A Longitudinal Approach to the Analysis of Social Media Engagement: The Case of Anger-Driven Climate-Skeptic Message Propagation During the 2021 German Elections

This study examines the advantages of longitudinal modeling over cross-sectional approaches for analyzing social media engagement, specifically exploring how emotional reactions—particularly anger—influence content sharing. While cross-sectional methods are commonly used, engagement patterns are inherently temporal, making longitudinal approaches more appropriate. These methods offer a middle ground between simple cross-sectional analyses and complex time-series techniques, reducing omitted variable bias and accounting for time-varying factors like algorithmic amplification while accommodating irregular social media data. We empirically compare these approaches using 1,137 environmentally themed Facebook posts from German political parties during the 2021 federal election, focusing on Alternative für Deutschland (AfD) and Die Grünen (The Greens), parties with opposing environmental policies. Cross-sectional estimates using three sampling strategies were compared with longitudinal results via Bayesian multilevel regression with negative binomial specification.Findings reveal that longitudinal modeling produces more conservative and precise estimates, while cross-sectional methods tend to inflate effect sizes and interparty differences. Although all models show anger positively associates with sharing for AfD and negatively for Die Grünen, longitudinal analysis provides greater inferential stability by controlling for time-invariant confounders. Such differences have important theoretical implications, underscoring the value of incorporating temporal dynamics into social media research while acknowledging data access challenges.