UK General Election - May 7th 2015

[You kip if you want to...]

Alastair Darling standing down in Edinburgh South West.

Labour people being much, much nicer about him than they would've been this time 5 years ago...

[You kip if you want to...]

Ashcroft has Labour at 29%, tying their 2010 result.

Hold on tight, here comes a sh**tstorm.

[Filuwaúrdjan]

Though to add totally meaningless context a different (though also somewhat sketchy) outfit shows 35%.

[You kip if you want to...]

Ashcroft's national polls have been so erratic as to deem them a bit pointless, although his marginal polls are helpful.

Of course, this is a psychological blow for Labour (terrible news for ed m). Although an election with the big two at 30-29 would be funny in the sense that the media narrative would be all about Labour's terrible result and not the Tory's sub-1997 result.

[ChrisDR68]

Roll on the ongoing weakening of the unrepresentative two party system

Unfortunately it's likely both parties ratings will pick up once the actual campaign starts in March next year.

[You kip if you want to...]

Unfortunately it's likely both parties ratings will pick up once the actual campaign starts in March next year.

That'd be a first.

[🦀🎂🦀🎂]

At this point UK polls are far too depressing to look at. Damn this 5-year long term! Just like Labour to implode the minute before they run the risk of winning.

[njwes]

Is there any consensus on whether Lord Ashcroft's polls are any good?

[You kip if you want to...]

Is there any consensus on whether Lord Ashcroft's polls are any good?

His national ones are very volatile and his marginal ones are the only ones of their nature. Of course, we won't know is these are any good until the election's been and gone.

His by-elections polls have been pretty mixed. Very good in Clacton, okay in Wythenshawe and Newark, terrible in Corby and Heywood.

[Filuwaúrdjan]

His national polls are untested and look pretty dodgy at this stage (they bounce around in a weird and illogical style).

[EPG]

Volatile doesn't mean worse. You're always going to get a lot of sampling error in a good opinion poll, and it's more obvious when you publish frequently like YouGov or Ashcroft, whereas a lot of companies release at larger frequencies that make normal movements indistinguishable from error. It's easy to reduce headline variance if you make the appropriate adjustments to undecided voter intention, while not necessarily making the poll more accurate.

What should always cause concern is a pollster who's never published during a general election.

[YL]

The high volatility of the Ashcroft polls is probably simply down to having a fairly small effective sample size.  The "headline" sample size may be 1000, but the actual voting intention figures in this Monday's poll are based on just 522 respondents.

Speaking of Ashcroft, he has just released some more constituency polls, in the Tory held seats ranked between 40 and 52 on the Labour target list.  Summary of the results (using his preferred question, the constituency specific one): he has the Tories narrowly holding Blackpool North & Cleveleys, Kingswood and Loughborough, but Labour gaining Northampton North, Bury North, Erewash, City of Chester, Croydon Central, Worcester, Keighley, Cannock Chase (three way marginal, with UKIP actually ahead on the standard question and second, only two points behind, on the constituency one) and Wirral West.

[YL]

For comparison, here are the YouGov polls for the last week (the date is the date of publication in the Scum or the Sunday Times):

31 Oct: Con 33 Lab 32 UKIP 15 LD 7 Green 7 SNP/Plaid 5
2 Nov: Lab 32 Con 31 UKIP 18 LD 7 Green 6 SNP/Plaid 5
4 Nov: Lab 34 Con 33 UKIP 15 LD 8 Green 5 SNP/Plaid 4
5 Nov: Lab 34 Con 32 UKIP 15 LD 7 Green 6 SNP/Plaid 5
6 Nov: Lab 33 Con 32 UKIP 17 LD 7 Green 7 SNP/Plaid 4

[EPG]

It depends on one's filters.

Unless they changed recently, YouGov is very generous in accepting vote intention at face value from past non-voters and those reluctant to answer. These each increase the sample size, making each day's sample more similar to the face-value answers from the population as a whole. Note that the population in this case is YouGov's polling panel, but this contains several hundred thousand people, so N-n is pretty negligible. As with almost every poll, N-n is less important a problem than either systemic sampling bias or mis-interpreting the underlying phenomenon from the sense data. Large samples from a self-selecting pool induce stability but do not reflect an unambiguously correct way to represent general election outcomes.

On the other hand, YouGov also weight the panel's responses to various characteristics even including newspaper readership. In principle, this should keep results more stable even when there is natural variance in the sample from the panel.

[YL]

YouGov's headline sample size is also larger, for example 2041 in yesterday's poll.  (Presumably it's easier for them to get a large sample than it is for a phone pollster.)  It does indeed seem as if they have a weaker turnout filter, as the actual voting figures are based on about 1600 respondents, around three times as many as Ashcroft's.  This may have something to do with apathetic non-voters being less likely to sign up for a polling panel as well.

I have some qualms about internet panel polling as a methodology, but my impression is that YouGov have a fairly good track record, which suggests that they have found ways of dealing with its intrinsic flaws.  (I'm not prepared to say the same about any of the other internet pollsters...)

And yes, political weighting (such as party ID or past vote, though I have issues with the former especially as used by Populus) or newspaper readership (likely to be correlated with party ID) will reduce volatility too.  So it's not surprising that Populus and YouGov give relatively stable figures.

[You kip if you want to...]

Alex Salmond rumoured to be eyeing up Inverness, Danny Alexander's seat.

The SNP came third on 18% here in 2010, but he'd have a good chance.

[ChrisDR68]

I've read about the possibility of a Lab / SNP coalition after the next election with Salmond as deputy prime minister.

Considering he's about as popular in England as Howard Webb was when he refereed a game at Anfield that sounds like electoral suicide for Labour in the years following May 2015.

[EPG]

If so, he might still be more popular in England than Ed Miliband!

[Filuwaúrdjan]

About as likely to happen as I a to join the Conservative Party.

[EPG]

It is not simple to see a situation in which the SNP would participate in the government at Westminster, for reasons such as the SNP's devotion of political, intellectual and other resources to the government at Holyrood, and its vaunted moral stance about not casting Scottish votes on English laws. Reversing each would be very awkward but necessary if participating in a parliamentary majority, let alone a coalition. Not that the latter is likely. Coalition with Labour clashes with the latter's recent outbreak of Gordon Brown fanboyism which opposes almost all economic devolution. Coalition with English Conservatives looks toxic in Scotland.

On the other side of the putative coalition, not only is Salmond unpopular in England (albeit in the sense that Hannibal was unpopular in Rome, rather than in the Ed M sense), but a deal even well-short of coalition would inevitably be portrayed as anti-English. So the incentives to coalesce would only work if the SNP were objectively to hold the balance of power, which won't happen in the emerging multi-party environment.

[You kip if you want to...]

It's not impossible that they'll give confidence when it matters, just as long as they get whatever they want, which they probably would if they add up to 326 with the Tories or Labour.

[EPG]

Yes. The more intractable problems are the domestic costs to be paid in such a deal to both the SNP (at Holyrood) and the other coalition party (in England).

[Filuwaúrdjan]

Returning to the real world for a moment, everyone is aware that the SNP cannot be relied on in a parliamentary situation and in all likelihood no one would even try. It's a waste of letters to even speculate.

[andrew_c]

The major parties are going to impose cordon sanitaire on the SNP, especially if the SNP wins a lot of seats.

[ChrisDR68]

Guardian/ICM poll out today:

Labour 32%
Conservatives 31%
UK Independence Party 14%
Liberal Democrats 11%
Greens 6%
SNP / Plaid Cymru 4%
Others 2%

http://www.theguardian.com/politics/2014/nov/10/icm-guardian-poll-support-labour-drops-criticism-ed-miliband

That sort of result in May 2015 could leave both main parties on between 280-290 seats and the Lib Dems on 30-35 seats.

Not enough for the Lib Dems plus one of the others to get to the magic 326 figure needed for a majority in the house of commons.

A three party coalition or a two party coalition plus a confidence and supply agreement with a third looks distinctly possible right now.