Hierarchical Linear Modelling (HLM) (45) was used to test the relationship between country-level variables and negative consequences from drinking by regressing consequences scores on country-level drinking pattern and economic/social development variables in models including individual-level frequency and quantity of alcohol consumption and interactions of country-level variables with individual-level consumption measures. Each country-level variable was assessed in a separate analysis because they were highly inter-correlated, except that country-level average frequency and quantity (which were not highly correlated) were included in the same model because, used together, they provide a measure of drinking pattern. These analyses did not control for age of respondents because of the high correlation between age and country measures of economic and social development. We used the over-dispersed Poisson version of HLM because the negative consequences scores were equivalent to count variables (summing of positively endorsed items divided by number of items). All person- and country-level variables were grand mean centered. All analyses included random intercepts to allow for unexplained mean country-level differences. For individual-level variables, the variance in slopes across countries was examined to determine whether it
