Friday 18 November 2016

Garbage in, garbage out

The Sheffield modellers have been in action once again, this time making some wild speculations about alcohol consumption in twenty years time. Cancer Research UK commissioned this little job which, I'm sure, is exactly what people who run marathons and buy pink ribbons want their money to be spent on.

You can read the report here but I doubt your life will be improved by doing so. I don't think it's even designed to be read. It has no academic merit and has no value as a policy paper. Instead, it seems to have been designed to get some cheap headlines and promote the Sheffield crew's pet cause of minimum pricing (while also furthering the cause of putting cancer warnings on alcohol, which is crucial in turning booze into the new tobacco). The timing of the report is revealing. It is surely no coincidence that today was the deadline for the Scotch Whisky Association to lodge its appeal to the supreme court in the minimum pricing case (which it has done).

The problem with predicting the impact of alcohol consumption is that nobody has any idea whether drinking rates will rise or fall next year, let alone over the next 20 years. We don't know what the economy will do, we don't know how the demographics will change and we don't know how tastes will change. Any forecast is guesswork and although the Sheffield team draw up four different scenarios, they caution against taking their guesses too literally...

The scenarios were selected to include increases and decreases in future consumption of varying degrees and, in some cases, reflect plausible explanations for the recent falls in consumption. None of these scenarios should be interpreted as predictions of what will happen in the future. Instead, they should be seen as illustrative examples of how the future could look under different conditions

That didn't prevent the CR-UK press release being headlined with...

135,000 alcohol-related cancer deaths predicted by 2035

And the media naturally followed suit with headlines like this...

More than 135,000 UK drinkers 'will die of cancer caused by alcohol by the year 2035'

One of the red flags indicating that the report was designed for propaganda purposes is the time-span of this estimate. Disease and mortality rates are nearly always measured in annual terms. The 135,000 estimate is not an annual figure. It is the combined total for the next twenty years. The only plausible reason why somebody would use such an unusual measure is that they want to create a BIG NUMBER for the media.

The annual figure of 6,750 doesn't sound scary enough and, with more than half a million people dying in the UK each year, it isn't scary, particularly since most of the cancer deaths occur in old age and extreme old age. And so they have multiplied it by 20 for effect.

The decision to model four different scenarios is commendable, but it is weird that there is hardly any variation in their estimates regardless of whether they assume alcohol consumption rises, falls or stays the same. Their alcohol-related cancer mortality estimates are all within a tiny bound between 133,213 and 134,636. 

In their best case scenario, they think there will be an average of 6,661 deaths per annum. In the worst case scenario, there will be 6,732 deaths. If that's the difference, it hardly seems worth worrying about.

The real purpose of the report is to push minimum pricing, which they reckon will prevent 674 cancer deaths but, again, this is over a twenty year period. In other words, we can carry on as we are and have 134,636 deaths between now and 2035 or we can bring in a regressive policy which will cost billions of pounds and have 133,962 deaths. The difference is a rounding error and, given the serious flaws in the Sheffield model, it is doubtful that even this marginal effect would be seen in practice.

There is more than could be said about this, but it's not worth the effort. It's hack work. CR-UK got their publicity. The Sheffield crew got their money. Let's move on.

No comments: