The Incidence of a Soda Tax, in Pennies and Pounds

Controversy over public health policies targeting carbonated soft drinks has catalyzed cross-disciplinary debate. Though beverage demand is characterized by purchases from multiple categories per trip, existing research using soda data predominantly employs discrete choice models, which restrict consumers to single unit choices. This paper instead combines a multiple-choice utility function with a hierarchical Bayesian model to estimate household-level preferences across beverages. I find that households in less healthy counties (as measured by mean body mass index) consume more regular soda per capita, have stronger preferences, and are less price sensitive than households in more healthy counties, suggesting that soda tax proposals should be county-specific. Because market players possess varying degrees of market power, I also study soda tax incidence. Simulations of equilibrium prices in various tax scenarios demonstrate that the assumption of 100% pass-through to prices, ubiquitous in soda tax research, underestimates the true change in prices, further underscoring the need for a county-level approach. Calculations then reveal that most of a counterfactual tax-induced shift in consumption is due to an income effect and low cross-price elasticities, so that while city governments are unlikely to raise enough revenue to cover the healthcare costs related to soda consumption, the downstream health benefits induced by a soda tax compensate most households for the reduction in utility