Saturday, June 3, 2017

Either this is a rogue poll, or the Tories are now in total meltdown

There will be several polls published tonight, but the one you're about to see may well be the most important of the lot, because the fieldwork was conducted at lightning speed after last night's so-called "debate" between Theresa May and Jeremy Corbyn.  Some commentators felt that the programme was relatively positive for May, but if these numbers are right it appears that the opposite is true.  Perhaps looking an underpaid nurse in the eye and using the fatuous phrase "magic money tree" isn't such a guaranteed vote-grabber after all.

GB-wide voting intentions (Survation, online fieldwork) :

Conservatives 40% (-6)
Labour 39% (+5)
Liberal Democrats 8% (n/c)
UKIP 5% (+2)
SNP 4% (n/c)

Scottish subsample : SNP: 44%, Labour 28%, Conservatives 21%, Liberal Democrats 6%, UKIP 2%

Obviously an individual Scottish subsample can't be regarded as remotely reliable, but nevertheless it's heartening to see a decent SNP lead, because we know from the experience of 2010 that high-profile debates excluding the SNP can have an extremely damaging effect.

UPDATE : Two other polls are more favourable for the Tories, although the fieldwork for both is not quite so recent.  ORB has the Tory lead increasing from 6% to 9%, while ICM have the lead only falling from 12% to 11% (although it should be noted that's still the smallest gap ICM have reported at any time during the campaign).

UPDATE II : The YouGov poll (which may be the final GB-wide poll of the evening) gives the Tories a lead of 4%, which suggests the state of play has remained very stable for several days.  However, it looks like the fieldwork either wholly or mostly preceded last night's "debate".  So the million dollar question for this evening : is Survation simply a random outlier, or is its more recent fieldwork highly significant?  Could it be picking up a Cleggasm-type effect?

The average Tory lead in tonight's polls is 7.2% - probably not quite hung parliament territory, but not too far away from it.

UPDATE III : ICM's Scottish subsample is : SNP 41%, Conservatives 29%, Labour 23%, Liberal Democrats 5%, Greens 2%.

Are you hiding behind the SOFA, Theresa? Now even Opinium say the Tory majority could be under threat...

Well, the first of tonight's many polls is about as reassuring as we could realistically have hoped for, with Opinium showing the Tory lead falling to its lowest level of the campaign, and with the SNP enjoying a reasonable lead in the Scottish subsample.  Bear in mind, though, that Opinium's numbers are always a little out of date by the time we see them - fieldwork was concluded on Wednesday.

GB-wide voting intentions (Opinium) :

Conservatives 43% (-2)
Labour 37% (+2)
Liberal Democrats 6% (-1)
SNP 5% (n/c)
UKIP 5% (n/c)
Greens 2% (n/c)

Scottish subsample : SNP 47%, Conservatives 31%, Labour 19%, Liberal Democrats 2%, UKIP 1%.

Opinium have thus joined the ranks of pollsters saying that a hung parliament is at least a credible possibility. I suspect most projections from a 6% Tory lead would still produce some sort of Tory majority, but the truth is that nobody really know how the regional variations will play out.  Most commentators will probably refuse to take the hung parliament scenario seriously unless tonight's polls from the most Tory-friendly firms (ICM and ComRes) at least show the Tory lead dropping to single figures.  An 8% or 9% Tory lead would obviously be very unlikely to produce a hung parliament, but at that point you'd be able to make the case that ICM and ComRes would only need to be over-compensating for the 2015 error by a small amount for it to become possible that the Tories might lose their majority.

My reading of the hints Martin Boon has been dropping on Twitter is that tonight's ICM poll probably isn't going to produce that kind of result, but we'll see.

*  *  *

UPDATE : The ComRes poll will probably dampen talk of a hung parliament, because it shows the Tory lead steady at 12%.  However, just to put in perspective how strong the ComRes turnout weighting is, the Tories are only 4% ahead before the adjustment is made.  The Scottish subsample is less favourable for the SNP than Opinium was, but crucially agrees with Opinium that there is no sign of a Labour surge : SNP 40%, Conservatives 32%, Labour 15%, Liberal Democrats 9%, Greens 2%, UKIP 1%.

We'll soon know a lot more than we currently do

Saturday nights are always the busiest time for opinion polls during an election campaign, because so many of the Sunday papers commission them.  We'll get the usual batch of GB-wide polls from Opinium, YouGov, ICM, ComRes, etc, but there's also going to be a full-scale Scottish poll from Survation.  I must say I'm just a tad nervous about this one, because it's presumably going to be the most up-to-date Scottish poll we've seen - the last one was Ipsos-Mori, which was conducted between the 22nd and 27th of May.  It's perfectly possible that Survation will pick up changes that have happened since then, with the most troubling potential scenario being a significant swing from SNP to Labour as a result of the Corbyn bandwagon.

As I always say, subsamples are our early warning system, albeit we never quite know whether that system is feeding us dud information.  There have been a couple of troubling subsamples over the last 24 hours : Ipsos-Mori put the SNP on 34%, Labour on 31% and the Tories on 25%, while a subsample of 18-24 year olds from ICM put Labour way out in front.  Both of those results can be explained away without too much effort - Ipsos-Mori's subsample was particularly tiny, and ICM may have just found a very weird sample, because it had Labour in the lead even on 2015 vote recall.  All the same, both polls are up-to-date, and are at least consistent with the possibility of a Labour surge at the SNP's expense.

On a more reassuring note, the YouGov projection model has remained relatively stable, and today's update suggests the SNP are on course to win 47 of the 59 seats in Scotland.  That's based on interviews over the last seven days, so it covers a slightly longer timescale than most regular polls, but it seems unlikely that the SNP would be as high as 47 if YouGov had picked up a sharp drop in their support over the last few days.  However, the floor for the SNP (ie. the minimum number of seats they're expected to win within a 95% confidence interval) has suddenly fallen from the low 30s to the high 20s - and that may well indicate some kind of boost for Labour, even if those votes aren't necessarily coming from the SNP.  There's no way the Tories on their own could ever be responsible for getting the SNP below 30.

Friday, June 2, 2017

Is it cos she is TYRANNICAL? Theresa rocked as Ipsos-Mori poll suggests she could LOSE HER MAJORITY

You know the drill by now, folks.

*puts on Canadian accent*

It's another terrrrr-ible afternoon for the Conservative party.

Britain-wide voting intentions (Ipsos-Mori) :

Conservatives 45% (-4)
Labour 40% (+6)
Liberal Democrats 7% (n/c)
SNP / Plaid Cymru 3% (-1)
Greens 2% (-1)
UKIP 2% (n/c)

As soon as they published this poll, Ipsos-Mori came under fire from commentators (and indeed from other pollsters such as ICM's Martin Boon) for producing the implausibility of a Labour vote share that isn't far off what Tony Blair achieved in his landslides of 1997 and 2001.  But it should be noted that it's happened in spite of Ipsos-Mori introducing a stricter turnout filter for this poll, which predictably works in the Tories' favour.  Without that, the swing would be even more extreme and the Tory lead would be down to just 3 points.  The other more general point is that the Lib Dems were much, much stronger in 1997 and 2001, which is a potentially reasonable explanation for Labour now having a vote share comparable to that period, even though their current leadership does not command the same public confidence that Blair's team did.

A familiar alibi for the Tories whenever they get a bad poll is that leadership ratings are often more predictive of election results than headline voting intentions are.  There's some truth in that, which is why they must be horrified that Jeremy Corbyn has almost caught up with the Prime Minister in net satisfaction ratings.  We're very used to Theresa May having a negative rating in Scotland, but this is the first time it's happened across Britain as a whole - 43% of respondents are satisfied with her, and 50% are dissatisfied.  Corbyn's ratings are almost identical - 39% satisfied, 50% dissatisfied.  It's true that May does better when respondents are asked to make a straight choice between the two leaders, but even on that measure her lead has dropped catastrophically since April (from 38 points to 15).

The combined 3% vote share for the SNP and Plaid doesn't look great, but as the same firm has only just produced a full-scale Scottish poll giving the SNP a massive 18% lead over both the Tories and Labour, we probably shouldn't be too alarmed.  The Britain-wide sample was 1046, so the Scottish subsample will have been less than 100 even before the turnout filter was applied.

As long-term readers of this blog will remember from the independence referendum, Ipsos-Mori always publish two sets of voting intention numbers side-by-side - the headline numbers filtered by turnout, and numbers without the turnout filter applied.  On the latter measure, this poll is truly jaw-dropping.

Britain-wide voting intentions (Ipsos-Mori, without turnout filter) :

Labour 43% (+6)
Conservatives 40% (-6)
Liberal Democrats 9% (+2)
SNP / Plaid Cymru 3% (-1)
UKIP 3% (+1)
Greens 1% (-1)

I suspect a number of politicians on the left will be starting to think about the potential virtues of Australian-style compulsory voting.  If we had it for this election, we'd currently be talking about the probability of a hung parliament, rather than the vague chance that it might just about happen.

Heartbreak for Ruth as SNP enjoy double-digit lead in a poll conducted prior to the Tory manifesto meltdown

With all due respect to the Herald, I fail to understand what earth-shattering insights we're supposed to glean six days before a general election from the publication of a poll that was conducted between two and three weeks ago.  Yes, the "new" BMG poll is part of a regular series and the long delay between fieldwork and publication is not untypical, but during the heat of a campaign everything becomes very compressed, and (as we've discovered) it's possible to get significant shifts of opinion over the course of a few days that might otherwise take weeks, months or years.  If it wasn't feasible for BMG to get their numbers out a bit quicker this time, I struggle to see the point of the exercise.  It even looks like almost all of the fieldwork preceded the disastrous Tory manifesto launch, which means the poll practically comes from a different political era.  However, for what it's worth, here is BMG's estimate of what the state of Scottish public opinion was quite a long time ago...

Scottish voting intentions for the UK general election (BMG, 12th-18th May) :

SNP 43%
Conservatives 30%
Labour 18%
Liberal Democrats 5%

There's nothing remotely surprising about those figures - they're pretty much bang in line with what other firms were showing during the earlier part of the campaign.  We know with a high level of confidence that there has been a Labour recovery since then, and that the Tories' second place is now under some kind of threat.  But what is much less clear is whether the extra Labour votes have come predominantly from the Tories themselves, from the SNP, or from both parties roughly equally.  Until that question has been answered, it's a bit redundant of the Herald to be using ancient polling data in an attempt to fuel speculation that Angus Robertson is under threat of losing his seat in Moray.  If there has been direct movement from Tory to Labour since the BMG poll was conducted, that threat may have receded somewhat.  The much bigger 18% SNP lead in the more up-to-date Ipsos-Mori poll would tend to support that interpretation, but isn't sufficient as proof.  What we really need are updated figures from a firm that has already polled during the campaign.

Thursday, June 1, 2017

So what does THIS mean for your crazed plan to destroy Scottish democracy, Ruth? New SHOCK polls reveal Tory support is dropping LIKE A STONE

In case you're wondering, the title of this blogpost is merely a fond tribute to a characteristically barking mad anti-SNP headline in the Express yesterday.  Nevertheless, it's quite true that the last 24 hours of polling have been pretty brutal as far as the Tories are concerned.  YouGov polls in Wales and London have shown Labour's already substantial lead in both places increasing slightly (and with the Tories further behind than they were in 2015).  There was also a thunderbolt Britain-wide YouGov poll last night which suggested the race had moved firmly into hung parliament territory.

Britain-wide voting intentions (YouGov) :

Conservatives 42% (-1)
Labour 39% (+3)
Liberal Democrats 7% (-2)
SNP / Plaid Cymru 5% (+1)
UKIP 4% (n/c)
Greens 2% (n/c)

The SNP lead the Tories by 42% to 28% in the Scottish subsample.  That's better for the SNP and worse for the Tories than the previous two YouGov subsamples, but only very, very slightly.  It's more than a little frustrating not to find clearer evidence that the Tories' extraordinary tumble across Britain has been replicated in Scotland.  Even the Ipsos-Mori poll yesterday didn't really help in that respect, because there were no recent baseline figures to measure from.

It's still a bit of struggle to get to grips with YouGov's seats projection model, which also provides daily percentages for the popular vote.  Today's update puts the Tories on 42% and Labour on 38% - if that can be regarded as a de facto poll, it may indicate that the position has stabilised for now.  The SNP's projected seats total has decreased from 51 to 47, which presumably means that the most recent sampling was either a bit less favourable for the SNP or a bit more favourable for the Tories.  If so, that may just be random sampling fluctuation, of course.

The seats projection has a huge margin of error, with YouGov only saying they are 95% confident that the SNP will fall somewhere between 35 and 54 seats.  That's a timely reminder of just how many current SNP seats may be vulnerable if the current polling estimates are overstating the party, or if differential turnout rears its ugly head.  It's also worth making the point that YouGov's central projection of 47 seats includes Orkney and Shetland, which stretches credibility somewhat.  The projection model does use actual polling in each constituency, but fills in any gaps in the small sample with respondents from other constituencies, and that's probably what is going wrong in the Northern Isles.  You can't assume that a 35-year-old, No-voting, Remain-voting male in a middle income bracket who lives in Lerwick is likely to choose the same party as someone from the same demographic groups who lives in, say, Dingwall.  The political culture in Orkney and Shetland is radically different, and I'm not sure either YouGov or Ashcroft are taking sufficient account of that (although obviously I would be delighted to be proved wrong).

Last week, Panelbase overhauled their methodology to make it much more similar to ICM's, and thus make it considerably more Tory-friendly.  In spite of that, their new poll today tells a familiar tale in terms of the direction of travel...

Britain-wide voting intentions (Panelbase) :

Conservatives 44% (-4)
Labour 36% (+3)
Liberal Democrats 7% (n/c)
SNP 5% (n/c)
UKIP 5% (+1)
Greens 3% (+1)

*  *  *

A couple of people have emailed me to ask why I'm ignoring the "Ashcroft constituency polling".  For clarity, there isn't any Ashcroft constituency polling in this campaign - he's instead using a projection model similar to YouGov's.

Wednesday, May 31, 2017

Massive boost for the SNP as Ipsos-Mori phone poll gives them whopping 18% lead

Who can now believe that we used to dread the arrival of Ipsos-Mori's quarterly "Hi, I'm John MacKay" poll for STV?  Recently, it's often been our salvation.  The new one confirms the message from other firms that the SNP are several points down on their vote share from 2015, but crucially it differs from earlier polls in suggesting that the unionist vote is split down the middle, leaving the SNP with an enormous lead.  In a first-past-the-post election, what matters most is not the leading party's vote share, but the size of the gap between that party and its nearest challenger.  It's suddenly no longer even clear who that challenger will be.

Scottish voting intentions for Westminster (Ipsos-Mori) :

SNP 43%
Labour 25%
Conservatives 25%
Liberal Democrats 5%
Greens 1%
UKIP 1%

And now for the bad news, or at least the slightly less good news.  Although the polls in general slightly underestimated the SNP in the run-up to the 2015 election, Ipsos-Mori were very much an exception - two of their three polls put the SNP on 52%, and the final one gave the SNP 54% and Labour just 20%.  The actual result was SNP 50%, Labour 24%.  Because Ipsos-Mori don't weight by past vote, there's no guarantee that any skew towards the SNP has been resolved.  So this poll does not necessarily remove the danger (suggested by SurveyMonkey's findings) that the SNP may slip to 40% or below. 

Nevertheless, an 18% lead gives the SNP an enviable buffer against any polling error, and as it happens the final Ipsos-Mori poll of the 2015 campaign actually overestimated the Tories as well. 
So who will win the battle for second place in the popular vote?  Labour catching up with the Tories is not a bolt from the blue - although the SurveyMonkey poll suggested the Tory vote was holding up, Labour were firmly within striking distance.  Recent subsamples have told a similar story - Labour have typically been in the 20s where previously they were in the teens.  The balance of evidence is that the Tories still have the advantage, but it does appear that one crucial aspect of Ruth Davidson's legend is now under genuine threat.  Gosh, wouldn't that be a tragedy?

YouGov projection model points to better SNP result than regular polls

YouGov seats projection :

Conservatives 310
Labour 257
SNP 50
Liberal Democrats 10
Plaid Cymru 3
Greens 1
"Others" 1
Northern Ireland Parties 18

Hung Parliament : Conservatives short by 16, Labour short by 69

I was going to do my customary 'shock, horror' headline, but my heart wasn't in it this time, because to be honest I just don't believe this projection - it doesn't pass the 'smell test' at all.  If anything, YouGov have until now been more pessimistic about the SNP's prospects than other firms, so it's hard to understand why they're suddenly estimating an outcome that would be very nearly as good for the SNP as the 2015 result.  And the explanation of the model used for the projection has so far been as clear as mud - the Times article refers twice to a YouGov "poll", and yet Sam Coates (who wrote the article) has loudly insisted on Twitter that it isn't a poll at all.  He also added that it's based on a 7000-strong sample interviewed over seven days, which makes it sound very much like a poll, or some sort of poll aggregate.  Apparently a more detailed explanation will shortly appear on the YouGov website, so hopefully it will make more sense then - although I suspect I'll still disbelieve the numbers even after I know more.

There has been some debate over whether the projected result would be enough to dislodge the Tories from government.  The answer is that it almost would be - but not quite.  A potential "progressive alliance" of Labour, the SNP, Plaid Cymru, Greens and the SDLP would probably have either 313 or 314 seats - the uncertainty is because the SDLP's result has not been estimated.  That would be sufficient to slightly outnumber the Tories on their own, but not to outnumber the Tories and Northern Ireland unionist parties in combination.  The DUP would always favour a Tory government over one led by Jeremy Corbyn, so in practice the Tory-led forces would have the edge, and could only be dislodged if the Lib Dems offered their full allegiance to an unstable-looking Corbyn-led alliance.  That seems highly unlikely, especially given that individual Blairites within the Labour ranks would probably be trying to sabotage any arrangement.

One thing's for sure, though - Theresa May's own position as Prime Minister would look utterly untenable.

Tuesday, May 30, 2017

The campaign has moved onto the SNP's ground

Just a quick note to let you know that I have a new article at the International Business Times to mark the SNP's manifesto launch.  It argues that the campaign's reversion to a traditional battle between civilisation and the Tories may have worked in the SNP's favour as polling day approaches fast.  You can read the article HERE.

Ice-cold ICM poll sends chill down Tyrannical Theresa's spine

I'm just back from the SNP manifesto launch in Perth.  I don't suppose I learned anything that I wouldn't have done from watching on TV, but there was that unmistakeable frisson of excitement in seeing superstars like Brian Taylor and Alan Cochrane in the flesh.  I also took some photos of Murdo Fraser abandoning the last vestiges of his dignity by leading a Dad's Army-style "protest" outside the hall - I'll have a look later and see if they're worth sharing with you.

Meanwhile, we have a new GB-wide poll from ICM...

Conservatives 45% (-1)
Labour 33% (+1)
Liberal Democrats 8% (n/c)
UKIP 5% (n/c)
SNP 4% (n/c)
Greens 3% (+1)
Plaid Cymru 1% (n/c)

The Scottish subsample is : SNP 43%, Conservatives 26%, Labour 22%.  That's reassuringly normal after Survation served up the SNP's worst subsample of the campaign last night (albeit based on a particularly tiny number of respondents).

Some people (including journalists who should know better) have an irritating habit of regarding only one poll as important - the last one they saw.  So we'll undoubtedly hear that this ICM poll has got the Tories back on track with a 12-point lead, and that Theresa May is once again heading towards a three-figure majority, etc, etc, etc.  That ignores the fact that ICM have clearly emerged as the most Tory-friendly pollster in this campaign, and that 12 points is the lowest gap they've shown since the election was called.  The direction of travel strongly suggests that any new poll from YouGov would be likely to continue showing a relatively tight race, possibly on the fringes of hung parliament territory.  As none of us will know until election night which polling methodology (if any) is the correct one, all that today's poll does is underscore the increasing and genuine uncertainty over the outcome of this election.

Monday, May 29, 2017

Indications that the new Survation phone poll may show Tory lead dropping to just 6%

I'm going out on a limb with this one because these figures aren't officially confirmed yet, but having done a bit of Kremlinology on Twitter I'm inclined to believe they're probably accurate.  They seem to have been published early by mistake and then quickly deleted.  (And I should stress that I'm not breaking an embargo, because I haven't been sent embargoed information.)

Britain-wide voting intentions (rumoured Survation phone poll results) :

Conservatives 43% (n/c)
Labour 37% (+3)
Liberal Democrats 8% (n/c)
UKIP 4% (n/c)

The apparent drop for "others" isn't necessarily a cause for concern, because there's still around 8% of the vote that the four parties listed above don't account for, which in theory leaves plenty of room for a decent SNP share.

Some people are attributing mystical significance to the Survation phone series, because it uses the same methodology as an unpublished poll late in the 2015 campaign which (unlike almost every other poll) was very close to the final result.  As that was just a single poll, though, it's unclear whether that good performance was achieved by chance (ie. random sampling variation) or because of superior methodology.  If the latter is the case, it may be significant that Survation seem to be producing results similar to YouGov and ORB, rather than similar to the more Tory-friendly pollsters ICM and ComRes.

Phone polls are rare in this campaign, but after the shock of the EU referendum we've probably moved past the point of assuming that phone polling is bound to be more accurate than online polling.  ICM have actually abandoned phone polling altogether because of the difficulty of getting a representative sample by phone.  Nevertheless, if Survation's figures are confirmed, it'll at least be reassuring to learn that phone polls don't appear to be producing better results for the Tories.

Sunday, May 28, 2017

Full-scale Scottish poll from SurveyMonkey gives SNP a 10% lead over the Tories

Today brings word of the first full-scale Scottish poll for quite a while, but frustratingly it's from SurveyMonkey - they have little track-record in Scotland and therefore it's very difficult to interpret their results.

SNP 39%
Conservatives 29%
Labour 25%
Liberal Democrats 4%

The percentage changes that are being reported in some quarters are from the actual 2015 general election result and not from any previous SurveyMonkey poll.  We don't know whether SurveyMonkey's methodology is SNP-friendly, Tory-friendly, Labour-friendly or essentially neutral, so it's impossible to say for sure whether 39% is 'bad' for the SNP or whether 29% is 'good' for the Tories.  The Sun are attributing epic significance to the fact that the SNP are below 40%, but in reality 39% is within the standard margin of error of the 41-42% recorded by the SNP in the two YouGov polls in the campaign so far.  A 10% gap between SNP and Tory is the lowest reported by any firm during the campaign, but not by much - Panelbase had the gap at 11%, and nobody (with the arguable exception of Google surveys) has had it higher than 15%.  So we'd be jumping to enormous conclusions if we assume that this single poll is proof that the SNP's position has worsened over the course of the campaign - it's perfectly possible that SurveyMonkey's methodology would have produced similar results (at least in respect of the SNP-Tory battle) a month ago.

Nevertheless, there are a few possible causes for concern here -

* The poll reinforces the impression that there has been a Labour recovery, and opens up the possibility that the extra votes are coming more from the SNP than the Tories.  If that is indeed what's happening, it's a tragedy twice over.  Voters moving from SNP to Labour are just making it more likely that the Tories will win several seats in Scotland, and thus potentially boosting Theresa May's overall majority.  In the vast majority of the Tories' target seats in Scotland, the SNP are the only party that can stop them.

* For what it's worth, SurveyMonkey's numbers are eerily similar to today's Scottish subsample from YouGov, which says : SNP 40%, Conservatives 30%, Labour 25%, Liberal Democrats 4%.  As far as I know, YouGov are the only firm who claim to weight their Scottish subsamples properly (although admittedly the margin of error on such a small sample will still be enormous).

* There's a distinct lack of evidence in any of this that the Tory slump south of the border has been replicated in Scotland.  Perhaps that evidence will never arrive - perhaps it simply didn't happen.

* If the SNP's lead over Labour has slipped to 15% or below, we have to accept that some seats may be lost to Labour (although it should be only a handful, unless the situation deteriorates).

My own interpretation is that the SNP are simply suffering from their traditional handicap in Westminster elections - this is an away fixture for them, and they are being largely ignored by the London broadcast media which unfortunately still enjoys such huge influence in Scotland.  (2015 was a freakish exception because of the paranoia in London over the Jocks actually having some say within government.)  However, the good news is that the SNP are still comfortably in first place, and if they can get their voters out (a vitally important 'if') the first-past-the-post system should convert that into a handsome victory.  Even The Sun acknowledge that the SurveyMonkey poll would see the SNP winning more than two-thirds of seats in Scotland.  And there are some advantages to this election being increasingly framed as a traditional Tory v Labour battle - it means that in some battleground seats, the SNP will be able to turn the tables on the unionists' favourite tactic by targeting Labour and Lib Dem supporters with the message : "only the SNP can stop the Tories here".