Would a soda tax increase alcohol sales?

Soda taxes (or taxes on sugar-sweetened beverages (SSB), if you prefer) have been in the news quite a bit recently. You can read some of the debate over whether soda taxes would be effective in fighting obesity here (Laurie Kubiak) and here (Boyd Swinburn). Eric Crampton also has a number of related blog posts (see here and here and here and here). The short version is that NZIER's report for the Ministry of Health suggests that soda taxes would be pretty ineffective.

In this post though, I want to focus on something else. Let's say a soda tax is effective, and people consume less soda. What will they consume instead? Will they switch to bottled water? Or to sugar-free beverages? Or maybe they will switch to alcohol? In other words, what will consumers substitute towards if soda is made more expensive with a tax?

A new paper in the Journal of Epidemiology and Community Health (ungated), by Diana Quirmbach (London School of Hygiene and Tropical Medicine) and others looks at exactly this question. Or rather, they look at whether consumers would purchase more alcohol if the price of soda increased. They used data on 6 million beverage purchases by nearly 32,000 UK households in 2012 and 2013, and estimate own-price elasticities for soda (separated into high-sugar, medium-sugar, and low-sugar varieties), and cross-price elasticities between soda (by type) and alcohol (beer, lager, cider, and spirits). They found that:

...own-price elasticities for non-alcoholic drinks are lower than for alcoholic beverages (that is, alcoholic drinks are more sensitive to price change), and that elasticities for all three SSB groups are inelastic (ie, smaller than 1), which means that there is a less than proportionate decrease in purchase following a price rise. This also compares with relatively inelastic (ie, insensitive) reactions to changes in the price of alcoholic drinks (except for lager for low- and medium income groups, and cider and wine for the high-income group)...

Increases in the price of high-sugar SSBs are associated with increased purchases of diet drinks, juice and lager (ie, they act as substitutes), whereas they decrease purchases of medium-sugar SSBs and spirits (ie, they act as complements). Increases in the price of medium-sugar SSBs impacts across a wider range (high-sugar SSBs, juice, water, beer, lager, wines and spirits), although all categories affected witness reduced purchasing (ie, a consistent complementary relationship of ~0.1% for a 1% price increase). Increases in the price of diet/low sugar SSBs increases the purchases for all other categories (with the exception of the two other SSB categories), ranging from 0.1% for juice to 0.7% for milk-based drinks and spirits per 1% price increase. 

In other words, a tax on soda (especially on high-sugar and diet/low-sugar sodas) would induce consumers to consume more alcohol. So, the overall effect of a soda tax on calories consumed (and hence, obesity and other health problems) is somewhat ambiguous, especially when you consider that alcohol is more energy-dense than sugary sodas. If people consume more alcohol as a result of a soda tax, the net impact on society may well be negative. Of course, this assumes that a soda tax is effective at all, which the NZIER report on the evidence thus far indicates is far from clear.

The Quirmbach et al. study does have some issues, which are common to most studies that try to estimate elasticities when prices are not directly observed. The main problems are well summarised in this article in The Conversation by Robin Room and Heng Jiang, and their main criticism is:

In principle, elasticity is about what happens over time when there is a change – such as a new tax – which results in a higher price.

But the study was not actually measuring the effects of change in price over time. Rather, it correlated how much one family bought of each beverage type when faced with a particular set of prices against how much another family bought of each beverage type with a different set of prices.

But because the study isn’t actually measuring and correlating the change that elasticities would measure – a new tax and the change in consumption over time – it offers no direct evidence of what would happen in case of a change like a new tax, and should not be interpreted as having done so.

In spite of the problems with the study, it does raise a valid question that requires further investigation. If we tax soda, should we be concerned about the negative impacts of what consumers would purchase instead?