05; maths +0.07 (p = 0.067) one year after, +0.22 (p = 0.014) four years after; under strict bans girls' GPA +0.12 (p = 0.032) and maths +0.25 (p = 0.028); no never-treated schools, 40% of schools linked (corrected 2026-11-13 from the discussion-paper full text; the published JHR tables not yet read). The split is by gender and father's education, not prior test scores. Evidence already on file that limits or points against the pattern, recorded here and not applied as an attack: Kessel, Hardardottir & Tyrefors 2020 (Economics of Education Review 77, 102009, DOI 10.1016/j.econedurev.2020.102009; IFN WP 1288 full text): Sweden grade 9, no average improvement; no individual test scores, so prior achievement could not be tested. Using school-level socioeconomic status as a proxy, the lowest quintile had test scores +0.251 (SE 0.101) and a 3.328 percentage-point drop in failing maths; merit points −0.063 (not significant). The authors say these may be chance and that the event graph is not convincing. Allcott, Baron, Dee, Duckworth, Gentzkow & Jacob 2026 (NBER Working Paper 35132, preprint, not peer reviewed when added; full text): US schools adopting lockable pouches; limited systematic heterogeneity in pre-registered subgroups; no positive pooled test-score effect for any subgroup examined; in middle schools, negative effects were more pronounced among higher-achieving schools (not pre-registered; school-level baseline achievement, not pupil-level). Shi & Villarroel 2026 (EdWorkingPaper 26-1413 / IZA DP 18426, preprint, not peer reviewed when added; full text): Chile, classroom-level teacher restrictions; test scores near zero; no systematic differential effects by prior achievement, gender or socioeconomic background. Goodyear et al. 2025 (The Lancet Regional Health – Europe 51, 101211, DOI 10.1016/j.lanepe.2025.101211; CC BY, full text read): cross-sectional comparison of 20 restrictive vs 10 permissive English secondary schools, 1,227 pupils aged 12–15; no significant difference in teacher-rated odds of being below target in English (adjusted OR 1.45, 95% CI 0.85 to 2.47, p = 0.18) or maths (1.01, 0.45 to 2.27, p = 0.98). Not a before/after or randomised design, and no split by prior achievement; attainment is a teacher rating, not a test score. Figlio & Özek 2025 reports year-two score gains but has no prior-achievement split in the published subgroups. Laboratory “mere presence” findings (Ward et al. 2017) did not replicate in a preregistered direct replication (Ruiz Pardo & Minda 2022, Acta Psychologica 230, 103717, DOI 10.1016/j.actpsy.2022.103717; CC BY full text; N = 383; OSpan location F(2, 377) = 0.10, p = .91). This claim is about school test results, not mere presence in the lab. Added 2026-11-12 on Jonatan's yes, as a distinct claim from C6, with an empty ledger; not attacked in the same turn.

Does-not-transfer-to

School-provided one-device-per-pupil programmes (C9). Average effects of phone bans, which are C6.

What would kill this

A well-designed field randomised trial, difference-in-differences study, or meta-analysis of school phone bans finds that lower-achieving pupils do not gain more on test results than higher-achieving pupils: either neither group gains, or any gain is as large or larger for higher-achieving pupils.

Tutor consequence

none yet

Attack record

No attacks yet

No attack is counted against this claim yet.

Corrections to our notes

These fix a number or a fact in our notes. They are not attacks and do not count toward the floor of 3 survivals.

• 13 November 2026 — In Beland & Murphy 2016 the average rise is significant only at the 10% level, the lowest quintile's gain comes from subjects other than English and maths, and in maths alone the lowest quintile fell; in Abrahamsson's Norwegian study girls' and boys' educational effects cannot be told apart statistically.

Linked sources

9 linked sources (3 supports, 4 limits, 2 challenges)

Supports (3)

Limits (4)

Challenges (2)

Review state: first-pass draft, pending human review.