Back in January we were told that Mexico’s tax on sugary drinks cut consumption by six per cent in its first year. This received global news attention at the time because it was seen as proof that sugar taxes ‘work’. Such taxes really only work if they reduce obesity, of course, but given the virtually non-existent impact of previous soft drink taxes, campaigners could be forgiven for their excitement about some empirical evidence finally coming their way.
The figure came from a study in the British Medical Journal conducted by one of the world’s most prominent sugar tax campaigners, Barry Popkin, which modelled soft drink sales against a theoretical counterfactual, which is to say it compared sales after the tax with what the authors thought would have happened if the tax had not been enacted. There is always a degree of ‘garbage in, garbage out’ with economic forecasting (which is essentially what this is, albeit retrospectively), but the results did not strike me as unrealistic at the time. Standard economic theory suggests that higher prices lead to lower demand.
It seemed to me that the problem with Mexico’s soda tax was not that it didn’t reduce consumption, but that it reduced it by a trivial amount at a significant cost to consumers. Here was a relatively poor country introducing quite a large tax on a fairly small component of the country’s calorie supply and making a little dent in it. At best, the BMJ study suggested that it had reduced calorie consumption by the equivalent of one sugar cube a day which, as Tom Sanders said at the time, is ‘a drop in the caloric ocean’.
In recent months there have been reports of sugary drink sales in Mexico bouncing back and some have even claimed that sales never fell to begin with. Getting the actual data has been surprisingly difficult up until now, but thanks to Mexico’s National Institute of Public Health (NIPH), we have it — and it tells a rather different story.
The tax was implemented in July 2014. According to the NIPH — which keenly supports the tax — annual sales of sugary drinks averaged 18.2 million litres between 2007 and 2013. In 2014, this rose to 19.4 billion litres, and it rose again in 2015 to 19.5 billion litres.
On the face of it, sugary drink sales were seven per cent higher last year than they were before the tax was introduced. This is not good news for Jamie Oliver, but these figures need to be adjusted for population growth. Using the correct measure of per capita consumption we get the following results:
2007-13: 160 litres
2014: 162 litres
2015: 161 litres
This is still not good news for Jamie Oliver and so the NIPH asks the public to trust regression models that have made further adjustments to the data for possible confounding variables such as climate and economic growth. As with the BMJ study, these models suggest that soft drink consumption would have been higher if there had been no tax. Perhaps it would, but there is clearly a difference between arguing that sales would have been higher if the weather had been colder or the economy had been sluggish and asserting, as the NIPH does, that ‘there was an average decrease of -6% in 2014 and -8% in 2015’.
Without these adjustments, the (erroneous) conclusion would be that there was an increase in sales after the tax, both in 2014 and in 2015, while the adjusted analysis shows that there was a reduction in sales after the tax.
At this point, words are beginning to lose their meaning. It is rather like the supporter of a losing football team claiming they would have won if their striker hadn’t been sent off with ten minutes to go. It could be true, but no one will ever know. The fact remains that they lost, and while you might be sympathetic to a supporter who claimed they should have won, you would worry about their state of mind if they claimed — as matter of historical fact — that they did win.
Imagine the hilarity at the Mexican tax office if Coca-Cola and Pepsi-Co requested only paying tax on what public health campaigners said they sold in 2015 rather than on what they actually sold. If they told the government that their recorded sales figures were ‘erroneous’ on the basis of a speculative regression model I suspect they would be met with a terse reply.
Modelling of this sort has its place, but we are in danger of replacing fact with theory. The appeal of the Mexican sugar tax as a global precedent is that it led to a reduction in sugary drink consumption and therefore, it is supposed, a reduction in calorie intake. Claiming that it led to a six per cent reduction in fizzy drink consumption is an unambiguous normative statement that has been repeated around the world.
We now know that statement would be more truthfully phrased as ‘consumption rose by one per cent after the tax was introduced but some researchers have hypothesised that it would have fallen by six per cent if x, y and z had occurred’. We also now know that there is no chance of the tax reducing obesity because, whatever the models might say, the empirical fact is that Mexicans are drinking more sugary drinks than they did before.