Social media’s link to youth mental health revealed by Australian panel data

Australias landmark social media ban for children under 16 has dominated headlines around the world, but a new study suggests that keeping young people away from the platforms may only be half the battle. The other half, the research indicates, lies inside the family home, where warm parenting, open communication and clear screen-time rules appear to buffer much of the psychological damage associated with the heaviest social media use. Drawing on nearly two decades of nationally representative longitudinal data, economists Alfred M. Dockery, Alan Duncan, Astghik Mavisakalyan and Loan Vu find that young Australians who share or post on social networking sites hourly or more often suffer measurably worse mental health than their peers who abstain, and that the harm is amplified among those who started using social media at the age of 12 or younger.

The study, published in the journal SSM – Mental Health, exploits one of the richest datasets of its kind anywhere in the world: Growing Up in Australia, the Longitudinal Study of Australian Children, which has tracked around 10,000 children and their families every two years since 2004. The authors pool data from survey waves conducted in 2018, 2020, 2021 and 2023, yielding 12,939 usable observations on young people aged 16 to 25. Mental health is measured with the Kessler Psychological Distress Scale, or K6, a clinically validated screening instrument that asks respondents how often over the past four weeks they felt nervous, hopeless, restless or fidgety, that everything was an effort, so sad they could not be cheered up, or worthless. Each item is scored on a five-point frequency scale, and the total ranges from 6 to 30 in the version used here, with higher scores indicating greater distress. A score above 19 on this scale is treated as a clinically meaningful screening threshold for probable serious mental illness.

The headline result is striking in its precision. In a two-way fixed effects model that controls simultaneously for time-invariant individual characteristics such as genetics, birth cohort, gender and cultural background, and for time-varying factors including age-sex effects, mothers birth year, postcode, state, and the month and year of the survey, young people who post or share on social media hourly or more often have K6 scores that are 0.75 points higher, on average, than those who never use the platforms. That corresponds to 0.14 standard deviations of the distress distribution, or roughly a 6 per cent increase relative to the sample mean of 13.06. To put the figure in context, the authors note it equals about 30 per cent of the average rise in psychological distress observed across the Australian population during the acute phase of the COVID-19 pandemic. Heavy users were also 3 percentage points more likely to cross the probable serious mental illness threshold, and 8 to 10 percentage points more likely to report specific symptoms such as feeling upset or worthless, or that everything was an effort.

Perhaps the most technically consequential finding is what the data do not show: a smooth dose-response gradient. The associations for moderate users, those posting several times a month or several times a week, are small and statistically indistinguishable from zero once the full battery of controls is applied. Only at the extreme frequency level, sharing or posting hourly or more often, does the relationship turn decisively negative. The authors interpret this as evidence of a threshold effect rather than a linear one, implying that ordinary, moderate engagement with social media is not the problem. The damage is concentrated among the roughly 10 per cent of young people who post several times a day or more, a pattern that remained robust when the researchers re-estimated their models with a series of alternative reference groups and with an eight-category version of the frequency variable.

Because correlational findings of this kind are vulnerable to reverse causality, the possibility that distressed young people simply retreat into their phones, the authors deployed several strategies to shore up the inference. They replaced current social media use with lagged measures, asking whether a young persons posting frequency one or two survey waves earlier, equivalent to roughly three to four years in the past, predicted their current mental health. It did. Using social media intensively one wave prior was associated with a 1.96-point higher K6 score, and the continuous frequency measure remained significant even at the two-wave lag. The association weakens as the time gap lengthens, which the authors caution could reflect either a genuinely short-term effect or the fading influence of unobserved confounders, and they are explicit that these estimates should be read as robust associations rather than proven causal effects. Robustness checks excluding the pandemic years entirely, controlling for a 22-item Stressful Life Events Index and the number of siblings, and converting frequency categories into an estimated count of daily use instances all reproduced the core result, in some cases with even larger coefficients.

Addiction-style symptoms tell a parallel story. Around 32 per cent of young people in the sample reported feeling a need to continually use more social media, 27 per cent said they had tried and failed to cut back, and 19 per cent reported becoming anxious when prevented from using the platforms. All three indicators were positively associated with psychological distress. Those who became anxious when prohibited from using social media scored nearly a full point higher on the K6 and were 5.2 percentage points more likely to screen positive for probable serious mental illness. The authors also pioneered a dimension no previous study has examined: the age at which a person first used social media. Using questions asked of children between 2012 and 2016, they classified respondents into those who began using social networking sites by age 12 and those who started between 13 and 16. The early starters were significantly more likely to report addiction symptoms and psychological distress years later, and for them the link between heavy use and poor mental health was substantially stronger. Age 12, the authors note, marks the transition from primary to secondary school in Australia, a period of major social and emotional upheaval, and the finding offers some of the first empirical support anywhere for age-based access restrictions of the kind Australia has now legislated.

The study also traces the behavioural ecology that surrounds heavy use. Compared with never-users, the heaviest users were 14.1 percentage points less likely to have daily contact with their parents, 8.5 percentage points more likely to feel lonely, 7.6 percentage points less likely to sleep at least seven hours a day, and roughly 10 percentage points more likely to have smoked in the past four weeks; they also consumed significantly more alcohol, around 1.8 additional units in the previous week. The authors read this cluster through the lens of the social displacement hypothesis, the idea that intensive digital engagement crowds out offline relationships and health-promoting behaviours, while the persistent loneliness among heavy posters is consistent with the social compensation hypothesis, in which people who struggle to connect offline seek validation online and fail to find it. Heavy substance use, they suggest, may reflect self-medication for the stress and disrupted sleep that accompany a hyperconnected digital life.

Where the study offers genuine hope is in its heterogeneity analysis by family environment. Young people whose parents had set rules about screen time showed a weaker association between social media use and psychological distress, as did those whose parents asked about their problems, displayed physical affection, expressed happiness around their children, and maintained warm, emotionally engaged relationships. In these families, the link between intensive posting and distress was visibly attenuated. The authors are careful to flag the caveats: parents who impose screen-time rules may differ systematically from those who do not, and reverse causality is possible if parents respond to a childs deteriorating mental health by imposing restrictions. These patterns, they stress, describe correlations within different family types rather than proven causal protective effects. Even so, the convergence of evidence across multiple parenting dimensions suggests that family context meaningfully shapes how social media exposure translates, or fails to translate, into psychological harm.

The policy implications land at a delicate moment. Australias Online Safety Amendment Act 2024, passed in November 2024 and in force since December 2025, prohibits children under 16 from holding accounts on major platforms including Facebook, Instagram, TikTok, Snapchat, Twitter and Reddit. Yet recent survey evidence suggests compliance has been low: nearly two-thirds of Australian children aged 14 to 15 continue to use banned platforms, and peer effects and social norms make enforcement notoriously difficult. Against that backdrop, the new findings suggest that regulatory bans alone may not be enough. If restriction cannot be reliably enforced, the authors argue, the priority should shift toward helping young people develop healthier digital habits, supported by parental monitoring, screen-time rules and warm family communication, alongside the broader regulatory architecture. The stakes are economic as well as personal. Related research cited in the study estimates that a one-point rise in psychological distress is associated with a 1.6 per cent increase in annual healthcare costs among young Australian women, while Australian workers with high distress lose around AUD 3,600 per worker per year in productivity, and US adolescents with clinically significant distress go on to earn roughly USD 5,700 less annually a decade later.

The authors are candid about the studys limits. Social media use and mental health are both self-reported, exposing the analysis to measurement error and recall bias, and the data record only the frequency of posting or sharing, not the platform, the content, or the distinction between active creation and passive scrolling, which prior work suggests may carry very different psychological consequences. Nor can the study test cognitive mechanisms such as upward social comparison or digital status seeking directly. The relationship between posting and distress may also depend on a young persons offline social world: for socially well-connected adolescents, frequent sharing may extend existing friendships and buffer distress, whereas for socially anxious teenagers the same behaviour may deepen a reliance on online validation. Understanding why some young people are more vulnerable than others, the authors conclude, is the next frontier for a field that, thanks to datasets like this one and a natural policy experiment now unfolding in Australia, is finally getting the longitudinal evidence it has long lacked.