Scotland in Detail: The Constitution’s Messy Maths

A favourite game in Scotland is trying to read the constitutional runes from party-based elections. Usually, this comes from tallying up the votes for each party in council by-elections according constitutional preference. It also sometimes happens on opinion polls, though those usually have Independence data in them anyway, and national election results – see bunfights over whether the narrow pro-Union party lead in constituency vote or equally narow pro-Independence party lead in list vote is more reflective of some settled view on the constitution.

Unfortunately for the folk doing so, this isn’t quite how the world works. For a first formal entry into what is likely to be a very infrequent series I’m titling “Scotland in Detail” (rebranding from “Understanding Scotland”, which by coincidence is a similarly timed project by some other folk), which will look at more of the detail of our electoral landscape, I thought I’d delve into the messy world of cross-constitutional voting patterns.

For one key problem with taking council by-elections as constitutionally useful, Scotland has (or rather had, in the 2017 elections) a mighty 354 local council wards. No two of these wards are identical in political composition. A by-election in Glasgow is likely to find quite a strong result for pro-Independence parties overall. Meanwhile, one in Berwickshire is going to look very good for the sum of the pro-Union parties. But neither of them would be representative of Scotland as a whole. 

There’s also the issue of turnout, both the level of turnout and who does so. A council by-election is lucky if it gets around 35% turnout. The Independence referendum in 2014 had a staggering 85% turnout. We know turnout is uneven across different groups, as is support for either side of the debate. So a low turnout council by-election is unlikely even to accurately capture constitutional support within its own bounds.

But most importantly, voters are a lot more complex than partisans assume. In fact, they can be downright bizarre! I’m thinking especially of the two voters in an Aberdeen ward in 2017 who marked the fascist National Front as their first preference, then UKIP second, and third… the Greens. Objectively, that makes no political sense whatsoever, yet two people did it. Though usually not so extreme, very few voters align 100% with the party they vote for, and as totemic as the issue may be, that includes on the constitution.

Subsamples? Suboptimal!

In fairness there is some awareness of this. However, that can be quite one sided, with people excitedly clutching the limited evidence that suits their views, and discarding more comprehensive evidence that doesn’t. For example, it’s often claimed that somewhere in the region of 25% of Scottish Labour voters support Independence, against their party’s preference for the Union. In fact, an even higher claim of 40% was investigated by the Ferret last year. You might hear this from both Independence supporters (keen to inflate their own side’s size) and Conservatives (keen to nab Labour’s voters by painting them as weak on the Union.)

What’s happened there is selective reporting on polling subsamples – those are the further breakdowns of party support by e.g. age, gender or constitutional preference. It’s the same story with the claim, on the other side of the equation, that fewer than half of Green voters support Independence. This even made it into a few columns when the Greens signed a cooperation agreement to enter into government with the SNP. 

In that case, it’s a misleading reading of just part of a poll from Lord Ashcroft. I don’t cover Ashcroft polls on BBS as he isn’t a member of the British Polling Council, and that alone is cause to tread a bit more carefully. Ashcroft’s MO appears to be to provide neat little reports for journalists to read so they don’t have to look at the tables, and he’s apparently selective about what he includes in them.

It turns out the figure in question is drawn from the constituency vote – yet, everyone knows Green support is primarily on the regional list side of things. Look at that part of the poll and Green voters were more than two-thirds (68%) in favour of Independence. The poll disproves itself, but you’d only know that if you looked at the tables. I’m not a journalist, yet I knew to go looking for more info when faced with a dramatic claim, an approach I’d commend to anyone who hopes to inform the public.

Basically it should go without saying that focussing on little bits of data you like or find exciting is no way to go about actually determining what the Scottish public as a whole think. It’s especially important not to hook entire theories on one subsample from one poll – by their nature, they aren’t very large samples. I’ve even got an entire section here on BBS explaining why subsamples are subject to massive margins of error and should never be used in isolation.

A quick google finds plenty of examples of otherwise reputable media sources centring entire articles on subsamples. This is just really bad use of data, and it’s frankly very disappointing. You’ll note that here on BBS I very rarely, if ever, make reference to subsamples. However, if you are going to play the subsample game, you can mitigate the issue somewhat by using multiple polls.

Party and Voter Alignment on the Constitution

With the caveats about subsamples still firmly in mind, I set about looking back at polls from towards the end of the Holyrood campaign. A total of seven polling agencies conducted polling in the last couple of weeks, so I’ve pulled the constitutional preference by party from each of these polls and averaged it out. That helps smooth out some of the error from just a single poll, and also from variations due to pollster-specific methodologies.

If we start by looking at the constitutional position within parties, what we see is that for each of the major parties, the overwhelming majority of their voters also share their party’s constitutional stance. For further clarity, we can display figures in terms of being on the same or other side of the question, rather than Yes vs No.

This perhaps makes clear why the Greens and Labour are most talked about and served as my examples earlier. With a ratio of around 4.3 to 1 (72% to 17%) in support of their party’s position and 12% unsure, the Greens are the least constitutionally settled party – albeit still pretty settled. Labour aren’t a million miles off that with a ratio of about 4.8 to 1 (76% to 16%), though their portion of don’t knows is more middle of the road on 8%. The claims about both parties I mentioned above aren’t borne out by this wider look.

If we then turn to the SNP and Lib Dems, they round to the same overall figures of 81% to 12%, albeit in mirror image. The Lib Dems are fractionally less settled with a ratio of 6.7 to 1 versus the SNP’s 6.8, and both have 7% unsure. Finally, the Conservatives are unsurprisingly the most certain party with a whopping 19.3 to 1 (93% to 5%) ratio, and just 2% of their voters saying they don’t know.

Let's Get Chunky

The above gives us a reasonable estimate of how constitutional preferences lie within each party’s voter base. But what does that mean for the electorate overall? 81% of both SNP and Lib Dem voters align with their party’s constitutional position, but there are a lot more SNP voters than Lib Dem voters. If we average out the raw No (46.4%), Yes (45.0%) and Don’t Know (7.6%) figures from those seven polls, then apply the previous section’s proportions to each party’s final 2021 result, the electorate seemingly chunks up like this:

What this shows us is that the largest bloc of voters in Scotland are SNP-Yes voters, who make up almost a third (32.5%) of the overall electorate. Although the Conservatives are the most constitutionally settled group of voters, they are a smaller proportion, so the Conservative-No group is the second largest slice of the electorate, at 21.9%. Third is Labour-No on 13.7%, and fourth is Green-Yes on 5.8%. This is where it gets messy however, with some fascinating cross-constitutional findings.

The fifth largest bloc is SNP-No, with 4.8%. That’s ahead of the Lib Dem-No chunk of 4.1% of voters. That means that amongst the electorate overall, there are more pro-Union SNP voters than pro-Union Lib Dem voters! Similarly, though Labour’s voters are more likely to back Independence than the SNP’s are the Union, at 2.8% of voters they are a smaller group in total. And although Greens are the least constitutionally settled, Green-No voters are only 1.3% of the overall electorate. That’s not all that much higher than Conservative-Yes voters on 1.1%. In other words, you’re roughly as likely to meet a Pro-Union Green as you are a Pro-Independence Conservative.

Similarly, although more Green voters than for any other party didn’t know how they’d vote on the constitution, at 1% of the electorate Green-Don’t Knows are a smaller proportion than Labour-Don’t Knows (1.4%) and SNP-Don’t Knows (3%).

However, we can also see from this some of the limitations in trying to use subsamples, polling data, and election data all together. Only 0.6% of the electorate is estimated to be Other-No here. However, the total vote for hardline pro-Union parties (All for Unity, Abolish the Scottish Parliament, Reform UK and UKIP) was 1.5%. Whilst some of those voters will indeed counterintuitively support Independence, like my early example of the National Front-UKIP-Green voter, and some will be unsure, it’s very unlikely to be so low as just two in five backing the Union.

Finally, we can squish all of that data back into a relative breakdown within each constitutional option. What we find is that almost half of No support comes from Conservative voters (47.2%), whilst less than a third (29.5%) is from Labour and less than a tenth (8.8%) from the Lib Dems. Meanwhile, SNP (10.3%) and Green (2.8%) voters sum up to 13.1% of the No vote.

On the Yes side, SNP voters are utterly dominant, contributing almost three-quarters (72.4%) of the Yes vote, whilst the Greens then give just over an eighth (12.9%). The total contribution of the pro-Union parties to the Independence camp is 10% of the Yes vote, mostly from Labour (6.2%), then Conservative (2.5%) and lastly Lib Dem (1.3%)

The SNP’s voters account for over a third (39.5%) of don’t knows, whilst both Labour and collective Other parties make up 18.4%, followed by Greens on 13.2%. The Other share here is likely to reflect some of the gap I identified previously.

Final Thoughts

As I’ve said myself though, we still need to be careful of subsample based analysis. What I’ve pulled together here is probably within a reasonable distance of the current reality, but it can’t be entirely spot on. Everything above should be viewed as having caveats such as “roughly this many…” or “according to this estimate…” attached to it – it’s just that makes for a very frustrating writing style if literally followed.

What’s likely to be much more accurate is the data coming out of the Scottish Election Study. As far as I can make out, the full 2021 data isn’t yet available, but they were quick out of the gates with some really interesting bits on Scotland’s “tribes” (along Yes/No, Leave/Remain axes) and the non-SNP pro-Independence parties. This might be a topic I revisit if and when I can go through that data (and if the good folk behind the SES don’t do so themselves!), but for the moment, this seemed like a useful and informative piece to have to hand.

It’s also important not to view this as anything more than a snapshot. This is all drawn from data at the end of April and early May. Polling since the election hasn’t really shown any major shifts, either in support for parties or Independence. In the longer term, we may find much more movement – both in terms of party support, and also that constitutional attitudes in one party’s base may harden, whilst another’s may soften.

Hopefully, this will encourage readers not to put any stock in dramatic findings based on a single poll subsample, and provide a better understanding of the complex interplay between constitutional and party preferences. The Electorate is a mysterious beast indeed, but it can be relatively well understood if you take the time to do so.