|
|
Post by woollyliberal on May 15, 2022 20:22:47 GMT
New Find Out Now / Best For Britain poll and MRP. The link only shows the raw percentages. I have the turnout adjusted figures as these: Lab 43% Con 34% LD 9% Green 5% It looks very high for Lab and a bit low for LD. Even with a high end result for Labour, it has them not getting a majority. The MRP results look odd for the LDs. 8 seats if there are no pacts, 7 if Reform does a deal with the Tories. 12 if the LDs do a deal with Labour. Even taking account of MRP models' odd handling of Scotland, that seems on the low side of everybody's expectations. I'm starting to the MRP has less value than often touted. I'd still trust it to show which seats are doing better or worse than UNS, but I think its' prediction of the winner leaves something to be desired. www.bestforbritain.org/may_2022_mrp_analysis |
|
iain
Lib Dem
Posts: 11,451 
|
Post by iain on May 15, 2022 20:34:50 GMT
MRP is just a projection method, it's not magic.
These guys have a terrible track record IIRC. As always, rubbish in, rubbish out.
|
|
|
|
Post by woollyliberal on May 15, 2022 20:49:15 GMT
New Find Out Now / Best For Britain poll and MRP. The link only shows the raw percentages. I have the turnout adjusted figures as these: Lab 43% Con 34% LD 9% Green 5% It looks very high for Lab and a bit low for LD. Even with a high end result for Labour, it has them not getting a majority. The MRP results look odd for the LDs. 8 seats if there are no pacts, 7 if Reform does a deal with the Tories. 12 if the LDs do a deal with Labour. Even taking account of MRP models' odd handling of Scotland, that seems on the low side of everybody's expectations. I'm starting to the MRP has less value than often touted. I'd still trust it to show which seats are doing better or worse than UNS, but I think its' prediction of the winner leaves something to be desired. www.bestforbritain.org/may_2022_mrp_analysisMaybe I'm looking in a different place to you but it seems to say the seats are Lab 336, Con 230, SNP 53, LD 8, PC 4, Green 1. You're right. One of their 3 predictions shows a Labour majority. |
|
|
|
Post by greenchristian on May 15, 2022 21:14:45 GMT
New Find Out Now / Best For Britain poll and MRP. The link only shows the raw percentages. I have the turnout adjusted figures as these: Lab 43% Con 34% LD 9% Green 5% It looks very high for Lab and a bit low for LD. Even with a high end result for Labour, it has them not getting a majority. The MRP results look odd for the LDs. 8 seats if there are no pacts, 7 if Reform does a deal with the Tories. 12 if the LDs do a deal with Labour. Even taking account of MRP models' odd handling of Scotland, that seems on the low side of everybody's expectations. I'm starting to the MRP has less value than often touted. I'd still trust it to show which seats are doing better or worse than UNS, but I think its' prediction of the winner leaves something to be desired. www.bestforbritain.org/may_2022_mrp_analysisMRP seems to be very good at predicting Labour/Conservative seats, but fairly poor at predicting seats where smaller parties are in contention. Which is exactly what you'd expect from the methodology. |
|
wysall
Forum Regular
Posts: 326
|
Post by wysall on May 15, 2022 22:09:26 GMT
New Find Out Now / Best For Britain poll and MRP. The link only shows the raw percentages. I have the turnout adjusted figures as these: Lab 43% Con 34% LD 9% Green 5% It looks very high for Lab and a bit low for LD. Even with a high end result for Labour, it has them not getting a majority. The MRP results look odd for the LDs. 8 seats if there are no pacts, 7 if Reform does a deal with the Tories. 12 if the LDs do a deal with Labour. Even taking account of MRP models' odd handling of Scotland, that seems on the low side of everybody's expectations. I'm starting to the MRP has less value than often touted. I'd still trust it to show which seats are doing better or worse than UNS, but I think its' prediction of the winner leaves something to be desired. www.bestforbritain.org/may_2022_mrp_analysisIsn't this Focaldata? It's funny that Best for Britain describes Reform standing down and all of its voters supporting the Conservatives as "the most likely scenario." |
|
|
|
Post by london(ex)tory on May 15, 2022 22:39:48 GMT
New Find Out Now / Best For Britain poll and MRP. The link only shows the raw percentages. I have the turnout adjusted figures as these: Lab 43% Con 34% LD 9% Green 5% It looks very high for Lab and a bit low for LD. Even with a high end result for Labour, it has them not getting a majority. The MRP results look odd for the LDs. 8 seats if there are no pacts, 7 if Reform does a deal with the Tories. 12 if the LDs do a deal with Labour. Even taking account of MRP models' odd handling of Scotland, that seems on the low side of everybody's expectations. I'm starting to the MRP has less value than often touted. I'd still trust it to show which seats are doing better or worse than UNS, but I think its' prediction of the winner leaves something to be desired. www.bestforbritain.org/may_2022_mrp_analysisIsn't this Focaldata? It's funny that Best for Britain describes Reform standing down and all of its voters supporting the Conservatives as "the most likely scenario." There’s no way that will happen. I would spoil my ballot paper before voting for this “Conservative” government. |
|
|
|
Post by tonyhill on May 16, 2022 5:39:54 GMT
Given that Reform only managed 119 local election candidates nationally this month you may have to.
|
|
YL
Non-Aligned
Either Labour leaning or Lib Dem leaning but not sure which
Posts: 4,915
|
Post by YL on May 16, 2022 6:24:12 GMT
MRP is just a projection method, it's not magic. These guys have a terrible track record IIRC. As always, rubbish in, rubbish out. Indeed, it will depend on the variables used and the sample size. The big YouGov MRPs before the last two elections had enormous sample sizes which allowed them to use constituency level information, which will help with predicting Lib Dem seats. Sometimes they actually overdid local effects; IIRC they sometimes had Claire Wright winning East Devon, for example. |
|
|
|
Post by andrew111 on May 16, 2022 9:00:05 GMT
MRP is just a projection method, it's not magic. These guys have a terrible track record IIRC. As always, rubbish in, rubbish out. Indeed, it will depend on the variables used and the sample size. The big YouGov MRPs before the last two elections had enormous sample sizes which allowed them to use constituency level information, which will help with predicting Lib Dem seats. Sometimes they actually overdid local effects; IIRC they sometimes had Claire Wright winning East Devon, for example. The "huge" sample size of 10k is still only 15 per constituency. They use constituency level demographics and possibly constituency level local election results but they can't account for constituency level campaigning by different parties and tactical voting easily. How they try and account for this is mysterious and Claire Wright referred to above suggests pure guesswork is their final adjustment! |
|
|
|
Post by bjornhattan on May 16, 2022 9:50:03 GMT
Indeed, it will depend on the variables used and the sample size. The big YouGov MRPs before the last two elections had enormous sample sizes which allowed them to use constituency level information, which will help with predicting Lib Dem seats. Sometimes they actually overdid local effects; IIRC they sometimes had Claire Wright winning East Devon, for example. The "huge" sample size of 10k is still only 15 per constituency. They use constituency level demographics and possibly constituency level local election results but they can't account for constituency level campaigning by different parties and tactical voting easily. How they try and account for this is mysterious and Claire Wright referred to above suggests pure guesswork is their final adjustment! What I do remember is that they had very wide confidence intervals for some of their estimates, particularly around independents. These seem to have vanished from the internet but I recall some constituencies had independent votes ranging from one or two percent up to 20% or more (with the central estimate being around 5%). |
|
The Bishop
Labour
Down With Factionalism!
Posts: 39,009
|
Post by The Bishop on May 16, 2022 10:37:11 GMT
New Find Out Now / Best For Britain poll and MRP. The link only shows the raw percentages. I have the turnout adjusted figures as these: Lab 43% Con 34% LD 9% Green 5% It looks very high for Lab and a bit low for LD. Even with a high end result for Labour, it has them not getting a majority. The MRP results look odd for the LDs. 8 seats if there are no pacts, 7 if Reform does a deal with the Tories. 12 if the LDs do a deal with Labour. Even taking account of MRP models' odd handling of Scotland, that seems on the low side of everybody's expectations. I'm starting to the MRP has less value than often touted. I'd still trust it to show which seats are doing better or worse than UNS, but I think its' prediction of the winner leaves something to be desired. www.bestforbritain.org/may_2022_mrp_analysisMaybe I'm looking in a different place to you but it seems to say the seats are Lab 336, Con 230, SNP 53, LD 8, PC 4, Green 1. Shockingly low figure for them from this outfit. |
|
YL
Non-Aligned
Either Labour leaning or Lib Dem leaning but not sure which
Posts: 4,915
|
Post by YL on May 16, 2022 12:04:41 GMT
Indeed, it will depend on the variables used and the sample size. The big YouGov MRPs before the last two elections had enormous sample sizes which allowed them to use constituency level information, which will help with predicting Lib Dem seats. Sometimes they actually overdid local effects; IIRC they sometimes had Claire Wright winning East Devon, for example. The "huge" sample size of 10k is still only 15 per constituency. They use constituency level demographics and possibly constituency level local election results but they can't account for constituency level campaigning by different parties and tactical voting easily. How they try and account for this is mysterious and Claire Wright referred to above suggests pure guesswork is their final adjustment! YouGov's was actually 100k, not 10k, which is enough to pick up hints of local effects but with a fair margin of error, hence the wide confidence intervals mentioned by bjornhattan. See Ben Lauderdale's explanation here, where he does acknowledge this. |
|
|
|
Post by John Chanin on May 16, 2022 13:17:22 GMT
The "huge" sample size of 10k is still only 15 per constituency. They use constituency level demographics and possibly constituency level local election results but they can't account for constituency level campaigning by different parties and tactical voting easily. How they try and account for this is mysterious and Claire Wright referred to above suggests pure guesswork is their final adjustment! YouGov's was actually 100k, not 10k, which is enough to pick up hints of local effects but with a fair margin of error, hence the wide confidence intervals mentioned by bjornhattan . See Ben Lauderdale's explanation here, where he does acknowledge this. MRPs can’t pick up local effects and shouldn’t try. However with a big enough sample they should be able to say something sensible about Conservative-Liberal contests generically. I’d want around 50 from each constituency, ie a sample size of around 30,000. This is way more than the standard national thousand or so, but not impossible for YouGov given the size of their panel. This would have large enough numbers for sensible segmentation. |
|
|
|
Post by andrew111 on May 16, 2022 15:22:04 GMT
The "huge" sample size of 10k is still only 15 per constituency. They use constituency level demographics and possibly constituency level local election results but they can't account for constituency level campaigning by different parties and tactical voting easily. How they try and account for this is mysterious and Claire Wright referred to above suggests pure guesswork is their final adjustment! YouGov's was actually 100k, not 10k, which is enough to pick up hints of local effects but with a fair margin of error, hence the wide confidence intervals mentioned by bjornhattan. See Ben Lauderdale's explanation here, where he does acknowledge this. Ah, correct, 105k. I read an incorrect press headline I think |
|
|
|
Post by woollyliberal on May 23, 2022 11:02:53 GMT
|
|
|
|
Post by bjornhattan on May 23, 2022 11:05:52 GMT
If the Greens win East Hampshire at the next general election, I will buy the entire forum a pint.
|
|
Tony Otim
Green
Suffering from Brexistential Despair
Posts: 11,913
|
Post by Tony Otim on May 23, 2022 11:50:17 GMT
|
|
|
|
Post by Pete Whitehead on May 23, 2022 11:52:12 GMT
Risible bollocks
|
|
|
|
Post by bjornhattan on May 23, 2022 12:22:11 GMT
Green gains Bournemouth E Bournemouth W Bury St Edmunds Dorset W Forest of Dean Hampshire E Isle of Wight Lewes Reigate Salisbury Somerset N Somerton & Frome Stafford Stroud Truro & Falmouth Vale of Glamorgan Worth noting that the Greens were only contesting those seats, plus: Cannock Chase New Forest West North Herefordshire Rutland and Melton South Staffordshire West Worcestershire Witham So essentially, wherever EC decided the Greens would contest, they won (unless the seat in question was incredibly safely Tory). |
|
|
|
Post by finsobruce on May 23, 2022 12:24:24 GMT
Green gains Bournemouth E Bournemouth W Bury St Edmunds Dorset W Forest of Dean Hampshire E Isle of Wight Lewes Reigate Salisbury Somerset N Somerton & Frome Stafford Stroud Truro & Falmouth Vale of Glamorgan The Greens got 4.6% in Stafford at the 2019 General election.
|
|
