In June 2016, the UK voted in an advisory referendum to leave the European Union.
As soon as the result was known, there was interest in the factors that correlated with voting intentions. Lord Ashcroft carried out research that suggests age was a good postdictor of voting choice.
Although the referendum was advisory, not mandatory, David Cameron said his Government would bind itself to the referendum result. However, no UK government can bind its successors, and he immediately resigned.
Then in 2017, we had another election in which no party won an overall majority, and during which the two largest parties offered radically different visions of Brexit.
Referenda haven’t been used much in the UK. After the 2017 General Election, and even more so after the 2019 European Elections, I asked myself the question, how might you decide the shelf-life of a UK referendum mandate? We don’t allow dead people to vote, and since it seems that 2016 Leave Voters tended to be older than Remain voters, I wondered if it might be possible to work out when the number of 2016 Remain voters surviving outnumbered the surviving 2016 Leave voters.
Here are my conclusions.
Sir John Curtice writes occasionally for the BBC Website, and he published graphs showing his research into how voting intention had correlated with age, and how participation correlated with age.
The UK Office for National Statistics (ONS) publishes:
- UK Population Details, men and women, segmented by age, in one year segments.
- UK Death rates, men and women, segmented by age, in 5 year segments.
I used the figures for 2015.
First, I applied Sir John Curtice’s graphs for participation and likelihood of voting by age to the ONS population segmentation to come up with a prediction of what the Brexit vote would be.
I got an answer of 17.6 million, which is pretty close to the actual 17.4 million. Bear in mind I haven’t ‘tuned’ the model at all. I was frankly amazed, and decided there was no need for anything more sophisticated (see e.g. http://www.statsguy.co.uk/brexit-voting-and-education/ for some very interesting detailed analyses).
I then applied the ONS segmented death figures to my modelled Leave numbers by age to arrive at a prediction of the number of first year deaths, and got a number well over half a million (the bar effect on the graphs is because the age segments are 5 year bands, whilst the population segments are years).
Obviously you can only die once, so I turned the number of deaths into a likelihood of a 2016 Leave voter surviving 1 year, which turns out, on my model, to be 0.97. Three years of this, and 1.5 million 2016 Leave voters have died.
Annual deaths in the UK are around 600,000. Subtract the dead Leave voters, and you’re left with 100,000 deaths, to be shared between Remain voters and ‘Did Not Vote’. ‘Did Not Vote’ includes all children under 18 who have a low chance of dieing, but I have not modelled them at all. Remain voters outnumbered Abstainers of voting age, 5 to 3, most deaths would be in people of voting age, the Abstain age profile is similar to that of the Remain age profile, so I would guess that perhaps 60,000 2016 Remain voters have died each year.
So the cross over point, when the number of living 2016 Remain voters exceeds the number of living Leave voters, must be sometime around now.
Should anyone wish to carry this further, here is my model, in the form of a LibreOffice spreadsheet.