Real-world effect of soda taxes

A new study in the journal Health Economics by Jason Fletcher and coauthors examines whether variation in soda tax policies across states leads to differences in weight and obesity.  

First, the authors note previous work on the issue:

studies using data on individual-level consumption and within-state variation in actual tax rates have found no net measureable effects on population weight. For example, Fletcher et al. (2010a) find that increases in soft drink tax rates decrease soda consumption among children, but do not influence total caloric intake, as children increase their consumption of other high-calorie beverages. This finding is consistent with a similar lack of effects for adults (Fletcher et al., 2010b). Other research taking this approach finds mixed results, demonstrating that average weight in some high risk populations may be more susceptible to soda taxes (Sturm et al., 2010).

Then, they point out a potential problem with this line of research: the variation in tax rates across locations isn't large enough to tell us what will happen if a state passed a "large" soda tax - or whether there are "non-linear" effects:

one concern with the ability of the results from some previous studies to predict the consumption response to large taxes, such as the 18% tax proposed in New York in 2008, and a potential reason for the differences in the results from the various strands of literature is that the existing soda tax rates are too low to be meaningful to most consumers because the average tax rate in 2006 was approximately 5% (Sturm et al., 2010; Todd and Zhen, 2010). Implicit in this argument is that substitution effects would also exhibit a threshold effect, where at high enough soda tax rates, individuals would substitute towards no beverages or low-calorie alternatives (e.g., water).

What did they find?

First, we examine whether there is any evidence of non-linear effects of current soda tax rates, with the idea that if very large taxes could have relatively larger effects, then we should see evidence consistent with this hypothesis based on the larger tax rates in our data, which reach 12%. However, using a variety of specifications, we find no evidence of effects on use or weight for a nationally representative sample of adults.

Our second approach uses a new comparative case-study method that leverages the sudden and large tax increase found in Ohio in the early 1990s. This method creates a ‘synthetic Ohio’ based on a weighted average of states that are most similar to Ohio’s population BMI before the tax was raised. Outside of simulation methods, this is the most informative approach to understanding the potential impact of recently proposed taxes, and it suggests very little evidence that the large tax imposed in Ohio had any detectable effect on population weight. Together, our results cast serious doubt on the assumptions that proponents of large soda taxes make on its likely impacts on population weight.