Image by Gunilla G CC 2.0 via Flickr

In a recent paper from the Grattan Institute, A sugary drinks tax: Recovering the community costs of obesity, Report No. 2016-15, Stephen Duckett and Hal Swerissen put their argument for an excise tax of 40 cents per 100 grams of sugar, on all non-alcoholic, water-based drinks that contain added sugar, or sugar-sweetened beverages, SSB for short. There may be a convincing case to be made for such a tax, but Duckett and Swerissen have not made it.

My paper does not attempt a comprehensive assessment of the proposed tax, but only to assess the quality of the case made by Duckett and Swerissen. Unfortunately, in their narrowly based assessment they made serious errors, both conceptual and quantitative; hence my judgment of ‘Not Proven’.

The usual way to assess such a proposal is to try to quantify the major effects, so as to test whether the benefits are likely to exceed the costs. For public policy purposes, consideration should also be given to what groups will bear the costs and who will enjoy the benefits. This is standard fare of cost-benefit analysis of a kind expected for any serious policy proposal.

The benefits of the tax would include the following, with the bold parentheticals indicating whether or not Duckett and Swerissen quantified the item; I have listed the items in order of my guess of likely magnitude:

  1. Better health and longevity for those who would be less obese because of the tax. (Not quantified.)
  2. Lower spending on prevention and treatment of obesity and related conditions, and on public welfare payments and pensions:
    a. Public spending (Quantified.)
    b. Private spending (Not quantified.)
  3. More engagement by those people in community activities, and participation in the paid workforce, with the latter reflecting in
    a. Higher earned incomes and, therefore, higher tax revenues. (Quantified.)
    b. A deduction from income in 3a, to account for reduction in enforced or voluntary idleness or leisure. (Not quantified.)
  4. The public revenue from the tax itself. (Quantified.)

Costs associated with the proposed tax would include:

  1. Loss of consumer enjoyment from consuming SSBs, with a deduction to account for consumer savings of spending on SSBs. (Not quantified.)
  2. Public and private administrative costs for the new excise tax. (Not quantified.)

Duckett and Swerissen focussed narrowly on the fiscal items 2a and 3a above. By neglecting some benefits, like the value of improved health and longevity, Duckett and Swerissen bias the outcome against their own proposal. However, the opposite is true for their neglect of the cost items, 5 and 6.

‘Third party’ or ‘community’ costs

Rather than attempt a comprehensive assessment, Duckett and Swerissen rely on estimates of two fiscal items, namely, the savings in public expending on pensions and health care, and the increase in income tax that could arise because the SSB tax reduces the prevalence of obesity and thus increases workforce participation. Considering only these two items, they find a reduction in the net fiscal burden on the non-obese, and thus conclude that the tax would contribute towards achieving the goal of the subtitle, Recovering the community costs of obesity.

But they neglected the costs that the tax itself imposes on the non-obese, which is item 5 above. The non-obese, after all, consume SSB because they enjoy them, and their consumption generates no or trivial ‘community costs’. Using Duckett and Swerissen’s data and their estimates of the parameters, I show that the non-obese would be made worse off financially by the SSB tax: their fiscal gain is less than the value of their loss as consumers, and so the tax fails by the very criterion that Duckett and Swerissen adopted.[i]

Two per cent of what

However, my back-of-the-envelope estimates employ an unexamined but questionable assumption in the Grattan paper: although Duckett and Swerissen explain the basis of their estimates of the effects of the tax on the prevalence of obesity itself, they do not attempt to justify their assumption about the flow-through effects that a fall in obesity would have on other variables of interest.

In particular, Duckett and Swerissen say that the imposition of their proposed tax on SSB would reduce prevalence of obesity by two per cent. They then assume that a two per cent reduction in prevalence would bring two per cent of the savings in fiscal outlays that would arise, were obesity to be completely eliminated. Similarly, they assume that a two per cent reduction in the prevalence of obesity would have two per cent of the effect on employment that would occur were obesity completely eliminated. This proportionality assumption about flow-through effects seems implausible.

By their own account, the SSB tax would cause a small fall in the weight of an obese person: they quote studies suggesting an eventual 0.41kg weight reduction. For a person 1.7m tall and weighing 116kg, such a small weight loss – 0.35 per cent – would reduce her body-mass index (BMI) from the lower limit of Obese Class III to the upper limit of Class II; however, it seems unlikely that this would have an effect on her health costs or labour-force participation equal to two per cent of the effect of her ceasing to be obese, which would require her weight to fall to 86kg – a fall of over 25 per cent, presumably requiring extensive and persistent changes in lifestyle.

Veerman, Sacks, Antonopoulos and Martin (2016: 8) find that a modest reduction in SSB intake, due to a similar 20 per cent ad val. SSB tax, would translate into the saving of around 170,000 disability-adjusted life years (DALY) for the 2010 Australian adult population. By my calculation, these 170,000 DALY would add at most 0.02 per cent to that population’s life years. However, the estimated effect on the prevalence of obesity itself would be a hundred times larger: ‘about 2.7% (0.7 percentage point) among men, and 1.2% (0.3 percentage point) among women, compared to business as usual’ (Veerman, Sacks, Antonopoulos and Martin 2016: 4. The simple average of 2.7 and 1.2 is around the two per cent fall in prevalence that Duckett and Swerissen used.)

The definition of prevalence is the percentage of population with MBI of 30 or above. Thus, most if not all of the two per cent reduction in prevalence would occur amongst those at the lower end of Class I Obese (30 ≤ MBI < 35), which contains over 3,250,000 adults, or 65 per cent of the obese. Arguably these are people with a condition that is unlikely to have had major effects on their health or labour force participation; thus, we would expect a second- or third-order effects from a tax-induced minor reduction in weight.

Thus, the proportional decline in the prevalence of obesity could be up to two orders of magnitude larger than the proportional effect on life expectancy and, presumably, similarly for workforce participation and earning power. (Veerman, Sacks, Antonopoulos and Martin did not discuss employment effects.)


Duckett and Swerissen put all their eggs into the fiscal basket, and claimed that a 40¢ excise tax per 100g of sugar in sugar sweetened beverages would benefit the non-obese financially and, therefore, was justified economically.  On their own data and assumptions, the reverse is true: the non-obese would suffer financial loss from the imposition of the tax.

At best, on Duckett and Swerissen’s own data, the tax would have small aggregate benefits and costs, with the net balance effectively zero (before the public and private costs of implementing and administering the tax). However, to reiterate, they did not consider an item that may tip the balance in favour of the tax, which is the private value of improved health, my #1 above.

Nonetheless, their whole exercise is problematical, because Duckett and Swerissen assumed without argument that a two per cent reduction in the prevalence of obesity would have two per cent of the effects that would occur, were obesity eliminated completely, whereas it seems much more likely that a two per cent reduction in the prevalence of obesity itself would lead to a much smaller proportional reduction in the ill effects of obesity, including on public outlays on health and welfare, and forgone income tax.

[i] Theirs is a paradoxical framework in which the public policy problem exists because some of the burden of obesity falls on the non-obese; logically, therefore, there would be no case for the tax if everyone were obese, because then none of the burden of obesity would fall on the non-obese.

This article has 5 comments

  1. Stephen Duckett

    Our report, which argues for the introduction of an excise tax on sugar-sweetened beverages (SSBs), deliberately did not undertake a complete cost-benefit analysis. Our report instead focused only on the ‘third-party’ costs of obesity. We excluded the individual health (and financial) benefits from lower SSB consumption and lower BMI, as this case has been well made by numerous public health advocates, including many Australian academics and the World Health Organisation. Recognising these health benefits, more than twenty countries and sub-national governments have implemented or announced an SSB tax.

    We recognise that an SSB tax by itself will not solve Australia’s obesity problem and will need to be combined with other anti-obesity policies, particularly those targeted at children. In the report, we acknowledge the shortfalls of the proposed SSB tax, including that non-obese people will pay the tax. However, non-obese people consume fewer SSBs than obese people. There are also very close substitutes to SSBs, meaning consumers can easily avoid the tax and so lose little utility. We also acknowledge that an SSB does not perfectly target the costs of excess unhealthy food consumption that contributes to obesity. However, it is a good second-best option as it effectively targets products contributing to the obesity problem, minimises the taxing of products with beneficial nutrients (most SSBs generally have no beneficial nutrients, unlike a broader sugar or junk food tax), and is simple so keeps administrative costs low (the Parliamentary Budget Office estimates administrative costs of $7 million per year). Including the private benefits make the case for an SSB tax even stronger.

    Stephen Duckett and Trent Wiltshire

  2. Stephen Duckett and Trent Wiltshire simply ignore Jonathan Pincus’ criticism of their work, which is that they assume–completely implausibly–that a 2% reduction in the incidence of obesity would lead to a 2% reduction in the effects of obesity.

    Moreover, as Pincus points out, they argue in their paper that a a full CBA is not needed as it is enough to show that the tax would reduce the burden on the ‘third party’. But the direct effect of the tax would be to increase the burden, not reduce it; and they have provided no proof that a fiscal gain for the non-obese is a sufficient condition for the benefits of the tax to exceed its costs. If it is not such a sufficient proof, then why is the question they address of interest?

  3. The sugar tax is a pretty straightforward pigouvian tax. These tend to have a small positive impact on welfare and certainly do better from a welfare enhancing perspective than most other taxes currently applied. The government has a revenue shortfall so it’s win-win.

    Most evidence would suggest There’d be a small drop in sugar consumption, a small increase in revenue, in a distributional sense it would impact the poor slightly more but we are talking tiny amounts of money here.

  4. Jonathan Pincus

    The sugar tax proposed by Duckett and Swerissen is not win-win, contrary to Ben Phillips: the 72% of adult Australians who are non-obese would lose on balance. Their consumption of sugar in beverages causes no Pigovian externality, but they would be taxed nonetheless.

    I estimate that the tax would have an average excess burden of 20c per dollar of revenue; and, by the usual rule of thumb, the marginal excess burden would be around 40%, which is high, not low. Considered purely as a revenue measure, it fails.

    When it is considered as a Pigovian tax, then I agree with Ben Phillips that it is a tax involving tiny amounts of money–and even much tinier than Duckett and Swerissen report: a 2% reduction in the prevalence of obesity is very unlikely to cause a 2% reduction in the detrimental consequences of obesity, which is what they assume–more likely, the effect would be one tenth or one hundredth of 2%. Therefore, this tax proposal would not deserve much attention, except that it does have a heavy weight behind it.

  5. Miranda Stewart

    I would not really go to the barricades to support the “non-obese” in accessing soft drinks – why not tax it – one reason is regressivity, or possibly, less social harm than substitutable alternatives. NB, in the UK, they add fighting tooth decay to the list of positives, as a sugar tax made it into the budget, and was passed by the Parliament before the early election> . The Institute of Fiscal Studies is agnostic about the effects which all depend on elasticities… See

Leave a comment

Your email address will not be published. Required fields are marked *