Screens and young people
Screen time and teenage mental health
The associations are real, consistent in direction and very small, and the two best studies of school phone bans point in opposite directions.
The association between adolescent screen use and mental health is real, mostly negative in direction and very small. The best-powered analyses find that digital technology use explains a fraction of a per cent of the variation in adolescent well-being, and meta-analytic correlations with depressive symptoms sit at the low end of what psychology counts as a detectable effect. Longitudinal cohort work points from social media use to later depressive symptoms rather than the reverse, with tiny coefficients. No study has established that smartphones or social media caused a population-level decline in teenage mental health. That case rests on two trend lines moving together, not on a design that can identify a cause. For a workplace-oriented comparison point on how digital activity is translated into metrics, see this page.
The effect sizes are small and they have stayed small
The most-cited analysis here ran every defensible combination of variables across three national datasets rather than the one producing the most interesting result. The same data supports a large negative finding or none at all, depending on the controls chosen.
Digital technology use explains at most 0.4% of the variation in adolescent well-being.
Direction: Decrease. Strength of evidence: Strong.
Caveat Cross-sectional secondary analysis of three existing datasets, all relying on self-reported screen time, which tracks logged use only weakly.
A review the following year reached the same order of magnitude and added a persistent problem: most of the field is cross-sectional and its screen-time measure is inaccurate.
Meta-analyses find correlations of roughly r = .07 to r = .13 between social media use and depressive symptoms or well-being, and only 14% of studies in the reviewed literature used longitudinal designs.
Direction: Increase. Strength of evidence: Strong.
Caveat A narrative review rather than a fresh meta-analysis; the authors note self-reported screen use correlates only about r = .20 with objective measures.
A correlation that size improves a prediction of a teenager's mood by an amount invisible in a class of thirty, and stable only in samples of tens of thousands.
Small does not automatically mean absent
The most recent synthesis argues the inconsistency is a power problem rather than a null result: an umbrella review of 72 reviews covering 1,070 effect estimates from 452 studies concluded that the mixed findings may reflect inadequate statistical power in studies of general social media use rather than a true absence of effect (Tølbøll et al., Child and Adolescent Mental Health, 2026). Only the abstract was accessible, so pooled effect sizes are unreported.
Direction of causation is beginning to separate out
Cross-sectional data cannot say whether unhappy teenagers use more social media or heavy users become unhappier. Following the same children for years can, and the largest US cohort found the arrow running one way.
Within-person increases in social media use predicted later depressive symptoms (β = 0.07 from year 1 to 2 and β = 0.09 from year 2 to 3), while depressive symptoms did not predict later social media use.
Direction: Increase. Strength of evidence: Strong.
Caveat Social media time was self-reported and the coefficients are very small; the JAMA page was CAPTCHA-blocked, so figures were verified through secondary reporting.
Heavy use looks different from average use
Averages hide the tail. In the UK Millennium Cohort Study the heaviest users reported more depressive symptoms, and the girls-boys gap came as much from who was in that group as from susceptibility.
Using social media three or more hours a day rather than one to three hours was associated with 26% more depressive symptoms in girls and 21% in boys, and 43% of girls versus 21% of boys were in the heavy-use group.
Direction: Increase. Strength of evidence: Mixed.
Caveat Cross-sectional at age 14 within a longitudinal cohort, and sleep, cyberbullying, self-esteem and body image accounted for much of the association, so heavy use partly stands in for those pathways.
Two of those mediators have entries here: what devices do to sleep and abuse that continues after the school day.
School phone bans have produced opposite results in two countries
Restricting phones is the one lever governments have pulled at scale, which makes the disagreement between the two best studies consequential. An English study found no wellbeing difference between restrictive and permissive schools despite a measurable drop in phone use.
Restrictive school phone policies showed no association with mental wellbeing (adjusted WEMWBS difference −0.48, p = 0.62) despite reducing in-school phone use by 0.67 hours.
Direction: No detectable effect. Strength of evidence: Mixed.
Caveat Cross-sectional and observational: schools chose their own policies rather than being randomised, and the design cannot detect effects of long-established bans.
A Norwegian event study using the timing of ban adoption reported the opposite: fewer healthcare consultations for psychological symptoms among girls, less bullying for both sexes and better grades for girls, with larger effects among the most disadvantaged (Abrahamsson, 2024). It is an unreviewed working paper whose coefficients could not be verified, so only the direction should be taken from it.
Global connectivity data points the other way again
The largest study of internet use and well-being measures access rather than social media, and finds the association positive nearly everywhere.
84.9% of associations between internet connectivity and well-being were positive and significant, the main exception being 4.9% of associations with community well-being that were negative, concentrated among women aged 15–24.
Direction: Increase. Strength of evidence: Strong.
Caveat A correlational multiverse analysis of Gallup World Poll data measuring internet access rather than social media; the authors explicitly disclaim causality.
The negative exception is small, but it falls on the same group the UK cohort flags — a coincidence worth more attention than either finding alone.
The short version
- Digital technology use explained at most 0.4% of the variation in adolescent well-being across three national datasets (Orben & Przybylski, 2019).
- Meta-analytic correlations between social media use and depressive symptoms cluster around r = .07 to r = .13 (Odgers & Jensen, 2020).
- The ABCD cohort found small within-person effects running from social media use to later depressive symptoms and not the reverse (Nagata et al., 2025).
- Most of the literature is cross-sectional and rests on self-reported screen time that tracks logged use weakly, so the true effect is unresolved.
- Evidence on school phone bans is split: an English study found no wellbeing difference, a Norwegian working paper found benefits concentrated among girls.