Published in Nature Human Behaviour in 2019, Oxford researchers Amy Orben and Andrew Przybylski applied specification curve analysis (running all possible analytical combinations rather than cherry-picking one) to three large-scale datasets totaling 355,358 adolescents.

The association between digital technology use and adolescent well-being was negative but extremely small, explaining at most 0.4% of the variation in well-being. To put this in perspective, the researchers compared it to other factors: regularly wearing glasses had a larger negative association with well-being than social media use, and eating potatoes had a similar-sized negative correlation.

The key insight was that researchers studying this topic can obtain widely varying results depending on which analytical choices they make (which variables to control for, how to measure screen time, which well-being outcome to use). The specification curve approach showed that across all reasonable analytical choices, the effect was consistently near zero, suggesting that alarming findings in individual studies may reflect selective analysis rather than real effects.

Published via Cambridge University Press, Stanford communication researcher Jeff Hancock and colleagues conducted a comprehensive meta-analysis of 226 studies involving 275,728 participants examining the relationship between social media use and various well-being outcomes.

The overall association between social media use and a combined measure of well-being was r = 0.01, with a 95% confidence interval crossing zero (-0.02 to 0.04), meaning it was not statistically significant. None of the specific associations with depression, anxiety, loneliness, or life satisfaction exceeded r = 0.20. The amount of variation in well-being attributable to social media use was described as "essentially zero."

The meta-analysis included both cross-sectional and longitudinal studies across multiple countries. The authors noted that the near-zero average effect does not mean no individual is harmed, but it does mean that social media use is not a meaningful predictor of mental health at the population level, which is what the claim of "causing" teen depression requires.

Published in Clinical Psychological Science in 2019, Canadian researchers tracked two samples longitudinally: 594 adolescents (mean age 12.21, surveyed annually for 2 years) and 1,132 university students (mean age 19.06, surveyed annually for 6 years).

Social media use at one time point did not predict depressive symptoms at the next time point for males or females in either sample. However, the reverse was true for adolescent girls: greater depressive symptoms at one time point predicted more frequent social media use later. This pattern is consistent with reverse causation, meaning depressed teens turn to social media as a coping mechanism rather than social media making them depressed.

This finding directly challenges the causal direction assumed in most alarming studies. If the arrow runs from depression to social media use (rather than the reverse), then reducing social media access would not reduce depression and could even remove a coping outlet for vulnerable teens.

Published in Computers in Human Behavior in 2020, Brigham Young University researchers followed 500 adolescents from age 13 to 20, surveying them annually for eight years about their social media use, depression, and anxiety.

At the between-person level (comparing heavy users to light users), heavy social media users were more depressed on average. However, at the within-person level (tracking the same individual over time), increases in an individual''s social media use did not predict subsequent increases in their depression or anxiety.

This distinction is critical. Between-person differences can reflect pre-existing factors (personality, family environment, socioeconomic status) that make some people both more depressed and more likely to use social media. Only within-person changes can suggest causation, and those showed no effect. The eight-year timeframe is one of the longest in the literature, providing substantial statistical power to detect effects if they existed.

Published in the Journal of Child Psychology and Psychiatry in 2020, UC Irvine developmental psychologist Candice Odgers and colleague Michaeline Jensen reviewed the accumulated evidence on adolescent digital technology use and mental health.

They concluded that "recent narrative reviews and meta-analyses do not support a strong linkage between the quantity of adolescents'' digital technology engagement and mental health problems." The review found that offline vulnerabilities (poverty, family dysfunction, pre-existing mental illness) tend to mirror and shape online risks, rather than digital technology creating new problems from scratch.

The authors specifically stated that "previous fears about screen time and mental health are not supported by the science" and warned that the moral panic around social media risks diverting attention and resources from known drivers of adolescent mental health problems such as poverty, adverse childhood experiences, and inadequate access to mental healthcare.

Published in Cyberpsychology, Behavior, and Social Networking in 2017, researcher Chiungjung Huang conducted a meta-analysis examining the relationship between time spent on social networking sites and various measures of psychological well-being.

The mean correlation between time on social media and well-being was r = -0.07, which is classified as a very weak effect in social science. Correlations with depression and loneliness were similarly small. An effect of r = 0.07 means social media use explains less than 0.5% of the variation in well-being.

To understand how small this is: by standard benchmarks in psychology, an r of 0.10 is considered small, 0.30 medium, and 0.50 large. An r of 0.07 falls below even the "small" threshold. At this effect size, thousands of other factors in a teenager''s life (family relationships, school performance, physical health, friendships) would each individually outweigh social media as a predictor of their mental health.

Published in Psychological Science in 2017, Oxford researcher Andrew Przybylski and Netta Weinstein analyzed data from 120,115 English adolescents aged 15 in a preregistered study (meaning the analysis plan was locked before looking at results, preventing cherry-picking).

The relationship between screen time and well-being was not linear but curvilinear (inverted U-shape). Moderate use was associated with higher well-being than no use at all. Negative effects appeared only at high levels, with turning points varying by activity: approximately 1 hour 40 minutes for gaming, 1 hour 57 minutes for smartphones, 3 hours 41 minutes for watching videos, and 4 hours 17 minutes for computer use on weekdays.

The finding that moderate social media use is associated with better well-being than zero use challenges the narrative that social media is inherently toxic. It suggests that digital technology provides genuine social benefits and that the concern should be focused on excessive use patterns rather than social media as a category.

Published in the Journal of Communication in 2021, University of Amsterdam researcher Patti Valkenburg and colleagues conducted a preregistered experience-sampling study of 387 adolescents aged 13-15, texting them 6 times per day for 3 weeks, producing 34,930 total assessments.

Using person-specific statistical modeling (N=1 Dynamic Structural Equation Modeling, which estimates a separate effect for each individual), they found that 88% of adolescents experienced no or very small effects of social media use on self-esteem. Only 4% experienced positive effects and 8% experienced negative effects.

This study revealed that population-level averages (the basis for most alarming claims) mask enormous individual differences. The finding that social media negatively affects only about 1 in 12 teens means that blanket statements like "social media causes teen depression" are misleading. The challenge for researchers and policymakers is identifying the vulnerable 8% rather than restricting a technology that is neutral or beneficial for 92% of its young users.