There is an emerging trend in alcohol research to give more attention to patterns of drinking (e.g. frequent binge drinking) as opposed to average consumption levels as predictors of various types of harm. We apply this focus here to population level data in an attempt to derive a measure of a community's alcohol use that is 'hazardous' as opposed to the traditional measure of total per capita consumption per year. Measures of alcohol consumption as predictors of rates of assault, alcohol-related morbidity and road crashes are compared across 130 geographical areas in WA for the financial year 1991/2. Based on epidemiological research from several countries, "hazardous" consumption was defined as alcohol consumed after having already consumed 6 standard drinks (75ml alcohol) for men or 4 (50ml) for women on any drinking occasion. A state-wide survey of alcohol consumption, data on liquor licence fees and from the 1991 Australian Bureau of Statistics Census were used to make various estimates of alcohol consumed for each geographical area. Estimates of total per capita consumption based on liquor licence data were highly correlated (r = 0.90, p<0.0001) with retail sales data and moderately with estimates based on regression analyses of survey and Census data (rho=0.42, p<0.001). Estimated 'hazardous' and total consumption were highly correlated when both were based entirely on survey and Census data (rho=0.85, p<0.001). All consumption measures were significantly correlated with rates of assault and acute alcohol related morbidity but not road crashes, with higher coefficients for hazardous consumption. Multiple regression analyses controlling for sociodemographic variables suggested that a) sales-based estimates of hazardous consumption were equal or slightly superior to those for total consumption for all harm indicators and b) survey/Census-based estimates of hazardous consumption were significantly superior to those for total consumption as predictors of both acute and chronic morbidity, though not assaults and crashes. It is suggested that estimates of hazardous per capita consumption for communities and larger populations may in future replace the traditional measure of total per capita consumption as a major target of public health policy.