Our objective was to examine the hypothesis that strengthening of tobacco policies is associated with decreased alcohol consumption. We employed a “differences-in-differences” quasi-experimental approach (Angrist & Pischke, 2008), and used data on U.S. state-level per capita alcohol consumption, derived from sales, tax and shipments data, from 1980–2009. The differences-in-differences approach exploits policy changes over time, and differences between states with different policies at any given time, to estimate the policy effect independently of stable state characteristics and national secular trends. To achieve this, state and year indicators were included as unordered categorical fixed effect covariates in all analytical models. As a result, associations are expected to be observed only if the magnitudes of within-state changes in policy correlate with within-state changes in risk of our outcomes.
