Do Sugar Taxes affect the right consumers?

Sugar taxes are often considered as a possible tool to tackle excessive sugar consumption. This paper estimates a dynamic multinomial Logit model of cola demand on a novel supermarket scanner dataset in order to study preference heterogeneity and state dependence in product choice. The model estimates allow evaluating the effectiveness of taxation in reducing demand for sugary colas across different consumer types. The results show that a sugar tax would be less effective among the targeted population of heavy sugar consumers. This policy, however, would be more effective among low-income households. Tax policy simulations show that a specific tax on sugar should be preferred to an ad-valorem tax on sugary colas on both corrective and equity grounds. This is because ad-valorem taxes can lead low-income households and heavy sugar consumers to substitute from expensive to cheaper sugary brands. Lastly, because households exhibit state dependence in cola choice, sugar taxes would be more effective in reducing sugar consumption in the long-run.