Focaldata

[andrew111]

Dec 26, 2021 19:38:20 GMT mattbewilson said:

Dec 26, 2021 18:00:00 GMT andrew111 said:

It would be similarly rubbish but might have some effect of its own

tbf many rubbished the MRPs prediction of libdem meltdown in 2019 but it turned out accurate

Well, there wasn't really a Lib Dem "meltdown" in 2019, getting 50% more votes than 2017 and moving back into second in a raft of seats. Failure to win a number of hoped for marginals such as Cheltenham was generally in line with UNS, and I would be interested to know if there is a seat that was predicted to be taken on UNS (with adjustment for regional swing) on the day that was not taken (maybe Sheffield Hallam?). With a fair voting system we would have had over 70 seats and no Tory "landslide".

Of course relative to the EU election generated hype it was poor, but nothing compared to the Labour "meltdown" which was relative to an actual previous result.

More generally, I managed to persuade the tables to open and as far as I can see the Labour vote is above 20% in every English seat in the model. 24% in Kingston and Surbiton and similar in Richmond Park for example. That did not happen in 1997 in Lib Dem won seats, and will not happen in 2024.

[graham]

Dec 26, 2021 20:01:53 GMT batman said:

Dec 26, 2021 19:39:53 GMT mattbewilson said:

i think what is meant is Labour aren't going to win seats in Glasgow but lose East Lothian. If Labour win seats in Glasgow they'll win East Lothian. If they don't win East Lothian they'll probably not pick up much more

exactly right. The only Labour target in Scotland which is easier than E Lothian is Kirkcaldy & Cowdenbeath and that's much easier. Glasgow Central wasn't even all that close in 2017, it was the second biggest SNP majority of the Glasgow seats.

Labour did win a seat in Glasgow in 2017 and was pretty close in every other city seat. I have read suggestions that Labour could wrest control og Glasgow from the SNP at the local elections. Were that to happen, it would be a useful launching pad for the parliamentary seats.

[gwynthegriff]

Dec 26, 2021 14:31:40 GMT graham said:

Dec 26, 2021 14:26:52 GMT Harry Hayfield said:

But remember it won't just be Ceredigion next time, it will be Ceredigion and Preseli, Pembrokeshire combined (minus a bit) and when the experts do their maths and declare it as a Plaid / Con battleground with Labour a distant third, you watch Labour deliberately tone down their campaign to ensure that a Conservatives doesn't get elected.

Only circa 20% of the Preseli seat will be combined with Ceredigion - the bulk becoming part of the new Pembroke seat.

Which parts of Preseli/Pembroke go into the "new" Ceredigion?

[clyde1998]

Been a while since I've posted on here, but I thought I'd breakdown some of what's wrong with this MRP analysis.

• As others have said, MRP polls cannot (adequately) take into account localised issues/campaigns nor tactical voting. All of the smaller parties depend on very concentrated levels of support and their campaigns reflect that.
• Compared to YouGov's previous MRP analyses, this model is significantly more basic (based on the limited information available). YouGov used the respondent's previous general election vote, 2016 referendum vote, sex, age, qualifications, housing tenure, ethnicity and martial status (possibly others I've missed) - notwithstanding any additional factors for the type of constituency the individual was voting in (such as Lib Dem-Conservative contests, region, etc.). Focaldata appear to only be using sex, age, previous vote and 2016 referendum vote for the individual factors - which is severely lacking for an analysis of this sort. A poll that weighted only on these measures would be insufficient, let alone MRP analyses.
• As with other MRP models in the UK, the different political context in Scotland should necessitate using 2014 referendum vote as a variable within Scotland. This would, additionally, help with the 'headline VI' for Scotland - which seems to massively underestimate SNP support, in this dataset, when compared to recent Scottish polling (37% compared to ~48%), a possible factor of not considering independence support.
• There seems to be a 'squashing' of party support ranges, as I've noticed from other MRP data in the past. As an example, Labour are down a lot in their strongest areas (9% in Walthamstow; 11% in East Ham; 8% in Liverpool Riverside; 7% in Leicester South), but up significantly in weaker areas (12% in Wealden; 15% in Witney; 17% in Newbury; 14% in Ceredigion). There are similar patterns for other parties. This will invariably have an affect on the seat estimates as any 'levelling out' of party vote shares across constituencies will overestimate the largest party's seat position compared to others and helps explain some of the more unrealistic results, such as Labour gaining Ceredigion and Wimbledon.
• We obviously don't know which candidates will be standing for election - major independent candidates, such as in Ashfield or East Devon, would heavily affect the results, as could not having Green or Reform candidates in a constituency.

[froome]

Dec 26, 2021 20:58:18 GMT clyde1998 said:

Been a while since I've posted on here, but I thought I'd breakdown some of what's wrong with this MRP analysis.

• As others have said, MRP polls cannot (adequately) take into account localised issues/campaigns nor tactical voting. All of the smaller parties depend on very concentrated levels of support and their campaigns reflect that.
• Compared to YouGov's previous MRP analyses, this model is significantly more basic (based on the limited information available). YouGov used the respondent's previous general election vote, 2016 referendum vote, sex, age, qualifications, housing tenure, ethnicity and martial status (possibly others I've missed) - notwithstanding any additional factors for the type of constituency the individual was voting in (such as Lib Dem-Conservative contests, region, etc.). Focaldata appear to only be using sex, age, previous vote and 2016 referendum vote for the individual factors - which is severely lacking for an analysis of this sort. A poll that weighted only on these measures would be insufficient, let alone MRP analyses.
• As with other MRP models in the UK, the different political context in Scotland should necessitate using 2014 referendum vote as a variable within Scotland. This would, additionally, help with the 'headline VI' for Scotland - which seems to massively underestimate SNP support, in this dataset, when compared to recent Scottish polling (37% compared to ~48%), a possible factor of not considering independence support.
• There seems to be a 'squashing' of party support ranges, as I've noticed from other MRP data in the past. As an example, Labour are down a lot in their strongest areas (9% in Walthamstow; 11% in East Ham; 8% in Liverpool Riverside; 7% in Leicester South), but up significantly in weaker areas (12% in Wealden; 15% in Witney; 17% in Newbury; 14% in Ceredigion). There are similar patterns for other parties. This will invariably have an affect on the seat estimates as any 'levelling out' of party vote shares across constituencies will overestimate the largest party's seat position compared to others and helps explain some of the more unrealistic results, such as Labour gaining Ceredigion and Wimbledon.
• We obviously don't know which candidates will be standing for election - major independent candidates, such as in Ashfield or East Devon, would heavily affect the results, as could not having Green or Reform candidates in a constituency.

Excellent analysis.

I do like the idea of a martial status.

[andrew111]

Dec 26, 2021 21:53:12 GMT froome said:

Dec 26, 2021 20:58:18 GMT clyde1998 said:

Been a while since I've posted on here, but I thought I'd breakdown some of what's wrong with this MRP analysis.

• As others have said, MRP polls cannot (adequately) take into account localised issues/campaigns nor tactical voting. All of the smaller parties depend on very concentrated levels of support and their campaigns reflect that.
• Compared to YouGov's previous MRP analyses, this model is significantly more basic (based on the limited information available). YouGov used the respondent's previous general election vote, 2016 referendum vote, sex, age, qualifications, housing tenure, ethnicity and martial status (possibly others I've missed) - notwithstanding any additional factors for the type of constituency the individual was voting in (such as Lib Dem-Conservative contests, region, etc.). Focaldata appear to only be using sex, age, previous vote and 2016 referendum vote for the individual factors - which is severely lacking for an analysis of this sort. A poll that weighted only on these measures would be insufficient, let alone MRP analyses.
• As with other MRP models in the UK, the different political context in Scotland should necessitate using 2014 referendum vote as a variable within Scotland. This would, additionally, help with the 'headline VI' for Scotland - which seems to massively underestimate SNP support, in this dataset, when compared to recent Scottish polling (37% compared to ~48%), a possible factor of not considering independence support.
• There seems to be a 'squashing' of party support ranges, as I've noticed from other MRP data in the past. As an example, Labour are down a lot in their strongest areas (9% in Walthamstow; 11% in East Ham; 8% in Liverpool Riverside; 7% in Leicester South), but up significantly in weaker areas (12% in Wealden; 15% in Witney; 17% in Newbury; 14% in Ceredigion). There are similar patterns for other parties. This will invariably have an affect on the seat estimates as any 'levelling out' of party vote shares across constituencies will overestimate the largest party's seat position compared to others and helps explain some of the more unrealistic results, such as Labour gaining Ceredigion and Wimbledon.
• We obviously don't know which candidates will be standing for election - major independent candidates, such as in Ashfield or East Devon, would heavily affect the results, as could not having Green or Reform candidates in a constituency.

Excellent analysis.

I do like the idea of a martial status.

The Sun Tzu of psephology has a nice ring to it! (provided you don't try to say it out loud!)

[graham]

Dec 26, 2021 20:50:47 GMT gwynthegriff said:

Dec 26, 2021 14:31:40 GMT graham said:

Only circa 20% of the Preseli seat will be combined with Ceredigion - the bulk becoming part of the new Pembroke seat.

Which parts of Preseli/Pembroke go into the "new" Ceredigion?

I believe that the initial proposals would see St Davids , Fishguard & Goodwick  and the rural northern areas of Pembrokeshire merged with Ceredigion.

[graham]

Dec 26, 2021 20:58:18 GMT clyde1998 said:

Been a while since I've posted on here, but I thought I'd breakdown some of what's wrong with this MRP analysis.

• As others have said, MRP polls cannot (adequately) take into account localised issues/campaigns nor tactical voting. All of the smaller parties depend on very concentrated levels of support and their campaigns reflect that.
• Compared to YouGov's previous MRP analyses, this model is significantly more basic (based on the limited information available). YouGov used the respondent's previous general election vote, 2016 referendum vote, sex, age, qualifications, housing tenure, ethnicity and martial status (possibly others I've missed) - notwithstanding any additional factors for the type of constituency the individual was voting in (such as Lib Dem-Conservative contests, region, etc.). Focaldata appear to only be using sex, age, previous vote and 2016 referendum vote for the individual factors - which is severely lacking for an analysis of this sort. A poll that weighted only on these measures would be insufficient, let alone MRP analyses.
• As with other MRP models in the UK, the different political context in Scotland should necessitate using 2014 referendum vote as a variable within Scotland. This would, additionally, help with the 'headline VI' for Scotland - which seems to massively underestimate SNP support, in this dataset, when compared to recent Scottish polling (37% compared to ~48%), a possible factor of not considering independence support.
• There seems to be a 'squashing' of party support ranges, as I've noticed from other MRP data in the past. As an example, Labour are down a lot in their strongest areas (9% in Walthamstow; 11% in East Ham; 8% in Liverpool Riverside; 7% in Leicester South), but up significantly in weaker areas (12% in Wealden; 15% in Witney; 17% in Newbury; 14% in Ceredigion). There are similar patterns for other parties. This will invariably have an affect on the seat estimates as any 'levelling out' of party vote shares across constituencies will overestimate the largest party's seat position compared to others and helps explain some of the more unrealistic results, such as Labour gaining Ceredigion and Wimbledon.
• We obviously don't know which candidates will be standing for election - major independent candidates, such as in Ashfield or East Devon, would heavily affect the results, as could not having Green or Reform candidates in a constituency.

Gaining Wimbledon is hardly particularly unrealistic , given that Labour did hold the seat 1997 - 2005 and performed well there in 2017.

[batman]

Dec 26, 2021 20:31:03 GMT graham said:

Dec 26, 2021 20:01:53 GMT batman said:

exactly right. The only Labour target in Scotland which is easier than E Lothian is Kirkcaldy & Cowdenbeath and that's much easier. Glasgow Central wasn't even all that close in 2017, it was the second biggest SNP majority of the Glasgow seats.

Labour did win a seat in Glasgow in 2017 and was pretty close in every other city seat. I have read suggestions that Labour could wrest control og Glasgow from the SNP at the local elections. Were that to happen, it would be a useful launching pad for the parliamentary seats.

Labour were very close in South West and East, but not all that close in Central - reasonably close one might say. However it is true that the swing to the SNP in 2019 was lower in Central than in most seats of its kind, although still significant.

[batman]

Dec 26, 2021 22:27:59 GMT graham said:

Dec 26, 2021 20:58:18 GMT clyde1998 said:

Been a while since I've posted on here, but I thought I'd breakdown some of what's wrong with this MRP analysis.

• As others have said, MRP polls cannot (adequately) take into account localised issues/campaigns nor tactical voting. All of the smaller parties depend on very concentrated levels of support and their campaigns reflect that.
• Compared to YouGov's previous MRP analyses, this model is significantly more basic (based on the limited information available). YouGov used the respondent's previous general election vote, 2016 referendum vote, sex, age, qualifications, housing tenure, ethnicity and martial status (possibly others I've missed) - notwithstanding any additional factors for the type of constituency the individual was voting in (such as Lib Dem-Conservative contests, region, etc.). Focaldata appear to only be using sex, age, previous vote and 2016 referendum vote for the individual factors - which is severely lacking for an analysis of this sort. A poll that weighted only on these measures would be insufficient, let alone MRP analyses.
• As with other MRP models in the UK, the different political context in Scotland should necessitate using 2014 referendum vote as a variable within Scotland. This would, additionally, help with the 'headline VI' for Scotland - which seems to massively underestimate SNP support, in this dataset, when compared to recent Scottish polling (37% compared to ~48%), a possible factor of not considering independence support.
• There seems to be a 'squashing' of party support ranges, as I've noticed from other MRP data in the past. As an example, Labour are down a lot in their strongest areas (9% in Walthamstow; 11% in East Ham; 8% in Liverpool Riverside; 7% in Leicester South), but up significantly in weaker areas (12% in Wealden; 15% in Witney; 17% in Newbury; 14% in Ceredigion). There are similar patterns for other parties. This will invariably have an affect on the seat estimates as any 'levelling out' of party vote shares across constituencies will overestimate the largest party's seat position compared to others and helps explain some of the more unrealistic results, such as Labour gaining Ceredigion and Wimbledon.
• We obviously don't know which candidates will be standing for election - major independent candidates, such as in Ashfield or East Devon, would heavily affect the results, as could not having Green or Reform candidates in a constituency.

Gaining Wimbledon is hardly particularly unrealistic , given that Labour did hold the seat 1997 - 2005 and performed well there in 2017.

believe me, I know the constituency very well, and have canvassed for Labour in it several times this year. It is pretty unrealistic even on the existing boundaries. If the boundaries change as mooted it would become completely unrealistic. On the basis of all recent polls, which includes those giving Labour national leads of 8 & even 9%, the Lib Dems would gain the seat, and I can assure you that local circumstances are considerably more in Labour's disfavour than the polls suggest. Labour has a much better chance in Cities of London & Westminster, and I'd argue Finchley & Golders Green though the latter is quite a bit harder.

[Adam in Stroud]

Dec 26, 2021 22:34:18 GMT batman said:

Dec 26, 2021 22:27:59 GMT graham said:

Gaining Wimbledon is hardly particularly unrealistic , given that Labour did hold the seat 1997 - 2005 and performed well there in 2017.

believe me, I know the constituency very well, and have canvassed for Labour in it several times this year. It is pretty unrealistic even on the existing boundaries. If the boundaries change as mooted it would become completely unrealistic. On the basis of all recent polls, which includes those giving Labour national leads of 8 & even 9%, the Lib Dems would gain the seat, and I can assure you that local circumstances are considerably more in Labour's disfavour than the polls suggest. Labour has a much better chance in Cities of London & Westminster, and I'd argue Finchley & Golders Green though the latter is quite a bit harder.

Agree with all of that. My gut feeling is that neither of the last two are realistic prospects for us.

[gwynthegriff]

Dec 26, 2021 22:25:02 GMT graham said:

Dec 26, 2021 20:50:47 GMT gwynthegriff said:

Which parts of Preseli/Pembroke go into the "new" Ceredigion?

I believe that the initial proposals would see St Davids , Fishguard & Goodwick  and the rural northern areas of Pembrokeshire merged with Ceredigion.

In which case I would expect Plaid to do better than their recent performance in PP would suggest.

Landsker and all that.

[graham]

Dec 26, 2021 22:50:57 GMT gwynthegriff said:

Dec 26, 2021 22:25:02 GMT graham said:

I believe that the initial proposals would see St Davids , Fishguard & Goodwick  and the rural northern areas of Pembrokeshire merged with Ceredigion.

In which case I would expect Plaid to do better than their recent performance in PP would suggest.

Landsker and all that.

Quite probably so - though there is likely to be a good Labour vote in St Davids , Fishguard & Goodwick - maybe Solva too. I cannot see the new seat going Tory myself.

[graham]

Dec 26, 2021 22:34:18 GMT batman said:

Dec 26, 2021 22:27:59 GMT graham said:

Gaining Wimbledon is hardly particularly unrealistic , given that Labour did hold the seat 1997 - 2005 and performed well there in 2017.

believe me, I know the constituency very well, and have canvassed for Labour in it several times this year. It is pretty unrealistic even on the existing boundaries. If the boundaries change as mooted it would become completely unrealistic. On the basis of all recent polls, which includes those giving Labour national leads of 8 & even 9%, the Lib Dems would gain the seat, and I can assure you that local circumstances are considerably more in Labour's disfavour than the polls suggest. Labour has a much better chance in Cities of London & Westminster, and I'd argue Finchley & Golders Green though the latter is quite a bit harder.

A few leaflets citing this poll - and having the LDs in third place - will not do Labour any harm. I have to say I have no knowledge as to the likely impact of the boundary changes here.

[johng]

Dec 26, 2021 22:50:57 GMT gwynthegriff said:

Dec 26, 2021 22:25:02 GMT graham said:

I believe that the initial proposals would see St Davids , Fishguard & Goodwick  and the rural northern areas of Pembrokeshire merged with Ceredigion.

In which case I would expect Plaid to do better than their recent performance in PP would suggest.

Landsker and all that.

I think the Plaid vote across PP is pretty suppressed as the party has basically no chance of winning. Rather foolishly, they haven't been putting up much effort in the seat either despite a merger of the north with Ceredigion being quite likely for some years.

The current proposed split doesn't really harm Labour and there's no realistic way we'd be so far back behind Plaid/ Conservatives we are uncompetitive.

[graham]

Dec 26, 2021 23:17:11 GMT johng said:

Dec 26, 2021 22:50:57 GMT gwynthegriff said:

In which case I would expect Plaid to do better than their recent performance in PP would suggest.

Landsker and all that.

I think the Plaid vote across PP is pretty suppressed as the party has basically no chance of winning. Rather foolishly, they haven't been putting up much effort in the seat either despite a merger of the north with Ceredigion being quite likely for some years.

The current proposed split doesn't really harm Labour and there's no realistic way we'd be so far back behind Plaid/ Conservatives we are uncompetitive.

I don't think the new seat as proposed will be so different to the Ceredigion & Pembroke North seat which existed 1983 - 1997 - though that did not extend as far south as St Davids and Solva.

[mattbewilson]

Dec 26, 2021 20:29:35 GMT andrew111 said:

Dec 26, 2021 19:38:20 GMT mattbewilson said:

tbf many rubbished the MRPs prediction of libdem meltdown in 2019 but it turned out accurate

Well, there wasn't really a Lib Dem "meltdown" in 2019, getting 50% more votes than 2017 and moving back into second in a raft of seats. Failure to win a number of hoped for marginals such as Cheltenham was generally in line with UNS, and I would be interested to know if there is a seat that was predicted to be taken on UNS (with adjustment for regional swing) on the day that was not taken (maybe Sheffield Hallam?). With a fair voting system we would have had over 70 seats and no Tory "landslide".

Of course relative to the EU election generated hype it was poor, but nothing compared to the Labour "meltdown" which was relative to an actual previous result.

More generally, I managed to persuade the tables to open and as far as I can see the Labour vote is above 20% in every English seat in the model. 24% in Kingston and Surbiton and similar in Richmond Park for example. That did not happen in 1997 in Lib Dem won seats, and will not happen in 2024.

I mean yes in terms of vote it wasn't a bad result for the Lib Dems but in terms of turning votes into seats it wasn't particularly good.

I do agree that it's unlikely any of the seats you mention will see such a surge in support for Labour and split the vote

[mattbewilson]

Dec 26, 2021 23:02:44 GMT graham said:

Dec 26, 2021 22:34:18 GMT batman said:

believe me, I know the constituency very well, and have canvassed for Labour in it several times this year. It is pretty unrealistic even on the existing boundaries. If the boundaries change as mooted it would become completely unrealistic. On the basis of all recent polls, which includes those giving Labour national leads of 8 & even 9%, the Lib Dems would gain the seat, and I can assure you that local circumstances are considerably more in Labour's disfavour than the polls suggest. Labour has a much better chance in Cities of London & Westminster, and I'd argue Finchley & Golders Green though the latter is quite a bit harder.

A few leaflets citing this poll - and having the LDs in third place - will not do Labour any harm. I have to say I have no knowledge as to the likely impact of the boundary changes here.

in the real world though it takes people to design those leaflets and deliver them. I don't know the situation in Wimbledon but one of the parties problems from my experience is in some seats we last won when Blair was leader we had people to do these things. But those people are old now or aren't with us either in the party or in life. We often talk about all these new young inspired people but outside a general election when it actually matters there's so few people to do any work

[mattbewilson]

Dec 26, 2021 20:31:03 GMT graham said:

Dec 26, 2021 20:01:53 GMT batman said:

exactly right. The only Labour target in Scotland which is easier than E Lothian is Kirkcaldy & Cowdenbeath and that's much easier. Glasgow Central wasn't even all that close in 2017, it was the second biggest SNP majority of the Glasgow seats.

Labour did win a seat in Glasgow in 2017 and was pretty close in every other city seat. I have read suggestions that Labour could wrest control og Glasgow from the SNP at the local elections. Were that to happen, it would be a useful launching pad for the parliamentary seats.

they also won East Lothian in 2017. If they hadn't I'd understand the argument that perhaps Glasgow is a better prospect that East Lothian

[YL]

Dec 26, 2021 9:04:39 GMT cogload said:

The map looks like predicting Portsmouth South as a Conservative gain (or the one I saw posted on twitter did). That isn't happening.

Just to be clear, the actual MRP figures do not show Portsmouth South as a Tory gain; indeed it shows Labour 20 percentage points ahead.

Of course Portsmouth South does show that LD/Con contests can turn into Labour seats, as does Sheffield Hallam. But in both cases something happened to break the tactical voting pattern, and I don't see why that would happen now in Bath; the most plausible explanation is indeed that this particular MRP methodology, especially this far out from an election, just doesn't pick up the dynamics in that sort of constituency.