Highway safety is an important issue in both urban and rural areas. Based upon a unique panel of incorporated cities over a 96-month period from January 1982 to December 1989, this paper evaluates the effects of the State of California's ban on common site sales of gasoline and alcohol to curb drunk driving. Geographic information systems (GIS) contribute spatial variables to an analysis of the impact of the ban on several categories of alcohol related crashes: fatalities, serious injuries, and property damage. By considering the types of alcohol-related crashes for the five-county Los Angeles area as interrelated, a seemingly unrelated regression (SUR) model locates more precise impacts. Overall the ban appears to reduce fatalities, serious injuries, and property damage crashes; but in relatively less population dense cities, the ban increases property damage and may increase more serious crashes in some urban neighborhoods. This suggests that the ban discourages highly impaired drivers but may motivate less impaired drivers to travel to an alternate alcohol purchase site. Diverse experiences in adjacent municipalities warn large regional or nation-wide evaluations of highway safety to take care when adopting large region or state 'fixed effect' adjusters in analyses that presume sub-region homogeneity.