Recent Posts

So an 18% swing again in the latest Blue wall, compared to around 15% on average in current polls.

Blue Wall comparison to GE2019:

Lab 36% (+15)
Con 29% (-21)
LD 25% (-2)

On these figures, the typical 'Blue Wall' seat of Woking, which is one of the 42 polled,  becomes Con 29, LD 29, Lab 31.

Yes, YouGov and Opinium both use panels for their polling. I don't think any others do, with the possible exception of Survation, who definitely have a panel, but am not sure if that is used for polls or just other, commercial work.

Both YG and Opinium prompt for RUK and Greens. 

One significant difference in their findings is that YouGov always have more respondents answering that they 'Would Not Vote'. They typically get around 15% of these to Opinium's 8%. Once these are excluded, their figures tend to be very similar indeed - both for their VI and regional data. if anything, Opinium's unadjusted' data shows slightly higher Lab leads. So it is very largely the 're-weighting' that Opinium have done for the past 18 months which makes the difference. 

With the Tories now on just 26% on average in the polls, they could well struggle to hold this seat, despite starting from a 60% share at the last General Election.

This would see them reduced to a 41% on UNS, or a 35% share on a proportionate swing.

As I pointed out 3 weeks ago when Dorries first said she was going to resign, the Conservatives have actually fared "worse than proportionately" in each of the 5 contested by elections where they have been defending during this parliament: Chesham and Amersham, Old Bexley and Sidcup, North Shropshire, Wakefield and Tiverton and Honiton. So using these as a pointer, their vote share could be halved to around 30%. 

Predicting where those lost votes will go is far more difficult, though. 

Savanta's figures are  the most difficult of any pollster to understand, as their tables run to more than 250 pages per poll, and I have never found any details of what their methods are. It appears from their figures that they do not prompt for RefUK or Greens, but beyond that it's all a mystery.

Unless they are following the re-weighting methodology used by Opinium, I think it's largely coincidental that they produce similar headline figures for Con and Lab - and they are very different for the other parties. Opinium's most recent poll of 23 June would have been a 25 point Lab lead (23/48) under their old methodology, which would as usual have put it much in line with YouGov who reported 22/47 on 21 June. 

Professor Rose's words are: " if the lead falls to 5 percent the Conservatives do not win the election but Labour loses its majority..."

The clear meaning is that a 5% lead is the point at which labour ceases to have a majority - not the point where Con and Lab are equal. But if you are in any doubt, then check Electoral Calculus itself where if you enter figures with a 5% Lab lead (and no tactical voting) it produces figures of Lab 314, Con 270. You would imagine that Professor Rose of Electoral Calculus's comments are in line with what his own tool is showing.

Besides, a Labour lead of 5%, and a swing of 8.3% from 2019 would be bound to give Labour more seats than the Tories, even using UNS. 

That equals the highest Lab lead that Savanta have shown for 4 months - dating back 17 polls to early February. 

Having checked the Conservatives' performance in recent by-elections where they have been defending, the pattern is clearly that they have fared even worse than a proportional swing would suggest. I'm fairly sure that this has been so historically when a Tory government has been at a low ebb in the polls, too, but there are only a few 21st century examples (e.g Corby in 2012).

The most recent exaamples are the two by elections on 22 June 2022 of Tiverton and Honiton and Wakefield. Both were Tory losses, and both with proportionately identical Conservative losses. The Tories held just 64% of their vote in each case, going down to 38.5% from 60.2% in T&H, and to 30% from 47.2% in Wakefield. Of course, the LDs gained far more votes in T&H than Lab did in Wakefield, but for now I am just trying to predict the Tory vote.

Looking at the polls from that time, the Conservatives were averaging 33%, so somewhat higher than their current 29%. This means that they were holding 74% of their 2019 vote share (@33/44.7), and those 64% by-election retention rates were a fair bit worse than proportional. 

Turning to a couple of other relatively recent Tory by-election losses: In North Shropshire in December 2021, they held just 50.4% of their 2019 vote, amid a LibDem surge. Polling at that time put them around 32%, so a 71.5% retention rate. Old Bexley and Sidcup was just a couple of weeks earlier on 1 Dec 2021, but pre-dated Partygate. The Con VI was then around 36.5%, so they were retaining 82% of their 2019 share. But in that by-election they again fared somewhat worse, retaining just under 80%.

I have not covered the 'Lab hold' by elections, but to be brief, the Con vote there has generally fallen in line with the proportion implied by polls at the time (and no more).

But in those seats where the Conservatives have been defending, their own vote has sufered a 'worse than proportional' loss every time. And considerably worse than proportional on some occasions, especially where the LibDems are the challengers. 

So to return to those 60% 2019 Con vote shares in Mid Bedfordshire and Selby& Ainsty: UNS says that they would fall to around 44%; proportional swing says they would be down to 39%. And the kind of losses we've seen in recent Con by-election losses suggests it could go down to the low-to-mid 30s. 

c-a-r-f-r-e-w

"My point however was also that the VI kept climbing AFTER the election, getting above the Tories, showing that more gains were possible."

If you're making that comparison, it's worth looking at post-election polls after other General Elections, too.

If you check the polls from a month or two after the GEs in 1987, 1992, 1997, 2001, 2005, 2010, 2015 and 2019, it is clear that in each case the winning party held a larger lead than thay had at the General Election itself. It's the 'halo effect' - and it does not mean that Thatcher, Major, Blair , Cameron, or Johnson could in each case have done better if only they had timed the election for a month or two later. It was the election win itself which provided the polling boost - much as a good by-election win can do.

To give a couple of recent examples: in the first 6 polls after GE2019, the Tories had a VI of 47% and an average lead of 16-17 points. And it was the same after GE2015 - the Tories were up by 2 points to nearly 40% in the first 6 polls of the new Parliament and led by 10% (compared to 6.6% at the GE). The most extreme example of all was Labour under Blair in 1997. The Lab VI averaged 57.5% in the 16 polls from May to Dec 1997. 

2017 was unusual, in that the 'halo effect' went to Corbyn rather than May. But this is unsurprising given how he had effectively 'won' the campaign by making some very unexpected net gains and May losing her majority in Parliament.

Not convinced of that....

Compare the swings in Boris Johnson's Uxbridge and South Ruislip constituency with those in the neighbouring Ruislip Northwood and Pinner.

Uxbridge and S Ruislip: 2017 Con to Lab 6.5%, 2019 Lab to Con 2.1%

Ruislip, N'wood & Pinner: 2017 Con to Lab 6.6%, 2019 Lab to Con 2.4%

So Johnson has fared 0.2% worse than Nick Hurd/David Simmons/whoever in the neighbouring seat.

en.wikipedia.org/wiki/Uxbridge_and_South_Ruislip_(UK_Parliament_constituency)

en.wikipedia.org/wiki/Ruislip,_Northwood_and_Pinner_(UK_Parliament_constituency)

Labour would certainly expect to win Reading West, as they only need a small swing.

But Mid Bedfordshire is not really safe on current polling, despite the 38% Tory majority. It requires more than the current average GB Con>Lab swing of 13%, but recent South of England cross-breaks average swings of 19% per the past 7 YouGovs and 18.5% per Opinium. So we could expect a close contest. 

It is also a seat which has (slightly) moved Labour's way in the 2 most recent elections. Comparing 2019 with 2015, the Conservatives are up 4 points,  Labour up by 6, and LDs up by 5 in 3rd place.

YouGov tables are here.

docs.cdn.yougov.com/g63pvg88hg/TheTimes_VI_230531_W.pdf

As usual, there are more 2019 Con 'Don't Knows' than Lab (by 23% to 11%). This would reduce the Lab lead by 3-4 points if you apply a re-weighting or reversion. 

Nothing particularly unusual in the cross-breaks. Labour still ahead in the South (and all other geographical areas). Conservatives still losing support most in the Midlands. Brexit tracker supports 'Wrong to leave' by 55 to 33 points. 

A good article from Callum Jones on Polling Report on the Ashcroft Constituency poll for Uxbridge and South Ruislip - he doesn't think it is likely to be accurate as it's a phone poll and hugely divergent in the under-sampled younger age-groups..

pollingreport.uk/articles/reason-to-doubt-uxbridge-constituency-polls-boris-bounce

As for Ashcroft's claims that "Johnson is able to command extraordinary affection and loyalty" ; this wasn't in evidence in his previous constituency results here. There was a larger than average swing to Labour in 2017 (6.5%) and a lower than average swing to him in 2019 (2.1%). 

Also, worth noting that the seat forms part of R&W's 'Blue Wall' sample, along with 41 other London and Southern constituencies. This typically shows a swing 2-3 points larger than general GB polling - for example, yesterday's was a 16.5% Con>Lab swing. 

That's not quite so - the Tories lost about 3 points over the 2017 campaign, from 46.5% to 43.5%. May called the election on 18 April. These are the comparative figures for the 7 polls on 18-20 April 2017 compared to the actual result.

Con 43.5% (down 3 from 46.5)
Lab 41% ( up 15 from 26%)  - Note polls averaged 36.5% for Lab and overstated UKIP
LD 7.6% (down 3.4 from 11%)
UKIP 1.9% (down by 6.4 from 8.3% av)
Green 1.7 (down 1.3 from 3%)

However, the Tories probably lost more than 3 points to Lab in the campaign, as their 3-point fall in VI coincided with a spectacular fall in the UKIP VI - from 8.3% in the early polls to a 1.9% share on election day. It isn't credible to suggest that three-quarters of those who had been backing UKIP just a few weeks earlier switched to Labour during the campaign. YouGov's final poll of the campaign showed just 18% of the UKIP2015 vote switching to Lab and 44% of it going to the Conservatives. 

en.wikipedia.org/wiki/Opinion_polling_for_the_2017_United_Kingdom_general_election

I quite often look at the comparative 'Don't Knows' in YouGov. It looks to me like an adjustment in line with Opinium (or using ICM's old 50% re-allocation which has a very similar effect) knocks about 3-4 points off a Labour lead, on average. 

As I have mentioned before, YouGov's and Opinium's fieldwork actually produce very similar findings. Hence an 'adjusted YouGov' looks very much like Opinium, and vice versa. The main difference is that YouGov have the Greens higher and Lab a little lower.