Inconclusive evidence for causal effects between beliefs in conspiracy theories: A reanalysis and comment on Williams et al. (2025)

Based on positive cross-lagged effects in random-intercept cross-lagged panel models (RI-CLPM), Williams et al. (2025) concluded that belief in a specific conspiracy theory causally increases future beliefs in other conspiracy theories. We reanalyzed the same data using latent change score models (LCSM) and stable trait, autoregressive trait, and state (STARTS) models, in addition to RI-CLPM. Our findings revealed contradictory patterns, where some models indicated increasing effects and others decreasing effects, and some models showed no effect at all. Given these inconsistencies, we suggest that the effects reported by Williams et al.’s (2025) may be spurious and that conclusions about causality were premature. It is important for researchers to bear in mind that correlations, including effects in the RI-CLPM and other models, do not prove causality when used to analyze observational data. We recommend researchers to analyze data with alternative models and compare their findings. If findings from alternative models converge, conclusions of causality are corroborated (although never finally proven). However, if findings diverge, as in the present case, conclusions of causality are questionable.