Wednesday, 10 June 2020

Alcohol sales after minimum pricing

I had a slight sense of déjà vu this morning when I saw this story...

Alcohol off-sales drop credited to minimum unit pricing

There has been a decline in alcohol consumption in Scotland's population, official data suggests. 

A report published by Public Health Scotland said there had been a reduction of between 4% and 5% in Scotland's shops in the year after minimum unit pricing was introduced.

Haven't we been here before? Ignore the reference to the recently formed Public Health Scotland. This is the MESAS evaluation again. 

This time last year, MESAS reported a 2.9 per cent drop in alcohol consumption in Scotland in 2018. Not a particularly impressive statistic, but the media nevertheless treated it as the final proof that minimum unit pricing (MUP) 'works'. 

What does the new study add? Well, it only looks at off-trade sales, for a start. And it looks at the full twelve months after minimum pricing was introduced, rather than the calendar year of 2018 (MUP began in May 2018). As it is an evaluation of minimum pricing (which only affected the off-trade), this is good.

It also makes various adjustments to the data, using England and Wales as a control group. We'll come to that in a moment, but let's first look at the findings.

The report finds that off-trade sales fell by 2.6 per cent in the first year of MUP (compared with the year before). Spirits sales fell by 2.4%, wine by 1.3%, cider by 17.4% and beer by 0.7%. Sales of fortified wine rose by 6.7% and the sale of ready-to-drink beverages rose by 12.3%.

In England and Wales, meanwhile, there was a 2.3% rise in off-trade alcohol sales. Put very simply, if you assume that the trend in Scotland would have been similar to that of England and Wales in the absence of MUP, the impact of MUP was to reduce alcohol sales by 4.2%. That is the headline figure in the report. Figures of -4.5% and -4.8% also appear, depending on adjustments. This is the source of the claim that there has been 'a reduction of between 4% and 5% in Scotland's shops'.

If you need a control group for a natural experiment of this kind, England and Wales are as good as it gets, culturally and economically, but if you look at the trend over the years, they are not a great counterfactual. There have been several years in the recent past when consumption has gone up in Scotland and down in England and Wales, or vice versa. The trends don't always go in the same direction, let alone at the same rate. 

Nor should you expect them to. Beer sales seem to have risen in the summer of 2018, for example, and the authors of the new report note that this was probably influenced by a heat wave. They argue that this helps explain why overall beer sales barely fell in Scotland in the first twelve months of MUP. However, they note that there was a 4.9% increase in beer sales in England and Wales and conclude that 'the legislation was associated with lower off-trade beer sales in Scotland than would have been the case in the absence of MUP.'

Perhaps so. It was certainly a hot summer, but it was hotter in England than it was in Scotland, and there was a World Cup that year that England was in and Scotland wasn't. You would therefore expect more beer sales in England regardless. You can't treat England as if it was just Scotland without MUP.  

A further problem is that sales in Scotland are not wholly independent of sales in England. Many people in Scotland have reported going on booze runs to Carlisle and Berwick since minimum pricing was introduced. It is difficult to estimate how much alcohol sold in England is consumed in Scotland, and nobody has really tried, but it has surely increased since May 2018. This will have inflated the sales figures for England and reduce the sales figures for Scotland.

The authors of the new report say it is 'highly unlikely that cross-border purchasing could account for the net effect of MUP reported here.' Of course it doesn't account for all of it, but how much does it account for? We will probably never know.

Finally, there is the issue of the sales figures themselves. They do not all come from till receipts, by any means. A lot of them are the result of modelling, estimating and adjusting by the authors. I wrote about that in this briefing paper. It is not a trivial issue.

All the same, based on the evidence published so far, I think sales probably went up a bit in the summer and went down a bit overall. The claim in the BMJ last year that sales fell by 7.6% in the first eight months was activist junk, but a decline of some magnitude is no more than you would expect from the law of demand. It doesn't seem to have been a large decline, and any divergence from England and Wales had no effect on alcohol-related mortality - which, let's not forget, is supposed to be real target of the policy.

But the authors go too far when they start making claims like this:

In unadjusted analysis, the introduction of MUP was associated with a 6.7% (3.1% to 10.5%) increase in the volume of pure alcohol sold as fortified wine per adult in Scotland. In England & Wales, there was a 6.0% (-8.2% to -3.8%) decrease over the same time period. In the unadjusted, controlled model, MUP was associated with a 4.8% (0.4% to 9.3%) increase in off-trade fortified wine sales in Scotland, which increased slightly after adjustment for disposable income and substitution (5.7% (1.3% to 10.3%)).

So there was a 6.7% rise in fortified wine sales in Scotland and a 6% decline in fortified wine sales in England and Wales. Given that England and Wales are supposed to be the control group, this implies that MUP led to a really big rise in fortified wine sales. And yet the Scottish figure goes down after adjustment??

As it turns out, even the 6.7% rise in the 'unadjusted analysis' was the result of adjustments:

...we were able to take into account underlying trends in the data series through the analytical method employed. This allowed us to strengthen the interpretations we made in our descriptive analysis and more confidently isolate the impact of MUP. For example, based on our descriptive analyses, we reported that off-trade sales of fortified wine increased by 16% in the year after MUP was introduced.

So it actually rose by 16%.

However, we also noted that this was a continuation of an already upward trend. Using SARIMA in this study, thereby adjusting for this existing trend, we found that MUP was associated with an increase of less than half of this magnitude.

Once they adjust for 'seasonal and secular trends', they manage to show that minimum pricing led to a 5.3% decline in the sale of fortified wine! 

I'm sorry, but that's just being silly. Even people who work in the drinks industry can't predict next year's category sales based on 'existing trends' and if there's one thing everybody agrees on, it has that minimum pricing in Scotland has led to a rise in fortified wine sales. Buckfast, in particular, is an obvious substitute for strong cider and was already sold at more than 50p per unit before minimum pricing was introduced. Its sales reached a record high last year and it has overwhelming dominance in the fortified wine category.

Minimum pricing campaigners lie, but sales figures don't. Any analysis that turns a 16% rise into a 5% decline is highly suspect.

I've been relatively impressed with MESAS evaluation so far (I have low expectations), but this kind of black-is-white overreach does it no credit.

No comments: