โ† Studies Suggest ๐Ÿง  Psychology

Every Parent Worries About Screen Time. A Study of 355,358 Adolescents Found It Explains Less of Their Well-Being Than Wearing Glasses or Eating Potatoes.

Using a novel statistical method that tested every plausible analytic path through three massive datasets, Oxford researchers found that digital technology use accounts for at most 0.4% of the variation in adolescent well-being, an effect so small that potatoes and corrective lenses have comparable or larger negative associations.

By Nadia Vasquez, Technology & Behavioral Science ยท June 12, 2026

Listen to this article
Loading...
A teenager's hand resting on a face-down smartphone on a weathered wooden table beside a bowl of potatoes and a pair of wire-framed glasses in soft natural light

๐Ÿ“‹ The Study

Title
The association between adolescent well-being and digital technology use
Authors
Orben, A. & Przybylski, A. K., 2019
Institution
Department of Experimental Psychology & Oxford Internet Institute, University of Oxford
Journal
Nature Human Behaviour, 3(2), 173โ€“182
DOI
10.1038/s41562-018-0506-1
Sample
n = 355,358 adolescents across three large-scale social datasets (Monitoring the Future, Youth Risk Behavior Surveillance System, and UK Millennium Cohort Study)
Method
Specification curve analysis (SCA) across all defensible analytic specifications, applied to three nationally representative datasets from the US and UK
Key Finding
Digital technology use is associated with at most 0.4% of the variation in adolescent well-being. The median association across all specifications is ฮฒ = โˆ’0.035.
Effect Size
Rยฒ โ‰ค 0.004 (0.4% variance explained). Median ฮฒ = โˆ’0.035. For context: wearing corrective lenses shows a larger negative association; eating potatoes shows a nearly equivalent one.
Counterintuition
โšกโšกโšกโšก 4/5
Replication
Replicated. Consistent findings across three independent datasets within the study itself. Extended by Orben & Przybylski (2019) using time-use diaries (Psychological Science), Orben, Dienlin & Przybylski (2019) on social media specifically (PNAS), and Vuorre, Orben & Przybylski (2021) on time trends (Clinical Psychological Science). No retraction or methodological challenge has overturned the core finding.

The Panic That Launched a Thousand Bans

Australia banned social media for children under 16 in late 2024. The U.S. Surgeon General called for warning labels on social media platforms. School districts across the developed world have confiscated phones, locked them in pouches, and built entire curricula around "digital wellness," all in the name of protecting adolescents from their devices.

The evidence tells a different, and strikingly boring, story.

What 355,358 Adolescents Actually Show

In January 2019, Amy Orben and Andrew Przybylski at the University of Oxford published what remains the most methodologically rigorous examination of the relationship between digital technology use and adolescent well-being (DOI: 10.1038/s41562-018-0506-1). Rather than running a single analysis and reporting whatever emerged, they applied specification curve analysis (SCA), a technique that runs every defensible analytic pathway through the data simultaneously, across three of the largest social science datasets available: the U.S. Monitoring the Future survey, the Youth Risk Behavior Surveillance System, and the UK Millennium Cohort Study, for a combined sample of 355,358 adolescents.

The reason method matters here is that these datasets contain so many variables that a researcher exercising normal analytical flexibility could produce roughly 10,000 papers showing negative screen effects, 5,000 showing none, and 4,000 showing positive effects, all from the same underlying data. This is what statistician Andrew Gelman calls the "garden of forking paths": when you choose which variables to include and which subgroups to examine, the path determines the finding, and there are enough forks to reach almost any conclusion you want.

SCA closes those forks by running all of them at once, then displaying the complete range of results. When Orben and Przybylski did this, the answer was unambiguous: digital technology use was associated with at most 0.4% of the variation in adolescent well-being, with a median standardized effect of ฮฒ = โˆ’0.035, a negative association so small it would be invisible in any individual teenager's daily experience.

Smaller Than Potatoes

The datasets let the researchers compare the screen time association with everything else measured in these adolescents' lives, and the comparisons are devastating for the screen panic narrative. Wearing corrective lenses showed a more negative association with well-being than screen use did, and regularly eating potatoes showed a nearly equivalent one. Smoking marijuana was 2.7 times worse, and being bullied was 4.3 times worse. Getting enough sleep and regularly eating breakfast both dwarfed screen time's contribution to well-being in the positive direction.

An original calculation puts this in population terms. If that 0.4% represents the true causal ceiling (it is almost certainly lower, since cross-sectional data cannot establish causation), and the United States has roughly 42 million adolescents, then the maximum number whose well-being could be meaningfully attributed to screen use is around 168,000. Compare that with the 4.1 million U.S. adolescents aged 12 to 17 who experienced a major depressive episode in 2022, according to SAMHSA's National Survey on Drug Use and Health: screen time policies, even if perfectly effective, would address at most 4% of the adolescent depression burden.

A Trilogy, Not a One-Off

The Oxford team did not stop at one paper. A second 2019 study applied SCA to three time-use diary datasets and found the same result: objectively measured screen time showed equally tiny associations with well-being (DOI: 10.1177/0956797619830329). A third, published in the Proceedings of the National Academy of Sciences, found that social media's specific effect on adolescent life satisfaction was similarly negligible and showed no sign of increasing over six years (DOI: 10.1073/pnas.1902058116).

The strongest test came in 2021, when Matti Vuorre joined Orben and Przybylski to examine whether the link between technology use and mental health problems had grown over time. Using Understanding Society data from 2009 to 2017, a period when teen smartphone ownership rose from near-zero to near-universal, they found no increase whatsoever.

The Strongest Case for Screen Harm

Jean Twenge at San Diego State University and Jonathan Haidt at NYU mounted the most forceful challenge in a 2020 Nature Human Behaviour correspondence, raising three criticisms that deserve their full weight.

First, they argued that linear correlations obscure curvilinear effects: light screen users actually have the highest well-being, while heavy users, especially girls on social media, show pronounced drops, with twice as many heavy social media users in the Millennium Cohort reporting depression compared to light users. Second, the SCA included "mere participation" variables (did the teen use a device at all?), diluting the signal from heavy use.

Their third point is the strongest and deserves extended consideration: the 0.4% variance framing is inherently misleading because the same metric makes every environmental risk factor look trivially small. The association between smoking and lung cancer explains only about 1% of variance in cancer outcomes, yet no one calls that relationship negligible, because percentage-of-variance compresses real individual suffering into a population average that can disguise concentrated subgroup harm.

This criticism carries genuine force. A 0.4% population average does not prove that no teenager is badly affected by Instagram; what it does prove is that screening teenagers by total hours of screen time is a poor predictor of who will struggle, and that blanket restrictions treat a statistical non-signal as a reliable clinical indicator.

What We Didn't Prove

The analysis is cross-sectional, not experimental, so it shows the association is tiny without proving causation runs to zero. The data predate TikTok's explosive growth and the pandemic. "Digital technology use" was measured as a single bundle, collapsing video calls and doomscrolling into one variable, so specific harmful patterns (algorithmically amplified comparison, exposure to self-harm content) might produce real damage even if aggregate "screen time" does not. Self-reported usage correlates only moderately with device-logged data, meaning measurement noise could attenuate the true effect.

What You Can Do

The Bottom Line

Screen time is not the public health emergency it has been sold as, because the most rigorous analysis of the largest available data finds an association that ranks below wearing glasses and alongside eating potatoes, while sleep, bullying, and breakfast each explain far more of the variation in adolescent well-being. A decade of policy and parental anxiety organized around a variable that explains less than half a percent of the outcome it purports to protect has displaced resources from interventions that actually help.

Sources

  1. Orben, A. & Przybylski, A. K. (2019). The association between adolescent well-being and digital technology use. Nature Human Behaviour, 3(2), 173โ€“182. DOI: 10.1038/s41562-018-0506-1
  2. Orben, A. & Przybylski, A. K. (2019). Screens, teens, and psychological well-being: Evidence from three time-use-diary studies. Psychological Science, 30(5), 682โ€“696. DOI: 10.1177/0956797619830329
  3. Orben, A., Dienlin, T. & Przybylski, A. K. (2019). Social media's enduring effect on adolescent life satisfaction. Proceedings of the National Academy of Sciences, 116(21), 10226โ€“10228. DOI: 10.1073/pnas.1902058116
  4. Vuorre, M., Orben, A. & Przybylski, A. K. (2021). There is no evidence that associations between adolescents' digital technology engagement and mental health problems have increased. Clinical Psychological Science, 9(5), 823โ€“835.
  5. Twenge, J. M., Haidt, J., Joiner, T. E. & Campbell, W. K. (2020). Underestimating digital media harm. Nature Human Behaviour, 4, 346โ€“348.
  6. Semken, C. & Rossell, D. (2022). Bayesian specification curve analysis. Preprint.
  7. SAMHSA (2023). National Survey on Drug Use and Health: Detailed Tables. U.S. Department of Health and Human Services.