You are all getting excited again about a model which is good for predicting the overall result but not for predicting either individual constituencies or smaller parties. Just look at PC last time in 2017 - predicted 2, actual 4. And look at the 95% ranges; Conservative 328-385, Labour 187-238. So YouGov are actually forecasting a Conservative majority of between 6 and 120 with a 95% confidence. And that is all assuming that the model adequately accounts for the oddities and local effects present in this election which have been absent or rare in past GEs.
I really don't know why so much attention is being paid to YouGov MRP. If it's 43-32 then of course we know Con is going to get a middling majority - whether it's 50, 60 or 70 well who cares - it doesn't matter. The big issue is will it be 43-32 - and ICM and ComRes are suggesting it may not be that right now, let alone on polling day.
Also, if the Lib Dems get 13 seats that is basically the end of the party.
Perhaps it is the feel up here in Westmorland but it would be hard to conceive - and has been all along that the LDs would really get 13 seats. 6 or 7 would seem more likely. Obviously they are doing better elsewhere ?
If these 30 or 40 ideas were meaning anything up here then Matt Severn would be in contention in Penrith and Border - remember David Maclean nearly lost it in 1983. I'm anything but Matt's best friend, but he is by far the most credible candidate the LDs have put up in Penrith since 1997.
I really don't know why so much attention is being paid to YouGov MRP. If it's 43-32 then of course we know Con is going to get a middling majority - whether it's 50, 60 or 70 well who cares - it doesn't matter. The big issue is will it be 43-32 - and ICM and ComRes are suggesting it may not be that right now, let alone on polling day.
Obviously the point is that the YouGov MRP came much closer to predicting the Tory lead last time than the conventional polls! (Albeit it was still a slight overestimate.)
They have a sample of 150ish per seat, and then modify it according to demographics...I think. There are still some assumptions in there - minor candidates are assumed to score 0, for instance.
But lots of interesting results. Labour losing Bolsover but holding Canterbury, for example. All the independents seem doomed.
Labour will hold Canterbury with an enlarged majority.
Also, if the Lib Dems get 13 seats that is basically the end of the party.
Perhaps it is the feel up here in Westmorland but it would be hard to conceive - and has been all along that the LDs would really get 13 seats. 6 or 7 would seem more likely. Obviously they are doing better elsewhere ?
If these 30 or 40 ideas were meaning anything up here then Matt Severn would be in contention in Penrith and Border - remember David Maclean nearly lost it in 1983. I'm anything but Matt's best friend, but he is by far the most credible candidate the LDs have put up in Penrith since 1997.
Near doubling of the LD vote share is not really compatible with halving the number of seats, even allowing for the viscitudes of FPTP. Is there a list of the forecast Lab to Con losses and near misses? I have some interesting iseas for searching out betting value.
YG say this is based on an 11% lead and LDs on 14%
Surely it's not "based on" an 11-point lead. That comes out of the model. The lead in this model in 2017 was out of kilter with conventional polls, wasn't it? Now it's very much in kilter.
No - if you look at the detailed explanation, they are doing both a national forecast and 650 constituency forecasts derived from the same data, and the national one, collected over 7 days, is 43-32. If Labour has reduced the lead to 7, as the two latest polls suggest, then the Con lead in seats goes down to around 25 (see p. 35ff in the detailed report). If, conversely, Opinium was right with the 19% lead, then the Con majority is over 150.
There will be a significantly better national vote share, and many strong second places. While I have been predicting a Con majority, so Parliament will probably go full term, it will be a chaotic shambles of a government.
If the Tories get a majority then it is Labour that is likely to rip itself apart.
The battle for the leadership of the party and a fight for the future policy on rejoining the EU for a start.
Nah, we've been down before. There will be a significantly better national vote share, and many strong second places. While I have been predicting a Con majority, so Parliament will probably go full term, it will be a chaotic shambles of a government.
If the Tories get a majority then it is Labour that is likely to rip itself apart. The battle for the leadership of the party and a fight for the future policy on rejoining the EU for a start.
I am LD not Labour, but such a result would be the end of Corbynism. While the Tory party would start united, and most likely stay so. Brexit is going to be wholly owned by the Tories, so if we are not in the sunny uplands by the next election, those Blue Labour seats are going to be hard to retain.
YG say this is based on an 11% lead and LDs on 14%
Surely it's not "based on" an 11-point lead. That comes out of the model. The lead in this model in 2017 was out of kilter with conventional polls, wasn't it? Now it's very much in kilter.
No - if you look at the detailed explanation, they are doing both a national forecast and 650 constituency forecasts derived from the same data, and the national one, collected over 7 days, is 43-32. If Labour has reduced the lead to 7, as the two latest polls suggest, then the Con lead in seats goes down to around 25 (see p. 35ff in the detailed report). If, conversely, Opinium was right with the 19% lead, then the Con majority is over 150.
Con lead in seats of 25 with a 7% lead? Surely you mean a Con majority of 25
The next "Star Wars" film comes out a few days after the election. It covers the ongoing failure of a once-unstoppable franchise to come to terms with its selection of a catastrophically unfit director and dependence on old themes. And the other has lightsabres.
A good example of the limitations of this model is the forecast for Finchley. It is still predicting Labour will come second to the Tories. I am not arguing against the forecast of a Tory hold, but to suggest Labour will come second clearly shows the failure of the model to adequately account for local effects. This is not a surprise or criticism of those who have built the model; they would need a much bigger dataset to handle such effects. But it does demonstrate how ludicrous it is rely on the model much beyond its prediction of the overall result.
LD 13 would be hilarious. Could easily end up down on Tim Farron 2017.
So Bollocks to Brexit becomes sod off to swinson
Right at the start of the campaign, after the very first canvassing session, I reported here how badly she was going down in the SW. Along with Revoke. Without revoke, she might have been given a hearing. But it was not happening. She was actively repelling people away. Not that the LibDems wanted to hear it, of course. Just that viciously anti-LibDem MarqueeMark, doing his usual routine. Whatever. And as for "Prime Minister Jo Swinson"..... *chortle*
And me too. When I reported this on Sunday I got a really bad back-bite for likening her to a character in children's literature. But, it seems that Farron is now fighting as an "Independent Liberal".
Nah, we've been down before. There will be a significantly better national vote share, and many strong second places. While I have been predicting a Con majority, so Parliament will probably go full term, it will be a chaotic shambles of a government.
If the Tories get a majority then it is Labour that is likely to rip itself apart. The battle for the leadership of the party and a fight for the future policy on rejoining the EU for a start.
I am LD not Labour, but such a result would be the end of Corbynism. While the Tory party would start united, and most likely stay so. Brexit is going to be wholly owned by the Tories, so if we are not in the sunny uplands by the next election, those Blue Labour seats are going to be hard to retain.
Depends on how hard the MPs work their seats. Now there are exceptions but the incoming candidates - if they are elected look to me as if most of them will be like Trudy in Copeland. That sort of really responsive MP will be a novelty in a lot of these seats.
Also, if the Lib Dems get 13 seats that is basically the end of the party.
Perhaps it is the feel up here in Westmorland but it would be hard to conceive - and has been all along that the LDs would really get 13 seats. 6 or 7 would seem more likely. Obviously they are doing better elsewhere ?
If these 30 or 40 ideas were meaning anything up here then Matt Severn would be in contention in Penrith and Border - remember David Maclean nearly lost it in 1983. I'm anything but Matt's best friend, but he is by far the most credible candidate the LDs have put up in Penrith since 1997.
Near doubling of the LD vote share is not really compatible with halving the number of seats, even allowing for the viscitudes of FPTP. Is there a list of the forecast Lab to Con losses and near misses? I have some interesting iseas for searching out betting value.
I should have added the Tory selection to replace the well entrenched Rory was extremely controversial and regardless of the qualities of the Tory candidate - presumably pretty good his connections with P and B or Cumbria are an order of magnitude less than Matt's.
A good example of the limitations of this model is the forecast for Finchley. It is still predicting Labour will come second to the Tories. I am not arguing against the forecast of a Tory hold, but to suggest Labour will come second clearly shows the failure of the model to adequately account for local effects. This is not a surprise or criticism of those who have built the model; they would need a much bigger dataset to handle such effects. But it does demonstrate how ludicrous it is rely on the model much beyond its prediction of the overall result.
The MRP to me doesn't seem to be producing that many results different from what you would be expecting on a UNS model based on the national vote shares YouGov have been polling.Perhaps UNS is best after all.
YG say this is based on an 11% lead and LDs on 14%
Surely it's not "based on" an 11-point lead. That comes out of the model. The lead in this model in 2017 was out of kilter with conventional polls, wasn't it? Now it's very much in kilter.
No - if you look at the detailed explanation, they are doing both a national forecast and 650 constituency forecasts derived from the same data, and the national one, collected over 7 days, is 43-32. If Labour has reduced the lead to 7, as the two latest polls suggest, then the Con lead in seats goes down to around 25 (see p. 35ff in the detailed report). If, conversely, Opinium was right with the 19% lead, then the Con majority is over 150.
Yes, they are doing a national projection, but it's being done by different methods than are used in traditional polling. Last time those methods predicted the national Tory lead more accurately than the conventional polls. Now they seem to be in almost exact agreement with the average from polls conducted over the same period. Also, the MRP figure was much more stable over time than the conventional polls.
And if you're suggesting the Tory lead has dropped from 11 points to 7 points in the last 3-4 days, on the basis of selecting the two lowest leads out of the last four polls (and ignoring two others with leads of 10 and 11) then I don't think that's a sound procedure at all.
Comments
If it's 43-32 then of course we know Con is going to get a middling majority - whether it's 50, 60 or 70 well who cares - it doesn't matter.
The big issue is will it be 43-32 - and ICM and ComRes are suggesting it may not be that right now, let alone on polling day.
If these 30 or 40 ideas were meaning anything up here then Matt Severn would be in contention in Penrith and Border - remember David Maclean nearly lost it in 1983. I'm anything but Matt's best friend, but he is by far the most credible candidate the LDs have put up in Penrith since 1997.
I can live with principled objections to Brexit but their campaigns in the 2017 GE removed any shred of integrity from their subsequent actions.
Is there a list of the forecast Lab to Con losses and near misses? I have some interesting iseas for searching out betting value.
The battle for the leadership of the party and a fight for the future policy on rejoining the EU for a start.
While the Tory party would start united, and most likely stay so. Brexit is going to be wholly owned by the Tories, so if we are not in the sunny uplands by the next election, those Blue Labour seats are going to be hard to retain.
And the other has lightsabres.
And if you're suggesting the Tory lead has dropped from 11 points to 7 points in the last 3-4 days, on the basis of selecting the two lowest leads out of the last four polls (and ignoring two others with leads of 10 and 11) then I don't think that's a sound procedure at all.