Holyrood projection - November 2018 update

More Scottish Parliament polling again

Five Holyrood polls were fielded in October and November, including two after the SNP conference: Survation for the Sunday Post (fielded 28 September to 2 October), Panelbase for the Sunday Times (fielded 28 September to 4 October), Survation for the SNP (fielded 3 to 5 October), Survation for the Daily Record (fielded 18 to 21 October) and Panelbase for the Constitutional Convention (fielded 2 to 7 November). As always there are noticeable discrepancies between their results so each taken individually might lead to different conclusions. So first here are the voting intentions from all five:

I will not draw hasty conclusions from just one poll as John Curtice did in The Daily Record based on the 21 October poll only. As always we must take a step back and look at the bigger picture. Though the SNP should be worried that the two polls fielded after the conference are also the ones with their lowest voting intentions. Current super-sample is 5,136 with 1.33% MOE. So, with all the usual caveats, here are the current weighted averages:

This is a clear warning for the SNP. Being 5% down on the constituency vote and 7% down on the list vote from the previous election means bad news. But bear in mind the SNP won 53 of their 59 constituency seats with double-digit margins in 2016 so most seats remain likely or safe holds. Then we all know the list results are largely pre-determined by the constituency results. So the odd ways AMS works make a 7% loss on the list vote less of a threat that you might think at face value. But the combination of both is a giant step into the danger zone.

The constituency results

On this polling here is how the 73 constituencies project on SLLM rating.

On current polling six seats would change hands, all from the SNP to the Conservatives: Angus North and Mearns (Mairi Gougeon), Aberdeen South and North Kincardine (Maureen Watt), Edinburgh Pentlands (Gordon MacDonald), Moray (Richard Lochhead), Perthshire North (John Swinney), Perthshire South and Kinross-shire (Roseanna Cunningham). The first five would be marginal gains with less than 2% lead so odds are the SNP could in fact hold them with strong campaigning. Perthshire South would be trickier as the projected Tory margin would be almost 6%. Then we would also have six marginal seats that might go either way, including five of the projected Tory gains.

Raw math says that Graeme Dey might have a harder time holding his seat than expected though a Tory gain remains a remote possibility. The drop in SNP vote also puts the current Labour seats out of the danger zone though Labour would still be unable to gain any extra one.

The regional lists results

As usual the bizarre workings of AMS would trigger more changes on the list seats. Here are the results we get by region on current polling. This time the Liberal Democrats and Labour would come out as the winners with three gains and two gains respectively and no losses.

For the other parties there would again be some sort of musical chairs with nine seats overall changing hands. Right now the list MSPs notionally losing there seats would be Ross Greer (Greens), Maree Todd, Paul Wheelhouse (SNP), Alison Harris, Jeremy Balfour, Alexander Stewart, Dean Lockhart, Peter Chapman and Tom Mason (Conservatives).

Interestingly five of the potentially defeated Tory list MSPs (Balfour, Stewart, Lockhart, Chapman, Mason) might even keep their day job if they're lucky enough to be selected in one of the constituencies Tories are projected to gain in their respective regions. Though one of them would have to throw Murdo Fraser under the bus to get selected in Perthshire North. And two of the potentially defeated SNP constituency MSPs (MacDonald, Gougeon or Watt) might be just as lucky if they top the list in their regions. The 'compensatory seats for FPTP losers' part of AMS.

The big picture

Finally here is what the next Scottish Parliament would look like on this polling, and how the parties would perform in terms of net gains and losses. The Liberal Democrats and Labour would score only gains, while the other parties would have a mix of gains and losses. And the end result is quite distressing.

These results would be an unmitigated disaster for both the SNP and the Yes Movement as a whole. Switching from a nine-seat pro-Independence majority to a one-seat Unionist majority would blow up two of the locks in the triple-lock mandate to seek Independence. Now just picture the electoral schedule if no snap GE happens in the meantime: Holyrood elections in 2021 and 2026, Westminster elections in 2022 and 2027. Provided the pro-Independence majority comes back in 2026, the first realistic window of opportunity for another Section 30 order would be 2028. Meaning independence pushed back to 2030 at best, if ever. Let that sink in and you will realize why the SNP definitely need to up their game and do it quick (more on this below). Of course that also means SNP voters need to get their asses out of the couch and get the fuck out to vote. With all due respect.

But as usual there is more here that meets the eye at first glance. So I tried to determine which alternate scenarios are possible under current polling. It's quite easy to figure out which constituency seats are marginals but a wee smitch trickier for marginal list seats. I simply added a feature to my model that swings the list vote shares to the outer limits of MOE, uniformly across all regions. It's admittedly a rather crude approach but it delivers a rough estimate of what could happen with both the FPTP and the list vote moving to the edges of MOE:

The pro-Independence parties could now be anywhere between 60 and 68 seats. Which doesn't mean a pro-Independence majority could be held that easily. Odds are more like 3-to-2 in favour of a Unionist majority. Labour and the SNP are the two most impacted by small swings and for the exact opposite reasons. Labour because they rely almost exclusively on list seats and the SNP because they rely almost exclusively on FPTP seats, so both are more sensitive to even small swings in either category. Conservatives have a more balanced representation so fall in the grey zone where changes in FPTP seats are roughly compensated by changes in the other direction in list seats.

Finally the usual breakdown of seats by region.

Just remember that some surprises may happen with the seventh list seats as some would be allocated with less than 1% margin. AMS delivers marginal list seats just like FPTP delivers marginal constituencies. For example Ross Greer would lose his West Scotland seat by barely 2k votes out of an estimated 300k turnout, while Greens would miss a seat in South Scotland by about 800 votes and hold one in Mid Scotland and Fife by about 700 votes. And the surprise SNP seats in Lothian and North East would be gained by 300 and 700 votes respectively.

Can election projection models be trusted?

Questions are often asked about the reliability of projection models and the results they deliver. All legitimate questions as the seat projections often make headlines and undoubtedly influence the parties' campaign strategies. Besides my own model I know of two sites who currently offer a prediction tool for Scottish Parliament: Scotland Votes and Election Polling. I have explained in a previous post how my own model works. I don't know the exact algorithms used by the other two models but the results they deliver give strong hints about how they work.

So here is what my model projects from the five polls taken individually and their weighted average:

Same using Scotland Votes:

Same using Election Polling:

Finally how the three models allocate constituency and list seats, based on the weighted average:

The results from all three models are quite similar as might be expected. The differences in constituency seats come from the relative weight of Uniform National Swing (UNS) versus Proportional National Swing (PNS) used in the algorithms. The differences between UNS and PNS are explained here. I have done various simulations that show Scotland Votes uses pure PNS and Election Polling uses pure UNS while my own model uses 30% PNS + 70% UNS. UNS is slightly more favourable for the SNP while PNS is slightly more favourable for the Tories.

Differences in the list seats results come from the way the raw data are processed. Both Scotland Votes and Election Polling appear to use full UNS across al regions. I rely instead on regional crosstabs when they exist (Survation) and full UNS only when the pollster does not publish regional crosstabs of the list vote (Panelbase). Then I can only state the obvious again: projections are only as good as the polls they use. And we all know UK polls in general and Scottish polls specifically are not exactly the most reliable in the galaxy.

Of course what really matters is that all three models show an Unionist majority. Whether it's by one seat or three seats is irrelevant. The point is that these results, plus the fact that the most recent polls are also the worst, should raise huge red flags for everyone in the SNP. The next Holyrood election is three-and-a-half years away and already Unionists are on the offensive on all fronts. So should the SNP lest they be doomed to suffer a major setback.

And now what can possibly go wronger?

I am less optimistic right now about the SNP's future prospects than I was a month or two ago. This is not just because of the less than stellar voting intentions in the most recent polls as we don't know yet if this is the start of a real downward trend or not. It has more to do with what happened in Holyrood two weeks ago and how the SNP fumbled it on two major issues. Two major political fuckups, one of which is already a PR disaster and the other a timebomb.

First we had the SNP voting down Labour and Green amendments to the Planning Bill. The story has already been told so I won't repeat it, and all the hard facts are in the Official Committee Report anyway. You could call that one an own goal. When even The National publishes a highly critical article on the issue, you should realize something went wrong. Unless of course if you think pissing off environmentalists and community activists, who otherwise might have felt inclined to support you, while handing out ready-to-use live ammunition to the oppositions is a viable option.

On the afternoon of the same day Scottish Parliament passed the infamous People's Vote amendment, which made an already bad day even worse. This story too has been told repeatedly with all sorts of spin by everyone from Patrick Harvie to Andrew Adonis. As usual all the gory details are duly recorded in the Official Chamber Report. If you read the full motion as amended it is quite obvious that Tavish Scott's amendement is not actually one. Rather it's a rider (non-germane amendment in parliamentary lingo). And quite obviously appending it to Richard Lochhead's motion was a trick to ensure it would pass, as SNP MSP's would find it more important to support the motion than to question the amendment. And it worked though some reportedly have buyer's remorse now. But too late. 

You might remember that I explained why the SNP supporting People's Vote would be very bad idea. That was three months ago. Others on Twitter sent strong warnings too at about the same time, long before the SNP conference. A lot of SNP MPs, former MPs, MSPs and local branches are on Twitter and quite a number of them follow me and other 'concerned Yes tweeps' who addressed the issue back then. So it's not like the SNP can pretend they haven't been warned. As much as I hate going into 'I told you so' mode, I guess that's just the thing to do this time. 

Then we had Pete Wishart and Angus MacNeil voicing concern about how support to People's Vote could harm the second Independence referendum. When both the 'cautious path to Indy' guy and the 'bold path to Indy' guy raise the same objections, it should flash a massive red light and give reason to think it through. Especially when you remember that Wishart tweeted early in August that he was 'open to the idea' of supporting People's Vote and then obviously fully assessed the downside. That all of this did not set off an alarm invites the obvious question: what are Sturgeon's ulterior motives here?

Because of course nothing beats pissing off SNP-Leave voters and Yes-Leave voters in a single blow. Except possibly drawing flak from your own parliamentarians while going squarely against public opinion who don't support the SNP's stand here. The National had an online poll on the very issue after the ScotParl vote and here is what it says:

I see a disturbing similarity with the 2010 coalition deal here. Just like Nicola Sturgeon with People's Vote, Nick Clegg had made up his mind beforehand to ram the coalition deal down his party's throat not matter what. And it did not exactly end well for the Liberal Democrats and Clegg himself. Let's just hope the similarity will not go that far for the SNP.

More than ever stay tuned for further upsets and further broadcasts.

Saor Alba Gu Bràth