|
|
Post by James E on Jul 12, 2024 9:14:28 GMT
It's interesting that Luke Tryl (above) comments that "it would take just a 6% swing to the Conservatives for Labour to lose its majority". If this is correct, then Labour could now maintain a majority when 1 or 2 point behind in the popular vote, which is really quite similar to the position in the Blair years. So Peter Kellner's "Grumblers and Defectors" piece in May 2023 has proved totally wrong - even more so than I had predicted. If a simple UNS approach is taken of looking at Labour's most vulnerable seats and assuming an equal swing to all of the second-placed parties, then Labour could withstand a swing of 4.35%. This is equivalent to a lead of 1.5%. My best estimate is that Labour could probably now maintain a majority provided that they have any sort of overall lead over the Tories in the popular vote. Conversely, the Tories would now appear to need a swing of lead of 7%-10% or more to achieve a majority, with a 0-10% lead resulting in a hung parliament. Looking back at similar calculations of Labour's apparent UNS-based target from previous elections, we now have : Post GE2005 -2% Post GE2010 2% Post GE2015 9% Post GE2017 7% Post GE2019 12% Post GE2024 0% (?) ukpollingreport2.proboards.com/thread/67/labour-win-outright-majoritySo these figures have proved to have very little predictive value in the past, though never to quite the same extent as in 2024. Having said that, the pattern has been for the 'bias' of FPTP to go towards whichever of Lab or Con is ahead - so to Labour post 1997 and to the Tories post-2010. In particular, both Lab and Con have benefitted from 'fist time incumbency' bonuses when defending their newly-won seats in 2001 and 2015 respectively. The kind of detailed polling we will need for predicting seats will involve several separate battlegrounds, including Lab V SNP, seats with a high Muslim population, and the 'new' and 'old' Labour-held marginals, which of course all followed very different patterns last week. |
|
|
|
Post by James E on Jul 4, 2024 20:39:05 GMT
Just for comfort's sake... I remember someone saying exit polls haven't been more than x% out for decades. Can anyone remember what it was? I remembered the 2015 exit poll understated the Tories seats by about 15. Most have been more accurate than that. However, this is an especially difficult election to model, and the margin of error for seats may be wider this time. I think the final You Gov MRP is the most trustworthy guide we have at the moment - it was posted by Shevii yesterday on page 10 of this thread. Their full write up is well worth a read. |
|
|
|
Post by James E on Jul 4, 2024 19:41:45 GMT
JamesE? Hope he's not ill. I am fine but not able to post much at the moment. Election was not ideally timed for me. My initial prediction of a Labour majority around 200 still looks possible although with a lower vote share than I was thinking a month ago. And thanks for your concern - I am very well. |
|
|
|
Post by James E on Jun 25, 2024 16:06:08 GMT
The Wikipedia page that shows opinion polls for the 2019-24 parliament has added a feature that lets you look at the local regression trend lines for the polls taken during the election period. Just press the appropriate button. Looking at this suggests that Labour had dropped from about 45% to 41% (-4), the Conservatives from 24% to 21% (-3), Reform has risen from 11% to between 16 and 17% (+6), the Lib Dems have risen from just under 10% to 11% (+1) and the Greens and SNP are absolutely dead straight at just over 5% and just under 3% respectively. Also all these movements have flattened out lately as if that is probably it - although I wouldn't be surprised to see Reform drift down a little as the GE approaches. My gut feel is that the Conservatives will get closer to 25% and Reform 15% on the day. en.wikipedia.org/wiki/Opinion_polling_for_the_2024_United_Kingdom_general_electionThe headline movements can be a little misleading as there have been methodological changes by several pollsters to their published figures: YouGov, MIC, Survation and JLP have all made changes which reduce the Lab VI, and should increase Con. My own 'like-with-like' comparison against the overall pre-election averages from various pollsters in the period from 1 Jan -22 May of this year is: Lab -1.7 LD +1.0 Con -4.4 Ref + 5.1 That's from the 10 most recent polls from JLP, Delta, R&W, MIC, WeT, Savanta, Opinium, Techne, BMG and YouGov, but using comparable data for those who have changed methods. Also, my comparison is against a lower base to pjw's as the Lab VI was at a higher point when the election was called than it had averaged in the previous 4 months. |
|
|
|
Post by James E on Jun 25, 2024 9:30:12 GMT
... .My understanding of this 'unwinding' adjustment made by YouGov and Focaldata is that this shifts the constituency results a bit towards Uniform Swing, and away from Proportionate Swing. By definition, Uniform Swing is done by comparison only to the previous General Election, and not earlier GE results. From what I have seen of YouGov's MRP constituency results for the South of England, it looks to me like the effect is about a 1.5% to 2% swing back to the Tories in some of their safest seats from 2019. This would have a similar effect to what Focaldata are doing, as described by James Kangasooriam, who estimates that their 'unwinding' pushes of 30-50 seats back to the Tories, because their raw findings indicate a swing which is 'beyond proportional'. If the actual swing is 'merely proportional' then the effect would be to overestimate the Tories by around 20 seats. Hmm. This is a longstanding issue by now. I'm not exactly sure from your description exactly how they are doing this, but going all the way back to 2010 the traditional voting pattern was distorted by brexit. Even before then Con had picked up being negative about the EU was a way to attract extra votes, particularly obviously people inclined towards UKIP. Ukip took votes 2:1 from con compared to lab, and then these transferred to con more and more as the issue became Brexit. Hence labour lost the red wall seats, as their brexit voters moved to con. The issue would be that the core loyalty of those seats might well not be as expressed in 2019, but you would need to look much further back more to 2005. Brexit wasnt called that back then, but even in 2010, I think it played a part in Cameron winning at all. Without Euroscepticism con would not have achieved the number of MPs they did then, and the balance with lab would have been more even, lab might have had more. and then we might have been looking at a lab-lib pact, which many thought more natural. Brexit has not totally unwound even now, but there is a majority to rejoin and presumably even bigger numbers who believe brexit was miss-sold to them. ..... I have provided YouGov's data some time ago about Voting Intentions and 'Brexit Attitudes', so to repeat what I said then: Voting Intentions are just as much falling along pro-and anti-Brexit lines as they were in 2019. If someone thinks Brexit was the right thing to do, they will probably be voting Con or Ref. If they think it was Wrong, they will probably be voting for one of Lab/LD/SNP/Green/PC. This holds true for about 88% of voters, but with around 6% of each group crossing over ( e.g. Pro-Brexit Lab or anti-Brexit Con). There is a large swing with Leave voters, but this is the product of the Tories' losses to Reform UK, rather than an exceptional level of Lab gains :Lab are up by about 10 points with Leave voters from 14% to 24%. This is a separate thing to 'uniform', 'proportional' or 'beyond proportional' swing. Let's work on the basis that the Conservatives are now on course for 22.3% of the GB vote, as opposed to 44.6% in 2019, so that their overall share is being roughly halved. Uniform swing shows them down by 22 points, so they would fall from 32% to 10% in London, and from 55% to 32% in the South or Midlands of England. Proportional swing shows their vote being halved, so they would get 16% in London and 27.5% in the South or Midlands. There is an awful lot of detailed regional data showing that the Tories are losing a proportional or greater-than-proportional share of their support in their strong areas - so per these figures, they are getting less than 27.5% in the Midlands, but more than 16% in London. Even with their 'adjusted' figures, which boost the headline Con share, it is clear that they are keeping a lower proportion of their support in their 'strong' areas, which would be disastrous for them if it really happens. However, several MRPs such as YouGov, Focaldata, and MIC adjust their seat figures away from proportionality and towards uniform swing. This is not a huge adjustment in constituency vote shares, and generally takes the results from being 'beyond proportional' to somewhat less than 'proportional', but still a long way from 'uniform'. And it means that there are a significant number of safe Conservative seats where the published MRPs show the Tories holding on narrowly where the actual data findings would show them losing narrowly, |
|
|
|
Post by James E on Jun 24, 2024 20:00:35 GMT
@james E Re-MRP surveys.Are you aware as to the extent to which these surveys rely on recent election results when making forecasts for particular constituencies? If they do so, do they only take account of the most recent GE - ie 2019 - and none at all of the GEs of 2017 and 2015? They must make use of the 2019 result to provide a baseline for each constituency. But it is also used when making the adjustment known by YouGov as 'unwinding' .My understanding of this 'unwinding' adjustment made by YouGov and Focaldata is that this shifts the constituency results a bit towards Uniform Swing, and away from Proportionate Swing. By definition, Uniform Swing is done by comparison only to the previous General Election, and not earlier GE results. From what I have seen of YouGov's MRP constituency results for the South of England, it looks to me like the effect is about a 1.5% to 2% swing back to the Tories in some of their safest seats from 2019. This would have a similar effect to what Focaldata are doing, as described by James Kangasooriam, who estimates that their 'unwinding' pushes of 30-50 seats back to the Tories, because their raw findings indicate a swing which is 'beyond proportional'. If the actual swing is 'merely proportional' then the effect would be to overestimate the Tories by around 20 seats. |
|
|
|
Post by James E on Jun 24, 2024 16:39:18 GMT
I have been critical of Focaldata's MRP figures before, with their nadir coming when they showed LDs, PC and Green winning a combined total of 7 seats in May 2023 ( their combined others of 25 included 18 from Northern Ireland). However, their write-up and reasoning for these figures is a very good read, and I particularly appreciate the section below about the 'unwinding adjustment', which is also a feature of YouGov's MRPs, and has a similar effect: 44 seats according to Kanagasooriam's previous article (also linked below) www.focaldata.com/blog/focaldata-prolific-uk-general-election-mrp"How might we be wrong?
#1 Unwinding adjustment is too aggressive
We deploy an adjustment - which YouGov coined as “unwinding”. This adjustment seeks to cater for some of the historical problems with MRP and regularisation and attenuation. We think it’s best to be really transparent about how material this adjustment is. Our model learns from past election distributions to suppress MRP tendencies to move beyond proportional swing. Our adjustment results in significant change in seat count of c.30-50 seats for the Conservatives, meaning that without it, we would be nearer the lowest ends of MRP forecasts for the Conservatives, and highest for Labour. Our political judgment remains that beyond proportional swing is unlikely - but accept entirely that we don’t know where the “slope” is likely to be between uniform national swing and proportional. The evidence from the 2024 local elections was firmly towards, if not entirely proportional. Please see a long-form discussion of this here. If our estimates were entirely proportional for these set of estimates we estimate the Conservative seat count would be c.20 seats lower.
However, even with unwinding, looking at the marginality of seats the Conservatives only hold c.60 seats safely with a margin of over 5% - and are acutely vulnerable to change in vote share up to election day. Labour by contrast hold c.410+ with a margin of over 5%. This is a staggering difference. We are not forecasting c.60 seats for the Conservatives but it is entirely reasonable to assume this could happen - if our unwinding adjustment has been too aggressive."
www.focaldata.com/blog/bi-focal-17-why-are-all-the-election-mrps-so-different |
|
|
|
Post by James E on Jun 24, 2024 14:46:10 GMT
Does anybody know if Reform didn't exist how many of its present voters would vote Tory/Labour/won't vote? There has been a fair amount of polling on this, and the short answer is that less than half switch to Tory, most going to 'won't vote' or another minor protest party (but not Labour in great numbers). However, I've haven't got time to look this up at the moment - I'm sure JamesE will have it at his fingertips!  Thank You pjw1961 . It took me a few minutes to find in my old posts, but I found two, somewhat inconsistent sets of figures: "The overall split of that 8.5% RefUK vote per Survation (with about 1,000 respondents) in April 2024 was: Con 37% Lab 18% LD 6% Green 5% SNP / PC 3% Others 31% This can be compared to YouGov's findings last December, when Reform were on around 8%. When asked what they would do if Reform did not have a candidate, their respondents divided as: Con 31% Lab 3% LD 4% Green 3% SNP/PC 0% Would not vote 27% Others 20%" This is very different to looking at the Reform UK VI by their 2019 vote, where they are overwhelmingly Con2019 (+the 2% BXP2019) [if there are any other requests for stuff I have posted before, it might be a good idea to get these in quickly, as I will not be around much for the final week of the campaign and the immediate aftermath - anyone who is logged in on this site should be able click on my name and look back over my old posts.  ] |
|
|
|
Post by James E on Jun 24, 2024 14:10:53 GMT
Reform polling at 8% London-wide when their appeal lies mostly in the outer suburbs, suggests that in some of those latter places it could be as high as 20%+. This could definitely bring Bromley and Biggin Hill into play for Labour (as some of the MRP polls have already shown). , b roadly agree with you, but it may be wrong to assume that Reform isn't taking votes from Labour, especially in areas where ULEZ may be a factor for some voters. Polling tables suggest that Reform's rise is coming overwhelmingly from the Tories and not Labour. YouGov's last 5 polls show an average of 3% Lab2019 voters switching to Reform, with close to 30% of Con2019s doing so. Savanta's most recent GB poll also show 3% of Lab 2019 going to Ref, so worth about 1 point in the overall RefUK VI. YouGov's most recent London poll in April had 0% Lab to Ref, and 25% Con to Ref. Incidentally to answer lens regarding Lab's performance in London: Labour was getting lower swings in London well before ULEZ became an issue. And where pollsters have asked for Westminster VI and mayoral polling in the same poll, they get very different results - for example, in this YouGov Khan led Hall 47/25 in a poll showing Lab leading Con for Westminster by 54/17. ygo-assets-websites-editorial-emea.yougov.net/documents/LondonResults_Mayor_240430.pdf |
|
|
|
Post by James E on Jun 23, 2024 11:02:06 GMT
Re Clacton. A few facts and figures as to where it stands as a Labour target. It is the Conservatives' 9th safest seat, won by 72% to 16% in 2019. In an ordered list of Labour targets, it ranks as Number 399, meaning that Labour might be looking at 600+ seats if winning there. It was also close to 75% Leave in 2016. Polling earlier this year showed it as Con 38% (-34), Lab 30% (+14), Ref 18% (new). This was in the absence of Farage as a candidate. The overall proportion of Con +Ref of 56% looks very plausible as this is roughly three-quarters of the 72% they got in 2019, in line with the overall fall in the combined ROC share from around 48% to around 36%. Labour's rise of 14 points looked about right to me too. While we can expect big swings in Leave voting seats, this is more a product of the Tories losing Leave voters than Labour gaining them: the overall rise in the Lab-leave vote may be around 8-10 points. Looking at that poll, it seems to me that Labour's best hope was in an even split in the ROC vote - something like Lab 30-34%, Con 30%, Ref 22-26%. With Farage as a candidate, this looks unlikely. The same poll in January showed Labour in 3rd place with Farage as a candidate, with figures of Ref 37%, Con 27%, Lab 23%. Two more recent polls also show Labour in 3rd place, on 18% or 24% compared to the Tories 21% or 27% and Reform's 48% or 42%. It is not really the point though. Party workers should not be treated in this way. I am afraid it is yet another example of Stalinist control freakery from Labour HQ.
I'm not seeking to discuss what you call 'Stalinist freakery', nor to deny Norbold's comments. My 'point' was to provide same facts and figures on Clacton as a possible long-range Labour target. If you are not interested in that then feel free to ignore my post. |
|
|
|
Post by James E on Jun 23, 2024 10:47:32 GMT
Re Clacton.
A few facts and figures as to where it stands as a Labour target.
It is the Conservatives' 9th safest seat, won by 72% to 16% in 2019. In an ordered list of Labour targets, it ranks as Number 399, meaning that Labour might be looking at 600+ seats if winning there. It was also close to 75% Leave in 2016.
A Survation constituency Poll earlier this year showed it as Con 38% (-34), Lab 30% (+14), Ref 18% (new). This was in the absence of Farage as a candidate. The overall proportion of Con +Ref of 56% looks very plausible as this is roughly three-quarters of the 72% they got in 2019, in line with the overall fall in the combined ROC share from around 48% to around 36%. Labour's rise of 14 points looked about right to me too. While we can expect big swings in Leave voting seats, this is more a product of the Tories losing Leave voters than Labour gaining them: the overall rise in the Lab-leave vote may be around 8-10 points.
Looking at that poll, it seems to me that Labour's best hope was in an even split in the ROC vote - something like Lab 30-34%, Con 30%, Ref 22-26%.
With Farage as a candidate, this looks unlikely. The same poll in January showed Labour in 3rd place with Farage as a candidate, with figures of Ref 37%, Con 27%, Lab 23%. Two more recent polls also show Labour in 3rd place, on 18% or 24% compared to the Tories 21% or 27% and Reform's 48% or 42%.
|
|
|
|
Post by James E on Jun 22, 2024 19:13:29 GMT
New polling with @observeruk The Labour lead is now 20 points •Labour 40% (n/c) •Conservatives 20% (-3) •Reform 16% (+2) •Lib Dems 12% (n/c) •Greens 9% (+2) •SNP 3% (+1) Fieldwork: 19 - 21 June. Changes from 12 - 14 June. Opinium. That's Opinium's equal largest Labour lead this year. By my preferred comparison to the Jan-May average by party for them this is: Lab 40% (-1.5) Con 20% (-6) Ref 16% (+5) LD 12% (+2) Green 9% (+2) Without their adjustment, Labour would lead by 23.5%. 42% Lab , 18.5% Con, and 13% Ref I think that the smaller adjustment is more due to the lower Tory vote retention (they only lead Lab by 36/16 with Con 2019 voters) rather than fewer 'Don't Knows'. |
|
|
|
Post by James E on Jun 22, 2024 13:31:18 GMT
“Somebody is going to be very embarrassed when the results come in. In this election it probably won’t matter so much because it’ll be a Labour landslide but in a closer campaign they could be calling it completely wrong. Yes, it will be Tory annihilation – but there’s a big difference between 50 seats and 150.”
That depends on what you mean by a "big difference".
Most decent models and MRPs show around 15 seats moving from Con to Lab for each 1% of swing. However, that does not include any further losses to the LDs, and it overlooks the fact that in the likely proportional swing, the Tories losses accelerate to a higher rate when they fall below around 25 points. So 20 Tory seat losses per 1% swing is quite realistic.
This in turn means that the current range of polls, with Labour's lead over the Tories varying from 14 to 24 points* really could make a difference of 100 seats. But that's a product of First Past the Post, not MRP.
* e.g. MIC's Lab 39% Con 25% V R&W's Lab 42% Con 18%
|
|
|
|
Post by James E on Jun 21, 2024 9:14:17 GMT
The latest 5 polls are fairly consistent in their movements relative to pre-election (1 Jan-22 May 2024) averages for the same pollsters*.
For the most recent polls from Techne, R&W, BMG, MiC and YouGov (17-18 June, not the full MRP) these movements are:
Lab -2%
Con -4.5%
LD +1%
Ref +6%
*for comparisons, I use the 2 pre-election YouGov MRPs, and the unadjusted figures from MiC's tables, as these are the only available like-with-like data in each case.
[16:30 Fri 21 Jun - we can add the latest WeThink (Lab 43, Con 22, LD 8 Ref 13) to this, at least as far as the Lab VI is concerned as it also 2 point lower than the Jan-May average for them, for the others, it's Con -2, LD -1, Ref +2]
|
|
|
|
Post by James E on Jun 20, 2024 19:27:28 GMT
More in Common 🔵 CON 25% (-) 🔴 LAB 39% (-2) 🟠 LIB DEM 11% (-) 🟣 REF UK 14% (-) 🟢 GRN 5% (-) 🟡 SNP 3%(+1) Dates 17-19/ Their unadjusted figures are: Con 22% Lab 41% LD 11% Ref 14% There is a huge anomaly in their East England figures: the 'raw' figures show Lab 31%, Con 28%, but after adjustment this becomes Lab 19%, Con 37%. |
|
|
|
Post by James E on Jun 20, 2024 18:06:26 GMT
|
|
|
|
Post by James E on Jun 20, 2024 12:26:41 GMT
I have lost all faith in MRPs to do anything except predict the overall position that the Tories will lose and Labour will win - which conventional polling is also doing. They are all over the place in terms of the exact result and individual constituencies. Maybe one will prove more accurate than the others when we see the real results, but for the moment pretty worthless as a guide to anything beyond the big picture. Agree PJW - and frankly the variance is so big that even if one MRP lands pretty close to the final result how will we say that's not just random luck given all the others will have missed? James E said earlier that YouGov at least were just about within the MoE in 2019 but even that seems incredibly generous - they were off by 8-9% on both the Tory and Labour seat totals, which given how many clearly safe seats there were to start with is a sizeable failure to translate what was a pretty accurate VI survey into seats won.
yougov.co.uk/politics/articles/26828-final-2019-general-election-mrp-model-small- Put another way, if Johnson's 80 seat majority was within their MoE, so too in the opposite direction was Jeremy Corbyn becoming PM. At which point it's hard not to say their MoE amounted to "yep, very confident there's an election happening". The final YouGov MRP underestimated the Conservatives' seats by 26. YouGov's central forecast was for them to win 339 seats, with a range of 311 to 367 - so within 28 seats in either direction. The main reason for this was the underlying polling of Con 43%, Lab 34%. While you describe that as 'pretty accurate', it underestimated the Tories' vote by 1.7% and overestimated Labour by 1.1%. If these had been the actual voting figures then the Conservatives would have won around 16 fewer seats. So YouGov's slight polling error was the main reason for them getting the result wrong by 26 seats. Or to come at this from another angle, if they had the vote shares absolutely accurate, they probably would still have underestimated the Tories seats total by about 10. Overall, their range of 28 seats above or below their central forecast was equivalent to about a 2% swing. That's quite a generous margin, but is close to what might be considered a 'normal polling error' for UK General Elections. |
|
|
|
Post by James E on Jun 19, 2024 20:57:53 GMT
jayblanc"Ideally, MRPs should report a histogram of likely results, but most often they report a seat count and then a backward calculated vote-share estimate. This is also why when doing a poll-of-polls average you should not include MRP results, because they are third or fourth order extrapolations.)" My understanding of YouGov's published VI percentage figures for MRPs is that these are the sums of the polling responses. These are different from the sums of the constituency estimates, which as you say are extrapolations. So in this case, YouGov's figures of Lab 39%, Con 22%, LD 12%, Ref 15% can be compared to their recent polls.
|
|
|
|
Post by James E on Jun 19, 2024 10:34:32 GMT
graham "That does not work for Wimbledon based on last night's MRP which came up with - LD 33 Lab 32 Con 24 - very much a tossup now." Wimbledon is nowhere near being a 'safe' seat, and is somewhat of an exceptional case. The Tories have almost no chance of holding it, even if the LDs and Lab take an even share of the vote as per Ipsos's MRP. But there are plenty of seats which can be used to test my advice above: as well as Fareham, there are Chelmsford, Salisbury, Runneymede & Weybridge, Reigate, Cotswolds North, St Neots & Mid-Cambs, and many others with a gap of less than 10 points between a 2nd place LD and 3rd place Lab from 2019 (or notional results).It will be interesting to see in 17 days' time how well this holds up. As a wider point: there are wildly varying MRP results by constituency for the kind of seats we are discussing, where the tactical position between LDs and Lab is in some doubt. I think that the figures I have shown above are relevant here - the LDs did outperform their general showing in 1997 where they were in second place, but this was concentrated into those seats where they were a relatively close second. This is also borne out by EC/FON's findings from earlier this year on the LDs' performance in their strong, medium and weak seats. www.electoralcalculus.co.uk/blogs/ec_lib2seats_20240226.html
|
|
|
|
Post by James E on Jun 19, 2024 10:08:17 GMT
Re Fareham & Waterlooville.
The notional 2019 result is Con 61%, LD 18.3%, Lab 16.5%.
For most seats, the simple approach to tactical voting is to support whoever was in 2nd place last time, but some allowance needs to be made for polling movements since then, as Labour are up by 9 points, and there has been significant movement between LDs and Lab - cross-breaks show around 25% of 2019 LDs switching to Lab.
Here, the 1997 precedent looks significant to me: the LDs outperformed their general polling position (down 1%) where they appeared to be in fairly close contention, but did not do so where they were in a distant second:
"Conservative held seats from 1992, LDs in second place:
Marginal (15 seats) :Con -9.7%, Lab + 3.3%, LDs + 2.6%
Possible (30 seats) : Con -12.1%, Lab +6.8%, LDs + 2.4%
Safe (98 seats) : Con -12.6%, Lab + 8.8%, LDs -0.4%."
Not how similar the pattern in those 'safe seats' was to the overall movements at the 1997 election ( Con -11.5%, Lab + 9.5%, LDs - 1.0%)
A "safe" seat here was one where the Tories had a majority of 20%+ from the previous election. But these figures do not distinguish between those with large and small gaps between LDs in 2nd and Lab in 3rd. My own analysis of 1997 suggest that Lab will do better where they were a close 3rd to the LDs than where they were distant. In fact, I have not been able to find a singe seat in 1997 where Labour failed to overtake the LDs with less than a 14-point deficit from 3rd to 2nd.
This time, as a simple rule, I would suggest that in 'Con held, LD 2nd, Lab3rd' seats, where Labour is less than 10 points behind the LDs they will probably at least overtake them. This is consistent with current polling movements, and the pattern of how such seats moved back in '97.
|
|
|
|
Post by James E on Jun 18, 2024 19:35:20 GMT
|
|
|
|
Post by James E on Jun 18, 2024 17:00:50 GMT
It's a decent sized sample of nearly 20,000. The VI figures match up with YouGov's most recent full MRP at Lab 43%, Con 25%. The LDs have one point less with Ipsos at 10%, and RefUK 2 more at 12% However, Ipsos has Labour converting the same lead into more seats than YouGov - its Lab 453, Con 115, LD 38, compared to YouGov's Lab 422, Con 140, LD 48. I'd normally believe YouGov's findings against other pollsters, but am inclined to think that the effects of their constituency prompt, and 'unwinding' adjustment may be a bit overdone. |
|
|
|
Post by James E on Jun 18, 2024 11:08:11 GMT
Scottish Westminster Voting Intention: LAB: 34% (-5) SNP: 30% (+1) CON: 13% (+1) LDM: 8% (=) RFM: 7% (+3) GRN: 6% (-1) Via @yougov , 3-7 Jun. Changes w/ 13-17 May. ygo-assets-websites-editorial-emea.yougov.net/documents/Internal_ScotlandVI_Tatical_240607_V2.pdfA 4 point Lab lead is close to the recent average for all Scottish Westminster polls. This is also true for the 4 Scottish YouGov polls so far this year, which have shown Labour leading by 2,1, 10 and now 4 points. We've now had 8 Scottish Westminster polls and MRPs from YouGov and Survation since the election was called. Nearly all of these are in the range of a Labour lead in Scotland of 0-6 points, but with two big outliers: an R&W with Labour 10% ahead, and the Survation MRP with SNP 7 points ahead. |
|
|
|
Post by James E on Jun 18, 2024 9:57:58 GMT
Thanks for the warm welcome everyone Here in Cambridge I've only had one leaflet, from Labour. It's such a stark contrast to the vibrant campaigning of the local elections where we had leafleting and visits from all the major parties except Tory. You can barely tell there's a general election on here. Thinking back to 2019, it was rather similar though. I have some thoughts about Cambridge, and MRPs, but I'm digging into some more data first. Related to it, I recall reading on the old UKPR that YouGov may have better reach to typical non-voters in its panel (being one of the first online-only pollsters). Does anyone know if that is/was true? And if in the years since does anyone know if other pollsters have improved their reach? As others have noted, YouGov are the only pollster who have a real track-record for their MRPs. Their 2019 one was not as successful as 2017, but was still within the outer limit of MoE. But they found that they under-polled those who paid low attention to politics, which in turn led them to understate the Con vote then. They have since then ensured that they include more 'low attention to politics' respondents in their polls, though unsurprisingly they tend to get a low response rate, and as a result generally need to upweight the responses they do get. This might be part of the reason for the volatility of YouGov's published figures compared to other pollsters. YouGov do also include the correct proportion (34%) of those who did not vote in 2019, as do Survation and Techne, for what that's worth. As for Cambridge ( I live just outside it) I think a low swing is likely. Electoral Calculus have Labour winning with 51%, up 3 points. This is likely to be the outcome in a lot of strongly Remain Labour seats, not just in cities such as London, Bristol or Liverpool. www.electoralcalculus.co.uk/fcgi-bin/calcwork23.py?seat=Cambridge |
|
|
|
Post by James E on Jun 18, 2024 9:40:49 GMT
This is interesting, some of the polling companies reallocate don't knows, which tends to favour the tories But there is now evidence when pressed more don't knows will break for Labour than the tories Survation Labour 22% Tories 14% Still undecided 43% x.com/TomHCalver/status/1802744645132132510There are remarkably similar figures to this in the most recent JL Partners poll. Asked of the Don't Knows which party they are most leaning towards voting for: Lab 23% Con 13% LD 10% Ref 10% Still Don't Know 36% Interesting to note that JLP's highly experimental methodology does not follow the logic of this, as it reduces Labour's lead by 5 points, and boosts Ref from 16% to 18%. |
|
|
|
Post by James E on Jun 17, 2024 19:27:51 GMT
We have now had 7 polls since YouGov's one with Labour on 37%, so enough data to check the overall figures against the pre-election averages (using the unadjusted figures for JLP and MIC). Labour's losses have disappeared; this means that they are now polling at the same level as their Jan-May average, as opposed to 1-2 points above at the start of the campaign, or around 2 points below around a week ago. Overall average net movements per the most recent Delta, R&W, MiC, JLP, Opinium, Savanta and Techne, compared to the same pollsters' 20-week pre-election figures are:
Lab +0.2%
Con -4.8%
LD +1.0%
Ref +4.5%
|
|
|
|
Post by James E on Jun 17, 2024 18:12:04 GMT
Redfield Wilton 10,000 mega poll Labour leads by 25%. Tied-lowest Conservative % (worse than Truss). Highest Reform %. 🇬🇧 Westminster VI (14/6-17/6): Labour 43% (+1) Reform UK 18% (+1) Conservative 18% (–) Lib Dem 12% (-1) Green 5% (–) SNP 3% (–) Other 1% (–) Changes +/- 12/6-13/6 Not for the first time, R&W's detailed figures show the Conservatives' strongest 'region' is now Wales (where they have 24% support) and their second strongest, London (23%). R&W's Con to Lab Swings by region are: East 28% South East 26% West Midlands 26% East Midlands 24% South West 21% North East 15% Yorks & H 13% North West 12% Wales 4% London 3% Scotland shows a 20% SNP to Lab swing (on a sample of 420), and note that the Welsh sample is only 265. Most regions are 500-1200. |
|
|
|
Post by James E on Jun 17, 2024 17:24:30 GMT
johntel "... So the headline could equally have been "BBC Question Time: analysis of guests over nine years suggests an overuse of LABOUR and GREEN voices".
It would have been much more revealing to list the TOTAL number of appearances by people of each political allegiance, including both politicians and non-politicians." Isn't it the BBC's normal practice to invite One Conservative, One Labour, One other party (LD/SNP/PC Green) and one 'other'. If that's so then the overall Con and Lab numbers will be the same. So it's really the choice of the '4th Panellist' which is an issue to address here (as well as perhaps the 'other party' one). [EDIT 20:50 - there is more complete data on Question Time panellists here: theconversation.com/bbc-question-time-analysis-of-guests-over-nine-years-suggests-an-overuse-of-rightwing-voices-232315?utm_source=twitter&utm_medium=bylinetwitterbutton
|
|
|
|
Post by James E on Jun 17, 2024 14:17:21 GMT
JLP Poll *Reform UK at record high, Tories at record low; Labour lead at 17 points* Change on last week in brackets LAB: 40% (-1) CON: 23% (-1) REF: 18% (+3) LDEM: 9% (-2) GRN: 5% (-) Figures with out reallocating don't knows Unadjusted this poll is: Lab 43% Con 21% Reform 16% LD 10% Green 5% ...and compared to the average of JLP's 2 pre-election polls (unadjusted), those figures are: Lab 43% (-0.5) Con 21% (-2.5) |
|
|
|
Post by James E on Jun 17, 2024 14:08:33 GMT
c-a-r-f-r-e-w That graph shows Tice and Farage's combined constituency visits as 78% Con held, 22% Lab held. Given that they have made few public appearances around the country , it may mean that they have visited as few as 2 Lab-held constituencies and 7 Con-held, for example. Or at least something in those proportions. I would not describe that as "prioritising Labour seats for campaigning".
|
|
]
