Categories
politicalbetting.com » Blog Archive » What went wrong with the polling? More starts to emerge but
politicalbetting.com » Blog Archive » What went wrong with the polling? More starts to emerge but few answers
In this News Statesman article James Morris, from Labour’s US-based pollster GQR, explain how its approach was different and is likely to produce far fewer don’t knows which, it is argued, add to accuracy. He also notes that I was one of those interviewed.
0
This discussion has been closed.
Comments
And the polls were wrong.
As for compouter2, anyone seen him/her recently?
Anyone remember wumper?
And, congratulations Rod, on getting it right.
On the other hand if you accept that the majority of the population doesn't really think about the issues until asked then you ask them to make a choice of support in a void. What the election campaign does is create a context when they are encouraged to reach views which then determine that choice. If the pollster can do that accurately their answers are more likely to be accurate too.
If this analysis is right it is not really a late swing as such, more a crystallisation of the inchoate and underlying thinking as decision time approaches that determines how people vote. To be more specific when asked who they are inclined to vote for they might say Labour because their default position is that they care and are nice. When they consider in a bit more detail who has the correct policies that are going to work in the real world they vote Conservative.
Is it really practical or possible for pollsters to force that choice at an earlier stage? I suspect not. It is the actual decision that is the driver to the choice they ultimately make.
(that's enough sociology for one morning-ed)
So we had Cameron rampaging around the West Country demanding 23 more seats (he got 24, a truly remarkable achievement suggesting incredibly accurate targeting and modelling) and Ed in Warwickshire North desperately and unsuccessfully seeking to bring "soft" target seats on board that according to the polling had already been won. We had Gove confident that the exit poll was about right and Harman not exactly decrying it either.
This suggests to me that the pollsters need to look very carefully at what the major parties did to get more accurate information. If it was canvassing returns etc there is not much that they can do about that because it is polling with sample sizes running into the hundreds of thousands but if the politicians were modelling the population in a different way it has proven to be more accurate, at least on this occasion.
Another useful exercise will be to do comparisons between Lord Ashcroft's constituency polls and the actual results. My perception, without having done this in detail, is that they were incredibly inaccurate but in fairness they were in line with national polling in the main.
We are at risk of reaching the wrong conclusions on small sample sizes here but if anything the forced "constituency" based polling proved to be even less accurate, at least for the Lib Dem constituencies where their supposed local popularity did them almost no good at all whilst Tory first timers, where no incumbency bonus was detected in the polling, did much, much better (sorry Nick).
What needs to be investigated is how all the polls ended up pretty much the same and what the pollsters did .
One thing is certain, good or bad, no pollster should be allowed to change their methodology in the ...say 6 months... before the election.
I just didn't believe the polls.. I doubt many will pay attention in future either and that's bad for the polling industry.
http://www.standard.co.uk/news/politics/greens-would-back-tory-zac-to-be-mayor-over-labour-says-peer-10241968.html
Given the lefty outrage on social media, I envisage shy Tories becoming even shyer next time around. My kids dare not tell anybody that they voted blue.
1970, 1992, 2015. You will note the gap between those is 22-23 years.
Given the Fixed Term Parliaments Act, the next wrong election is due by May 2040.
Does anyone want to bet me 100 UN Credits that the polls will be right then? I look forward to coming back on here with the ghost of OGH and a robot representing Peter Kellner and saying I told you so.
Several possibilities:
The leadership issue - only Ed or Dave would be PM.
The Scottish issue - it did seem to be an issue in several conversations that I had.
Shy Tories- I am not convinced. We did not get evidence of shy kippers or LibDems or of Green "virtue signaling" so why would these factors affect the more mainstream parties?
Late swings in the polling booth. Possible, but only via my other points and if anything the final days trend was the other direction.
It may seem a bit contrarian but I think the polling was actually fairly accurate, but interpretation needs to add the leader approval figures into the mix. It became a PB joke, but when even Labour posters here would post EICIPM, there is a real leadership issue.
On 'shy Tories' - just because people don't feel the need to tell the world and his wife how they will vote does not make them shy. It is for many a private matter. Group affirmations on twitter and Facebook make me cringe with embarrassment.
My hunch? The accurate results for UKIP, the Greens and the Lib Dems were a coincidence. The one thing we saw almost nothing of in the last Parliament was switching between the two main parties. It would be odd if that started on polling day.
It was Nicola wot won it ... for the Tories.
Parties are saying that many business and similar people felt excluded/threatened by EdM/Labour, but do not recall any polls targeting that area specifically?
FPT:
Chris g - late of this parish is worth following on Twitter - a few tweets from the past couple of days:
'The death of Tory Scotland' (sic)
Con vote in Scotland"
2001 360,658
2005 369,388
2010 412,905
2015 434,097
Con vote total:
2001 8.3 M
2005 8.7 M
2010 10.7 M
2015 11.3 M
Lab Vote Total:
2001 10.7 M
2005 9.5 M
2010 8.6 M
2015 9.3 M
I know, we're supposed to look at 'share'.......but almost all those Tory pensioners from 2001 must be at least 80 by now......
1992 25.6%
1997 17.5%
2001 15.6%
2005 15.8%
2010 16.7%
2015 14.9%
2001 Election - Conservative 360,658
Census 2011 - White: Other British 417,109
2010 Election - Conservative 412,905
The only one which still stands is the the PB Tories were and are always right
What you can say is that, in absolute terms, the Tories did better (increasing their number of votes by c. 5% since 2010), but they were overwhelmed by the SNP tsunami. So it's not reasonable to say that a low share of the vote on this occasion was a failure by the Tories.
What will be interesting is whether in 2020 (a) they retain/grow those extra votes and (b) the tsunami retreats at all and to what extent
Labour lost because their geek king had no clothes,
Lib Dems lost because they were seen to be a very slippery bunch..which they were.
Farage lost because his party are seen as bonkers..along with the Greens and assorted others.
..but apparently it is all the fault of the pollsters...Salmond..Sturgeon..the EU..Greece..The San Andreas Fault..the rain in Aberdeen..Kim il Yong..Uncle Tom Cobley and his mates..Libyan Submarines..The price of Lobsters..and the gross stupidity and ignorance of the great unwashed aka The British Voters...dadumdedum
But even the canvassers were reporting huge numbers of people who were genuinely befuddled about how they were going to vote. I suspect the big issue for the polling was how these don't knows were weighted. I think many were not don't know between Labour and Conservative. The Labour vote was as high as it was going. It was a tussle between voting Conservative and LibDem; between Conservative and UKIP; between Conservative and Can't Be Arsed. And perhaps between Labour and Can't Be Arsed.
This isn't being wise after the event. I said this on here, several days before polling day. I said if each of those groups breaks even 50:50 for the Conservatives, they were home and dry.
Plus there was always going to be a reversion to the status quo. The Tories offered Prime Minister Cameron; Labour offered Prime Minister Ed. No contest. The Tories offered a track record on turning the economy round. Labour offered no case for change - and no apology for their previous epically bad mismanagement. In the end, their manufactured outrage over an NHS they had weaponised, their manufactured outrage over wealth (when the highly-mobile top 1% of tax payers already pay 28% of the tax), their manufactured outrage over a manufactured tax - the so-called Bedroom Tax, an extension into the public sector of a principal Labour itself had introduced to the private sector - and their manufactured outrage over food-banks (were Labour REALLY making an issue out of the poor actually being fed??) did not constitute enough of a manufacturing base for people to turn to them in their millions.
Conservative +14.49%
That 96% of Other White British used to vote Con & its now 99%?
What's wrong with "New Liberal Democrats?"
The possibility of Labour getting fewer seats than the Tories and being propped up by the SNP was a major factor in the Tories winning.
Labour need to stop taking their heartlands for granted.
This leaves us late swing or poor sampling (and its corollary, poor weighting).
It looks like we have too many polls with not enough resources allocated to any of them. The internal ones that both main parties did seem to have been a lot more accurate.
I've mentioned this before but one interpretation of an old nursery rhyme is that it refers to the accession of James I (James VI of Scotland) when a few of his hangers-on descended on the English Court looking for favours ...
"Hark hark the dogs do bark
The beggars are coming to town
Some in rags and some in jags
And one in a velvet gown."
Not that Nicola has ever worn a velvet gown.
Spooky.
Bloody NHS! Can't do anything on time!
A majority of 12. That will do them. Oh - and the SNP - let's give them a great big block vote of irrelevance. Tee hee hee....
They somehow co-ordinated that result over 650 constituencies. I mean - how the fuck do they do that?
33% of people intended to vote Conservative, 33% Labour.
The Cons got 24.39% , Labour 20.09%.
So the Conservatives got a 73% turnout and Labour got a 60% turnout. At the end of the day, having Militwump for leader failed to get the Labour vote out whereas the fear of the tartan hordes drove Tories to the polling booth.
Survation: - “there was just a very late swing.”
Hmm, this ‘theory’ I find hard to accept, - IIRC the polls began to clustering together during the final week, with those pollsters showing the highest blue/red leads, both contracting to a median level of ~33/35%? – There was simply a systemic failure by the pollsters to report blue leads in the month prior to the exit poll. – Actually YouGov did in April, but tweaked their methodology and it disappeared.
*titters*
A simple way of checking that is to go and survey now a large sample of the postal voters in a thorough way. If the postal voters (weighted by demographics and past vote etc) were already voting in a way different to the way the demographic polls were predicting then there was no late swing.
Labourlist had a source that indicated some samples of the postal votes were not in line with the polls.
Think they would break up quickly and ditch the lead vocalist.
I don't buy the late swing theory. I don't know about Labour, although the article is interesting, but I certainly think the Conservatives knew the polls were wrong.
In the run up to the election and, indeed, on the day we heard that senior Conservatives were saying that 290 seats was at the top end of their expectations and that the party would have done well if they got close to that. I, along with most other people, looked at that through the prism of the polls and thought they were right - 290 was going to be a spectacularly good result. But in the 40+ years I've been voting I cannot remember either Labour or the Conservatives make an optimistic prediction. I have occasionally known them to not be pessimistic enough but not optimistic. They manage expectations. Which is another way of saying that they give themselves a low target so that they are unlikely to fall short and can claim a win, no matter how poor the result. If the Conservatives were sticking with previous form that means they knew 290 was pretty much nailed on and were expecting to do better. Hence my conclusion.
I may, of course, be completely wrong. Maybe the senior Conservatives talking to the press really did believe that 290 was the most they could expect. And I'm sure they were nervous about it given the stubborn refusal of the polls to move away from a ConLab tie. But I think they were managing expectations. That, after all, is what you expect them to do.
I think very few.
That surely is another important lesson from 2015.
http://www.thewestmorlandgazette.co.uk/news/12933309.UPDATE___There_is_no_question_if_I_felt_the_local_people_didn_t_want_me_to_do_it__I_wouldn_t_do_it____Tim_Farron_on_Lib_Dem_leadership_bid/
I was baffled that it was not shifting the opinion polls. But then, it was driven by the polls. Lord Ashcroft giveth and Lord Ashcroft taketh away.
Especially since results like 285-265 would have been very precarious from a government-forming perspective. Thus if 285 had been seen to be falling short of expectations that would not have helped.
On Panorama last night where J Vine completely lost control, noticed the arrogance of the SNPers who claim they could advise the rest of GB how to be independent and avoid Austerity - but SNP quickly retreat from financial independence - which means they haven't a clue of how to conduct the Scottish economy.
My conclusion
"Either Lord Ashcroft is right, they are a snapshot not a prediction or more likely they were very wrong, just look at the ComRes South West Marginals poll that showed the Lib Dems were getting the dockside hooker treatment in contrast to the Ashcroft polls which should the Lib Dems holding on"
If as I suggested at the time, Q1 would be more accurate than Q2, then Lord A was right.
For all the talk of what happened in England, that is truely labours biggest disaster. I see no reason why they will ever go back to labour.
You can be sure that if by some quirk of electoral fate the SCons had increased their vote share but lost votes, it would still be portrayed as evidence of their wonderfulness.