|
|
Post by James E on Feb 18, 2024 13:34:34 GMT
Re "Turnout" at the by elections in Kingswood (37.1%) and Wellingborough (38.0%).
There have been previous 50 by-elections since 2010, so here's a bit of analysis of the turnouts of those.
Firstly, by-elections where the LibDems are in contention consistently show a significantly higher turnout than the average*. And the two uncontested by elections of Southend West and Batley & Spen (Oct 16) produce low turnouts. But leaving these contests aside, we have had 41 by-elections between Con/Lab (plus SNP or UKIP).
The overall average of turnout of these has been 38.3%. And this has varied very little over the past 4 parliaments, going back to 2010. Breaking this down into the 4 Parliaments, the averages are:
2010-15 38.0% average turnout
2015-17 40.3% average turnout
2017-19 39.3% average turnout
2019-24 37.7% average turnout
Admittedly, there were few by-elections in the 2015-19 period, but the comparison between the current parliament and 2010-15 is based on 16 by-elections in each. So those "low" turnouts of 37.1% and 38.0% are in fact right in line with the long-term average of 38.3%. And while this is lower than for many contests before 2010, and especially before 1997, it has been a consistent norm now for the past 14 years.
* no by-election in the past 14 years in which the Lib Dems have been in contention has had a turnout lower than 44%, and the overall average is just over 50%, compared to 38% for those where they finished 3rd or lower.
|
|
|
|
Post by James E on Feb 16, 2024 18:33:25 GMT
Kingswood in full Lab 11176 (44.9%) Con 8675 (34.9%) Reform 2578 (10.4%) Green 1450 (5.8%) LD 861 (3.5%) .....etc... Reform's vote is in line with, or even slightly above. the polls, whereas LibDems are way down. Tactical voting?Yes, definitely. It has been a feature of nearly all English by elections in the past few years that whichever of Lab or LibDems is not challenging sees their vote fall. And this has been happening to a greater extent than in previous parliaments. In fact, the LD vote was a bit better in yesterday's by election than it has been in some that took place last year - for example, the small LD vote was reduced to under 2% in Uxbridge. Yesterday's by-elections saw the LDs' vote share fall by 3.5% and 3.2% in two constituencies where they have never been in contention. I would imagine that they might have lost a point or so anyway, in line with general polling, so that's probably about 2.5% of their vote lost 'tactically'. This would then have added about 1 point to the swings. When the opposite happens, the LDs themselves tend to gain by far more votes from Labour (but in far fewer constituencies). Looking at the Somerton and Froom by election last year, the Labour vote fell by 10.3%. And in Tiverton & Honiton in 2022 it fell by 15% . Given that Labour are well up on their 2019 share, the 'tactical' vote accruing to the LDs may be even higher than those figures. en.wikipedia.org/wiki/2023_Somerton_and_Frome_by-electionen.wikipedia.org/wiki/2022_Tiverton_and_Honiton_by-election |
|
|
|
Post by James E on Feb 16, 2024 14:22:54 GMT
Kingswood in full Lab 11176 (44.9%) Con 8675 (34.9%) Reform 2578 (10.4%) Green 1450 (5.8%) LD 861 (3.5%) UKIP 129 (0.5%) Wellingborough in full Kitchen (Lab) 13,844 (45.8%) Harrison (Con) 7,408 (24.5%) Habib (Ref) 3,919 (13%) Gunn (LDm) 1,422 (4.7%) Turner-Hawes (Ind) 1,115 (3.7%) etc... My predictions were not so good this time, as I predicted swings of 22% and 23% as opposed to 16% and 28%. The Conservatives' losses in Wellingborough are quite staggering, holding just 39% of their GE2019 share; to be fair, in Kingswood, they held a far more credible 62%. Looking at the overall dynamics across these two contests, the Tories lost nearly half of their 2019 vote share, with Labour gaining to the tune of 53% of their losses. If these proportions are applied to the overall 2019 vote shares, we get Con 23%, Lab 45%. Very close to what several pollsters are currently showing. As for the 37-38% turnout: the average in by elections over the past 6 years has been just over 40%, so this should not surprising. We have seen lower turnouts in recent by elections at: Tamworth (36%), West Lancs (31%), Stretford and Urmiston (26%), Birmingham Erdington (27%), Old Bexley & Sidcup (33%), and in the 2017-19 Parliament at Newport W (37%) and Lewisham E (33%). Higher by-election turnouts tend to occur where there is a perception that it is a genuinely close contest and/or the LibDems are in contention to win. |
|
|
|
Post by James E on Feb 16, 2024 12:16:10 GMT
The Reform UK vote share in the two by-elections doesn't look to me to be particularly impressive for a would-be right wing insurgency party. If disaffected Tory voters wouldn't vote for it in a by-election it seems unlikely that they are going to do so in any greater numbers in the UK GE ( even if Farage returns to politics). This won't, of course, stop the far right of the Tory Party arguing that the party must move even further right. In both of these by-elections, the Reform UK vote share is about two-thirds of what UKIP achieved in GE2015, when they took 12.6% of the overall GB vote. 13.0% in Wellingborough compares to UKIP's 19.6% in 2015, and 10.4% in Kingswood compared to UKIP's 14.8% in 2015. Part of the reason is that RefUK are drawing votes almost exclusively from the Conservatives: YouGov show them typically taking around 22% of the decided Con 2019 voters, but just 1% of Lab2019 and 1.5% of LD2019. In contrast, YouGov's figures just before the 2015 election showed UKIP drawing support much more widely: around 15% of Con2010, 10% of LD2010, and 5% of Lab2010. |
|
|
|
Post by James E on Feb 14, 2024 18:54:00 GMT
hireton (or anyone else who may know) re "fourth, and largest, poll in a row showing the SNP well in front" Any idea what Easton is referring to? The 5 previous Scottish Westminster polls so far this year show 2 Lab leads, 2 SNP leads and a tie. Adding in YouGov's running average, it looks neck-and-neck overall with SNP and Lab on 34-35% each. en.wikipedia.org/wiki/Opinion_polling_for_the_next_United_Kingdom_general_election#ScotlandThe previous Westminster MRP we have had was from YouGov in Jan with the SNP winning 25 seats; this might just have made them still the largest party for seats in Scotland, but certainly not 'well in front'.
|
|
|
|
Post by James E on Feb 14, 2024 14:22:17 GMT
This is now a month old, but I don't think anyone has picked up on a constituency poll done by Survation in Clacton. .... Result was: Con 38% (-34) Lab 30% (+16) Ref 18% (new) LD 6% Others 9% So a swing of 25% to Labour (or 26% to RefUK), and the Conservatives retaining just a little over half of the 72% vote share they had in 2019. It is yet more evidence that the swings at the next election will be highest where the Conservatives have most votes to lose. . The caveat is that Clacton is a distinctly odd place in terms of its social make up and political history and I would be wary of drawing any wider lessons from it. Jaywick - most recently in the news for a woman being mauled to death by two XL Bullies - although improved from its nadir, still has staggering levels of deprivation and dismal life expectancy. A genuine one off as a constituency. I'm working the other way round to what you are suggesting here: these figures are very much what I would have expected for Clacton, given a large amount of other relevant polling data. So while I acknowledge all that you say about the constituency, this poll does not require some special local explanation. Clacton's GE2019 result of Con 72%, Lab 15% was very similar to the figures for Leave voters as a whole. These were Con 74%, Lab 14%, per YouGov. But in their first 6 polls of 2024, and with an aggregate sample of around 5,000, YG now show Leave voters as a whole dividing as: Con 37% (-37) Lab 27% (+13) Survation's Clacton poll (above) is remarkable similar to this, and in fact shows an identical swing of 25%. As a slight caveat to this, Survation's GB polls show the Tories faring somewhat better than YouGov do, so based on their overall figures we might have expected the Conservative vote to come in just over 40% in Clacton. However, there is another trend which may be relevant here. I have noted before that YouGov show the Tories losing a higher-than-average proportion of their 2019 vote in the South of England, and retaining more votes in London or Scotland. So while YouGov have the Tories currently retaining 49% of their 2019 vote overall, they show them retaining just just 45% of their 2019 vote share in the South. I have run the same check over Deltapoll and Opinium's figures for their first 4 polls each of 2024, and these show the same pattern to YouGov. Opinium have the Conservatives retaining 59% of their 2019 vote in their headline figures but only 55% in the South of England. Deltapoll have the Conservatives retaining 61% of their 2019 vote overall , but only 57% in the South. So it should not be a surprise that Survation too have the Tories faring a few points worse than expected in a Southern constituency poll. Now a straightforward 'proportional' swing produces much higher swings in the safest-looking Conservative seats, as would a 'semi-proportionate' swing, which is what we had in 1997. But these figures show a 'larger-than-proportional' fall in the Tories vote share in their safest areas, so the effect is likely to be quite extreme. Conversely, I would expect to see some very low swings in safe Labour seats, particularly in places where the Greens have some support. With apologies for repetition, this is what I posted on the same subject on 23 January: "I have posted numerous times on 'proportional swing' and in particular how the Tories are losing more support in their stronger demographics. But it's actually worse than that: per YouGov, they are losing an above-average proportion of their voters in their strong regions, and retaining more of their small level of support wherever they were weak. These are the average Vote-retention proportions per the average of the past 6 published YouGov polls (7th Dec onwards):
1 Remain Voters 12/19% = 63% retention
2 London 18/32% = 56% retention
3 Scotland 14/25% = 56% retention
4 Wales 19/36% = 54% retention
5 North of England 20.5/39% = 53% retention
6 Leave Voters 38/74% = 51% retention
7 C2DE Social class 24/48% = 50% retention
8 ABC1 Social Class 21/43% = 49% retention
9 Midlands 27/55% = 47% retention
10 South of England 26.5/55% = 45% retention "
|
|
|
|
Post by James E on Feb 14, 2024 12:28:39 GMT
Savanta showing a fairly big drop in Labour support LAB 41% (-5) 🔵 CON 29% (+2) 🔶LIB 11% (+1) 🔷 REF 8% (-1) 🟢 GRN 3% (=) 🟡 SNP 3% (+1) (February 9-11) That's a large drop in Labour support. Imagine if the Tories actually turned it around and we had to put up with this lot for 5 more years? Is Savanta one of the polling companies that already takes into account those going back to the Tories? I think that Savanta do not re-weight in the manner of Opinium or MiC, but their methodology has consistently produced lower Labour leads than the overall average, typically by around 4 points. It's been a while since I looked at their detailed figures, but when I last did so Savanta's raw figures showed a far higher lead for Labour, as they found disproportionatly too many respondents who 'recalled' voting Labour in 2019, and too few Tories. As ever, it's wise to look at the fieldwork dates of any poll. For this one, we have two polls from R&W and Deltapoll with fieldwork partly or wholly after the range of 9-11 Feb. Both of these show increased Labour leads. Nevertheless, Labour's average lead over 10 polls has slipped a bit. It crept just over 20 points recently and is now 18.2%. |
|
|
|
Post by James E on Feb 13, 2024 11:14:27 GMT
More in Common bucking the trend, difficult to know if it's just noise, a backlash against the Green pledge U Turn or Rochdale, or of course a combination of all three Labour 40% (-3) 🌳Conservatives 29% (-) 🔶Liberal Democrats 11% (+1) 🟣Reform UK 10% (+2) 💚Greens 6% (-) N=2050 7/2-11/2 This looks to me like just normal variation ('noise'). Taking the 6 most recent polls, we have Labour leads of: Deltapoll 8-12 Feb 18% (+2) R&W 11 Feb 25% (+4) MiC 7-11 Feb 11% (-3) WeThink 8-9 Feb 16% (-6) Opinium 6-9 Feb 18 (+3) Techne 7-8 Feb 20 (-2) so 3 up, 3 down, and an overall average of a 1/3rd of a point lower. |
|
|
|
Post by James E on Feb 13, 2024 10:58:27 GMT
This is now a month old, but I don't think anyone has picked up on a constituency poll done by Survation in Clacton. This comes with a few riders: the sample is only 506, the constituency boundaries are a bit changed from 2019, and it is not Survation's standard VI question : they do not normally prompt by party names at all, which produces lower Refuk and Green VI than most other pollsters. Here, they asked respondents to choose between 5 options: the named Con and Ref candidates, generic Lab and LD candidates, or 'any other candidate or party'. It was also commissioned by Arron Banks.
Result was:
Con 38% (-34)
Lab 30% (+16)
Ref 18% (new)
LD 6%
Others 9%
So a swing of 25% to Labour (or 26% to RefUK), and the Conservatives retaining just a little over half of the 72% vote share they had in 2019. It is yet more evidence that the swings at the next election will be highest where the Conservatives have most votes to lose.
It is also a good guide as to how well Reform UK might perform in the next GE. Something similar to UKIP in 2015 might be expected, with them performing 7-8 points above their GB share in their target seats.
|
|
|
|
Post by James E on Feb 12, 2024 12:44:01 GMT
Survation poll of rural constituencies reported in The Independent but no more details yet. It is for the Country Land and Business Association (what I think used to be called the Country Landowners Association) Labour will beat the Tories in the most 100 rural constituencies in England, according to a shock new poll.
The poll found Conservative support had fallen by 25 points since the 2019 election, with just 34 per cent of voters in the 100 most rural constituencies in England saying they would vote for the party.
The Conservatives currently hold 96 of the 100 most rural seats, but now face losing more than half to Labour and the Lib Dems, including those of senior Tories Jacob Rees-Mogg, Jeremy Hunt, Thérèse Coffey, Andrea Leadsom, Mel Stride, Mark Harper and Liam Fox.
Labour support has risen over the same period, going from 20 per cent in 2019 to 37 per cent at the start of this year – giving the party a narrow lead in what has traditionally been considered Conservative territory. This looks to me consistent with Survation's GB polling, the most recent example of which shows Con 27%, Lab 44%. So the Tories are down to around 0.6 of their 2019 vote, with Labour gaining around 65% of the share lost by the Tories (not all of that directly from them). This latest poll of rural voters appears to be a little worse than that for the Conservatives, with them down to 0.58 of their 2019 vote, and Labour gaining by 68% of their losses. But that's probably just Margin of Error. They did a similar poll last year in which the figures fell just on the other side of a proportional loss. Given that we also have consistent data showing the Tories faring worse than average with ABC1s and voters in the South of England, these figures should not now be too much of a surprise. |
|
|
|
Post by James E on Feb 12, 2024 12:23:56 GMT
I haven't seen the yougov poll this relates to but apparently the Tories are now registering around 10% support in all voters under the age of 49. youtu.be/9e53CeCpjbA?si=uZDhXiZGZRWDronwIf you want to skip straight to the graph it's about 5 minutes in. I've checked the last 5 YouGov polls, and they show the Tories averaging 7.6% with the 18-24s and 12.6% with the 25-49s. That's between 11-12% as an overall average for voters under 50. |
|
|
|
Post by James E on Feb 11, 2024 14:25:17 GMT
A couple of predictions for this Thursday's by-elections. WellingboroughLab 44% (+17.5) Con 34% (-27.5) Ref 10% ...etc No quibble with your Kingswood prediction but the media are busting a gut to talk up Reform UK Limited in Wellingborough. I am not entirely convinced, but there is a possibility that RefUK might do a bit better that 10% at the expense of the Tories. Being in the Anglian region the local BBC has done a fair bit of vox-pop stuff in Wellingborough. Not scientific, but for what its worth the choice of Tory candidate has gone down really badly and I haven't seen any evidence of a RefUK bandwagon for all the media chatter. To date, Reform UK have performed below their GB polling average in every by-election they have contested . West Lancashire 4.4% (v 5.9% polling average) Somerton & Froom 3.4% (v 6.4% av) Rutherglen 1.3% (v 6.5% av) Selby & Ainsty 3.7% (v 6.4% av) Mid Bedfordshire 3.7% (v 6.1% av) Tamworth 5.4% (v 6.1% av) (the same is true, and by an ever greater margin, in the earlier by elections of Stretford, Chester, Tiverton and Wakefield) They are currently on an average of 10.8%, so 10% would be a good performance for them both in absolute and relative terms. |
|
|
|
Post by James E on Feb 11, 2024 10:56:25 GMT
A couple of predictions for this Thursday's by-elections.
These are based closely on the dynamics of other Con v Lab by elections in the past year or so, where there has been remarkable consistency* : the Conservative vote falls proportionately in line with current polling. Labour benefit by more than polling suggests, with a strong squeeze on the LDs** and perhaps the Greens, too.
On this basis, I am going for the Tories retaining just 0.55 of their 2019 vote share in both Wellingborough and Kingswood, which is consistent with their current average polling figure of about 25%. If they are reduced below half of their 2019 share, then this may be more in line with recent YouGovs, in which they are averaging just 21-22% typically.
Wellingborough
Lab 44% (+17.5)
Con 34% (-27.5)
Ref 10%
LD 5% (-3)
Grn 4%
Oths 3%
Swing 22.5%
Kingswood
Lab 53% (+19)
Con 31% (-25)
Ref 6%
Grn 5%
LD 4%
Oth 1%
Swing 22%
So, slightly smaller swings than the 23.9% in Tamworth and the 23.7% in Selby, but these are less safe seats. It's a little early for a detailed prediction for Rochdale, where I suspect that labour will stand still at around 50-54% of the vote. Even if it were not for the hit to Labour's support with Muslims (down to only 60%?) the Rochdale contest would have been expected to produce a lower swing, in line with other Labour held contests such as Chester or West Lancashire in late 2022.
* the obvious exception being Uxbridge.
**the converse applies in recent Con v LD by elections, where the Con vote falls even further than polling suggests, and the Lab vote is heaving squeezed by the LDs.
|
|
|
|
Post by James E on Feb 10, 2024 15:45:33 GMT
...
I gather there is some turbulence in the Green Party of England & Wales : the Green Party of Ireland (in its NI region) made a brief appearance in Stormont in 2007, but then lost its seats and shows little sign of resurgence. The Scottish Green Party, on the other hand, continues to show slow, but steady, improvement in polling and results - on the basis of the Ipsos poll, BBS suggests that they might win Hillhead in the next UK GE.Ipsos polled a total of 116 people in the Glasgow region. 75 described themselves as likely to vote, and of these 8 said they would vote for the SGP. See page 5 (of 55) in their tables. www.ipsos.com/sites/default/files/ct/news/documents/2024-02/ipsos-scottish-political-monitor-february-2024-tables.pdfAlso, the Westminster Constituency covering Hillhead is named Glasgow North. |
|
|
|
Post by James E on Feb 9, 2024 14:41:44 GMT
There is a wide divergence in Scottish Westminster polling at the moment, with Ipsos showing the SNP 10% and 7% ahead in their two most recent polls, while the most recent YouGov (in the wake of the Rutherglen by-election in October) put Labour ahead 38/32. However, Ipsos's last 4 polls in Scotland have all shown much better figures for the SNP than any other pollster. As I have noted above, YouGov's Scottish cross-breaks for Jan/Feb 2024 are not far from that showing Labour 4 points ahead, suggesting that their 6 point lead was not a total one-off. The difference between Ipsos and YouGov is stark when it comes to predicting seats, as it could shift about 20 seats in either direction. However, there are other pollsters whose full Scottish figures are in the middle ground. An average of the 4 most recent polls we have from R&W, Ipsos, Survation and Norstat (formerly Panelbase), plus those YG cross-break figures gives us: SNP 35% Lab 34% Con 15%. en.wikipedia.org/wiki/Opinion_polling_for_the_next_United_Kingdom_general_election#Scotland |
|
|
|
Post by James E on Feb 9, 2024 14:14:41 GMT
As we now have 6 YouGov polls since the start of 2024, it seems like a good time to give some updated averaged cross-breaks. These need to be seen in the context of YG's headline figures showing Labour 24-25 points ahead, so about 6 points higher than the rest of the polling industry, with a particularly strong performance by RefUK at the expense of the Tories. Pollsters such as Survation and Savanta, who do not include Ref in their prompt, have Ref about 4 points lower and the Tories 4 points higher.
London
Lab 54% (+6)
Con 18% (-14)
Swing Con to Lab 10%
South England
Lab 40% (+17)
Con 25% (-30)
LD 13% (-4)
Swing Con to Lab 23%
Midlands
Lab 45% (+12)
Con 27% (-28)
Swing Con to Lab 20%
North England
Lab 55% (+12)
Con 19% (-20)
Swing Con to Lab 16%
Scotland
Lab 36% (+17)
Con 14% (-11)
SNP 32% (-13)
Swing SNP to Lab 15%
Wales
Lab 47% (+6)
Con 20% (-16)
Swing Con to Lab 11%
Remain
Lab 59% (+10)
Con 12% (-7)
Swing Con to Lab 9%
Leave
Lab 26.5% (+12)
Con 37% (-37)
Swing Con to Lab 25%
ABC1 Voters
Lab 49% (+16)
Con 20% (-22)
Swing Con to Lab 19%
C2DE Voters
Lab 40% (+7)
Con 24% (-24)
Swing Con to Lab 16%
The pattern of the Conservatives' losses is consistent and proportionate: their share is roughly halved across all demographic and geographical groups, with the rider that they are doing a little less badly in their weak demographics, such as London, Scotland, Wales and 2016 Remain voters. Wales aside, this probably reflects the lower appeal of Reform UK to 2019 Tory voters within those groups.
|
|
|
|
Post by James E on Feb 9, 2024 9:33:55 GMT
Yougov - getting there Dave 😀 Labour lead at 25 points in latest YouGov poll for The Times CON 21 (-2) LAB 46 (+2) LDEM 9 (=) RefUK 12 (=) GRN 7 (+1) Fieldwork 7 - 8 February These figures are very close to the average per YouGov from their first 6 GB polls of 2024: Con 21.3% Lab 45.8% LD 9% Ref 11.3% Grn 7% |
|
|
|
Post by James E on Feb 5, 2024 14:34:45 GMT
And given the tendency of 'Don't Knows' to revert to the party they previously voted for Labour would most likely be on course to take 65% plus of the Muslim vote in an actual General Election.... [etc] Ir was me pointed that out not Steve!!! I think Survation tables suggest their 83% Lab vote was based on 75% turnout? Tables are here if you want to drill deeper (which is always appreciated with your posts): www.survation.com/https-www-survation-com-archive-2024-2/If that's true then like for like the 60% does seem reasonable and we can assume a lot of the don't knows are non voters, but I'm doubtful about the theory of reverting back to previous party this time as we saw with LD don't knows in 2015 and may see with Tory don't knows this year. Also I'd be querying eventual lower turnout among those who offered a voting intention. Looking at some of the clues in the background questions the favourable on Lab and Starmer doesn't really get any further than that headline figure (which as we agree is not comparing like with like)- 48% favourable Lab and 30% favourable Starmer and options of don't know of neither favourable/unfavourable a bit higher than the don't knows in the headline figure. Given the likely large Labour majority I can't see this impacting the overall General Election result much other than the same sort of random things mentioned in the BBC report and the ongoing loss of another "core" vote. On the reversion of the 'Don't Knows'. You may know that I am a firm believer in some measure of reversion of Don't Knows, but this needs to treated proportionately. Clearly not all DKs revert to the party they voted for last time, and they are more likely to not not vote than those with a firm Voting Intention. The old ICM practice of adding back a net half of each party's DKs seems to me a useful predictor. At present, that would cut about a net 4 points from Labour's lead, working on approximate figures of 25% Con2019 and 10% Lab2019 answering 'Don't Know'. Of course, this cannot apply to polls from Opinium or MiC, which already include a slightly larger adjustment. Incidentally, the same kind of calculation would explain why the LibDems got little net benefit in 2015. If they had (say) 25% DKs from their 24% 2010 vote-share, and the Conservatives (say) 16% from their 37% 2010 vote, then the actual numbers of each would be almost the same for each. (and apologies for attributing your own comments to Steve!) EDIT: Conveniently, courtesy of LL, we have actual figures for the relative levels of 'Don't Knows' going into 2015, which can be used to explain why the LDs did not get much of a boost from their higher DKs. These show as an average of 20% 2010 LDs, 12% 2010Con and 10% 2010Lab. If you calculate a 50% reversion for each, this would add 2.4 to LDs, 2.2 to Con and 1.4 to Lab, after which all the figures would need to be re-based to 100. So this would be consistent with a very small improvement for both LDs which is what happened in the last 10 months of the 2010-15 Parliament. |
|
|
|
Post by James E on Feb 5, 2024 11:01:35 GMT
As Steve has pointed out, this is not a like-with-like comparison, as the 'current VI' figures only add up to 72 with 28% either undecided or some other category. If the figures of those actually giving a VI are re-based to 100, the Lab VI is 60% (with 14% Grn, 8% Con, 8% LD) . But I would also doubt their figures for GE 2019, which appear to be Lab 86%, Con 10%, SNP 2%, LD 1%, Grn 1%. And given the tendency of 'Don't Knows' to revert to the party they previously voted for Labour would most likely be on course to take 65% plus of the Muslim vote in an actual General Election. There is a clear precedent for what might happen now, as Labour's Muslim vote fell substantially in the 2005 election as a result of the war in Iraq. But nearly all of the seats affected were very safe Labour. The only Lab/Con marginal where this probably made a difference was Peterborough, which the Tories won on an unusually high swing ; this was reversed in 2010 and 2015 when Labour then performed above expectations. news.bbc.co.uk/1/hi/uk_politics/vote_2005/frontpage/4520527.stm |
|
|
|
Post by James E on Jan 29, 2024 19:14:44 GMT
Interesting, suggests 'false recall' is responsible for a significant amount of the variation in the Tory vote share. Yougov, which shows some of the biggest Labour leads, apparently notes how people voted immediately after an election, so less likely to suffer from false recall I assume that the argument is that because pollsters weight their numbers to reflect the party strength at the prior election, then false recall will lead to that weighting being done incorrectly and so lead to an incorrect calculation of current VI. Have I understood that right, or is it something else? I have noted the same pattern in the tables of several pollsters who use 'recalled past vote' for the past couple of years. There are too many who say they voted Labour in 2019, and too few Tories and LDs. A good recent example is the Deltapoll from 29 Dec, in which they unusually provided figures by past vote. Their 'unweighted sample' of 1,163 who say they voted in 2019 is 504 Con (43.3%) 422 Lab (36.3%), 107LD (9.2%). See page 2 of link. deltapoll.co.uk/wp-content/uploads/2024/01/Deltapoll-Mirror-231229.pdfIn the same poll, the 'raw' figures work out as Con 23.5%, Lab 46%, whereas the weighted headline figures were Con 28%, Lab 42%. However, the past vote data can be hard to find, and it is clear that some (such as R&W) have made changes to address this. Techne say that they close their samples to certain demographic groups once the required number of that group has been reached. presumably, they just keep going until they find enough who remember voting Conservative in 2019 |
|
|
|
Post by James E on Jan 27, 2024 20:05:51 GMT
Westminster Voting Intention:
LAB: 42% (+2)
CON: 27% (=)
LDM: 10% (+1)
RFM: 10% (=)
GRN: 6% (=)
SNP: 3% (=)
Via
@opiniumresearch
, 24-26 Jan.
Changes w/ 10-12 Jan.
15 points is just above the norm for a Labour lead with Opinium.All their polls in the past year have been in the range of 10-18 points.
|
|
|
|
Post by James E on Jan 25, 2024 19:33:39 GMT
Those age cross-breaks quoted above are intersting too, especially in comparison to YG's figures for 2019: 18-24 Con 9% (-12) Lab 60% (+4) Swing 8% 25-49 Con 14% (-19) Lab 57% (+14) Swing 16.5% 50-64 Con 20% (-31) Lab 49% (+22) Swing 26% 65+ Con 35% (-29) Lab 25% (+9) Swing 19%
With that in mind, and being Burns night [1] Labour partisans may enjoy the 3rd verse of "Address to the Haggis" -
"His knife see rustic Labour dight,
An cut you up wi ready slight,
Trenching your gushing entrails bright,
Like onie ditch;
And then, O what a glorious sight,
Warm-reekin, rich!"
[1] We are having Haggis, Bashed Neeps and Champit Tatties tonight - Tradition! (Looks up to see the Fiddlers on the Roof, but it's just Barrowman, Mone and assorted other Tory criminals).You are right in this case, as the 50-64 age band here produced an unusually high Labour lead. The average of 8 YouGovs is Lab 42, Con 25, but that's still a swing of 20% on 2019, the highest . And the average for over 65s is Con 40, Lab 25, so a 16.5% swing. In the 2 younger groups, the averages are Lab 58, Con 9 for the 18-24s, and Lab 57, Con 13 for the 25-49s. So there, the figures in the latest poll are fairly normal. It looks like the difference in the latest poll is the increased rate of their older voters deserting the Conservatives for Reform UK. [We had (vegan) haggis tonight as well- quite nice!] |
|
|
|
Post by James E on Jan 25, 2024 11:30:42 GMT
Yougov, no change from last week, but that was a very good week for Labour Labour lead at 27 points in latest YouGov poll for The Times CON 20 (=) LAB 47 (=) LIB DEM 8 (=) REF UK 13 (+1) GREEN 6 (-1) Fieldwork 23 - 24 January This is a rare case where 'no change' comes as a slight surprise. This poll and others now confirm that the Tories' support has taken a further hit in the past week or two. Those age cross-breaks quoted above are intersting too, especially in comparison to YG's figures for 2019: 18-24 Con 9% (-12) Lab 60% (+4) Swing 8% 25-49 Con 14% (-19) Lab 57% (+14) Swing 16.5% 50-64 Con 20% (-31) Lab 49% (+22) Swing 26% 65+ Con 35% (-29) Lab 25% (+9) Swing 19% |
|
|
|
Post by James E on Jan 24, 2024 23:52:56 GMT
It would be interesting to know the 2019 vote percentages in those 150 seats. Presumably if they are Tory held it must have been around 45%, in which case 18% suggests they might have shed 60% of their 2019 vote! I think that would be unprecedented for the Conservatives unless anyone knows otherwise? It seems from the quotes above that the seats were top 150 Labour targets. As you say, not all of these were held by the Tories in 2019. Using the old boundaries, those 'top 150' would include 2 PC held and 26 SNP held. As a whole, these seats have relatively high numbers of Northern (48), Welsh (13) and Scottish seats (26), and lower numbers in the South of England (30). London has 10 on the list and the Midlands 23. While the Con-held ones range from 0-26% majorities, averaging around 13%, the Tories may be as low as 15% in those where they were not in contention. Because of this, I would estimate that the Conservative vote might be only 40% in the 150 labour target seats as a whole, perhaps around 7% ahead of Labour. The swing would then be 21.5% compared to 18% for GB as a whole. If this is right, it's not as spectacular as the comparison to the GB headline figures might suggest. Also, the Tories' vote retention would be 0.45 of 2019 (18/40%), right in line with the most recent YouGov poll where they were on 20%. |
|
|
|
Post by James E on Jan 23, 2024 16:17:29 GMT
Thanks to neilj for the analysis of Con2019 voters per YouGov above. But the pattern of their losses continues to go unnoticed in the media, and here there is something really remarkable going on. I have posted numerous times on 'proportional swing' and in particular how the Tories are losing more support in their stronger demographics. But it's actually worse than that: per YouGov, they are losing an above-average proportion of their voters in their stong regions, and retaining more of their small level of support wherever they were weak. These are the average Vote-retention proportions per the average of the past 6 published YouGov polls (7th Dec onwards): 1 Remain Voters 12/19% = 63% retention 2 London 18/32% = 56% retention 3 Scotland 14/25% = 56% retention 4 Wales 19/36% = 54% retention 5 North of England 20.5/39% = 53% retention 6 Leave Voters 38/74% = 51% retention 7 C2DE Social class 24/48% = 50% retention 8 ABC1 Social Class 21/43% = 49% retention 9 Midlands 27/55% = 47% retention 10 South of England 26.5/55% = 45% retention The overall average Con VI is around 22-23% now with YouGov compared to 44.7% at GE2019, so the norm with a proportional swing would be 50% retention across the board. But of course most models of how seats might change hands are based on the the simple 'Uniform national Swing' under which each party gains or loses by the same amount in each seat. This produces a pattern where the proportion of votes lost (or gained) can vary hugely dependent on the level of support at the previous election. So for a comparison to the table of losses above, this is how Uniform National Swing would predict the Conservatives' vote share, per the same groups: 1 Leave Voters 52/74% = 70% retention 2 South of England 33/55% = 60% retention 3 Midlands 33/55% = 60% retention 4 C2DE Social Class 26/48% = 54% retention 5 ABC1 Social Class 21/43% = 49% retention 6 North of England 17/39% = 44% retention 7 Wales 14/36% = 39% retention 8 London 10/32% = 31% retention 9 Scotland 3/25% = 12% retention 10 Remain Voters 0*/19% = 0% retention *UNS currently puts the Conservatives on a -3% vote-share with this group. The two tables are not quite in reverse order, but getting close to that. For the ABC1 and C2DE groups, YG recorded the Conservative vote in 2019 as close to their 44.7% overall share for each, so the fact that these come out in the middle is entirely what we should expect. This really is unprecedented: even in the 1997 General Election, the detailed results ended up close to the mid-point between a uniform and proportional swing, so that while the Tories lost a higher number of votes in their safer areas, they still held a higher proportion of their 1992 vote in the same areas and groups. The difference between the two tables above explains the huge divergence between UNS and MRP/ proportional models. And any narrative that claims that 'Labour need a record breaking 12.7% swing for an overall majority' is based on the same assumptions behind the second table above, rather than the detailed findings of the first. |
|
|
|
Post by James E on Jan 23, 2024 11:32:33 GMT
Per Electoral Calculus - on that poll 72 Conservatives, 41 Lib Dems, 487 Labour. Labour overall majority of 324. They have plenty to be sleepless about already. We're getting close to Canadian Conservative territory here. They were polling 15% when they were reduced to 2 seats (-154). Their leader Kim Campbell lost her seat in the rout, which does not augur well for Sunak if the Tory support continues to decline. I can't really see a Canada 1993 result happening, as the Conservatives are still likely to make some recovery from their 2019 'Don't Knows'. Probably not on the scale of the 'adjusted' Opinium or MIC polls (or the YouGov MRP) which show around a 14% lead, but more likely to around 17% - roughly halfway between the 'adjusted' polls (@14% av) and the 'unadjusted' ones ( @21% av). This might be something like Lab 43%, Con 26%. If that happened, the Tories could be reduced to around 120-140 seats. However, if their vote were really to be reduced much below that 26%, their seat-losses would accelerate well above my '15 seats per 1% swing' rule of thumb. This is due to a combination of their rather even vote-spread, proportionate losses, and the effect of tactical voting. Where they held 60% in 2019, current detailed polling suggests that they would be reduced to just 0.55 of this - so 33%. Anti-Tory voting would then ensure that the better-placed of Lab or LD takes the seat. A true 'meltdown' could occur if they were reduced to just 20% of the vote - or 0.45 of their 2019 share. At that point, they would be down to just 35% in their 'safest' seats, and with tactical voting against them, they would be reduced to single figures in Parliament. But I would see a figure of around 125 seats as a reasonable worst-case scenario for them. |
|
|
|
Post by James E on Jan 22, 2024 20:15:38 GMT
Delta and R&W's polls complete a sequence of 5 with fieldwork in the last week, of which 4 have shown their largest Labour leads for some time; the odd one out is Techne, whose 18% was their equal-lowest for 14 polls. Delta (22 Jan) 17 % Lab lead, their largest since early Nov 2023 (8 polls) R&W (21 Jan) 23% Lab lead, their largest since March 2023 We Think (19 Jan) 25% Lab lead, their largest since July 2023 Techne (18 Jan) 18% Lab lead, their lowest since Sept 2023 YouGov (17 Jan) 27% Lab lead, their largest since Oct 2022 en.wikipedia.org/wiki/Opinion_polling_for_the_next_United_Kingdom_general_election# |
|
|
|
Post by James E on Jan 22, 2024 14:08:44 GMT
Those interested in how party support and tactical voting are likely to affect the number of seats that parties win at the next General Election, which must be a few on this board to judge by the number of footie-related and covid posts on here, may like to read Ben Ansell's latest missive on Substack: How Bad Can It Get for the Conservatives?I am not going to attempt to summarise it; just read it. Ben also has a new version of his app which includes the new boundaries, so you can see what will happen in your own constituency, depending on what assumptions (including tactical voting) you put in. Ben also makes a comment about tactical voting, which may seem obvious to us aficionados but is not easily captured in models. In those constituencies where they weigh the Labour vote rather than count it, there is no incentive for Liberal Democrat and Green supporters to vote tactically as Labour will win anyway and the only interest that the Liberal Democrats and the Greens have is saving their deposits. So Liberal Democrats and Greens voting tactically will be mainly in those Tory-Labour marginals where the outcome is uncertain on polling day and tactical voting will have a larger effect than would be expected from the assumption it is spread uniformly across the country. Thanks for that. It is an interesting article, and I agreee that Tactical Voting will be a significant factor in GE2024/5. It is clear from all recent by-elections that voters are willing to line up behind Lab in their target seats and LDs in theirs, and the squeeze on whichever of Lab and LDs is third has never been greater. However, I cannot agree with his contention that tactical voting is the sole or main reason for the YouGov MRP model to produce a far better outcome than UNS for Labour. And that is what he claims at one point: "The MRP gives Labour a 120 seat majority, with 385 seats to the Conservatives’ 169. Now if you plug the poll numbers into most election calculators you don’t get a seat difference anywhere near that large. So something is going on. What is that? Very high levels of tactical voting." The real 'something else' is proportional losses by the Tories. To take some of those by-elections we have had in the past 13 months: the recent by-elections in Tamworth and Selby saw Labour achieve huge 23-24% swings at a time when the polls were showing them around 18% ahead (or a 15% swing). However, somewhat larger Labour poll leads in Dec 2022 saw them hold Chester and Stretford and Urmiston on swings of just 13.5% and 10% respectively. But these results were entirely consistent: The Conservatives' vote in each of Chester and Stretford &U was 0.58 of its 2019 share. And then when defending Selby they held 0.57 of their 2019 and Tamworth it was 0.61. 'Tactical Voting' can explain why Labour or LDs receive a larger-than-average boost in a target consituency, but it can never explain why the Tories losses are larger where they are defending, as is clearly happening in both by-elections and detailed polling. |
|
|
|
Post by James E on Jan 19, 2024 18:11:10 GMT
Not entirely. No one should get excited over individual polls, but the trend over a a long period is a very reliable indicator of the GE outcome. Government leads or small opposition leads (averaging under 7 to 8%) indicate that the government will be re-elected. Consistent large opposition leads (in excess of 8%) indicate a 'change of government' election is coming, although they don't tell you anything about the size of majority. That's what my 'average monthly opposition leads exercise is about. That model predicts the supposedly 'surprise' results in 1970 and 1992 correctly as well as all the more obvious ones. Average Subsequent 9 months prior to election When election called Final % share Lead election Lab Con Lab Con Lab Con 1992-97 25.8 Opposition 49.6 30.3 51.2 31 43.2 30.7 2019- 14.8 TBC! 44.0 26.0 2005-10 10.0 Opposition 25.6 42.0 30.6 37.8 29.7 36.91987-92 8.8 Govt 42.6 38.3 40.2 38.8 35.2 42.82010-15 6.1 Govt 36.6 34.2 34.2 35.2 31.2 37.8 1979-83 6.0 Govt 29.0 43.5 31.4 48.4 27.6 42.41983-87 5.4 Govt 38.0 33.8 31.6 42.0 30.8 42.21997-01 5.0 Govt 46.0 31.3 49.4 30.6 40.7 31.7 2001-05 2.7 Govt 34.3 31.6 37.6 34.0 36.2 33.2 2017-19 2.4 Govt 33.3 38.0 27.6 37.6 32.1 43.62015-17 0.0 Govt 30.6 39.6 24.6 45.6 41.0 43.5Overall, in 7 out of 10 cases , the Party that has led 9 months prior to polling day has gone on to win. Rather depressingly for Labour, 9 months out they have led 6 times, but only won 3. 4 times they have led at the time of dissolution, but once went on too lose. Whenever the Tories have led six months out or at the start of the campaign, they have won/formed the gvt.
What should be of concern for Sunak and the Tories, is that in general there is less variation between their VI % both six months out and at the start of the campaign than that for Labour. The polls seem to be a more accurate indicator of level of Tory support - with the exception of '87, they rarely add more than a couple of % points to where they are 9 months prior to polling day. With the exception of '97, the best they tend to do is gain 4-6%, which would put them in the low 30's come polling day.
The most dramatic gain in VI during a campaign, is the insurgency one run by Corbyn, no one else gets close. But it is likely a total outlier, and unlikely to be repeated. 9 out of 10 times, the party that has led when the election is called, has gone on to win. There is also only one case where a party seems to have actually turned it round during the campaign, and that was '92. If Sunak is to stand a chance, he really needs to close the gap in the next couple of months, but there appears little evidence of this happening.
I genuinely think, given the information we have, its about 90% certain Labour will win with an OM, 9% chance of Labour being the largest party in a hung parliament, and 1% chance of the Tories being the largest party.
Many thanks for this, and apologies for a quibble over your findings, but this is an important point: to evaluate the effect of the campaign period, we need to look at the polls before against the final polls - not the result. So for example in 1992 the polls moved just a little during the campaign period, so it is reasonale to conclude that the Conservatives were really 6-7 points ahead at the start, and had already 'turned around' the Labour leads from earlier in the parliament, which were in any case probably several points lower than the polls were showing. |
|
|
|
Post by James E on Jan 18, 2024 10:33:32 GMT
Wonder how much of that YG is just moe of a lead around just above 20 on YG? Right at the edge but 1 in 20 or so will be, IIRC. Dont we have to really wait for the campaign proper jimjam before polling really means anything ? All the media exposure-interviews & debates etc ? .... With just a couple of exceptions, the 'campaign proper' rarely produces more than a few points movement in the polls. Other than Labour's surge in 2017, and the Lib/SDP Alliance in 1983, I can't recall any other case of a party gaining a really significant increase in VI. That's comparing the polls at the start with the polls at the end. Polling error, as seen in 1992, 2015, and to a lesser extent 2017, 1997 and 2001, is another matter. And to answer JJ's point: YouGov's recent average has been a 22 lead, so this is 5 points above that. The most striking thing about this latest poll is the Con to Ref movement. As the fieldwork was 16-17 Jan, it is the most recent poll, so the 17 points with with Savanta a few days earlier does help to show it as a likely outlier. And the average Labour lead across all polls is up by around 2 points for the first polls of 2024, so we might have expected a lead of around 24. |
|
