addition of both policy variables compared to a more restricted model including only the other covariates. In each model, all three alcohol excise tax variables (beer, wine, and spirits) were included as covariates. Some states have state-run retail and/or wholesale systems for wine/spirits (3 states for wine and 18 states for spirits). The status of these states does not change over time in our data and is collinear with state fixed effects; thus the effects of these states are accounted for by the state fixed effects in the models. Analyses were weighted by population estimates for each state in a given year, using the U.S. Census intercensal estimates. Models were run for the full time period (1980–2009), and then for each half of the time period (1980–1994 and 1995–2009). Parameter estimates and standard errors were calculated using the SAS (Version 9.2, SAS Institute, Cary, NC) procedure “surveyreg”, employing state as the clustering unit to account for correlation of residuals within states in estimating standard errors (Angrist & Pischke, 2008; Arellano, 1987; Bertrand et al., 2004). Hawaii was excluded from analyses as alcohol excise tax data was not available for all years.
