Reported rates of anxiety and depression among adolescent girls have risen across many countries over the past fifteen years. The rise is larger for girls than boys in most datasets, and it is consistent enough across measures and countries to be taken seriously.
What produced it is contested, and the contest is not merely academic — different explanations imply different responses.
Separating the possible explanations
At least four things could produce a rise in reported rates, and they are not mutually exclusive.
A genuine increase in the underlying condition. More adolescents are experiencing clinically significant symptoms.
Increased recognition and help-seeking. The same underlying rate, better detected, because stigma has fallen and services have expanded.
Changed thresholds and vocabulary. Experiences previously described as stress or unhappiness are now described in clinical terms, both by adolescents and by clinicians.
Measurement change. Different instruments, different survey methods, different sampling.
Distinguishing these matters enormously. If the rise is mostly recognition, the appropriate response is service capacity. If it is mostly a genuine increase, the appropriate response is to identify what changed.
The evidence that it is at least partly real
The strongest argument against a pure recognition effect comes from measures that do not depend on self-report or diagnosis.
Rates of self-harm presentations to emergency services have risen in several countries, with the steepest rises among adolescent girls. Presentations of this kind are less sensitive to changes in vocabulary and stigma than survey responses.
Hospital admission data shows similar patterns in several jurisdictions.
These measures have their own problems — service availability affects presentation rates — but they are less susceptible to the recognition explanation than questionnaire data.
The social media debate, fairly stated
This is the most prominent proposed explanation and the most contested.
The case for: the timing of the rise corresponds reasonably well with the adoption of smartphones and image-based social platforms; the gender pattern matches, with girls being heavier users of the platforms most implicated; and the mechanism — appearance comparison, social exclusion made visible, displacement of sleep — is plausible and partially supported.
The case against: large-scale analyses of the association between screen use and wellbeing have generally found effects that are small in magnitude, comparable in size to associations with things nobody considers a crisis; longitudinal studies have produced mixed results on direction of causation; and cross-country comparisons find the rise in some places with different technology adoption patterns.
The most defensible current position is that there is a real association, that it is small on average, that it is considerably larger for some individuals than the average implies, and that the causal direction is not fully established.
Anyone stating this more confidently in either direction is going beyond the evidence.
Other candidate explanations
Sleep. Adolescent sleep duration has declined, insufficient sleep is causally linked to mood disturbance in experimental studies, and school start times are demonstrably misaligned with adolescent circadian biology. This is among the better-evidenced contributors and receives less attention than social media.
Academic pressure and the structure of assessment. Reported academic stress has risen in several countries.
Reduced unsupervised time and independent mobility, which have declined substantially over the same period in many countries.
Broader economic and political conditions, which affect adolescents' expectations of their own futures.
Most likely several of these operate together, which is unsatisfying but is what the evidence supports.
Why the gender difference
The larger rise in girls is consistent across datasets and its explanation is not settled.
Proposed factors include differential exposure to appearance-focused platforms, greater sensitivity to relational and social exclusion stressors, earlier pubertal timing, and differences in how distress is expressed and therefore detected.
The last of these deserves a note: internalising symptoms are more readily identified as mental health problems than externalising ones. Some of the gender difference in measured rates may reflect a difference in how distress presents rather than in how much exists, and boys' distress presenting as conduct problems or substance use is captured in different statistics.
What this means practically
Two things follow that are less contested than the causes.
First, service capacity is inadequate in most systems. Waiting times for adolescent mental health services are long in most countries with published data, and this is true regardless of which explanation for rising rates is correct.
Second, the interventions with the best evidence — discussed in the article on what works — are not new or speculative. Sleep, activity, structured therapy for those who need it, and reducing exposure to specific identified stressors all have better evidence than any of the more dramatic proposals.
The framing risk
There is a real cost to over-medicalising ordinary adolescent distress.
Adolescence involves genuine difficulty, and the ability to experience and tolerate difficulty without it becoming a clinical matter is developmentally important.
Research on some awareness campaigns has raised the possibility that broad messaging can increase symptom reporting without improving outcomes, by encouraging clinical interpretation of ordinary distress.
This is not an argument for dismissing adolescent mental health. It is an argument for precision — that a system which treats everything as clinical will have less capacity for what actually is, and that being unhappy about something genuinely bad is not a symptom.