UK General Election - May 7th 2015

[136or142]

Would the SNP back a Labour government?  Otherwise, according to the Ashcroft poll, no party can put together a majority.

Answering my own question: http://www.telegraph.co.uk/news/general-election-2015/11404828/Nicola-Sturgeon-price-of-helping-Ed-Miliband-into-Number-10-is-spending-spree.html

I'd assume Ed Miliband is going to have to turn this down lest Cameron run on it.

The Ashcroft poll didn't mention the local candidate names.  I'd guess that at least a number of incumbent Labour and Liberal Democratic M.Ps would have received higher numbers if their names were included.

[ObserverIE]

So we could be looking at quite a few 'Portillo moments'. The Alexanders, Murphy, Kennedy, Clegg, Hughes...

A bit harsh on Kennedy, who has never been a fan of the coalition, to bracket him with Portillo 1997.

[Filuwaúrdjan]

For some reason a new YouGov/Evening Standard poll of The Great City Babylon, even though the last one is still fresh. Money to burn? Anyway, modest pro-Labour movement from said earlier poll. Figures reported to be:

Labour 44, Con 32, UKIP 10, LDem 7, Greens 5

Which is a swing of around 5pts from 2010; YouGov's national polls are usually showing national swings of between 2pts and 4pts currently.

[Clyde1998]

For some reason a new YouGov/Evening Standard poll of The Great City Babylon, even though the last one is still fresh. Money to burn? Anyway, modest pro-Labour movement from said earlier poll. Figures reported to be:

Labour 44, Con 32, UKIP 10, LDem 7, Greens 5

Which is a swing of around 5pts from 2010; YouGov's national polls are usually showing national swings of between 2pts and 4pts currently.

London may be more likely to swing to Labour than the rest of the country, as they're pledging not to hold an EU referendum - and London doesn't seem to want to leave the EU (based on national poll sub-samples). The Conservatives are offering an EU referendum.

[afleitch]

A gentle reminder to all;

Uniform National Swing is still the best indicator (rather than all that subsample stuff and algorithms) at what is going to happen. It bet all other models in 2010 based on the eve of poll vote shares and does equally as well when you key in the actual votes received. One worth thinking about. All that has to change this year is a separate model for Scotland.

[Silent Hunter]

Quite frankly, you need less a swingometer and more a swing hypercube to model all the possibilities.

[YL]

A gentle reminder to all;

Uniform National Swing is still the best indicator (rather than all that subsample stuff and algorithms) at what is going to happen. It bet all other models in 2010 based on the eve of poll vote shares and does equally as well when you key in the actual votes received. One worth thinking about. All that has to change this year is a separate model for Scotland.

Not true.  Martin Baxter's (Electoral Calculus) Strong Transition model outperformed UNS in 2010.

Anyway, the fact is that we didn't have anything like the amount of constituency polling that we do this time.  If Al's right, and it's basically all junk, we might as well use a model based on national (plus Scotland-specific) movements like UNS or Strong Transition, but if there's actually some useful information in those Ashcroft and Survation polls then models like electionforecast.co.uk which use them might do better.  Eventually, we will find out.

One thing UNS and Strong Transition will not do well is predict UKIP performance in individual seats, but that may not matter very much looking at the overall result.

[afleitch]

A gentle reminder to all;

Uniform National Swing is still the best indicator (rather than all that subsample stuff and algorithms) at what is going to happen. It bet all other models in 2010 based on the eve of poll vote shares and does equally as well when you key in the actual votes received. One worth thinking about. All that has to change this year is a separate model for Scotland.

Not true.  Martin Baxter's (Electoral Calculus) Strong Transition model outperformed UNS in 2010.

The total seat error (if you add up the seats too high and too low by party in the prediction) with Baxter was 66. With UNS based simply on a 'poll of polls' on the last day, the error was only 48. Ultimately Baxter's transition model is a modified UNS anyway.

[YL]

A gentle reminder to all;

Uniform National Swing is still the best indicator (rather than all that subsample stuff and algorithms) at what is going to happen. It bet all other models in 2010 based on the eve of poll vote shares and does equally as well when you key in the actual votes received. One worth thinking about. All that has to change this year is a separate model for Scotland.

Not true.  Martin Baxter's (Electoral Calculus) Strong Transition model outperformed UNS in 2010.

The total seat error (if you add up the seats too high and too low by party in the prediction) with Baxter was 66. With UNS based simply on a 'poll of polls' on the last day, the error was only 48.

I was going on the actual shares cast (and, admittedly, Baxter's own figures).  Given how far out the polls were, I'm finding it difficult to believe that an election eve poll of polls actually did better on UNS: do you have a source for that?  Certainly Baxter's models gave worse results based on the polling than the actual figures.

Edit: I think I misunderstood which way you were measuring the performance.  But still, when testing them, I think it makes more sense to use the actual vote shares as input to the models than some rather erroneous polls.

You must be logged in to read this quote.

This is true, but my other point is that in 2010 UNS and its variants were the only game in town, because there was essentially no constituency polling.  The other model I was aware of, 538's, was IIRC essentially also a modified UNS, but the modifications it made didn't work very well.

This time, we do have constituency polling, so the models which use it (whether in a crude way, like May2015, or a more sophisticated one) and those which don't can be compared.  As I said, if Al's right about the constituency polling then not using it will presumably turn out better.

[afleitch]

http://news.bbc.co.uk/1/hi/uk_politics/election_2010/8609989.stm

I used the model on the BBC and plugged in the 'Poll of Polls'. It get's Labour out by only 1 seat. The main difference is between the Lib Dems and the Tories

Using the actual results, it underestimates the Tories by 15, Labour is overestimated by 10 and the Lib Dems by 5. UNS is a simple model (and I'm a fan of Baxter's), but it still does the job if you want to guess a seat total and don't mind about what the details.

[136or142]

http://ukpollingreport.co.uk/

LATEST UNS PROJECTION
Labour short by 19

[Phony Moderate]

Labour 4% ahead with YouGov, their biggest lead with them since December. The overall position right now is probably something between a tie and Labour 1% ahead.

[Clyde1998]

Kezia Dugdale is showing why people are turning their back on Labour [in Scotland] on tonight's Question Time. She is really out of her depth.

[Thomas D]

Hey, Clint Eastwood, since you love empty chairs so much, you're going to love these UK debates.

[afleitch]

http://www.bbc.co.uk/news/election/2015

Sexy logo.

[Filuwaúrdjan]

As it is Monday, there are more polls:

Populus: Labour 33, Con 32, UKIP 15, LDem 9, Greens 6, Other 6

Ashcroft/BouncyBouncyBall: Con 34, Labour 30, UKIP 15, Greens 8, LDem 5

[You kip if you want to...]

Sorry, but who's advising the candidates who're rejecting Blair's donations in Labour's 106 marginals?

[afleitch]

It would be nice for the Tories to be 'bawhair' ahead of Labour before the Budget.

[afleitch]

YouGov

CON 35
LAB 31
LIB 8
UKIP 14
GRN 6

For some reason Mondays are recently good polling days for the Tories

[Filuwaúrdjan]

Another poll of Wales (usual caveats apply, blah, blah) done by YouGov for ITN:

Labour 39, Con 25, UKIP 14, Plaid 10, Greens 6, LDem 5

[Clyde1998]

Another poll of Wales (usual caveats apply, blah, blah) done by YouGov for ITN:

Labour 39, Con 25, UKIP 14, Plaid 10, Greens 6, LDem 5

Compared to the last YouGov poll of Wales.

Lab - Up 2
Con - Up 2
UKIP - Down 2
Plaid - N/C
Green - Down 2
Lib - Down 1

Good to see UKIP continue to fall.

[Clyde1998]

YouGov polls only.

It seems that most of the Lib Dem 2010 votes moved to Labour early on, before some moved to UKIP more recently. They seem to be moving back to Labour again.

[Zanas]

YouGov polls only.

It seems that most of the Lib Dem 2010 votes moved to Labour early on, before some moved to UKIP more recently. They seem to be moving back to Labour again.

That is not how votes work.

[Clyde1998]

YouGov polls only.

It seems that most of the Lib Dem 2010 votes moved to Labour early on, before some moved to UKIP more recently. They seem to be moving back to Labour again.

That is not how votes work.

I know it won't be only Lib Dem voters moving to Labour, but a sizeable amount of their support will have moved towards Labour for their support to fall almost as fast as the Labour vote rose. (As an example)

Annoyingly, YouGov don't publish 2010 Vote in their Wales tables (they do for the UK-wide and Scotland polls though). The links at the bottom are an example of this.

This means that the assumptions I've made are based on the way each party gains and losses support compared to each other.

Wales Jan 2015 Poll, UK-wide 9 Mar 2015 Poll

[DL]

What seats in Wales would change hands with a uniform swing in Wales like that?

Another poll of Wales (usual caveats apply, blah, blah) done by YouGov for ITN:

Labour 39, Con 25, UKIP 14, Plaid 10, Greens 6, LDem 5

Compared to the last YouGov poll of Wales.

Lab - Up 2
Con - Up 2
UKIP - Down 2
Plaid - N/C
Green - Down 2
Lib - Down 1

Good to see UKIP continue to fall.