What the evidence really says about social media’s impact on teens’ mental health

The kids are not all right — and the device you are probably reading this on is to blame.

So argues the social psychologist Jonathan Haidt. In his new book, The Anxious Generation: How the Great Rewiring of Childhood Is Causing an Epidemic of Mental Illness, Haidt insists that smartphones and social media are fueling a “surge of suffering” that’s inundating teens all across the Western world.

By Haidt’s account, smartphones and the addicting social media apps we download onto them have lured the world’s youths away from those activities that are indispensable to healthy child development — such as outdoor play, face-to-face conversation with friends, and sleep — and trapped them in a digital realm that saps their self-esteem, drains their attention spans, and forces them to put on a perpetual, high-stakes performance of their own personalities.

Smartphones have even hurt kids who don’t use them much, according to Haidt, because they’ve restructured communal life in harmful ways. Teenagers’ rates of anxiety, depression, self-harm, and suicide have all skyrocketed as a result.

The author’s argument has resonated with many parents, and The Anxious Generation has topped the New York Times Best Sellers list in nonfiction. But his theory has also attracted no small number of critics.

For years, prominent psychologists have been accusing Haidt of fueling a moral panic. In their view, there is “no evidence that using these platforms is rewiring children’s brains or driving an epidemic of mental illness.” They insist that Haidt’s case against smartphones and social media apps owes less to the available data than to humanity’s perennial anxieties about new technologies and kids these days.

This dispute has generated reams of commentary as the media’s middle-aged social media users argue about whether their favorite apps are poisoning their kids. But few if any of these pieces have closely scrutinized how Haidt’s evidence stacks up against that of his detractors — and when one digs into these competing data points, the fundamental premises of the smartphone debate are thrown into question. In truth, it’s not entirely clear that there even is an international decline in teen mental health that requires explanation.

Ultimately, both Haidt and his critics overstate their evidence. The former’s case isn’t strong enough to prove that iPhones “destroyed” Gen Z, but it also isn’t so weak that it can be dismissed as the mere byproduct of a moral panic.

How smartphones are (supposedly) warping the minds of young people throughout the wealthy world

The foundational premise of Haidt’s argument is that something has been eroding the psychological well-being of young people over the past 14 years. The Anxious Generation substantiates this claim with a barrage of grim data points:

• The percentage of US teens who say that they’ve had one “major depressive episode” in the past year has increased by more than 150 percent since 2010, with the bulk of that increase coming before the Covid-19 pandemic.
• The share of US undergraduates with a diagnosed anxiety disorder increased by 134 percent between 2010 and 2020, while those with a depression diagnosis increased by 106 percent.
• Among American girls between the ages of 10 and 14, emergency room visits for self-harm grew by 188 percent during that period, while deaths by suicide increased by 167 percent.
• Among boys of the same age, ER visits for self-harm increased by 48 percent and suicide by 91 percent.
• The suicide rate among US teens between the ages of 15 and 19 also increased over this time period, though seemingly to a lesser degree.

But this crisis is not confined to the US, Haidt says.

Canadian teens’ self-reported mental health also started nosediving around 2012, and their rates of emergency room visits due to self-harm jumped as well. And adolescents in Britain, Australia, New Zealand, the Nordic countries, and most of Protestant Europe followed a similar trajectory.

According to Haidt, one international survey of 15-year-olds found that feelings of loneliness at school spiked in every world region but Asia between 2012 and 2015. Notably, in almost all cases, declines in mental health were much more pronounced among girls than boys.

From this, he concludes that something rewired young people’s brains beginning in the early 2010s, all across the wealthy world.

The rollout of smartphones fills the bill. Haidt notes that by 2013, a majority of US households owned an iPhone, an Android phone, or their like. And smartphones became similarly ubiquitous in many other rich nations shortly thereafter.

So, his prime suspect was in the right places at the right time. And he insists that the smartphone left its fingerprints all over anxious Zoomers’ brains.

His case for the phone’s culpability rests on four main pieces of evidence.

1) In surveys of adolescents, respondents who heavily use social media are more likely to report anxiety and depression than those who use the apps less. Haidt acknowledges that this correlation does not appear in every survey. But he insists that there is a broad consensus across studies that heavy social media use correlates with anxiety and depression, and that this association is especially strong for girls.

This indicates that there is an association between social media use and worse mental health outcomes. But it does not tell us that the former causes the latter. After all, depression might lead young people to develop Instagram addictions, rather than vice versa.

But Haidt’s second piece of evidence speaks to causality.

2) When researchers instructed a random sample of college students to abstain from social media for a period of time, those students tended to report improved psychological well-being (relative to themselves at the start of the study and/or to a control group). Conversely, when researchers forced college students to consume social media in a laboratory setting, they tended to report worse mental health. For example, when a team of psychologists assigned a group of college women to scroll Instagram for seven minutes, they expressed lower satisfaction with their bodies and a more negative emotional state afterward than those who were assigned to a control group.

Of course, laboratory experiments measure artificial conditions, and they only gauge the effect that social media has directly on an individual. The latter is a significant limitation for Haidt’s purposes, since he believes that the mass adoption of social media may be most damaging for its indirect effects on entire communities: In a school where virtually all young people spend the bulk of their leisure time online, a kid who doesn’t use social media apps will still suffer from their consequences. For example, such a student will have fewer opportunities to socialize with classmates in the real world than they would’ve had in the pre-smartphone era.

But Haidt has a third piece of evidence that arguably captures smartphones’ impact on collective life.

3) When communities gain access to high-speed internet — and, thus, better access to social media — their teens’ psychological well-being tends to decline. The Anxious Generation cites five studies that found when fiber optic cables arrived in an area, local rates of teen mental health diagnoses and hospitalizations started rising.

Finally, Haidt casts the gender gap in teen mental health as a fourth piece of evidence.

4) Girls are more vulnerable to social media’s harms and have registered a larger increase in mental health problems since 2012 than boys. Haidt substantiates that first claim by noting that, relative to boys, girls are more prolific users of social media, more likely to express aggression by trying to destroy their ex-friends’ reputations and relationships (a task that social media greatly facilitates), and more vulnerable to body image disorders (which platforms like Instagram tend to exacerbate).

He weds this empirical evidence to a theory of how smartphones and social media undermine teens’ mental health.

He notes that phones now claim most of young people’s leisure time, leaving them little room for myriad activities that are indispensable for healthy development. According to Gallup, American teens spend an average of almost five hours a day on social media platforms. When you account for streaming, video games, and other screen-based activities, adolescents are devoting nearly nine hours a day to digital media consumption, as the Common Sense Census found.

That screen time comes at the expense of sleep, exercise, and spending face-to-face time with friends, all of which are associated with mental well-being and healthy child development.

The case against the case against smartphones

Haidt’s theory has attracted many critics, who have collectively raised objections to virtually every premise in his narrative. Their case can be distilled into four arguments.

1) It’s not clear that there has actually been an international increase in mental illness among young people. There is no question that adolescent suicide rates have risen in the United States in recent years. But, as Stetson University psychologist Christopher Ferguson told Vox, America’s recent suicide increase is not a phenomenon specific to teens — suicide has been increasing in the US among virtually all age groups.

Across the Western world more broadly, there was no uniform increase in teen suicide during the 2010s in the World Health Organization’s data. In fact, between 2012 and 2019, the suicide rate among people aged 15 to 19 fell by 36 percent in France; 20 percent in Denmark; 64 percent in Ireland; 59 percent in New Zealand; 11 percent in Norway; 28 percent in Spain; and 36 percent in Belgium. Across Europe as a whole, the suicide rate among older teens fell by 26 percent in this period.

This is a big problem for Haidt’s thesis, according to his skeptics, because suicide rates are the most reliable indicator of mental health trends. Most other data on psychological well-being can be distorted by changes in social norms.

We have witnessed globe-spanning campaigns in recent decades to destigmatize mental health problems. At the same time, many clinicians have expanded the definition of various psychological disorders, a process that critics dub “diagnostic inflation.”

For this reason, just because teens in a given country have become more likely to say that they feel “anxious” or “depressed” — or be diagnosed with a mood disorder — that does not necessarily mean that their actual subjective experience has grown worse. Rather, those with mood disorders may be more likely to know that they suffer from such a condition or more likely to admit that they do. Meanwhile, some teens who suffer from sadness or stress that does not rise to a clinical level may be more inclined to conceptualize their feelings in clinical terms.

Data on hospital visits for self-harm, suicidal ideation, and mental health problems are vulnerable to similar distortions, University of Oxford psychologist Andrew Przybylski told Vox.

This is because hospital systems’ recordkeeping protocols can change over time. In 2015, the International Classification of Diseases (ICD) — a World Health Organization guide that instructs hospitals how to code diagnoses in official records — implemented a new edition, which recommended multiple major changes to coding practices.

For example, before 2015, a clinician needed to enter two separate codes in order to log a case of self-harm: They needed to first enter the type of injury (such as a Xanax overdose) and then separately mark whether the injury was intentional or accidental. Clinicians often neglected to enter the second code, according to researchers, which led to an undercount of self-harm incidents.

After the new ICD was adopted in October 2015, however, clinicians were able to record an intentional benzodiazepine overdose with a single code. Recorded instances of self-harm promptly spiked in many health systems. But this was a consequence of the coding changes rather than a reflection of any actual increase in cases.

The ICD’s revisions affect hospitals worldwide and have distorted some data on adolescent mental health. Przybylski noted that New Jersey hospitals saw a large spike in suicide-related visits among young people during the 2010s — a development that, on its face, might seem to support Haidt’s thesis.

And yet, despite this apparent jump in youth suicidality, the actual teen and child suicide rates in New Jersey were essentially flat over the same period. When researchers from Princeton and North Carolina State University investigated this disparity, they concluded that the spike in suicide-related visits was largely the product of coding changes.

In view of these measurement distortions and the falling youth suicide rates in Europe, Haidt’s critics argue that it’s unclear whether there even is a deepening teen mental health crisis that requires explanation.

2) The survey data doesn’t actually support Haidt’s thesis. Yes, there is a correlation between self-reported social media use and poor mental health in some surveys of adolescents, Haidt’s critics acknowledge. But taken together, they argue, the empirical literature shows a weak to nonexistent association between screen time and mental health. Further, in their view, Haidt’s own cherry-picked figures imply that social media use only explains about 15 percent of the variation in mental health between teens, meaning that other factors must account for the other 85 percent.

3) Haidt’s experimental evidence is weak and unreliable. To prove that social media causes mental distress, Haidt points to experiments in which researchers prohibited a random sample of college students from using social media for weeks and then evaluated their mental health. In a review of such experiments that will soon be published in the journal Psychology of Popular Media, Stetson’s Christopher Ferguson found that “evidence for causal effects was statistically no different than zero.”

Ferguson’s paper acknowledges that some experiments have produced results consistent with Haidt’s hypothesis. But he argues that these studies are generally plagued by methodological problems.

For one thing, many test subjects never report back to researchers after their prescribed period of abstinence. This could bias the studies’ results: A student who found that abstaining from social media was bad for their mental health might be more likely to drop out of a study than a student who benefited psychologically from logging off.

4) Haidt’s natural experiments are contradicted by better data. Haidt’s skeptics are similarly unimpressed by his citation of quasi-experiments in which the spread of broadband internet was linked to teen mental health problems. After all, there are only five of these studies, and each examines only a small subset of the developed world’s population.

Last year, Oxford’s Przybylski and Tilburg University researcher Matti Vuorre decided to see whether the correlation between high-speed internet access and worse teen mental health held up at scale. In their study, they examined changes in mobile broadband subscriptions and teen mental health outcomes in 202 countries over 19 years — and found that access to the internet (and, thus, social media) was not consistently linked with negative psychological outcomes.

“There’s nothing here that isn’t present in any of the past panics about video games, Dungeons & Dragons, or silent movies,” Przybylski told Vox. “Each of these, you have a new technology, a vulnerable group and a new mechanism. It’s always ‘This time it’s different,’ but there’s nothing in these claims that actually distinguishes it in terms of scientific evidence.”

So, are smartphones really poisoning Gen Z’s brains?

I’m not a social psychologist, nor a parent of a TikTok-obsessed tween. So I shouldn’t be trusted as the final arbiter of this dispute.

But in my view, the case for believing that smartphones are undermining teens’ mental health is significantly weaker than Haidt suggests — yet stronger than his most ardent skeptics allow.

It’s not clear that an international crisis of teen mental health even exists.

In its account of a global surge in adolescent distress, The Anxious Generation engages in a bit of cherry-picking. Haidt’s metric for whether a given country has witnessed a decline in teen mental health since 2012 shifts from one nation to another.

In the United States, he holds up an increase in teen suicides as evidence of a deepening crisis. In the Nordic countries, he cites survey data showing elevated rates of psychological distress but does not mention that teen suicide declined in Denmark, Finland, and Norway in the wake of the smartphone revolution.

Nor does Haidt’s book grapple with the fact that multiple international surveys of young people’s life satisfaction and well-being have failed to detect any clear decline since 2012.

Even Haidt’s own preferred international survey shows no signs of an increase in adolescent distress in Asia, the world’s most populous region. Given that Asians are 1) human beings with the same child development needs as everyone else and 2) prolific users of social media (at least in many national contexts), it is unclear how Haidt reconciles this survey finding with his thesis.

Ultimately, the absence of a clear pattern in global suicide data — combined with the various changes to diagnostic criteria and coding guidelines that could be inflating mental health statistics — should temper Haidt’s claims about the post-2012 “surge of suffering” more than they have.

To say that Haidt’s evidence for a global decline in teen mental health isn’t conclusive is not to say that it’s insignificant.

Haidt acknowledges that changes in social norms and coding practices may have contributed to the apparent increase in adverse teen mental outcomes over the past decade. But his research assistant Zach Rausch pointed out to me that, at least in the US, the uptick in adolescent cases of self-harm began years before the ICD’s coding changes.

More broadly, it remains the case that a wide variety of alarming indicators — from mental health hospitalizations to anxiety and depression diagnoses — did abruptly increase starting in the early 2010s in a wide range of Western countries. And in virtually every case, the increase was significantly larger among girls than among boys. If we were looking at random noise, we would not expect to see this largely uniform gendered pattern.

However, the empirical evidence linking social media to teen mental distress is much weaker than Haidt suggests.

Haidt acknowledges that the association between social media use and mood disorders in survey data appears small on its face. But he maintains that surveys are flawed tools for capturing the harms of social media in any case.

He told me that smartphones have hurt teen mental health in a wide range of different ways — some that have primarily affected boys, some girls, and some both. Analyzing the correlation between heavy social media use and depression in survey data might illuminate one of these mechanisms: the way that social media addiction corrodes mental well-being.

But, Haidt explains, you don’t need to be a heavy user of social media to suffer a public shaming, fall prey to Instagram-induced body dysmorphia, or have few opportunities for outdoor play because everyone else is staring at their phones. Thus, correlational studies don’t tell the full story.

“It’s kind of like if someone was trying to explain obesity growth in the 1980s, and they were to argue that, well, we measured french fry consumption, and there’s a bit of an effect, but it really isn’t enough to explain everything,” Haidt said. “Then I would say, well, okay, a part of it is french fries — we do eat a lot of french fries. But what about corn sweetener?”

This is a reasonable argument. But it also amounts to an admission that the correlational studies don’t do all that much by themselves to substantiate Haidt’s thesis.

Meanwhile, Ferguson’s paper persuasively argues that experimental studies on social media and mental health have collectively produced mixed results and suffer from major methodological flaws.

The natural experiments that Haidt cites are more compelling. But they should be interpreted cautiously in light of Przybylski and Vuorre’s inability to detect any strong association between broadband subscriptions and teen mental health internationally in their 2023 study.

Taken together, it seems clear that Haidt’s rhetoric is much stronger than his evidence.

I still suspect the phones are bad

All this said, I can’t bring myself to accept Przybylski’s view — that there is no reason to believe that social media is harming teen’s mental health, and that claims to the contrary merely reflect an age-old cycle of moral panics about new technologies.

In my view, “It is not healthy for kids to spend five hours a day staring at social media feeds that invite negative social comparisons, reward pile-ons, and induce addiction by design” strikes me as a pretty reasonable default assumption — not least because I probably spend about that much time staring at X (formerly Twitter) daily and surely suffer mentally as a result.

Granted, unlike me, the typical tween probably doesn’t spend much time sparring with @stalinlover69 and @auschwitzenthusiast1488 while their girlfriend says things like “Please put the phone down” and “Is this really how you’re going to spend our anniversary?”

But it still seems doubtful to me that kids are benefiting from spending less time hanging out with close friends in person and more time performing their selves on social media platforms in an often invidious competition for approval and attention.

Haidt’s work on teen mental health might be biased by his preexisting contempt for social media; he had initially intended to write a book lamenting Twitter’s impact on our politics before shifting his focus to Gen Z’s psychological woes. But his critics may have biases of their own. Ferguson and Przybylski have both spent much of their careers debunking alarmist theories about video games’ corruption of the youth and are inclined to view the discourse around social media through that lens.

In our conversation, Przybylski said he doubted that using social media shortens people’s attention spans. To me, this is a bit like doubting that chewing broken glass causes oral discomfort. And I imagine most of my fellow heavy X users would agree.

Definitively determining how social media affects adolescent mental health is going to take more time and research. Fortunately, there’s plenty we can do to help psychologically distressed teens in the interim, whether or not they log off.