Associations between youth's daily social media use and well-being are mediated by upward comparisons

The study addresses ongoing debates about whether social media use is detrimental, inconsequential, or beneficial for youths’ well-being. Prior evidence is heterogeneous and largely based on cross-sectional, between-person designs. Responding to calls for within-person research and mechanistic tests, the authors examined daily associations between use of Instagram, TikTok, and YouTube and four indicators of subjective well-being (positive/negative self-worth; positive/negative affect) in children aged 10–14. The study tested upward social comparisons—youths’ impressions that others are better off—as a mediator, and explored moderators (sex, social media self-control failure, and social comparison orientation). The rationale for focusing on ages 10–14 includes early initiation of social media, the developmental salience of identity and social comparisons, and evidence that younger adolescents may be especially susceptible to media effects. The key hypotheses were that higher social media use would be associated with poorer well-being primarily when youths engage in upward comparisons, and that person-level factors might moderate these links.

Prior research documents mixed associations between adolescent social media use and well-being, with meta-analyses typically finding small negative average effects but considerable heterogeneity. Many studies are cross-sectional; within-person evidence is limited and often shows small or null average effects, with person-specific variability. Upward social comparisons—judging oneself against seemingly superior others—are common on image-centric platforms and tend to reduce self-esteem and affect. Adult and older adolescent studies suggest upward comparisons mediate links from (especially passive) social media use to lower life satisfaction, self-esteem, and higher depressive symptoms; however, prior adolescent findings on mediation are inconsistent. Evidence also suggests potential moderators, including sex, self-control failure related to social media, and social comparison orientation, with some studies indicating stronger negative effects for girls and for those high in comparison orientation. The present study integrates these strands by testing upward comparisons as a mediator and exploring moderators in daily life among 10–14-year-olds.

Design and setting: A 14-day daily diary study (ambulatory assessment) conducted in Germany between April 6 and June 4, 2021 (during the COVID-19 pandemic). Participants completed evening surveys between 7–10 pm each day. Compliance rate was 85%, yielding 2,382 of 2,800 possible daily reports. Participants: N=200 children and young adolescents (103 girls), ages 10–14 (M=11.71, SD=1.02). Most attended Gymnasium (75.5%) and had German as native language (80%). Parents: 163 mothers participated; employment data reported. Inclusion criteria: smartphone ownership with internet access and German proficiency. Ethics approval obtained (DIPF_EK_2021_11); informed consent from children and parents. Procedure: Four parts—parent questionnaire (~10 min), baseline child questionnaire (~30 min; instructions video; measures of typical social media use, personality, well-being), 14-day daily diary (~10 min per day), and post-questionnaire (~10 min; including measures such as pathological social media use and emotion regulation). Recruitment via social media, schools, parent council, clubs, and word-of-mouth. Compensation: €5 each for baseline and post; €1 per daily survey plus €10 bonus for ≥12 diaries. Measures (daily):

• Social media use: Self-reported extent (1=not at all to 5=very much) of use that day of Instagram, TikTok, and YouTube on smartphones; aggregated across platforms. Prior work showed good convergence with objective use.
• Upward social comparisons: Six items capturing general impressions that others are better off (e.g., happier, prettier, more popular, cooler things/stuff); acceptable reliability (within/between ω≈0.83/0.96).
• Positive self-worth: Four items (e.g., liked myself; satisfied; felt good about myself; proud); reliability ω≈0.86/0.97.
• Negative self-worth: Four items developed for this study (e.g., disappointed in myself; wished I were different; angry at myself; felt worthless); reliability ω≈0.82/0.96.
• Affective well-being: Positive affect (three items: felt good, fantastic, content; ω≈0.78/0.94) and negative affect (four items: unhappy, sad, miserable, afraid; ω≈0.71/0.94). All items rated 1=not at all true to 5=completely true. Psychometrics: Multilevel confirmatory factor analyses (MCFAs) supported separating positive and negative self-worth at both levels; upward comparison scale showed acceptable fit and reliability. Factor correlations: positive vs negative self-worth r_between=-.79, r_within=-.48. Statistical analysis: Two-level multilevel structural equation modeling (MSEM) in Mplus 8.8 with Bayesian estimation (uninformative priors), two MCMC chains, 50% burn-in, 3000 iterations, thinning=50. Within-person effects modeled as random; random effects and Level-2 variables allowed to covary. Reported posterior medians and 95% credible intervals (CrIs), and model-implied R². Models: (1) SM use → four well-being indicators; (2) SM use → upward comparisons; (3) upward comparisons → four well-being indicators. Moderation tested via cross-level interactions predicting random slopes by person-level moderators (sex, social media self-control failure, social comparison orientation—abilities and opinions). Mediation: Multilevel mediation with random slopes for within-person paths (aw: SM use→upward comparisons; bw: upward comparisons→well-being; cw: direct SM use→well-being). Indirect effects computed as aw*bw + covariance(aw,bw) within-person; ab products between-person. Total effects = indirect + direct. Significance inferred when 95% CrIs excluded zero.

• Between-person associations (averaged across 14 days): Higher average social media use predicted lower positive self-worth (β=-0.28, 95% CrI [-0.41, -0.14]) and positive affect (β=-0.31 [-0.45, -0.17]), and higher negative self-worth (β=0.21 [0.07, 0.35]) and negative affect (β=0.26 [0.12, 0.40]).
• Within-person associations (day-to-day): On days with higher-than-usual social media use, youths reported lower positive self-worth (β=-0.08 [-0.13, -0.04]) and higher negative self-worth (β=0.07 [0.02, 0.11]); links with positive affect (β=-0.03 [-0.09, 0.01]) and negative affect (β=0.04 [-0.01, 0.09]) were not significant.
• Social media use → upward social comparisons: Significant at both levels (between: β=0.39 [0.27, 0.51], R²≈15.5%; within: β=0.09 [0.05, 0.14], R²≈5.6%).
• Upward social comparisons → well-being: Significant detrimental associations at both levels. Between-person: lower positive self-worth (β≈-0.61 [-0.69, -0.50]) and positive affect (β≈-0.56 [-0.65, -0.43]); higher negative self-worth (β≈0.68 [0.59, 0.76]) and negative affect (β≈0.67 [0.57, 0.75]). Within-person: lower positive self-worth (β=-0.26 [-0.31, -0.20]) and positive affect (β=-0.24 [-0.29, -0.18]); higher negative self-worth (β=0.27 [0.22, 0.33]) and negative affect (β=0.27 [0.22, 0.32]).
• Mediation by upward comparisons: • Between-person: Total and indirect effects of social media use on all four well-being indicators were significant; direct effects were not, indicating full mediation (e.g., upward comparisons paths: positive self-worth β=-0.67 [-0.68, -0.46]; negative self-worth β=0.69 [0.58, 0.78]; positive affect β=-0.48 [-0.60, -0.35]; negative affect β=0.62 [0.50, 0.72]). Total effects (selected): positive self-worth -0.43 [-0.66, -0.21]; negative self-worth 0.26 [0.09, 0.43]; positive affect -0.39 [-0.56, -0.22]; negative affect 0.25 [0.12, 0.38]. • Within-person: Indirect effects significant for positive self-worth (-0.06 [-0.12, -0.01]) and negative self-worth (0.05 [0.001, 0.10]); indirect effects for affect were not significant. Direct within-person effects were non-significant except a small negative direct effect on positive self-worth (β=-0.05 [-0.10, -0.004]). Total within-person effect significant only for positive self-worth (-0.16 [-0.27, -0.05]).
• Moderation: Of 36 tested cross-level interactions, only one reached significance—children higher in social comparison orientation (opinions) showed a weaker negative link between upward comparisons and positive affect; however, given multiple testing and modest reliability, this result is tentative. No credible moderating effects were found for sex, self-control failure, or comparison orientation (abilities).

The findings indicate that youths who use more Instagram, TikTok, and YouTube tend to report poorer average well-being and, on days of higher-than-usual use, show lower positive and higher negative self-worth. Upward social comparisons emerged as a consistent correlate of diminished well-being and as a key mechanism linking social media use to well-being: between-person effects were fully mediated and within-person effects on self-worth were partially mediated by upward comparisons. These results clarify prior heterogeneity by highlighting that the impact of social media use is particularly tied to how much youths feel others are better off, rather than to use per se. The daily-life, within-person evidence is especially relevant for interventions: reducing upward comparison tendencies or bolstering comparison coping strategies may mitigate adverse effects on self-worth. The absence of robust moderation by sex, self-control failure, or social comparison orientation (abilities) suggests the observed processes may generalize across diverse youths in this age range, though individual variability remains possible.

This study contributes evidence from a 14-day diary in 10–14-year-olds that daily social media use relates to lower self-worth and that upward social comparisons are both consistently linked to poorer well-being and mediate the association between social media use and well-being. Between-person links were fully mediated by upward comparisons, and within-person links to self-worth were partially mediated. These insights help explain mixed prior findings and point toward mechanisms amenable to intervention (e.g., fostering media literacy about curated content, training comparison coping). Future research should: (1) disentangle platform-specific effects and content types; (2) differentiate active versus passive use; (3) refine measurement of upward comparison processes versus resultant feelings; (4) incorporate objective use metrics and daily measures of problematic use; (5) test causal and reciprocal dynamics with intensive longitudinal and experimental designs; (6) recruit more socioeconomically and educationally diverse samples; and (7) examine contextual influences such as pandemic-related changes in usage and well-being.

• Sampling and generalizability: Convenience sample from Germany with overrepresentation of higher-education track (Gymnasium) families; limits generalizability.
• Platform aggregation: Usage of Instagram, TikTok, and YouTube was aggregated; potential platform-specific effects were not examined.
• Type of use: Active versus passive social media use was not distinguished; different usage types may relate differently to upward comparisons and well-being.
• Measurement of upward comparisons: Items captured general impressions/feelings that others are better off, which may not directly index active comparison processes.
• Correlational design: Causal inferences are limited; unmeasured confounders (e.g., offline social interactions) may influence observed associations and indirect effects.
• Possible reciprocity: Relations among social media use, upward comparisons, and well-being may be reciprocal; within- and across-day bidirectional dynamics were not modeled.
• Contextual timing: Data collected during the COVID-19 pandemic, when social media behaviors and mental health may have differed from typical patterns, potentially affecting effect sizes.