Aims: To assess an empirical basis for cross-national and cross-cultural comparisons of four aspects of the association between alcohol consumption and homicide: the relative strength of the association, the fraction of homicide rates attributable to alcohol consumption, possible gender differences in the association between consumption and victim rates and possible variations in beverage-specific effects on homicide rates.

Design, setting and participants: Time series analyses on differenced series of annual aggregate-level data on alcohol sales and homicide rates for the period 1950-95 were performed for each individual country. Estimates were pooled across countries within three regions of alleged differences in drinking pattern: southern Europe, central Europe and northern Europe.

Findings: Total alcohol sales were positively and statistically significantly associated with homicide rates in five countries. Beer sales were positively and statistically significantly associated with homicide rates in four countries, wine sales in another two countries, and spirits sales in two countries. The effect of alcohol sales was stronger for male homicide rates than for female homicide rates, and the estimated fraction of homicides that could be attributed to alcohol consumption appeared to be of the same magnitude in the three regions. When estimates were pooled across countries, the strongest association between total sales and homicides was found in the northern European countries and the weakest, but still statistically significant, in the southern European countries. Pooled estimates showed that beer sales were positively and significantly associated with homicide rates in all three European regions, whereas wine sales were positively and moderately associated with homicide rates only in the traditional wine drinking cultures in southern Europe.

Conclusion: The findings support the hypothesis that homicide rates are influenced by alcohol sales and more so in the northern European countries where the drinking culture is, to a larger extent, characterized by heavy drinking episodes. Moreover, the findings are suggestive of beverage-specific effects on violent behaviour being contingent upon characteristics of the drinking culture.