The evidence for the beneficial health effects of moderate drinking is weaker than commonly perceived. No randomised controlled trials have been done. Observational studies suffer from unavoidable methodological limitations, chiefly from confounding and misclassification. Clinical advice to patients as well as public health recommendations should discourage initiation of alcohol consumption, as well as recommend the reduction of excessive drinking. The absence of health benefits strengthens the the arguments for effective population-level policies, e.g. raising alcohol prices and restricting the physical availability of alcohol.