Drinking in the lockdown

The Office for National Statistics put out two sets of data yesterday which shine a bit of light on a couple of natural experiments, both involving The Drink.

The first is minimum pricing, which we have discussed many times. As previously reported, there was a fall in alcohol-specific deaths in Scotland in 2019 following a slight rise in 2018 (minimum pricing was introduced in May 2018). The new ONS data show that in England, where minimum pricing was not in place, there was a slight fall in 2018 and a slight rise in 2019.

 

All told, the alcohol-specific death rate fell from 20.5 to 18.6 per 100,000 in Scotland between 2017 and 2019. In England, it fell from 11.1 to 10.9 per 100,000. This is consistent with minimum pricing having a modest effect, but it is also consistent with random fluctuation; there were bigger falls in 2006-07, 2008-09 and 2011-12, and other evidence suggests that minimum policy didn't have much impact on heavy drinkers.

It will be interesting to see the figures for 2020 when lockdowns switched the market towards off-trade alcohol in a dramatic way. And that brings me to the second dataset which provides provisional figures of alcohol-specific deaths in the first nine months of last year (England and Wales only). It is quite a striking picture, with a sharp rise in the first quarter of 2020.

Both sexes were equally affected and, as the graph below shows, a similar increase can be seen among every age group over 30.

And it happened in every region of England, except the north-east where there was a steep rise in the second quarter.

 

What should we make of this? The easy answer would be to assume that people were drinking themselves to death during lockdown. Contrary to popular belief, the impact of heavy drinking can often be seen in aggregate figures very quickly. Although it can take many years to develop some alcohol-related diseases, people with chronic disease can be tipped over the edge by heavy episodic drinking while other causes of death, such as alcohol poisoning, are acute and rapid. 

The trouble with the lockdown theory is that the lockdown began on 24 March. The rise in alcohol-specific deaths took place in a quarter when lockdown was only in place for one of the thirteen weeks. Rates remained high in the second quarter but did not rise in the way that you might expect if these were deaths of lockdown despair.

Nevertheless, it is difficult to ignore the link. There was a certain amount of self-isolation before the lockdown officially began and many people may have been affected by stress at the thought of what was to come, not just in terms of lockdown but of the pandemic itself.

Whatever the cause, it undermines the single distribution theory of alcohol harm. As I said in September, different people responded very differently to lockdown. Some drank more, others drank less. Crucially, moderate drinkers tended to drink less and heavy drinkers tended to drink more.

Alcohol sales fell during lockdown and per capita consumption in 2020 seems to have been similar to previous years (we don't have comprehensive figures yet). Orthodox 'public health' theory suggests that we shouldn't see a rise in alcohol-related harm without a rise in per capita consumption. Moreover, the amount of alcohol advertising fell dramatically during lockdown and availability was greatly limited by bar closures. Since these are (supposedly) two of the main drivers of consumption, we should have seen a substantial decline in per capita consumption and a commensurate decline in alcohol-related harm.

We saw none of that. Instead we saw a spike in alcohol consumption among some (presumably heavy) drinkers which seems to have led to an almost immediate increase in harm. All of this leads to the common sense conclusion, which is alien to the neo-temperance lobby, that harmful drinking is driven by personal circumstance rather than 'commercial determinants'. Efforts to tackle harmful drinking should therefore focus on harmful drinkers rather than the whole population.

 

One day, perhaps, this message will get through.