BackgroundAccording to the self-medication and biopsychosocial models, individuals with social anxiety disorder (SAD) are at increased risk of developing an alcohol use disorder (AUD) as alcohol represents a maladaptive coping mechanism for some sufferers of SAD. The SAD-to-AUD causation was earlier supported in Norwegian longitudinal twin data and later questioned using longitudinal data from the USA. MethodsWe re-analyzed partly the same USA-based data (National Comorbidity Surveys, n = 5001), conducting theoretical and simulation analyses on different formulations of temporality and using real-data Logistic regression analysis to investigate whether baseline SAD was associated with AUD at the follow-up. ResultsUpon proper analysis of temporality, SAD preceded AUD. Specifically, SAD was the only one of the seven anxiety disorders that predicted 10-year later AUD after adjusting for all other anxiety disorders and AUD at the baseline (odds ratio was 1.70% and 95% confidence interval 1.12–2.57). SAD was also associated with incident AUD (OR = 1.64, 95% CI = 1.14–2.37). We provide formal, simulation-based, and data-based arguments on how certain flawed models of incidence attenuate the temporal association. ConclusionsWe demonstrated temporality and specificity in SAD-to-AUD association, which are considered signs of causation. We further identified and discussed problems in previous statistical analyses with different conclusions. Our findings add support for models positing causal effects of SAD on AUD, such as the self-medication and biopsychosocial models. The available evidence suggests that treating SAD should incur better chances of preventing AUD compared to treating other anxiety disorders, which lack comparable evidence on causation.