What Is the Relationship Between Social Media Use and Adolescent Well-Being?

Singh, Zhou, Curtis, Maher, and Dumuid (2026) examined how after-school social media use is associated with adolescent well-being across development. They asked whether the relationship is nonlinear and whether it differs by age and sex. The study analyzed data from the South Australian Wellbeing and Engagement Collection, including 100,991 students (173,533 observations) from grades 4–12 between 2020 and 2022. They found a U-shaped pattern: moderate users generally showed the best well-being. Compared with moderate use, the highest use increased odds of low well-being (e.g., OR ≈ 3.13 for girls and 2.25 for boys in grades 7–9). Nonuse also raised risk in later adolescence (OR ≈ 1.79 for girls; 3.00 for boys).

Why This Article Was Selected for The Policy Scientist

This article addresses a central policy question: how digital environments relate to adolescent well-being across developmental stages. The topic has broad relevance for education systems, public health, families, and youth services, particularly as social media saturation continues to rise. The study is timely given ongoing debates over screen time guidelines and youth mental health trends. The authors have an extensive publication record in youth well-being and movement behaviors, situating this work within a mature research program. The dataset is unusually large and population-based, strengthening measurement stability. Findings are plausibly generalizable to similar school-based populations, though cross-national transferability warrants caution. Mixed-effects models are appropriate; future work would benefit from causal inference designs or randomized trials.

Full Citation and Link to Article

Singh, B., Zhou, M., Curtis, R., Maher, C., & Dumuid, D. (2026). Social media use and well-being across adolescent development. JAMA Pediatrics. Advance online publication. https://doi.org/10.1001/jamapediatrics.2025.5619

Central Research Question

This study investigates how after-school social media use is associated with adolescent well-being across developmental stages. The authors focus on whether this relationship is linear or nonlinear and whether it varies systematically by sex and school grade. Rather than assuming that more use is uniformly harmful, the analysis evaluates three usage categories—none, moderate, and highest—to determine whether well-being outcomes differ across the full spectrum of engagement. The central question is therefore twofold: (1) What is the shape of the association between after-school social media use and well-being? and (2) Do these associations differ by age and sex during adolescence?

Previous Literature

Prior research on social media and adolescent mental health has produced mixed conclusions. Much of the early evidence relied on cross-sectional designs, which consistently reported that heavier social media use correlated with poorer psychological outcomes, including depressive symptoms, anxiety, reduced life satisfaction, and sleep disturbance. However, cross-sectional studies cannot establish directionality, leaving unresolved whether social media use contributes to poorer well-being or whether adolescents experiencing distress are more likely to increase their digital engagement.

Longitudinal studies have offered a more nuanced picture. Several cohort investigations have found weak or inconsistent predictive effects of social media use on later mental health after adjusting for baseline characteristics. Some studies instead observed reverse or reciprocal associations, where poorer mental health predicted increased use. Other research highlighted that problematic or addiction-like use, rather than time spent per se, was more strongly associated with declines in mental health. These findings suggested that the relationship is dynamic, heterogeneous, and potentially contingent on individual susceptibility factors.

Sex differences have emerged as one of the most stable patterns. Girls often report stronger negative associations between high social media use and well-being, possibly reflecting differences in platform preferences, social comparison processes, and feedback-seeking behaviors. Developmental stage also appears relevant. Evidence indicates that associations may emerge earlier for girls than boys, and that vulnerabilities may shift across adolescence. Despite these advances, gaps remained: few studies span the full adolescent age range, many assume linear exposure effects, and relatively few explicitly compare nonusers with users. This study addresses those limitations by examining nonlinear associations across grades 4 through 12.

Data

The analysis draws on the South Australian Wellbeing and Engagement Collection (WEC), an annual census of students in government schools. The WEC combines cross-sectional breadth with a cohort-sequential structure, enabling repeated observations for a subset of students across years. The study uses pooled data from 2020 to 2022, with participants contributing up to three waves. The final analytic sample includes 100,991 unique students and 173,533 observations.

Well-being is operationalized as the mean score across eight validated domains: happiness, optimism, life satisfaction, worry, sadness, perseverance, emotional regulation, and cognitive engagement. These indicators are adapted from established psychometric instruments and scored on Likert scales. Sadness and worry are reverse coded so higher scores indicate better well-being. The composite measure is dichotomized into high versus low well-being using a threshold consistent with prior WEC reporting standards.

After-school social media use is self-reported and calculated based on frequency and typical duration between 3 PM and 6 PM on weekdays. Weekly usage estimates are categorized into none (0 hours), moderate (>0 to <12.5 hours/week), and highest (≥12.5 hours/week). Sociodemographic covariates include sex, school grade, parental education, region of residence, and calendar year.

Methods

The authors employ mixed-effects generalized linear models with binomial outcomes to estimate associations between social media use categories and low well-being. Models include random intercepts at both the school and participant levels to account for clustering and repeated measures. Moderate use serves as the reference category, enabling direct comparisons with nonuse and highest use.

Analyses are stratified by sex and grouped into developmental stages: senior primary (grades 4–6), junior secondary (grades 7–9), and senior secondary (grades 10–12). Cross-sectional models pool observations across years while adjusting for grade, parental education, region, and calendar year. Developmental trend analyses introduce interaction terms between social media use and grade, allowing predicted probabilities of low well-being to vary across age.

Sensitivity analyses assess robustness. These include models estimated separately by grade and year, models using each individual well-being domain as an outcome, cohort-specific longitudinal analyses, and supplementary regressions using continuous measures of well-being and usage duration with polynomial terms to capture potential nonlinearity.

Findings/Size Effects

Across analyses, the association between after-school social media use and well-being is consistently nonlinear. A U-shaped pattern emerges, where moderate users generally display the lowest probability of low well-being, while both nonusers and highest users show elevated risk. This pattern appears across sexes but varies by developmental stage.

Highest social media use is associated with substantially greater odds of low well-being relative to moderate use. In grades 7–9, the odds ratios are particularly pronounced, reaching approximately 3.13 for girls and 2.25 for boys. These effects diminish somewhat in later grades but remain statistically significant. The magnitude indicates that adolescents reporting the highest usage levels have two- to threefold higher odds of low well-being compared with moderate users.

Nonuse presents a more complex pattern. Among girls in younger grades, nonusers do not differ significantly from moderate users. However, in grades 10–12, nonuse is associated with higher odds of low well-being (OR ≈ 1.79). Among boys, nonuse is associated with elevated odds across nearly all grade groupings, with the strongest effects in senior secondary school (OR ≈ 3.00). Developmental trend analyses show that for boys, nonuse becomes increasingly associated with poorer well-being beginning around grade 7 and eventually exceeds the risk associated with highest use.

Predicted probability plots illustrate distinct trajectories. For girls, the highest use category consistently corresponds to the greatest probability of low well-being, peaking during middle adolescence. Moderate use becomes increasingly favorable from grade 8 onward. For boys, moderate use remains associated with the lowest probabilities across grades, while nonuse shows a steep increase in predicted low well-being during mid-to-late adolescence.

Continuous models corroborate the categorical findings. Polynomial regressions reveal curvilinear associations, with optimal well-being occurring at intermediate levels of use (approximately 4 hours/week for girls and 6 hours/week for boys). These results reinforce the interpretation that both minimal and excessive engagement are associated with less favorable outcomes.

Conclusion

The study concludes that after-school social media use exhibits a complex, nonlinear association with adolescent well-being. Moderate use is generally associated with the most favorable outcomes, while both highest use and nonuse are linked to elevated risk, depending on sex and developmental stage. Highest use shows consistently adverse associations, particularly during middle adolescence. Nonuse is not uniformly protective and is associated with poorer outcomes in later adolescence, especially among boys.

The findings challenge simplified linear models and underscore the importance of developmental context. Associations vary across age and sex, suggesting that risks and benefits of social media engagement are not static. The authors emphasize that results are observational and should be interpreted cautiously, noting that unmeasured confounding and self-report bias remain possible. They suggest that future research employing causal inference frameworks or experimental designs would strengthen the evidence base and clarify mechanisms underlying these associations.