The Digital Whale

Screens and young people

Bullying that follows children home

The headline prevalence figures are artefacts of question wording, and the causal link between cyberbullying and self-harm is not established.

Two things need stating plainly before any number is quoted. First, published cyberbullying prevalence estimates range from roughly 4% to 46% of adolescents, and the spread is almost entirely a product of how the question was asked: lifetime experience of any listed behaviour, victimisation in the past couple of months, or bullying connected to school. All three are defensible measurements of different things. Second, victims show consistently raised rates of self-harm and suicidal behaviour, but the observational literature cannot establish that cyberbullying causes them. Every study in the relevant review was observational, and the reviewers name overlap with face-to-face bullying and prior mental-health difficulty as their finding's largest limitation. For a workplace-oriented comparison point on how digital activity is translated into metrics, see this reference.

Three careful surveys give three different answers

The most widely quoted US figure counts any experience of any listed behaviour, at any point, with no severity or repetition threshold.

2022

46% of US teens report at least one of six cyberbullying behaviours, most commonly offensive name-calling at 32%, with rates highest among girls aged 15–17 at 54%.

Direction: Increase. Strength of evidence: Strong.

Pew Research Center, Teens and Cyberbullying 2022, 20221,316 US teens aged 13–17, fieldwork April–May 2022

Caveat The headline counts a single instance of any listed behaviour with no severity, frequency or intent threshold, so it does not measure sustained bullying.

The largest international survey uses a recent window and a standardised item, and reports a figure a third the size.

2024

15% reported being cyberbullied at least once in the past couple of months (16% of girls and 15% of boys), up from 13% of girls and 12% of boys in 2018.

Direction: Increase. Strength of evidence: Strong.

WHO Regional Office for Europe, Health Behaviour in School-aged Children study, 2024279,000 adolescents aged 11, 13 and 15 in 44 countries and regions, 2021/22 round

Caveat The recall window and standardised item make these figures incomparable with lifetime "any behaviour" measures, and the round overlaps the pandemic.

The US federal series, which asks only about bullying connected to school, produces a figure an order of magnitude below Pew's.

2022

19% of students reported being bullied at school and 22% of those said it happened online or by text, implying roughly 4% of all students.

Direction: Cuts both ways. Strength of evidence: Strong.

National Center for Education Statistics, Condition of Education: Student Bullying, 20222021–22 School Crime Supplement to the National Crime Victimization Survey, ages 12–18

Caveat The survey asks only about school-connected bullying, so abuse occurring entirely outside school is excluded by design — the clearest illustration that prevalence is an artefact of question wording.

An earlier round of the same survey found 15% reporting online or text-based bullying, with three times as many girls as boys, 21% against 7% (Institute of Education Sciences, 2019). The wording changed between rounds, so the two are not a trend.

Definition, not country, drives most of the variation

That wording drives the number is not merely an inference from the discrepancy above; it has been tested directly.

2014

Mean cyberbullying prevalence was 15.5% for perpetration and 15.2% for victimisation against 34.5% and 36.0% for traditional bullying, and how a study defines bullying significantly predicts the prevalence it reports.

Direction: Cuts both ways. Strength of evidence: Strong.

Modecki, Minchin, Harbaugh, Guerra & Runions, Journal of Adolescent Health, 2014Meta-analysis of adolescent bullying prevalence studies

Caveat Now over a decade old and predating current platforms; the reported confidence intervals are implausibly narrow given the heterogeneity of the included studies.

Cross-country league tables are therefore close to unusable, and any comparison between surveys using different instruments is measuring the instruments.

Cyberbullying has not overtaken bullying in person

On the same evidence, traditional bullying is roughly twice as prevalent, and the two overlap heavily: most children bullied online are also bullied offline. The claim that online abuse has displaced or exceeded bullying in person is not supported by the prevalence data. What differs is continuity — it does not stop at the school gate — which is a claim about character rather than volume, and has not been quantified separately.

The association with self-harm is consistent, the causal claim is not

The association is among the more robust findings here. Its interpretation is not.

2018

Cyberbullying victims had 2.35 times the odds of self-harm (95% CI 1.65–3.34) and 2.10 times the odds of suicidal behaviours (95% CI 1.73–2.55).

Direction: Increase. Strength of evidence: Strong.

John et al., Journal of Medical Internet Research, 2018Systematic review of 26 studies covering 156,384 people under 25

Caveat All included studies were observational, the authors name overlap with traditional bullying and mental-health comorbidity as perhaps the biggest limitation, and reverse causality was untested.

An umbrella review six years later put the correlational picture in a wider frame, with effect sizes moderate rather than extreme.

2024

Median correlations were r = 0.24 between victimisation and depression and r = 0.29 with suicidality, while anti-cyberbullying programmes showed a median of only r = −0.07 for reduced victimisation.

Direction: Increase. Strength of evidence: Strong.

Kasturiratna, Hartanto, Chen, Tong & Majeed, Nature Human Behaviour, 2024Umbrella review of 56 meta-analyses covering 296 effect sizes

Caveat The review reports that 53.6% of the included meta-analyses conducted no quality appraisal of their primary studies, so evidence quality is uneven.

Those correlations are much larger than the associations between screen use and adolescent well-being, a reason to treat victimisation as a more tractable target than screen time. It is not a reason to call the relationship causal.

School programmes reduce it by small amounts

Interventions have been tested properly and they work, at a size that should shape expectations.

2022

School-based programmes reduced cyberbullying perpetration by g = −0.18 (95% CI −0.28 to −0.09) and victimisation by g = −0.13 (95% CI −0.21 to −0.05).

Direction: Decrease. Strength of evidence: Strong.

Polanin et al., Prevention Science, 2022Meta-analysis of 50 studies, 45,371 participants and 320 effect sizes

Caveat Moderator analyses produced few significant findings, so the review cannot identify which components are doing the work.

Because the moderator analyses were largely null, the evidence supports running a school programme but not choosing between them. Related: what schools do about devices and the teachable part of behaviour online.

The short version

  • Prevalence estimates range from roughly 4% in the school-linked US federal survey to 46% in Pew's lifetime measure, and the gap is the question, not the children.
  • A meta-analysis showed that a study's definition of bullying predicts the prevalence it reports (Modecki et al., 2014).
  • Traditional bullying remains about twice as prevalent, so claims that online abuse has overtaken it do not hold.
  • Victims show 2.35 times the odds of self-harm in pooled data (John et al., 2018), but every contributing study was observational and none ruled out prior vulnerability, so this is an association and not an established cause.
  • School programmes reduce perpetration by g = −0.18 (Polanin et al., 2022), a real but small effect, and the evidence cannot say which components work.