Friday, 28 November 2014

A belated reply to the minimum pricing modellers (part two)

Continued from yesterday...

Benefits of moderate consumption

Duffy and I say in the ASI paper that the Sheffield model ignores the health benefits of moderate alcohol consumption which, if accounted for, could reduce the putative benefits of minimum pricing. In their rebuttal, they deny this and say that various protective effects are included in one of their Lancet papers.

I dare say this is true. Six years of modelling has produced so many pages of estimates that it is rather like the Bible in that you can find support for any position if you look hard enough. Protective effects have certainly not been consistently accounted for. I forget which of the numerous editions of the model the graph below comes from, but it indicates no health benefits from moderate consumption. The curve below should be J-shaped, with a risk below 1.0 for much of the right-hand side of the graph.

Petty swipes

This is a relatively minor point, but it is indicative of the general tone of the Sheffield response:

When discussing price elasticities, Duffy & Snowdon also state "minimum pricing will raise the cost of every type of drink", and link this assertion to claim in a blog (rather than a peer-reviewed article) that "heavier drinkers are least responsive to aggregate changes in price" of this kind because they simply substitute their previous purchase for a cheaper option.

The blog post in question was written by Dr Eric Crampton, a professional economist who was then teaching at New Zealand's University of Canterbury. I quoted it because Crampton nicely explained an issue about price elasticities in words that a layman might understand. As anyone who has read the 500+ pages of minimum pricing reports knows, the Sheffield team feel no such obligation to speak plainly, but so what if an economist has explained things in a blog post? It's 2014.

Needless to say, the Sheffield response does not respond to the substance of what Crampton says.

Basic misunderstanding

Ignoring Crampton's point about cross-elasticities, the Sheffield team instead build a straw man to evade addressing my point:

When discussing price elasticities, Duffy & Snowdon also state "minimum pricing will raise the cost of every type of drink" ... Duffy & Snowdon appear to misunderstand both the policy and the modelling they are attempting to critique. The proposed minimum pricing policy would not enforce increases in the cost of every type of drink; it only directly affects alcohol sold below a given price per unit.

Obviously I understand what minimum pricing is. When I said that it will increase the price of every type of drink, I meant that it would affect spirits, wine, beer and alcopops, not that it would increase the price of every brand on the market. My meaning is perfectly clear from the context.

Bald assertion

They write:

Duffy & Snowdon assert that there is "oddly enough not enough information for a third party to rerun bits of the model". We are unsure if our critics have actually tried to reproduce any of our work, but we reject the view that this is not possible.

They can reject it all they like, but it is true. This criticism was penned by John Duffy but I had also noticed that the Sheffield model, though awash was figures, always breaks down when it comes to showing how the figures fit together to produce the final estimate. If the model could be rerun based on the available data, Duffy would be able to do so. If I may briefly resort to the kind of appeal to authority that riddles the Sheffield response, Duffy is a professional statistician, not someone who took a two week course in epidemiology as part of a public health degree.

Regressive impact of higher prices

They write:

Duffy and Snowdon contend that the effects of MUP on "people on low incomes" are important and we would agree that it would be useful to undertake further research to examine this. Of course this would only be possible by undertaking a model-based appraisal.

This is like saying that it is only possible to work out whether minimum pricing will increase the price of cheap alcohol by conducting a model. Of course it will. Likewise, it is very obvious that minimum pricing will act like an indirect tax and therefore be regressive. This is doubly true because it will affect beverages that are disproportionately bought by people on low incomes. This is really undeniable and in the end the Sheffield response resorts to the familiar 'public health' rhetorical device of redefining 'regressive' to mean any non-economic benefit to the poor, ie. putatively better health. This is clearly not the issue when people say that minimum pricing will be regressive. It is another diversion. 

Since publishing their response, the Sheffield researchers have published yet another version of their model. This predicted that the financial impact on low income and moderate drinkers will be less than previously thought. The 2010 version of the model estimated that a 45p minimum price would cost drinkers an average of £29.30 per year. The 2014 model, by contrast, claimed that the average cost per drinker would be just £2.12 per year.

The 2010 model projected that minimum pricing would cost 'harmful' drinkers, who purchase approximately ten units a day, £137.40 per year. By contrast, the 2014 model claimed that the same 45p minimum price would save them £4.01 per year! Low income harmful drinkers did best of all from the new projections, saving £34.63 per year. These are radically different projections for a model which is suppose to be solid, but they had the happy effect of producing headlines claiming that minimum pricing would not hurt the poor after all (which had been the major objection in Westminster).

The claim that drinkers will save money as a result of higher prices appear is counter-intuitive and almost certainly wrong. Although it is plausible that higher prices might make drinkers cut down their intake somewhat, it is difficult to imagine them deciding to spend less on alcohol overall. Even if heavy drinkers do not maintain their alcohol intake by spending more on alcohol - they already spend £2,685 per year according to the model - there is no obvious reason to think that they would actually reduce their alcohol budget.

The 2014 model was only able to show savings for some drinkers by relying on exceptionally high price elasticities for beer and cider. The Sheffield researchers note that ‘as off-trade cider accounts for a sufficiently large proportion of the alcohol spend by "low-income male harmful drinkers", the high elasticity of this beverage type leads to these consumers’ overall spending on alcohol falling under a 45p MUP’. It certainly does, but the model makes no distinction between the price elasticity of super-strength white cider and traditional scrumpy. Clearly, the two products have very different customer demographics whose demand for alcohol differs greatly. It is, to put it mildly, questionable whether drinkers of White Lightning are as price sensitive as the model assumes.

Low income moderate drinkers - the group which politicians express the greatest concern for - went essentially untouched in the 2014 model, paying a mere 4p a year more. Moderate drinkers as a whole were only projected to spend an extra 78p more per year, in contrast to the £8.70 projected in 2010. This requires some peculiar assumptions. In the new Sheffield model, a ‘moderate drinker’ is someone who drinks just 5.5 units per week, the equivalent of two pints of lager. A low income moderate drinker consumes even less, 4.6 units. This is a fraction of the government’s rather austere recommendations for healthy drinking (21 units per week for a man) and more accurately describes a very light drinker rather than a moderate drinker.

The claim that minimum pricing will have a negligible impact on moderate consumers therefore relies on a misleading definition of moderate consumption and unrealistic price elasticities. It also relies on the assumption that the minimum price will be 45p. In reality, no one is talking about a 45p minimum price any more. The costs would rise considerably if the model looked at a 50p price, as is being proposed in Scotland.

On being confounded by reality

In our ASI report, we point out Britain has seen a much bigger drop in alcohol consumption since 2006 than the Sheffield model predicted would result from a 50p minimum price and yet we have seen a much smaller decline in alcohol-related deaths than the model predicted:

All the projections in the SAPM are based on a 50p minimum price reducing per capita alcohol consumption by 6.7 percent from the 2006 level. But we know exactly what would happen if alcohol consumption fell by 6.7 per cent from the 2006 level because we have lived through it. But we know exactly what would happen if alcohol consumption fell by 6.7 per cent from the 2006 level because we have lived through it. Indeed, the decline in per capita consumption since that year has been closer to what the SAPM predicts would happen under a 70p per unit regime (ie. a 17.5 per cent decline). According to the model, the kind of reduction in alcohol consumption that Britain has already experienced should have reduced the number of alcohol- related deaths by 1,273 (28.3 per cent) in the first year, rising every year until 7,263 deaths (62.4 per cent) are prevented each year by 2015 (Purshouse, 2009; pp. 109-111). None of this has happened.

The ONS figures show that there were 8,221 alcohol-related deaths in the UK in 2004 when alcohol consumption was at its peak. By 2011, when we wrote the report, there were 8,748. The rates per 100,000 people were 12.95 in 2004 and 12.7 in 2011. In other words, there was essentially no change.

There was more of a decline in 2012 (the latest year for which we have data) which may or may not turn out to be a blip. The 2012 figure is 11.85 per 100,000, so the best that can be said is that there was a decline of 8.5 per cent in alcohol-related deaths (as a proportion of the population) between 2004 and 2012 during which there was a 17.5 per cent decline in alcohol consumption. This is a far cry from the claim that a 6.7 per cent reduction in consumption would lead to a halving, or more, of alcohol-related deaths.

The Sheffield team responds to this in three ways. Firstly, they say that the type of decline they predict is different from the decline that has actually taken place. Their decline, they say, would be the result of 'larger reductions in high-risk groups and smaller reductions in low-risk groups'. This is all well and good, but an 17.5 per cent decline in really rather large and it should, if Tim Stockwell is correct, have a significant effect on these 'high-risk groups'. As mentioned in the previous post, he says...

'...when total consumption of alcohol in the population declines, consumption among heavier drinkers is reduced and, further, rates of alcohol-related mortality also decline. The Sheffield Model applied these general principles specifically to the UK and provided numerical estimates of the benefits.'

Their second argument is that there is a lag effect between alcohol consumption and mortality, ie. the effects of reduced drinking do not all show up immediately. This is true as far as it goes and some of the Sheffield team wrote quite a good article about the time lag in 2012. Reviewing the literature, they found that, for liver cirrhosis, 'In all cases the greatest single year effect was seen in the first year with effects of declining magnitude seen in subsequent years.' Data shown below:

The effect on heart disease shows a similarly short time lag, with the biggest effect in the first year and the full effect shown by year 5.

With regards to suicide, Holmes et al. looked at ten studies and found that 'In all studies, an immediate effect of changes in aggregate consumption on the suicide rate was found. In five of these studies, the total effect occurred in the first year and no lagged effects were identified'. For other harms, all studies found an immediate effect and none found an effect after more than five years.

Between 2004 and 2009, there was a 15 per cent decline in alcohol consumption. We are surely entitled to have seen a large and positive impact on health outcomes from this decline by now if the fundamental beliefs of the medical temperance movement are correct. Indeed, we should have seen most, if not all, of the effect by 2011. We haven't.

The final argument of the Sheffield team is that some alcohol-related diseases are multi-causal.

Their proposed methodology implicitly assumes any change in mortality for causes partly related to alcohol consumption must be wholly due to changes in alcohol in alcohol consumption... For many diseases (such as colorectal, breast cancer and oesophagal cancers), alcohol is one of several risk factors (such as obesity, diet and smoking) that can affect the mortality rate. So, for example, if there were an increase in obesity then there might be an increase in the number of deaths from colorectal cancer even if alcohol consumption had decreased.

This sounds rather like the Sheffield researchers are getting their excuses in early in case minimum pricing is introduced and alcohol-related mortality stubbornly refuses to fall. In any case, their argument is not germane to the data we look at which, as the ONS says, 'only includes those causes regarded as being most directly due to alcohol consumption [and] does not include other diseases where alcohol has been shown to have some causal relationship, such as cancers of the mouth, oesophagus and liver.

And that's it. Sadly, the Sheffield team didn't respond to our mention of the Panorama balls up...


Christopher Snowdon said...

You beat me to it, DP. That was exactly what went through my mind as I was reading that paragraph. Of course all prices will go up, most significantly at the lower end of the market and gradually tapering off as the products get more exclusive and expensive. My daily tipple, a local Merlot, costs me the equivalent of about €1.50 a bottle. MUP would triple that cost. Thank the Gods that I live in a country that is singularly lacking in busybody killjoy puritans!

I struggle to comprehend how 50 units a week is considered 'hazardous'. I consider myself a fairly moderate drinker. I almost never drink during the day, and usually have my first glass at 6ish, and yet my weekly intake would be 60 - 70 units, which according to them is beyond hazardous and into full-blown alcoholism! I know they plucked the figures out of thin air, but their recommended weekly units intake is just risible.

Christopher Snowdon said...

Excuse my ignorance, but when we talk about 'models'....

...are we in fact talking about Excel spreadsheets or something more sophisticated?

I've done amazing things with Excel in the past and see how I could in fact 'model' the effect of MUP with it.

I'm just thinking, if the Sheffield 'model' is in fact an Excel (or open office) spreadsheet, would it be helpful to mention it in debates...?

Christopher Snowdon said...

Yes, this is rarely mentioned. I guess they think most people won't realise and they are probably right. To spell it out. My cheap and nasty £3.99 table wine goes up to £4.50. So those with better taste who spend £4.50 on a bottle will not be affected....Oh, er, yes. Something's not right .

Christopher Snowdon said...

This annoys me intensely each time it's repeated. Reasonable (to me and many others) wine is available for £3.79 a bottle, which works out at 40p a unit. A moderate 21 units a week at 45p a unit represents an increase of £1.05 a week or £54.6 a year. A minimum price of 50p a unit represents an increase of £109.2 a year.This is the effect on a low income moderate drinker. God knows where they get £2.12 a year from. Presumably by including everyone who drinks less than 21 units a week, including those who don't touch alcohol.

Christopher Snowdon said...

And probably including all the consumers of already more expensive alcoholic beverages which they blithely postulate not to be influenced by the minimum pricing ...

Christopher Snowdon said...

Yes, it was indeed an excel spreadsheet to begin with and then got ported to another application. GIGO in any event....

Christopher Snowdon said...