The Effects of Alcohol Taxes – New Evidence from Scanner Data

This paper uses scanner data from several thousand stores across the US in order to investigate the detailed effect of an alcohol tax increase on alcohol prices and sales and product substitutions. Specifically, we exploit a natural experiment in Illinois where spirits and wine taxes were unexpectedly and substantially increased in September 2009. Deploying a difference-in-difference design that uses stores in states other than Illinois as the control group, we find that consumers substitute towards beer which had only a very minor increase in tax. This response is so strong that the reductions in spirits and wine consumption are offset by increased beer consumption. In addition we find that alcohol taxes are over-shifted to consumers. In particular, the pass-through is around 1.5 for spirits and 1.3 for wine. Consumers also substitute towards cheaper products in the spirits category, but not for wine. This suggests that minimum spirits pricing would enhance the effects of tax increases. We also investigate - but find little evidence for - cross-border shopping. Finally, we find that the tax increase was not associated with a reduction in traffic fatalities. This result is plausible because there was actually no reduction in total alcohol consumed. Our research also advances the literature along several dimensions methodological dimensions. First, by using detailed scanner data, we overcome the difficulties of the ACCRA alcohol price dataset that has been commonly in the past. For example, we calculate our measure of alcohol prices based on more than 200 products that are being consumed across the US, rather than a single product. Furthermore, our store sample comprises several thousand diverse stores. Second, we combine these detailed data with plausibly exogenous variation generated by the Illinois tax increase. The tax increase went into effect within a few weeks of a new governor taking office and was part of a job market package rather than being driven by anti-alcohol sentiment. Both raw data and an event-study specification also strongly indicate that prices and sales in control-stores followed similar trends as those in Illinois. We also deploy several innovative methods to achieve reliable inference. Together, our approach ensure that we are estimating credibly causal effects rather than just correlations.