Sweet talk

Last night, I chatted with Bryan Crump on Radio New Zealand Nights about sugar taxes. When I pitched the topic a couple of weeks ago, I hadn't pegged that it would have been quite so topical in New Zealand. I was rather happy to be talking about sugar taxes the same night that the government here shot them down. I had worried that they might follow Britain's lead.

Next week, Jenesa Jeram launches her report on lifestyle regulations. Her report covers smoking and vaping, drinking, and sweet sweet sugar. The panel discussion includes IEA Fellow Jaime Whyte, Maori Party co-leader Marama Fox, and Treasury's Chief Economist Girol Karacaoglu. There may still be a couple of seats available; do come and join us if you're interested.

I usually put together a few notes for myself before my chats with Bryan. I've copied these below. They provide some of the references to studies I mentioned in the chat, and bits that we didn't get to talk about as well.

Sweet thinking: Notes for discussion with Bryan Crump, 12 April 2016

Tax and Choice

What’s the evidence on the effects of sugar taxes on consumption in the first place?

It depends on what you mean by ‘effect’. If what you want is a reduction in consumption of the taxed thing, taxes can do that. But if you want a reduction in obesity or improvement in other health outcomes, that’s harder to find.

Why? People respond to taxes in ways you might not want. If you tax sugar-sweetened beverages, but not other sugary things, then you can get a drop in SSB consumption but a rise in consumption of other sweet things, or other tasty and fattening things. You can then succeed in reducing SSB consumption but not in affecting obesity. Is it then a success or not?

Evidence? Work by Fletcher and co-authors found that soda taxes led to a moderate reduction in soft drink consumption by children and adolescents, but that this was entirely offset by increases in consumption of other high-calorie drinks.

This then leads to big problems when people start talking about the potential effects of soda taxes. It’s easy to find studies showing some effect of soda taxes on soda consumption. And that then leads folks to figure that the tax would reduce total caloric intake by that amount, but that’s just wrong: people eat other things instead. Could be other sugary, but not fizzy drinks; could be other sugary non-drink things; could be other stuff entirely. The effect of the tax on health will then depend on how much people respond to the tax, and what they shift to. Good studies will look not just at the effect of taxes on consumption of the taxed thing, but rather on obesity outcomes. And there just isn’t much there there.

But, it gets even worse.

There aren’t that many surveys out there that measure how much soda people drink. But what we do have is the Household Economic Survey. Respondents there say how much their household spent on different categories of expenditure. If you know how much a household spent on soda, and divided that by the going price for soda, you might think you have a measure of how much they consumed. And some studies use that to measure the effect of price changes on consumption. But one thing that people can do when prices change is decide to buy a discount brand instead of a named brand. The supermarket’s own brand of soda costs a lot less than Coke or Pepsi. If people shift to a lower cost product when prices go up, then you can’t really infer consumption from expenditure any more. If, after a price change, you see expenditure went down, you can’t tell if that’s because people are buying less in total, or because they’re consuming just as much as before (or possibly even more) of a discount brand instead.

If we want a specific NZ example, a HRC-funded study in the NZMJ estimated that a 20% tax on soda would avert or postpone 67 deaths per year. But that was all based on an assumed reduction in the quantity of soda consumption that comes with a 20% price change, out of that HES data. We can't really say what's happened to total consumption of something in response to a price change if the only thing we know is how much they spent on it – if they can shift to lower quality and cheaper options.

All of this means that people strongly overestimate the effectiveness of soda taxes, and do not think hard enough about what effectiveness means.

Mexico's sugar tax has been more comprehensive, with a peso-per-litre tax on sugar-sweetened beverages (not including alcoholic ones) and a relatively small tax on other sugary solids. There are a few holes in how they estimated things, but let's just take the results at face value because the technical stuff depends on whether you believe Duan smearing is appropriate in heteroskedastic data [seriously – just too technical to get into].

First, a peso per litre sounds like not very much, but if we compare it to the daily salary earned by the poorest workers in their study, it works out to 1.7% of the daily wage. The equivalent in New Zealand, as compared to a day's work at the minimum wage, would be $2 per litre. A $2 per litre tax here would likely have pretty large effects on consumption, especially among poorer people.

What did the study find? People in poor households reduced their consumption by a little over a litre of soda per month: 600 mL every 17 days. There wasn't really much effect among richer households.

And they have no way of knowing whether poor households shifted to buying sugar (which while still taxed, is more lightly taxed than sugar in a litre of soda) and making lemonade at home.

Under that kind of tax, the first thing I'd do if I were a manufacturer would be to start selling soda concentrate for people to mix at home. Why? Because 1 litre of concentrate that could make, say, 10 litres of soda attracts the same tax as 1 litre of diluted product.

What happens then when we look at the studies that do properly look at the effect of prices on obesity?

A paper in Health Economics looked at soda taxes across U.S. states. Now there aren’t really “soda taxes” per se but soda is often, but not always, exempted from generalised food exemptions from state-level sales taxes. Because of that, soda can have tax rates, relative to other food products, that range from 0 to 12%. They found no relationship between soda taxes and soda calorie intake but a small increase in caloric intake from other beverages.

BioMed Central in 2013 published a metastudy (compilation of the effects reported in lots of different studies) looking at the effects of SSB taxes. Most studies found no effect distinguishable for statistical noise. But the biggest one there reported that did find an effect argued that a 20% tax on soda would result in a 0.065 reduction in average BMI. Now the healthy BMI range goes from 18.5 to 24.99. A 0.065 change, in a range where "healthy" covers a range of 6.5 units, really isn't much. If we take someone of my height, it would be the equivalent of about 180 grams. That's maybe the difference you get if you weigh me after a meal instead of before a meal: not enough to make any kind of difference.

But even that is very likely an overestimate. Why? They looked at how a panel of households changed what they bought at the supermarket when the prices of things changed. If Coke were on special, people would buy more Coke. And so on. And that would make sense if people bought things for immediate consumption. But you can store the stuff after buying it on special, and that will look like people are highly responsive to prices – but it's not their consumption behaviour. It's purchasing when cheap for consumption over time.

So, why does the data show so little effect?

First, taxes are pretty low. Everybody says "Oh, well, taxes worked to reduce smoking, didn't they?" And they did – although the most recent tax hikes haven't done much in New Zealand. But remember that a $20 package of cigarettes has about $8 in cigarette costs and $12 in tax: the tax is 150% of the cost of the cigarettes. Is anybody talking about soda or sugar taxes anywhere near 150%? No. There's "Well, we have to start somewhere" talk, citing the tobacco precedent, but if the ultimate goal is to get to a soda tax that's up around the $1.50-$2.00 per litre mark, then I wish they'd be honest about it.

Second, soda's just different from tobacco. If you're a smoker and you're addicted to nicotine, and the price goes up, there aren't really other things you can easily switch to if you want to keep getting nicotine. You can switch down to cheaper cigarettes, and if you can navigate the crazy regime New Zealand has for importing nicotine cartridges for vaping, then you can do that. But otherwise, there aren't things people can easily shift to, so reductions in consumption mean real consumption drops. But if there's a tax on soda, there are plenty of other sweet things that people can switch to.

So, what happens if we just tax all the unhealthy things and subsidise all the healthy things?

There was a great field experiment run by Cawley, Hanks et al. Participants got a purchase card they could use at one grocery store, where all their purchases would be tracked. To encourage everyone participating to do all their shopping at that store, they gave everybody a 5% discount on any food that had any kind of health-star rating (healthy or unhealthy). During the experimental treatment period, some people then got a 15% rebate on everything combined with a 10% tax on unhealthy food, or a 5% rebate on everything combined with a further 10% subsidy for healthy foods. They amount to the same thing but the researchers wanted to see if framing mattered.

They found no significant effect on actual purchases overall, but an interesting puzzle among poorer households. Those poorer households wound up spending much more on both nutritious and less nutritious foods (spending on both went up by about $7 per week): the subsidy meant they could afford more of both, and so there wasn't any shift towards healthier eating.

This suggests that, among poorer households, the main effects of these kinds of taxes or subsidies runs through their ability to afford things in general: subsidising healthy foods may let people afford more unhealthy stuff, and taxing unhealthy foods may reduce consumption of healthy foods as people tighten their belts.

A bigger puzzle

Obesity is increasing in New Zealand and has been for a while. But if we look at the data from the 2008/2009 New Zealand Adult Nutrition Survey, the most recent one, and compare it with the 1997 survey, total caloric intake is down, and sugar consumption was either down or unchanged, depending on which kind of sugar you looked at and whether it was for men or women. See Table 9.3 at the linked report. But mean BMI increased by about 1.5 points over the period, and the proportion of overweight and obese people increased from about 50% to about 60%. If sugar is the villain, why did obesity rates jump while sugar consumption stayed stable or declined a bit?

But what is the justification?

If the problem is costs to the health care system, then we need to be careful: there's some decent international evidence that, over the long term, healthy people are the most costly: they live longer and collect more in social security.

But even without that, is the government really justified in forcing you to be part of a public health system, and then taxing and regulating you to make sure that you impose the smallest cost possible on that system? Why do we only ever seem to target health behaviours that are disproportionately engaged in by poor people? Smoking, soda, and the kinds of cheaper alcohol that are targeted by public health campaigners – those are predominantly consumed by poorer people. But consider rich people activities that have rather high per-activity costs: horseback riding, mountaineering, scuba diving, sailing, mountain biking and adventure kayaking. ACC costs of this kind of thing add up to less because fewer people do them. But 2007 stats had almost 3% of horseback riders making an ACC claim.

Further, if it's obesity and obesity's costs that are really the thing people want to worry about, and if people respond differently to different diets, then does it make more sense to tax sugar or to tax obesity directly? It sounds terrible, but the Washington Post's Catherine Rempell made the case last month. She wasn't making the case seriously: she was pointing out an underlying problem in the arguments around fat and soda taxes. If taxing obesity directly is unfair because obese people cannot control their weight, then taxing sugar can't get rid of obesity.

But all of this comes back to what works, and what you think works means. Suppose you think other people consume differently than you do because they just don't know any better. If you explain things slowly, and they don't change their behaviour, should you conclude that your information campaign failed and that you then need taxes or other harsher measures? Or that the people you were talking to never really expected that soda was a health food in the first place and didn't really want to change their behaviour?

Crossposted from The Sandpit