Maxime Tranchard (INRAE – AgroParisTech, TSE)
The Soft Drinks Industry Levy (SDIL), implemented in the United Kingdom in 2018, taxes soft drinks in tiers of sugar content and was explicitly designed to provide manufacturers with incentives to reduce sugar content of their products. In this paper, I develop a tool to evaluate and design such a tax. I separate the levy’s effect on sugar into a price channel and a recipe channel, and predict the response of recipes, prices and welfare to alternative schedules, including the reform of 2028. Using household scanner data, I estimate a random-coefficients demand system in which price sensitivity varies across households and each recipe carries its own fixed effect. Combined with a multiproduct pricing model, observed reformulation decisions yield bounds on product-level reformulation costs. Changes in recipes account for about a third of the levy’s effect on sugar. About 15% of the recipe channel runs through consumers substituting across products in response to the new recipes. The tiered design worked as intended: products near a threshold reformulate without losing demand, and products further away still find it worth avoiding the levy. Whether higher rates would raise welfare depends on the social cost of sugar. In this application at £3 per kilogram no counterfactual schedule does, while at £4 most do.