A new analysis of more than 183,000 American adults over 11 years finds that higher daily social media use is consistently linked to a greater likelihood of self‑reported depression. Using machine learning to examine demographic and media habits from 2014 to 2024, researchers report that social media use sits among the strongest predictors of depression, with risk rising notably once daily use crosses about two and a half hours. The study, published in npj Mental Health Research, suggests the connection extends beyond teenagers and has persisted as digital habits have evolved.
Social media refers to online platforms designed for interpersonal communication, sharing content, and public commentary on community life, services or products. Over the past decade, platforms such as Instagram, Snapchat and TikTok have reshaped how people socialise, express themselves and compare their lives with others. Public health experts have repeatedly questioned whether heavy engagement with these apps contributes to depression, a mood disorder marked by persistent sadness, low energy and diminished interest in normal activities.
Earlier findings offered a mixed picture. Some Danish research in 2016 found that a one‑week break from Facebook boosted life satisfaction for heavy users or those prone to envy online. A 2018 PsyPost report noted that teens who spent more free time on screen activities reported lower happiness than peers who spent more time interacting in person. Other studies argued that it is unclear whether social media triggers distress or simply correlates with pre‑existing struggles.
The current team—led by Martin Block of Northwestern University with Vikram Suresh and Hans Breiter of the University of Cincinnati and colleagues—built on Block’s earlier analysis of 2014 data. They analysed 11 waves of data from the Media Behaviour and Influence Study, an annual survey conducted each January and February from 2014 through 2024 by Prosper Analytics. Respondents were drawn from corporate databases, social media and direct mail, with yearly samples balanced to U.S. Census benchmarks for age and gender. After excluding incomplete responses and those finishing the survey in under two minutes, the analytic sample comprised 183,300 adults across the 11 years, with about 15,410 to 17,869 participants each year.
Depression was measured via a self‑report question listing 31 health conditions, to which respondents answered yes or no if they experienced depression. Participants also estimated the likelihood that they engaged in various media activities during typical weekdays and weekend days, from which the researchers derived average daily minutes spent on social media, internet use, television, gaming, email and news.
The researchers began with logistic regression to assess the association between each media activity and the odds of reporting depression while adjusting for demographic factors. They then trained balanced random forests—machine learning models that compare equal samples of depressed and non‑depressed individuals—for each survey year, using 15 demographic and media‑use variables.
To interpret the models, the team employed Shapley Additive Explanations, a technique from game theory that gauges how much each variable pushes a person’s predicted probability of depression up or down.
Across the 11 years, both daily social media time and the share of adults reporting depression rose. Depression prevalence increased from 15.1% in 2014 to 22.0% in 2024, a rise observed both before and after the onset of the COVID‑19 pandemic. After adjusting for demographics, each additional minute of daily social media use was associated with a 0.26% increase in the odds of reporting depression, equating to roughly 17% higher odds per extra hour per day.
Similar positive links emerged for daily time spent on video games, general internet use, television watching and email, with gaming and internet use showing per‑minute associations as strong as or slightly stronger than social media. News consumption showed a weaker relationship, appearing as a predictor in only two of the 11 waves.
In terms of predictive performance, the balanced random forest models achieved modest ability to distinguish depressed from non‑depressed adults, with a standard measure ranging from 0.66 to 0.74 across the decade. They correctly flagged about 40% to 53% of those reporting depression. A model trained on 2014–2018 data performed similarly when predicting 2019–2024, with scores from 0.68 to 0.72, indicating the observed relationships remained stable over time.
Additional sensitivity analyses showed that omitting the 2020 and 2021 waves or treating the pandemic as a separate factor did not alter the link between social media use and depression. When considering predictor importance across the decade, personal income ranked highest, followed by social media use, general internet use and age range.
The study also traced how social media’s predictive prominence evolved. It rose from ninth place in 2014 to consistently sit within the top four predictors from 2019 through 2024. While daily social media use is a readily modifiable habit, the researchers cautioned that high importance in their models reflects variation across the population rather than proving that reducing the behaviour will definitively lessen depression risk.
Subgroup analyses revealed notable patterns. Among 18 to 25‑year‑olds, depression prevalence among women ranged from 24% to 40% across the decade, compared with 15% to 25% for men. The models also performed better among adults aged 45 and older than among the youngest group.
The findings echo prior PsyPost coverage of related work on adolescents and younger populations, while also contrasting with studies that found no association when examining longer‑term changes in daily social media time. The researchers acknowledge that, because the data are cross‑sectional—sampling different individuals each year—the study cannot establish causation. It remains possible that people with depressive symptoms spend more time online, or that the relationship runs in both directions through mechanisms such as social comparison or sleep disruption. The use of a single self‑reported depression item and the January–February timing of media‑time estimates are noted as limitations, along with the commercial nature of the survey panel and its limited demographic balancing beyond age and gender.
The authors call for future longitudinal and experimental work to track individuals over time and to tease apart how different online behaviours—passive scrolling versus active interaction—affect mental health. They also suggest exploring whether keeping daily social media use below a potential 150‑minute threshold could offer meaningful benefits, though the current evidence does not establish a safe limit.
The study—Social media use and depression across 11 years of U.S. adult data—was authored by Martin Block, Vikram Suresh, Johnathan Avant, Sumra Bari, Nicole Vike, Felix Zhan and Hans Breiter.
