By Eric Crampton 06/04/2018


Otago’s public health people, perhaps unsurprisingly, didn’t like NZIER’s take on sugar taxes. They’ve blogged on it here, but they seem to have missed a few important points.

A Report commissioned by the Ministry of Health, written by NZIER, has recently been getting air-time as an argument against taxing sugary drinks.  However, the Report seems to us to be seriously flawed.
For example, the Report argues that soft drinks do not impose a negative externality.  A negative externality is when consumption of a “good” (product) imposes costs on others.  For example, it is universally accepted that tobacco smoking results in costs to society (eg increased health care spending), and therefore has negative externalities that justify tax.  The same is true of sugary drinks increasing obesity/diabetes/tooth decay rates that then results in health costs to society.  It is well recognised by economists that soft drinks impose such a negative externality, eg 2 ,5.  Just as with alcohol and tobacco in nearly all high-income jurisdictions, this negative externality issue is a strong rationale for the state imposing a tax to help internalise the cost to society and cover some of the future costs to the health system.

This is the minor point in their argument, but it is worth noting potential confusion on externalities because even some economists mess this one up. The ability to offload health costs onto others, either through insurance where the premiums don’t distinguish between smallish changes in risk, or through a public health system, only generates technological inefficiency to the extent that it changes consumption.

Think of it this way. There is a total amount that somebody’s health care is going to cost. Imagine a world without a public health system and where private insurance premiums were perfectly adjusted to risk factors. Think about how much a person’s health care will cost in that world in total. Then add back in the public health system or a less granular private insurance system (or one where they’re compelled to use community ratings or somesuch). By how much does the total spending on that person’s healthcare increase because they’re able to offload the costs of unhealthy diets and the like onto other people? The increase in the risky behaviour associated with the cost-offloading imposes a technological externality.

But not all of the cost associated with that is a net social cost either! The person engaging in more of the risky activity gets some benefit from it, and that benefit needs to be netted out. The net social cost is the excess of the extra bit of cost over the extra bit of benefit. If you want a measure of the “social cost” of sugar or whatever, you should be looking to that little triangle rather than the total quantum of cost.

So we should never be looking at the total health costs associated with some kind of consumption – just at the cost associated incremental change caused by the ability to offload cost. All of the rest of the health cost is a pecuniary effect: it changes the identity of who pays for something rather than the amount paid. And even when we’re looking at the incremental increase in the cost, part of that too is a transfer rather than deadweight cost.

Ok, enough about that. We covered it in Jenesa’s Health of the State report, and I’d covered it here before.

The more substantial critique from Otago is that NZIER’s literature review was incomplete. At page 18, NZIER lays out its review method:

The literature reviewed for this report was identified by first searching for English-language peer-reviewed papers published in the last five years with evidence of an impact of a tax or levy on sugar-sweetened beverages, sugar or sugary foods through the following databases: Econlit, Pubmed/Medline, Google Scholar, National Bureau of Economic Research (NBER), Research Papers in Economics (RePEc), Te Puna. The search used a combination of keywords and phrases, including “sugar”, “soda”, “sugary”, “sugar-sweetened”, “beverage”, “drink”, “tax”, “levy”, “impact”, “effect”, “evidence”, “consumption” and “intake”. As searching is an iterative process, other keywords were introduced later, including “elasticities”, “price”, etc. and additional targeted searches were added for authors with multiple relevant publications, for papers that were already known to the reviewers, or for papers identified in the references of other included papers where these appeared relevant. Opinion pieces, letters to editors, media articles, presentations, and authors’ replies to comments on
published papers were excluded.

Keep that in mind. They’re only looking at English-language, peer-reviewed papers published from 2012-2017.

Now here’s Otago’s critique:

The literature reviewed by NZIER also seems rather incomplete. Eg, for the 9 studies that we are aware of which have examined the impact of real world sugary drink taxes on health – the NZIER Report refers to just one of them.6 The missed ones include studies suggesting health favouring associations for BMI/obesity7 8 9 10and for reduced cardiovascular disease11; along with studies showing no benefit for health.12 13 14 Yet even for two of the latter studies finding no association – the authors suggest the null finding is probably because of very low tax rates in the studied settings and they recommend higher tax rates.12 14 We are also surprised as to why the well-publicised report from the Australian Grattan Institute on sugary drink taxes published in 2016,2 was also missed by NZIER. Perhaps as a consequence of a suboptimal literature search strategy, NZIER have missed some key information and this may have limited the value of their conclusions.

Now let’s check those references.

6 is Fletcher et al, 2015, published in Health Economics. This was included in NZIER’s review.

7 is Fletcher et al 2010, published in Contemporary Economic Policy. Remember that NZIER’s review included papers published in the last five years. Since NZIER’s work was in 2017, the earliest inclusion date is 2012. So it was excluded.

But let’s have a closer look, just for fun. Because this one is fun.

Fletcher et al 2010 one does indeed suggest that while there is no effect of sugar taxes on outcomes, it could be because observed taxes are too low.

Now what Otago either didn’t know or didn’t want you to know is that Fletcher et al 2015 is an answer to Fletcher et al 2010. The 2010 work posited that there could be bigger effects at higher tax rates – that there’s a nonlinear response. The 2015 paper tests for that nonlinear response and rejects the hypothesis that the 2010 paper suggested for its null finding – there’s no evidence of nonlinear effects within the range of observed taxes.

So, the question for the Otago people then: is there any good reason to highlight a mechanism that the authors disprove in later work? I can think of a few bad reasons. And the only good way to include it would be with a health warning that the mechanism was kinda torn apart by the authors’ later work.

But it’s all kinda moot since the study was restricted to work in the past five years.

8 is Kim et al 2006. That’s six years before NZIER’s window.

9 is a master’s thesis from 2013. I think that’s excluded by the restriction to peer-reviewed papers, but it’s perhaps debatable for a thesis that’s made it through committee.

10 is a 2010 paper, so two years before the window.

11 is a 2015 paper published in the Journal of the American Heart Association. So it meets the criteria. So it’s arguably a fair call that it should have been caught in the search. It wouldn’t have much changed the result since it’s just a cross-sectional study across US states suggesting that states with higher taxes on soda also have lower odds of poor cardiovascular health, but it would be a pretty big stretch to draw anything causal from that. You need at least panel data work identifying on state-level changes in soda taxes. Otherwise how could you tell that other things aren’t driving both policy and health outcomes?

12, 13, and 14 were all published in 2010 or 2009, and so are outside of the review window.

And 2 is a Grattan Institute report that would have been excluded as it wasn’t a peer-reviewed journal article. Otago didn’t chide NZIER for missing the Initiative’s report on sugar taxes, but it would also have been properly excluded as it also wasn’t a peer-reviewed journal article.

Ok. So the Otago People are mad that NZIER excluded a bunch studies that the Otago People know about. They didn’t bother to check the exclusion criteria in the NZIER’s paper, and instead jump to the next paragraph claiming that NZIER is beholden to Big Industry interests even though this report was MoH funded.

On checking, it looks like NZIER failed to include one study that they should have included, but including it wouldn’t have made a darned bit of difference. I suppose the Otago People could have complained that the review window was too narrow – I don’t know why it was set at five years but that seems a reasonable window where sugar taxes are pretty new. But the complaint as it stands seems more than a bit off.

My disclosure: the Initiative is funded by a broad range of corporate members including ones the Otago People wouldn’t like. Their membership has zero influence on my views of the Otago People’s work.