Bad stats confound sugar tax battle

Even the consultants producing the research that the soft drink industry has used in its campaign against the sugar tax don’t think the statistics are realistic.

The vociferous attack on Treasury’s proposed tax by the Beverage Association of SA (Bevsa) is based on work commissioned from consultancy firm Oxford Economics.

Over the past month, Bevsa has loudly touted the “potential” economic damage as amounting to 60 000 job losses and a R14 billion drop in GDP.

The report from Oxford Economics was released this week and shows a far more cautious interpretation.

The GDP number is not correct simply because it assumes the potential loss to the soft drink makers will not be mitigated at all by consumers buying other things.

There is, however, a far more fundamental problem with trying to predict what the tax will do, Oxford’s lead economist, Nick Stewart, told City Press.

“Do we think this will be the impact on jobs overall? Probably not,” said Stewart.

The flip side was that the pro-tax side of the debate also could not predict any concrete health benefits, he said.

Stewart emphasised that the work had to be done in a rush due to the short timelines set by Treasury. More realistic research is possible, but would take time and more data.

“This is not a forecast, it is not an estimate, but it is transparent in the approach.”

Data allowing economists to actually try to predict what the sugar tax would do simply did not exist, he said.

François Conradie, research head of NKC African Economics, a local subsidiary of Oxford Economics, said: “The main conclusion of the work is that we do not know enough and more work is needed.

“You’d expect some effect, but it is impossible to credibly forecast what it would be. We don’t have the elasticities.”

At the heart of the research Bevsa bases its claims on are a set of what economists call elasticities – the expected reaction to a change in price.

The ones used by Oxford were borrowed from a research article published by local academics last year and cited by National Treasury in its tax proposal.

The largely pro-tax academics who used it then to model the potential reduction in obesity due to the tax explained at length that it has major shortcomings.

The anti-tax researchers say the same, creating a situation where a bad economic model is consciously used to counter another bad economic model, both suffering the same flaws.

The elasticities come from studies in other countries, include illogical numbers and also fail to fully measure where spending will get diverted to if the tax succeeds in making people buy less soda.

The basic assumption is that a 1% price increase leads to a 1.3% decline in demand.

Stewart says it is very high.

“It is the first time I saw something of that magnitude.”

The report for Bevsa also uses a “counterintuitive” elasticity for diet soft drinks, whereby their sales also fall, and has no estimates for switches to any number of other things people might buy instead of soda.

“While such behavioural responses seem plausible, they are difficult to quantify in a statistically robust sense,” reads the Oxford report.

“What is crucial for this is cross-price elasticities. That basically drives it – but we don’t know what they are,” said Stewart.

“If consumers shift away from Coca-Cola, do they shift to a local pop, do they shift into a diet cola, do they shift into 100% juice, do they shift into cordials?

“What is crucial for this analysis is the shift to a smaller or larger volume size.”

An attempt to massage the implication of these uncertainties out of view are apparent in the Oxford report as well.

City Press was given a slightly earlier version of the document under embargo. In that version, the authors note that the research that exists could be read to “imply that overall consumer spending on soft drinks would remain broadly unchanged after the tax”.

It was also more upfront about the questionable elasticities, saying they “may not be appropriate for the South African soft drinks market”, an admission that undermines the whole model.

These passages were removed from the final version.

The Bevsa numbers are also based on a massively simplified assumption about how the industry would react to the tax.

Stewart admits that the model assumes that the industry does nothing to mitigate the effect and so the analysis starts at retailer sales. It is assumed that drink makers don’t lower their profit margins, and don’t push smaller units that would carry lower tax rates. It is also assumed that they won’t reformulate their drinks to use less sugar.

“It is a starting position ... given the length of the consultation period that was opened,” said Stewart.

“If that [reformulation and portion reduction] is going to happen, how quickly can that happen? It is not something we’ve looked at.”

Then there is another fundamental problem with the industry forecast, which would take a lot of extra work to address. All the forecasts are based on the assumption that sugary drinks retail at on average R10.50 a litre and that the tax would raise their prices by
about 25%.

This is true, but doesn’t reflect how the tax will actually work.

Because small units like cans are sold at a massive premium, their tax rates will be low – about 9% for a can of Coke. The opposite is true for 2-litre bottles, where the rate would be about 33%.

If you average this out and then apply the model, you get a completely different result from what you would get if you did each size separately and then added it up afterwards, which is more realistic.

The result is that the Oxford report likely overstates revenue losses, but understates volume losses.

“Ideally, what you would want to know is what the response is for each product line per pack size,” said Stewart

“It would be a huge number-crunching exercise. We would be delighted to do it, if we had the time. A lot of the information we are using here is not even South African.”

With the absence of a recent large-scale national study of dietary habits, it is hard to counter the industry’s arguments with numbers rooted in reality.

“How can you even measure the outcome of this if you don’t know where you are starting from?” asked Stewart.

Brent Crude
All Share
Top 40
Financial 15
Industrial 25
Resource 10
All JSE data delayed by at least 15 minutes morningstar logo
Company Snapshot
Voting Booth
Do you think it was a good idea for the government to approach the IMF for a $4.3 billion loan to fight Covid-19?
Please select an option Oops! Something went wrong, please try again later.
Yes. We need the money.
11% - 937 votes
It depends on how the funds are used.
74% - 6261 votes
No. We should have gotten the loan elsewhere.
15% - 1286 votes