When the antidote is the poison: Investigating the relationship between people’s social media usage and loneliness when face-to-face communication is restricted

COVID-19 lockdowns restricted mobility and face-to-face interactions, raising concerns about psychological well-being and loneliness. Digital channels, especially social media, remained accessible and usage increased. Two opposing perspectives exist: (1) a compensation view where social media supports social connectedness and reduces loneliness when in-person interactions are limited; and (2) an internet paradox view where online interaction displaces richer, more beneficial interactions, potentially increasing loneliness. The study asks whether changes in social media usage during a government-ordered lockdown, when face-to-face interactions are restricted, are related to people’s loneliness, and whether this relationship is positive or negative. Leveraging longitudinal panel data from Germany before (February 2020) and during (April 2020) the initial national lockdown, the study examines both changes in usage and baseline usage in relation to loneliness, contributing beyond prior cross-sectional work.

Theoretical development contrasts two streams. Compensation logic: Social media can foster social connectedness and reduce loneliness, supported in non-lockdown settings and among populations with limited face-to-face interaction; some cross-sectional studies during early lockdowns linked higher social media use to lower loneliness (e.g., adolescents and older adults). Internet paradox/displacement logic: Digital media, including social media, may offer low social presence and media richness, facilitating low-effort, distant ties that displace richer interactions (in-person or video chat), potentially increasing loneliness. Some cross-sectional studies during COVID-19 found higher social media use associated with higher loneliness, while richer media (e.g., video conferencing) did not show such undesirable effects. The authors propose competing hypotheses: H1 (compensation): Increased social media usage during lockdown is associated with decreased loneliness. H1A (social media paradox/displacement): Increased social media usage during lockdown is associated with increased loneliness.

Design and context: Longitudinal panel study in Germany spanning the initial nationwide COVID-19 lockdown (restrictions began March 22, 2020; first easing April 20, 2020). Two focal survey waves: pre-lockdown (T1: Feb 19–29, 2020) and during-lockdown (T2: Apr 8–15, 2020). Baseline panel established earlier (T0: Sep 2018; additional wave May 2019). To avoid anticipation or easing effects, analyses restricted to responses completed before March 1 (pre-lockdown) and before April 15 (during-lockdown); 98 and 5 cases excluded, respectively. Final matched sample N=825 (48 days average between T1 and T2). Sampling and fieldwork: Conducted with Harris Interactive AG; panel representative of Germany for gender, age, education; incentives ~€2 per survey; quality checks and exclusions applied. Measures: Loneliness (DV) assessed with three items from the revised UCLA Loneliness Scale (7-point frequency: never to very frequently; items: feeling alone, lonely, left out). Social media usage: Self-reported average weekday hours for social media (e.g., Facebook, Twitter, Instagram). Key predictors include change in social media usage (T2–T1) and baseline usage (T1). Other media usage: Sum of hours for newspapers/websites, radio, TV, blogs, podcasts; both baseline and change included. Communication channels: Face-to-face communication in hours (baseline and change); video chat for non-occupational communication on a 5-point intensity scale (1=never to 5=multiple times/day; baseline and change). Baseline F2F and video chat prior to lockdown taken from T0 where necessary. COVID-19 impacts: Personal restraints index (mean of 5 items on changes to routine, lifestyle, hobbies, travel, financial prosperity; 7-point agree-disagree). Work-life restraints: 3-point single item (no effect, some effects, cannot work). Demographics: Age (5 categories), gender, education (3 levels), employment status (student, employed, unemployed, retired), living alone. Personality: Big Five single items (7-point agree-disagree) and social media confidence (single item). Analysis: Structural equation modeling (SEM) in AMOS 27 with longitudinal specification controlling for pre-lockdown loneliness. Invariance testing across time (configural, weak, strong); weak invariance selected (best fit). Model fit (main analysis): RMSEA=.037 (90% CI .030–.044), CFI=.986, SRMR=.017; chi-square=198.13, df=93. Bootstrapped standard errors (n=10,000). Robustness checks included alternative specifications and OLS regressions. Descriptives: Average loneliness decreased slightly from T1 mean=2.69 to T2 mean=2.56 (1–7 scale). Social media usage increased marginally (0.70 to 0.73 hours/day; not significant). Face-to-face communication decreased significantly; video chat use showed no significant average change.

- Main SEM results (DV: loneliness during lockdown; R²=.593): - Change in social media usage (T2–T1): B=0.131, β=0.099, SE=0.040, p<.05 (higher change associated with higher loneliness). - Baseline social media usage (T1): B=0.083, β=0.067, SE=0.039, p<.05 (higher baseline associated with higher loneliness). - Face-to-face communication: - Change: B=−0.057, β=−0.113, SE=0.016, p<.01 (more F2F associated with lower loneliness). - Baseline: B=−0.044, β=−0.074, SE=0.019, p<.05. - Video chat communication: Change and baseline not significant. - Other media use: Change and baseline not significant. - COVID-19 impacts: - Personal restraints: B=0.231, β=0.217, SE=0.028, p<.01 (associated with higher loneliness). - Work-life restraints: ns. - Personality and demographics: - Neuroticism: B=0.084, β=0.092, SE=0.026, p<.01 (higher loneliness). - Agreeableness: B=0.065, β=0.069, SE=0.024, p<.01 (higher loneliness). - Openness: B=−0.055, β=−0.058, SE=0.025, p<.05 (lower loneliness). - Conscientiousness, extraversion: ns. - Age: B=−0.119, β=−0.081, SE=0.044, p<.01 (older associated with lower loneliness). - Retired: B=0.261, β=0.074, SE=0.120, p<.05 (higher loneliness); other demographics mostly ns. - Pre-lockdown loneliness strongly predicts during-lockdown loneliness: B=0.654, β=0.614, SE=0.033, p<.01. - Post-hoc moderation analysis (R²=.600): Interaction of change in social media usage with change in video chat use: β=−0.088, SE=0.018, p<.01 (richer digital communication attenuates the positive association of social media with loneliness). Interaction with change in face-to-face communication: ns. - Overall pattern supports H1A (social media paradox): greater social media use during lockdown is associated with greater loneliness; no evidence supports H1 (compensation).

Findings indicate a social media paradox during lockdown: increased social media use is associated with higher loneliness, consistent with displacement of richer interactions (face-to-face or high social presence media) rather than compensating for reduced in-person contact. Richer digital channels (video chats) attenuate this association, suggesting that when social media is complemented by high-presence communication, the negative link is weaker. The relationship appears contextual; variations across settings (e.g., culture, lockdown severity) may explain conflicting prior findings. Personality factors also relate to loneliness under lockdown, with neuroticism and agreeableness linked to higher loneliness and openness to lower loneliness. Implications suggest users should prioritize richer communication channels during periods of restricted in-person contact, and policymakers should consider the differential social value of communication media when designing or communicating restrictions and supports.

This study provides longitudinal evidence from Germany’s initial COVID-19 lockdown that both increases in and higher baseline social media usage are associated with higher loneliness, supporting a social media paradox wherein social media may displace more beneficial interactions. Complementary use of richer digital communication (e.g., video chats) can soften this association. Contributions include clarifying competing theories with longitudinal data, highlighting the contextual nature of social media’s social value, and offering guidance for individuals and policymakers to emphasize rich communication channels during restrictions. Future research should investigate causality with digital trace data, examine different lockdown phases and cultural contexts, explore younger populations, and study nuanced roles of personality and distinct loneliness dimensions.

- Context/time-specific: Focuses on Germany’s initial lockdown; practices and platforms evolved thereafter, so effects may differ in later phases. - Generalizability: Results from Germany may not extend to other countries or to contexts where societal loneliness increased; sample excludes individuals under 16. - Self-reported measures: Media use and constructs are self-reported and subject to biases, though prior work suggests comparable predictive validity to digital trace measures. - Personality timing: Big Five measured during lockdown only; assumed stable over short interval but potential reporting bias cannot be fully ruled out. - Correlational design: Despite longitudinal controls and invariance testing, causality cannot be definitively established. - Measurement and sample constraints: Some pre-lockdown measures (e.g., F2F, video chat baselines) drawn from earlier panel wave (T0); panel, while broadly representative for age 16+, may still differ slightly from population benchmarks.