Some limitations should be considered when interpreting our results. We were unable to examine the per capita consumption of alcohol among individuals with different levels of smoking or drinking, or among different demographic groups. State level estimates of per capita alcohol consumption are based on sales data collected directly by the Alcohol Epidemiology Data System and shipments data from beverage industry sources. There could be inaccuracies in estimates of per capita alcohol consumption due to reporting error, random fluctuation over time, estimation of ethanol conversion coefficients, cross-border sales, tourists’ consumption, and variation in state reporting practices for sales (Lavallee et al., 2011). However, the use of alcohol sales data is also a strength, as it is not subject to self-report bias like survey data. Furthermore, the use of aggregated panel data based on sales allowed us to examine the consumption of different types of alcoholic beverages, which is often not available in survey data, over a 30 year time period. This study was also the first to incorporate effects of local cigarette tax and smoke-free air policies.
