Social Media Use Subgroups Differentially Predict Psychosocial Well-Being During Early Adolescence

Abstract

Despite the salience of the social media context to psychosocial development, little is known about social media use patterns and how they relate to psychological and social functioning over time during early adolescence. This longitudinal study, therefore, identified subgroups of early adolescents based on their social media use and examined whether these subgroups predicted psychosocial functioning. Adolescents (N = 1205; 11–14 years; 51% female; 51% white) completed surveys at baseline and a six-month follow-up. There were three social media use subgroups at baseline: high overall social media use (8%); high Instagram/Snapchat use (53%); and low overall social media use (39%). The high social media use subgroup predicted higher depressive symptoms, panic disorder symptoms, delinquent behaviors, family conflict, as well as lower family and friend support, than the High-Instagram/Snapchat and low social media use subgroups. The high Instagram/Snapchat use subgroup predicted higher delinquent behaviors and school avoidance than the low social media use subgroup, but also higher close friendship competence and friend support as compared to both the high social media use and low social media use subgroups. Social media use patterns appear to differentially predict psychosocial adjustment during early adolescence, with high social media use being the most problematic and patterns of high Instagram/Snapchat use and low social media use having distinct developmental tradeoffs.

Appendix

Post-Hoc Exploratory Analyses: Gender Differences

Analytic Plan. To examine whether there were gender differences in the relationships between T1 social media use subgroup membership and T2 psychosocial functioning, adolescents were assigned to a latent subgroup on the basis of the which subgroup had the highest posterior probability. Then, analyses of covariance (ANCOVAs) were conducted to examine gender × social media use subgroup interactions for all T2 psychosocial outcomes. Each model included the main effects of gender and of social media use subgroup membership and their interaction effect, as well as the T1 covariates of age, race/ethnicity, perceived socioeconomic status, and psychosocial variable of interest. Separate models for girls and boys were conducted when the gender × social media use subgroup interaction term was statistically significant to understand the nature of the gender differences (ps < 0.05), Bonferroni-Hochberg corrections were applied when examining pair-wise differences among latent social media use subgroups/classes in girls and boys.

This case assignment approach was necessary because it is currently not possible to conduct a multiple group analysis using the modified correction method of Bolck et al. (2004; BCH method) to examine gender as a moderator. It is important to acknowledge that the case assignment approach does not capture the probabilistic nature of the latent class model and the reality that latent subgroup membership is not fixed, often producing attenuated estimates (Bray et al. 2015). However, the case assignment method may be considered acceptable for latent class solutions with very high classification accuracy (>0.95) because classification error is minimized and individuals can be assigned to latent classes with a high degree of certainty (Masyn 2013), which was observed in the current study (0.97–0.98). To evaluate the possibility that classification uncertainty impacted analyses, the BCH method was conducted separately in girls and boys. For all models, the pattern and significance of the findings were highly comparable, suggesting minimal bias in using the case assignment approach to examine gender differences.

Results. There was no significant gender × social media subgroup interaction for any psychosocial outcomes in this study, including anxiety disorder symptoms, F(2, 1103) = 0.87, p > 0.05, η² = 0.000, depressive symptoms, F(2, 1125) = 0.01, p > 0.05, η² = 0.000, delinquent behaviors, F(2, 1076) = 0.33, p > 0.05, η² = 0.001, family conflict, F(2, 1140) = 0.10, p > 0.05, η² = 0.000, family support, F(2, 1130) = 0.29, p > 0.05, η² = 0.001, close friend competence, F(2, 1170) = 0.63, p > 0.05, η² = 0.001, and friend support, F(2, 1126) = 2.21, p > 0.05, η² = 0.005.

Discussion. Surprisingly, no gender differences were found in the extent to which social media use subgroups predicted psychosocial functioning. It is possible that some aspects of peer relationship processes have become more similar among girls and boys within the social media context, accounting for the lack of gender differences observed for internalizing problems, delinquent behaviors, and social functioning. The immediate, quantifiable, and public nature of peer feedback that occurs through relentless content updates and tools such as “likes,” comments, and sharing as well as exposure to “drama” and the stressors of others may be equally salient for both girls and boys, whereas boys may not be as attuned to these issues during in-person interactions. This hypothesis aligns with a recent study indicating that there were no gender differences in adolescents’ perceptions that social media use led them to experience negative emotions (e.g., feeling overwhelmed due to online drama, feeling worse about their own life, pressure to post idealized content of themselves) and positive emotions (e.g., feeling more connected to friends, pleasure from expressing creativity) (Pew Research Center 2018a, 2018b). Alternatively, gender differences may become more apparent when examining adolescents’ interpersonal processes within the social media context, as technology-based feedback seeking and social comparisons predicted increases in depressive symptoms to a greater extent for adolescent girls relative to boys (Nesi and Prinstein 2015). Gender differences in the relationship between social media use subgroups and psychosocial functioning also may become more robust during middle-to-late adolescence, when gender differences in internalizing and externalizing problems become more stable (Evans-Polce et al. 2015).