An Evaluation of the Soda Tax with Multivariate Nonparametric Regressions

This research extends past work by Shonkwiler and Yen (1999) by allowing for distributional flexibility and nonlinear responses in the form of established semiparametric and nonparametric regressions. The proposed models are shown to outperform the parametric version typically used in demand analysis to characterize a system of censored equations in terms of model fit and prediction power. Using the developed models, we derive elasticities associated with different individual-specific scenarios with regard to the recently proposed “penny-an-ounce” tax on soft drinks sweetened with sugar.