A clarification on sugar taxes

A few weeks ago, Rachel Thomas at Stuff included a short bit from me on sugar taxes and a new WHO metastudy. 

The Spinoff provided a perhaps too short a summary of that in its daily newsletter. I'd sent them a brief note of clarification, figuring that it was too late to change anything since it's an emailed newsletter. But it's online, here. 

The question of a sugar tax cropped up again earlier this month. Rob Beaglehole from the NZ Dental Association argues we need to be proactive with a sugar reduction strategy because of the woeful state of water fluoridation in New Zealand. The New Zealand Initiative’s Eric Crampton says the evidence doesn’t support the argument that a tax would reduce consumption.

I'd sent a note through to them, in case the issue came up again. I'll copy it here, mildly edited:

Thanks for the mention in today’s bulletin.

No need to do anything about it, but more for any future ones.

The problem with a ton of sugar-tax studies is that they assume that any drop in spending on soda (often the only data they have – they know spend shares, but they don’t know consumption) means a drop in consumption. Some of the drop could be a drop in consumption, but some of it could be a shift to downmarket product or product that’s lower cost-per-unit.

Again – this problem comes in when researchers are using household spend data that surveys people about how much they spent on various things over the past period, then combines that spend data with some measure of average prices to try to estimate consumption. 

If you see that someone shifted from spending $20/week on soda before a tax to spending, say, $18 in total after a tax, is it because:

a) They had been spending $1/can before the tax and bought 20 of them per week, and reduced that to 9 cans at $2/can after the tax – huge decrease in consumption!

b) They had been spending $1/can before the tax and bought 20 of them, and shifted to buying four 2-litre bottles of home-brand soda at $4.50 after the tax – small increase in consumption!

c) Some unknown combination of the two?

The studies that have only spending data effectively assume that (a) is the only thing that’s happening. (c) is the most likely thing, but the studies can’t see it. 

So it isn’t that a tax doesn’t decrease consumption. It is likely to decrease consumption somewhat. It’s more that the campaigners assume gigantic effects on consumption, because when many people are doing (c) and some are only doing (b), the studies assume it’s all (a).

It’s a technical issue in how the estimates are undertaken. They don’t handle things well when there’s adjustment not just in quantity but also in price-point. John Gibson over at Waikato has demonstrated the problems with it. But so much of the literature just ignores the problem. And the WHO meta-study is just a complicated average across studies that did not exclude studies known to have this problem. An average that includes overestimates is going to be an overestimate.