On the associations between adolescent social media use and health outcomes: An exploratory specification curve analysis and comparison with other concurrent predictors.
While most model specifications indicated a negative association between social media and health outcomes, most associations fall below accepted thresholds for reliability or clinical significance, and time on social media emerged as one of the least influential predictors of adolescent health compared to other relevant factors.
Key Findings
Results
Most associations between social media use and adolescent health outcomes were negative but fell below accepted thresholds for reliability or clinical significance.
Study used specification curve analysis, a computational method designed to reduce bias from selective analytical choices
Four health outcomes were examined: psychological well-being, mental health, health behaviours, and risk behaviours
Sample consisted of 15- and 16-year-olds in the west of Ireland (N = 2876)
Most associations were described as 'indistinguishable from statistical noise'
The paper notes findings remain inconclusive 'largely due to methodological and conceptual oversights' in the broader literature
Results
Some associations between social media use and health outcomes were positive, including the association between time on social media and vigorous physical activity for boys.
The positive association with vigorous physical activity was specific to boys
This finding illustrates that the direction of associations between social media and health is not uniformly negative
The specification curve analysis revealed variation across model specifications in both direction and magnitude of associations
Results
Associations between social media use and anger management and alcohol consumption were among the few that reached accepted thresholds for reliability.
The paper explicitly singles out anger management and alcohol consumption as exceptions to the general pattern of non-significant associations
These two outcomes were described as meeting accepted thresholds for reliability unlike most other associations examined
All other associations were characterized as potentially not clinically relevant at the population level
Results
Time spent on social media was one of the least influential predictors of adolescent health when compared to other concurrent predictors.
The study compared social media use against 'other relevant factors' as predictors of adolescent health outcomes
Social media use ranked among the least influential of the predictors examined
The comparison was conducted across all four health outcome domains: psychological well-being, mental health, health behaviours, and risk behaviours
This finding directly challenges narratives positioning social media as a primary driver of adolescent health outcomes
Results
Overall effects of social media use on adolescent health at the population level are small and may not be clinically relevant.
The study explicitly distinguishes between statistical reliability and clinical significance
The authors conclude that 'some specific associations between social media and adolescent health are statistically reliable' but that 'overall effects at the population level are small'
The sample was drawn from 15- and 16-year-olds (N = 2876) in the west of Ireland
Specification curve analysis was used to address the problem of selective analytical choices that may have produced conflicting findings in prior research
Methods
Specification curve analysis was applied to examine associations between social media use and health outcomes across a range of analytical specifications to reduce researcher bias.
Specification curve analysis is described as 'a computational method which reduces bias from selective analytical choices'
The method addresses methodological and conceptual oversights identified in the existing literature on social media and adolescent health
The analysis covered four distinct but related health outcomes: psychological well-being, mental health, health behaviours, and risk behaviours
The study population consisted of N = 2876 adolescents aged 15 and 16 in the west of Ireland
What This Means
This research suggests that the relationship between how much time teenagers spend on social media and their health is much weaker than public debate typically implies. Using a rigorous statistical approach called specification curve analysis—which tests hundreds of different ways of analyzing the same data to avoid cherry-picking results—researchers examined over 2,800 teenagers aged 15 and 16 in Ireland. While most analyses did find a slight negative link between social media use and health outcomes like mental well-being, most of these associations were so small they were essentially indistinguishable from random noise and unlikely to be meaningful in real-world terms.
The study also found that social media use was actually one of the weakest predictors of adolescent health when compared to other factors examined at the same time. Only two specific outcomes—anger management and alcohol consumption—showed associations with social media use strong enough to be considered reliable. Interestingly, for boys, more time on social media was actually associated with more vigorous physical activity, showing the relationship is not always negative.
This research suggests that the intense public and policy focus on social media as a major driver of adolescent health problems may be disproportionate to the actual evidence. While social media is not entirely harmless, the data indicate its effects are generally small at the population level and that other factors in teenagers' lives likely matter more for their health and well-being. The findings highlight the importance of using rigorous analytical methods that account for the many different ways data can be analyzed, since prior conflicting findings in this area may partly reflect methodological choices rather than true differences in effects.
Whelan E. (2026). On the associations between adolescent social media use and health outcomes: An exploratory specification curve analysis and comparison with other concurrent predictors.. Acta psychologica. https://doi.org/10.1016/j.actpsy.2026.106783