Aims: To use the rs1229984 variant associated with alcohol consumption as an instrument for alcohol consumption to test the causality of the association of alcohol consumption with hay fever, asthma, allergic sensitization and serum total immunoglobulin (Ig)E. Design: Observational and Mendelian randomization analyses using genetic variants as unbiased markers of exposure to estimate causal effects, subject to certain assumptions. Setting: Europe. Participants: We included a total of 466 434 people aged 15–82 years from 17 population-based studies conducted from 1997 to 2015. Measurements: The rs1229984 (ADH1B) was genotyped; alcohol consumption, hay fever and asthma were self-reported. Specific and total IgE were measured from serum samples. Findings: Observational analyses showed that ever-drinking versus non-drinking, but not amount of alcohol intake, was positively associated with hay fever and inversely associated with asthma but not with allergic sensitization or serum total immunoglobulin (Ig)E. However, Mendelian randomization analyses did not suggest that the observational associations are causal. The causal odds ratio (OR) per genetically assessed unit of alcohol/week was an OR = 0.907 [95% confidence interval (CI) = 0.806, 1.019; P = 0.101] for hay fever, an OR = 0.897 (95% CI = 0.790, 1.019; P = 0.095) for asthma, an OR = 0.971 (95% CI = 0.804, 1.174; P = 0.763) for allergic sensitization and a 4.7% change (95% CI = –5.5%, 14.9%; P = 0.366) for total IgE. Conclusions: In observational analyses, ever-drinking versus not drinking was positively associated with hay fever and negatively associated with asthma. However, the Mendelian randomization results were not consistent with these associations being causal.