Yes, Social Media is Harming Young People's Mental Health
A stylized literature review
Note: I am in the process of getting set up at a new job, and so will have less time to write over the next few weeks. In the meantime, I thought I’d repost a lightly edited version of the section on social media and mental health from my previous essay, “The Leviathan, the Hand, and the Maelstrom.” While a lot has been written on this issue, I am aware of no article besides this one that gathers much of the strongest research on social media’s mental health harms in one place for easy reference. Haidt, Twenge, and Rausch’s sprawling ‘collaborative review’ is a partial exception, but collects much more evidence of varying quality and relevance (ditto for other academic reviews). I hope this short essay can act as a more curated and accessible supplement.
In the aftermath of the historic rise of social media, it is easier than ever to make what social scientists call ‘unfavorable,’ or ‘upward’ social comparisons. Online, we are beleaguered by ostentatious displays of wealth, status, and power. It is only natural that we humans, who thrive and falter on social hierarchy, would react with class-conscious anger, anger directed not just outward, but also inward, at ourselves.
Hence the recent uprising in Nepal, which was led by members of Gen Z, and motivated in large part by online content showing off the opulent lifestyles of the country’s so-called “nepo-kids,” an unsurprising firestarter in a country with a per capita GDP of around 1,450 USD.
Not everyone reacts to the glossy irrealism of the digital world, though, with anger and contempt. Some, and in particular those yet to graduate from the angst of youth into the brashness of adulthood, react instead with self-loathing and hopelessness—that they’re not the person they should be, that they can never hope to compare to what they’re seeing on their screens. This explains why, since the rise of social media, an unprecedented amount of adolescents have fallen into mental health crisis.
Despite remaining more or less constant through the aughts, mental illness among young people in the U.S. began to sharply increase starting around 2010. To give a small, but grim sampling from Jonathan Haidt’s popular book The Anxious Generation, from 2010 through 2020, depression increased by 106% among college students, anxiety by 139% and 103%, respectively, among those ages 18-25 and 26-34, and suicide rates by 167% and 91%, respectively, among girls and boys aged 10-14 (p. 26-31).
There is of course a lively debate as to whether these developments can be attributed to social media. Critics often cite large meta-analyses which they claim show the existing evidence to be equivocal or inconclusive. For instance, Candice Odgers, a prominent critic of the alleged link between social media and youth mental health claims, in her critical review of Haidt’s book, that efforts by researchers to search for the “large effects suggested by Haidt … have produced a mix of no, small and mixed associations” (p. 29). Odgers’s phrasing would seem to suggest a blend of statistically significant and insignificant effect sizes with both positive and negative signs.
But even a basic review of the literature shows that suggestion to be inaccurate—and to be frank, Odgers should know better. Of the 4 meta-analyses and review articles cited by Odgers in her review of Haidt, every single one, but her own openly acknowledges that the predominant result in the correlational research under review, with respect to adolescent anxiety and/or depression in particular, is a small, but statistically significant association of about r = .1, in other words a persistent, although modest association between social media use and mental illness. Several similarly comprehensive meta-analyses Odgers does not cite all find the exact same thing. The most recent of these—Liu et. al. (2022)—reviewed 21 cross-sectional and five longitudinal studies covering 55,340 participants and found that “the risk of depression increased by 13% for each hour increase in social media use in adolescents” (p. 1). The notion that the correlational research is equivocal or inconclusive on this point is flat-out false (you can read Haidt’s own rebuttal to Odgers here.)
This doesn’t prove much in itself. Those correlations may reflect causality running in the other direction—that more depressed or anxious kids use more social media, rather than the other way around. But, at a minimum, they mean that the hypothesis of a causal link passes initial muster. Social media might well have been associated with reduced anxiety and depression. It might well have only ever had a statistically insignificant association. But instead, most of the correlational research points the same way at a first pass.
Fortunately, we don’t only have the correlational research to rely on. A growing number of experimental and quasi-experimental studies support the existence of a genuine causal link. Some are social media ‘reduction’ studies, in which regular social media users take a break from social media for a period of time, and some are social media ‘exposure’ studies, which assess the impact of gaining access to social media. Both point in favor of genuine causation.
In a recent meta-analysis of 20 social media reduction RCTs covering 10,106 (adult) participants, Liu et. al. (2025) found that the interventions had a common effect size of .233 on overall well-being, a small, but significant positive effect (p. 10). The largest reduction experiment to date, Alcott et. al. (2020), assessed the effects on 2,743 people of deactivating Facebook for the four weeks before the 2018 U.S. midterm election, and found they experienced .09 SD and .1 SD decreases in measures of anxiety and depression, respectively. Another large study, Tromholt (2016), with 1,095 participants, found that quitting Facebook for only one week significantly improved life satisfaction and emotional positivity, and that both effects were larger for both heavy and passive users. Brailovskaia et. al. (2020), Lambert et. al. (2022), Reed, Fowkes, and Khela (2023), and Davis and Goldfield (2025), though smaller in size, all found similar, significant effects in their own reduction experiments. Findings of decreased symptoms of depression were particularly common. Though a number of other small experiments show conflicting results, the balance of evidence, especially when adjusted for statistical power, favors a small, but positive causal effect of reducing social media use on mental health and general well-being.
Reduction studies, though, can at best tell us about the likely effects of reducing social media use in a world where it is already very common. Exposure studies, on the other hand, can help us understand the impact of its initial spread. As one would expect given the evidence reviewed so far, exposure studies provide further confirmation of a causal link between social media use and mental health. Two in particular are worth highlighting.
Braghieri, Levy, and Makarin (2022) took advantage of a natural experiment—the staggered early rollout of Facebook on college campuses between 2004 and 2006—to assess the differential impact of the platform on mental health among college students through 2008. They found that the introduction of Facebook led to a .085 SD increase in an index of poor mental health among college students, a magnitude equivalent to around 22 percent of the effect of losing one’s job, and a 2 percentage point increase in the share of students suffering from depression. They also calculated (on the basis of some relatively strong assumptions), that the introduction of Facebook accounts for approximately 24 percent of the increased prevalence of severe depression among college students over the last two decades (p. 3662).
Guo (2022) also took advantage of a natural experiment—spatial variability in the introduction of high-speed wireless internet to British Columbia—to assess the impact of greater social media access on teenage students. She found both that access to high-speed wireless internet was a good proxy for greater social media use (p. 22-23), and that, while boys did not experience a significant change in mental health in areas with access to high-speed wireless (and if anything saw slight improvement), teen girls in those areas saw a 90% increase in severe mental health diagnoses from 2010 through 2019 (p. 28). This is consistent with previous research finding associations between digital media use and mental health to be larger in girls than in boys.
Guo explains her findings by pointing to research that suggests girls are more sensitive to how social media might reshape their social landscape and interpersonal lives. She also cites internal research done by Meta which found, among other things, that Instagram made “’body image issues worse for one in three teen girls’” and that “’constant comparison on Instagram is the reason why there are higher levels of anxiety and depression’” (p. 9).
Meta’s in-house research team isn’t alone in this conclusion. A number of recent papers have honed in on unfavorable social comparison as a key mechanism explaining the link between social media and mental illness among young people, and in particular girls. Adolescents are (in)famously attuned to social hierarchies, and whereas they once would’ve only had to worry about local ones, those at their school or in their community, social media thrusts them at an early age into a substantially larger, more competitive social pool, one in which the ‘popular kids’ are not just the richest or most attractive in their school, but the richest and most attractive in the world.
Not only that, but engagement metrics in effect serve as ‘popularity scores’ that allow young people to compare where they or their peers fall in the pecking order. It makes a lot of sense that 57% of Gen Zers would “become an influencer if given the opportunity,” considering that, for them, this is more or less equivalent to becoming ‘the most popular type of person’ in their now oversized reference group. What is often missed in overspecific debates about the hyper-competitive nature of college admissions is that all of childhood is now much more competitive, including and especially the childish parts. Call it ‘free trade’ in the teenage popularity market. Once localized status competition—subject to various forms of ‘infant industry protection’—has now gone global. Unionized American machinists versus sweatshop workers in Bangladesh? Try being a teenage girl competing for likes with Charli D’Amelio.
Jokes aside, it is straightforward to see how, for social media’s heaviest and most sensitive users, constant contact with curated displays of high status might lead to distress. Exposure experiments like Guo’s further reinforce the conclusion implied by every other line of evidence: social media causes small, but statistically significant increases in mental illness among young people.
Those who continue to stubbornly oppose this conclusion are going to find themselves with fewer and fewer intellectual allies in the years to come. Strong evidence continues to pile up, and that is before the effects of the much more recent advent of short-from video content have made it into datasets. Though a few academic holdouts will likely continue to harp on the issue of small effect sizes, they would do well to consider the following:
a) much of the literature is cross-sectional or dependent on short interventions, meaning it won’t capture the compounding effects of social media use over time (remember: teens spend almost 5 hours on social media per day—and have been doing so for over a decade);
b) effect sizes are averages, and so consistent with vulnerable sub-populations who fare much worse than the norm, and in turn impose large second-order costs on others (financial, emotional, etc.);
c) the baseline number of adolescents with any one mental illness is low, and so not especially difficult to increase in large percentage terms with an intervention (i.e. social media) applied to 100% of adolescents; and
d) many adolescents may be on the threshold of mental illness, and only need a small nudge to be pushed over.
For these reasons and others, small effects on the individual level can mean very large ones on the national level (for more on the “indispensability” of small effects, see Götz, Gosling, and Rentfrom 2021).
While numerous questions of course remain as to the exact nature of social media’s effects on mental health, the evidence of a genuine causal link is strong. Social media has done serious, and sometimes irreversible harm to America’s young people. Yes, the details matter, and will continue to be teased out by careful research, but there is little room for debate over the broader strokes. Given the often glamorous, but warped image of the world that young people consume on social media, it is no wonder that many of them feel as if they cannot possibly measure up.