Poll Analysis: Savanta 24th – 28th of May 2024

It was all quiet on the Savanta front until earlier this month, when a new poll emerged following all the drama surrounding the Bute House Agreement and the end of Humza Yousaf’s time at the top. Not even three weeks later and we’ve got another one, this time presumably in reaction to the sudden call of a summer election no one was expecting. As usual, this poll (link to tables not available yet) was as the behest of the Scotsman (link to original writeup).

The previous Savanta covered the 3rd – 8th of May 2024. Changes are shown as (vs that poll / vs last election).


Nothing ever happens. This is all the same as the last poll, except for the numbers after decimal points taking “Others” from 6% to 5%. That’s it, that’s the commentary.


As outlined in a (now very, very, yeesh, look at the lack of quality) old post, I’ve generally avoided providing Westminster projections here on BBS. First Past the Post is really difficult to get right, and even small errors can give huge differences in seats. At Holyrood, that’s moderated by a proportional element that means you can’t be that far off, but pure FPTP at Westminster is a different beast. In addition, plenty of others do those projections much better than I ever could, so why bother?

That said, I have pulled together a simple projection model for this campaign. Unlike Holyrood, I won’t be providing seat-by-seat projections, and I’ll provide the Electoral Calculus estimate on the votes too for comparison, but this just helps provide a little bit more context to the vote figures.

A notable difference between the BBS and Electoral Calculus models in this scenario, with mine clearly viewing the current Conservative seats as more secure. Notably, both of our models think the SNP hold the Caithness, Sutherland and Easter Ross seat that the Lib Dems are generally expected to pick up; I think that’s a simple artefact of modelling rather than reflective of reality, personally.

Regional Vote

Nothing happened with the Westminster figures, but the Holyrood side of things has shifted in rather odd ways. I’d noted with the poll earlier this month that Savanta have a long history of finding the SNP lower than the average, but this is a startling if modest boost in their share, and one that’s outside the margin of error. That puts them a single point ahead (or in other words, statistically tied) with Labour. The only other change here is a single point dip for the Greens, but unlike the shifts for the SNP and Labour, that’s within margin of error.

I’m not sure if Savanta have maybe had a tiny little look under the hood and figured out what was causing those worse-than-average SNP figures, or if this is an outlier against their own trend. Either way, it does feel very jarring under the current circumstances to see SNP +4, though I have no doubt that Swinney will have breathed a sigh of relief, however small, on seeing that when the poll dropped.

Constituency Vote

What makes the Holyrood figures even weirder is that the single point lead is to Labour’s benefit on the constituency ballot, with the only changes on this side of things being a near-meaningless single point dip for both the SNP and Conservatives. It has typically been the case that the SNP have been more likely to hold a lead over Labour on this vote than the list vote, so whilst caveating again that this is a statistical tie allowing for error, purely as a headline figure it’s very slightly odd!

Seat Projection

Projecting that into seats might give us something like this:

Please see this page for how projections work and important caveats.

The effect of all this is that the seat projection is especially weird. That is a much, much bigger seat advantage for Labour than they should have for a point behind the SNP on the list vote; indeed, strictly speaking, you’d expect an SNP seat lead. What’s causing that? An absolutely enormous Labour overhang across the Central Belt. If you scroll down to the hypotheticals chart at the bottom of the piece and look at the “Pure List” segment, that’s effectively the seats with the “correct” proportional entitlement in each region.

That gives the SNP 44 and Labour 38; that is also a larger gap than I’d expect (see the “Scandi” segment for ideal proportionality), and comes down to vote distribution and in particular that the Lib Dems being in double digits makes things chaotic in Central and Glasgow. Or, in other words, Labour end up with a 9 seat advantage from overhang; truly wild! That’s also why despite not losing any votes the Lib Dems are down seats versus the last poll – Labour’s overhang blocks those already highly unlikely west central belt MSPs.

The usual constitutional question combined with the ordering of this piece effectively creates a Weird Burger, or perhaps a Weird Sandwich with No Change Bread. The only thing different here again comes down to some decimal place-based shifting of Don’t Knows, as otherwise the No and Yes figures are identical to the last poll, coming out at the cursed numbers (52:48) after excluding those Don’t Knows. You all know the story by this point: no room for either side to be complacent. 


As ever, the last little bit of analysis concerns those hypothetical and more proportional voting systems that BBS likes to play about with. The use of pure FPTP at Westminster is an affront to democracy, and though Holyrood fares far better, AMS is still deeply imperfect. The examples here simply transpose the poll findings onto more proportional voting systems – the reality is that different systems would of course result in different voter behaviour.

Scandinavian Style Westminster

If you find this or other Ballot Box Scotland output useful and/or interesting, and you can afford to do so, please consider donating to support my work. I love doing this, but it’s a one-man project and takes a lot of time and effort. All donations, no matter how small, are greatly appreciated and extremely helpful.
(About Donations)