Association for Public Policy Analysis & Management

Taxes on sugar-sweetened beverages (SSBs) are increasingly being used as a policy lever to decrease consumption of sugary beverages, reduce diet-related chronic conditions, and raise revenue to support various public programs. Now that the first wave of taxes implemented at the city level in the U.S. has been in effect for several years, evidence regarding the efficacy of the taxes is beginning to grow, with several studies providing evidence for individual cities. To date, however, there has been no multi-city analyses of SSB taxes that consider the impact of the taxes across multiple settings and policy parameters. Studies also typically use only one of two possible control groups: 1) an adjacent area just outside the treated city; or 2) another, nearby, city. This analysis contributes to the literature by: 1) being the first to analyze data from several cities at once; and 2) using multiple comparison groups and assessing the robustness of the results. Specifically, we use household receipt data to examine the impact of sweetened beverage taxes on household purchases of taxed and untaxed beverages in the four largest cities currently with such taxes: Philadelphia, Pennsylvania; San Francisco, California; Seattle, Washington; and Oakland, California. We estimate the impact of these taxes by comparing changes in household monthly purchases in the treatment cities to changes in one of two comparison groups: 1) areas adjacent to the treatment cities; or 2) a matched set of households nationally. We find that an increase in the beverage tax rate of 1 cent per ounce decreased household purchases of taxed beverages by 53.00 ounces per month (12.2 percent). There was a particularly large decline in Philadelphia, a 1.5 cent per ounce tax, like the tax in Philadelphia, translates to a decrease of 126.11 ounces (27.7 percent). In addition, there is evidence of a slight decline in purchases in the areas adjacent to the treatment cities, which leads to smaller impact estimates when using this comparison group.