7 October 2014

Modelling Neverendum

You might wonder why a blog apparently linked to the part of England which is furthest from Scotland takes any interest in the independence referendum held there on 18 September. Firstly, and obviously, all the regions of England would have been affected had the Yes vote won. Secondly, a particular regional interest was the subject of a post here in 2012, Would Scotland’s Loss be the South West’s Gain. This looked at the possible relocation after Scottish independence of the Royal Navy’s nuclear submarines to SW England, a subject which will be revisited in a future post. In the days after the referendum, two views seemed to emerge. One, delivered outside 10 Downing Street on the morning after by David Cameron, was that
… now the debate has been settled for a generation or as Alex Salmond has said, perhaps for a lifetime. So there can be no disputes, no re-runs – we have heard the settled will of the Scottish people.
But the outgoing SNP leader stated:
… the majority of Scots up to the age of 55 voted for independence, and a majority of Scots over 55 voted against independence … When you have a situation where the majority of a country up to the age of 55 is already voting for independence then I think the writing is on the wall for Westminster.
And his likely successor, currently his deputy, Nicola Sturgeon, was quoted by BBC News as saying when launching her bid to replace him:
… the country could only become independent if the electorate backed the move in a referendum. But she did not rule out the possibility of the SNP including a commitment to hold a second referendum in a future election manifesto.
The leader of the Scottish Conservatives, Ruth Davidson, subsequently warned that:
The SNP lost. Scotland demonstrated her sovereign will. And yet, people at the top want to run this race again as soon as possible. They want Scotland locked in a cycle of neverendum.
So it’s worth looking again at the actual result on 18 September:

Fig 1: Referendum results
The turnout, as has been widely pointed out, was remarkably high and unlikely to have been improved upon. So, for Yes to have succeeded, about 192,000 No votes would have had to come across. The main sources of insight as to how this could come about in another referendum at some point in the future are provided by opinion polling.

The polls were not startlingly accurate in their predictions in the run-up to the 2014 vote. In March this year a post here looked at one particular Ipsos-MORI opinion poll in detail. It had shown that at that time those certain to vote were divided 32% Yes, 57% No and 11% undecided. As the September vote approached, opinion polls started to indicate that the Yes/No balance had shifted to be much closer – closer in the polls it turned out, than in the actual result. Afterwards Professor John Cutis in his What Scotland Thinks blog concluded that there had been a systematic problem of underestimating the No vote beforehand. However, two essentially retrospective polls were carried out on referendum day by YouGov and Lord Ashcroft Polls and these had overall results close to the outcome and also provided a breakdown of voting behaviour:

Fig 2: Opinion polls post-voting
The two polls’ age data is easier to compare graphically, and tends to bear out Salmond’s view of older voters:

Fig 3: No voters by age from polls
The Ashcroft poll also asked:
Q.8 If it turns out that a majority has voted NO in the referendum, for how long do you think the question of whether Scotland should be independent or remain in the UK will remain settled?
Fig 4: answers to Q.8 by age
From this it is clear that nearly half of those who voted think the issue will not remain settled for more than 10 years, though, as in the 2014 referendum, enthusiasm for changing the status quo is less apparent among the older age groups. So what might happen in another referendum in five or ten years’ time? An impossible question to answer but it is possible to make an assessment, albeit crude, as to how demographics might have an effect. After all, the 16 year olds of 2019 are today’s 11 year olds and the 16 year olds of 2024 are today’s 6 year olds, today’s 11 year olds being 21 by then. A key source is Mid-2013 Population Estimates Scotland produced by National Records of Scotland in June 2014. First of all, their Infographic 1:
Fig 5: From Mid-2013 Population Estimates Scotland
This explains why the population of Scotland increased by 14,100 between mid-2012 and mid-2013. Using the data from YouGov quoted in Fig 2 above on voting preferences by place of birth (blue data below) it is possible to make a crude assessment of the voting impact of migration:

Fig 6: Possible effect of movements to and from Scotland in Fig 5
There are some obvious flaws. Firstly ‘People’ include children under 16 and secondly immigrant and emigrant preferences may be different from those by place of birth: Scottish No voters may be more outward-looking and therefore be more likely to be found in those leaving for the rest of the UK (RUK) and elsewhere than among their stay-at-home compatriots. There are other uncertainties: how many of those emigrating were originally from outside Scotland, went there and then didn’t like it? However, it isn’t easy to see how the net flow of inward immigrants combined with the pattern of the YouGov data could support hopes of a net increase in the Yes vote over the years ahead.

The ‘natural increase’ of 909 conceals much larger underlying population flows. Figure 5 from the Mid-2013 Population Estimates Scotland shows the facts of life and death:

Fig 7: From Mid-2013 Population Estimates Scotland 
(Anyone puzzled by the Male and Female notches at 65/66 should read this post!) Most of the nearly 56,000 deaths each year will be among the older members of the population, who, for this purpose, will be assumed to be over 65. The earlier comparison of No voters by age in the two polls (Figs 2 and 3) suggests that about 65% of the 65+’s are No voters. Table 1 of Mid-2013 Population Estimates Scotland contains details of Scotland’s age breakdown from which the numbers of new voters in 2019 and 2024 can be predicted (ignoring early deaths and migration). If the voting preferences of the new voters continues to be about 50% No, as the polling seems to suggest, there could be a shift to Yes of about 85,000 after 10 years due to ‘natural change’:

Fig 8: How 'natural change' might affect No vote
To describe this sort of modelling as simplistic is to be kind. Some of the underlying assumptions have already been made apparent, and there are others such as a lower turnout in future referenda differentiated by preference – fewer Nos, say. Also, the errors in the polls for subgroups are considerable – look at the two polls’ different estimates for Male and Female preferences (Fig 2 above) – and this can only get worse the smaller the subgroup, eg 16-18 year olds. As a sensitivity test (in italics in Fig 8) the assumptions made above can be altered in favour of Yes. If the No support among the group of emerging voters were to fall to 30% (unlikely), and the deaths were assumed to be among the oldest voters, and they had a No preference of, say, 75% (again unlikely, given female relative longevity and No preference), the shift to Yes would increase to about 190,000 – the number that Yes needed in 2014. This assumes, of course, that the majority of those who voted in 2014, who will still be alive in 2024, keep voting the same way and that the immigration/emigration effect in the meantime is nil.

So even if the emerging voters were Yes enthusiasts, it seems likely that a substantial number of those who voted No in 2014 would have to change their minds for Yes to get a majority in 2024. What might lead them to do so? This really is peering into the Scottish mists, but an unsatisfactory 'Devo max' (further devolution to Scotland as promised post-referendum), a successful Eurozone, a sovereign area solution for the Clyde naval base (keeping jobs and not antagonising NATO) all might help, as might Catalan independence.

On the other hand, the concentrated location of the Yes vote in 2014 is now well understood and the Yes vote might fall in those parts of Scotland not green in this map:

- not everyone in Scotland may warm to the prospect of being governed from the Republic of Glasgow.


  1. Can you run this again using data from the Edinburgh university study?

  2. Could you provide a link to the EU study - I can then think about it!