Warning. This article contains (historic) references to the Brexit referendum!
Some time ago now, the then deputy Prime Minister Nick Clegg made what on the face of it was an astute observation about the Brexit referendum.
“If you look at the demography, the oldest voters voted for Brexit in large numbers, the youngest voters did the opposite.” and “How can I put this politely? Actuarially, I suspect that the high point of the Brexit vote has already passed”. This statement was met with very little opposition.
Some time after that, this article claimed that “demographic factors alone are causing the Leave majority to shrink by around 1,350 per day, or almost half a million a year.”. Leaving the issue of spurious precision aside, this seems on the face of it a “sure thing”. But is it?
Whether the authors were aware of it or not, both of these claims make a strong assumption that the tendency to vote Leave was what epidemiologists call a “cohort” effect. In other words, the propensity to vote leave was a characteristic of the baby boomers and generation X’s, and not the Millennials and Gen Z’s. Age effects are different as I will now try to explain.
It may help to use an analogy. Imagine we were only able to assess the incidence of cancer in a single cross-sectional sample - and know nothing else about it. Like the referendum result we would observe a strong positive association between cancer and age. If we were to make the same assertion as Mr Clegg and Mr Kellner, then as time progressed the occurrence of disease would decrease, eventually to zero. Cancer as with most diseases, is subject to age, period and cohort effects. We know this through many years of experience. Age is probably the dominant factor in most cancers but period and cohort effects can play an important role as well.
Period effects affect all age groups at a particular point in time - typically these are wars, economic crashes and pandemics! Now, here’s the rub! Age, period and cohort effects mask each other – they are confounded and so what might appears to be, a cohort effect could in fact be a mixture of age and period processes and vice-versa. Without making some (strong) assumptions, you cannot say for sure.
In conclusion, it is easy to be misled by data and to mislead others by ignoring alternative explanations, willfully or otherwise. If the tendency to vote leave is predominately an age effect then a second referendum would most return the same result as the first time not withstanding possible influences from period effects. If there is a strong cohort effect then the result could be reversed. But even if that were to happen, we could not rule out period effects. Time may not tell after all.
Postscript: I wrote this article a long time ago and sat on it. Part of me thought we might in time know the answer. As of 2024, I am not sure we are any-the-wiser, as in my opinion, the period effect of the pandemic is further muddying the waters. It maybe that we may never get to know which is correct.