Why the 5% Lib Dem gap between ICM and YouGov?
January 29th, 2006
..
-
A tale of two polls and two polling methodologies
A new poll by ICM for the Sunday Telegraph this morning should ease some of the jitters within the Lib Dems. After Friday’s YouGov survey putting their share down at 13% the ICM numbers out today are CON 37: LAB 36: LD 18. The only change on Tuesday’s Guardian ICM survey is a one point drop for the Lib Dems.
The timing of the polls meant that all the interviewing by ICM took place after the Simon Hughes admission in the Sun while only about a third of the YouGov respondees would have aware of that development.
So how is it possible that the two pollsters most watched by political gamblers should be giving such a different view of the Liberal Democrat position? This gap does not make sense. I think it can be put down to three key factors.
The weightings that are used. In order to ensure that their samples are representative ICM asks how people voted last time and then builds in an allowance to deal with “misremembering”. In the last poll for which there is data ICM’s past vote weighting was; LAB 38.5%: CON 32.2% LD 21.9%. YouGov uses what it describes as a “political party identifier” to weight its samples on the following basis: LAB 47.1%: CON 34.7%: LD 13.9%. By comparison the May 2005 General Election result was LAB 36.2%: CON 33.2%: LD 22.7%. Also YouGov does not weight by the likelihood that people will vote.
“Spiral of silence” adjustment. A top-line adjustment is made by ICM to deal with those who say they do not know. The pollster assumes that 50% of don’t knows vote the same way they did last time. In last week’s Guardian poll this adjustment added one point to the Lib Dem share. This correction was originally introduced to deal with “shy Tories” who, it was thought, were reluctant to admit their allegiance to an interviewer.
The actual question that is asked. The YouGov internet panel members is presented with the following on his/her PC screen: “If there were a general election tomorrow, which party would you vote for? Conservative, Labour, Liberal Democrat, Scottish Nationalist/Plaid Cymru, some other party, would not vote, don’t know”. ICM has a much tougher approach that seeks to weed out non-voters by stating initially: “Some people have said they would not vote in a new General Election, while others would go and vote at their polling station. I would like to know how certain it is that you would actually go and vote in a general election?” Then the interviewer asks: “The Conservatives, Labour, the Liberal Democrats and other parties would fight a new election in your area. If there were a General Election tomorrow which party do you think you would vote for?”
Each of the three factors would, I suggest, explain the differences between the two pollsters. I have never been able to understand YouGov’s political identifier weightings and if Peter Kellner is reading this it would be great if he could add a comment. So which pollster is right? Who knows but what seems to be happening at the moment is that many more Lib Dems are now saying that the do not know.
The reasons I press the pollsters for their detailed data is that you cannot judge a survey by its headline figures alone.
Mike Smithson
Mike Smithson
MessageSpace Advertising
A fourth reason would be that one or other of the polls is a statistical outlier.
Whilst your last sentence, Mike, is clearly a truism - if political pollsters had honest clients, the British Polling Council would be unnecessary - it is also the case that it is precisely that dishonesty that creates profitable betting opportunities!
O/T, some wise words on politicians and alcohol from Shirley Williams (whom God preserve) here: http://news.bbc.co.uk/1/hi/uk_politics/4659128.stm
And to all of you who are thinking of standing in an election this year or later on, be guided by Nelson Algren: Never play cards with any man named “Doc.” Never eat at any place called “Mom’s.” And never, never, no matter what else you do in your whole life, never sleep with anyone whose troubles are worse than your own.
The first ICM question is an interesting one - but it is not well worded.
“Some people have said they would not vote in a new General Election, WHILE OTHERS WOULD GO AND VOTE AT THEIR POLLING STATION.”
In this part of West London one in five voters do not now go to a polling station but vote by post. There is also a very big differential between the two main parties (over half the postals are Conservative voters).
Perhaps they should look at this question again.
An alternative explanation is that pollsters such as ICM have become so good at fine tuning their data in a remarkably stable political climate that they have been unable to adapt to significant movements in the “tectonic plates”. You have already highlighted, Mike, the large increases in the underlying Tory figure, which is to some extent being “adjusted away” (almost as if there is an inherent assumption that Cameron is experiencing a temporary bounce), and now what is happening with the Libdems is perhaps even more revealing. If Libdem support is down sharply in favour of “don’t knows” then we are entering uncharted territory in making assumptions that they will come back.
Headline poll figures only serve to miss the true story anyway
4. The part of today’s Telegraph coverage which caught my eyes was Michael Brown talking about being seduced by a whip and having oral sex in Douglas Hogg’s Commons ministerial office with a lobby journalist.
His conclusion that when political sex scandals will end, UK parliamentary democracy dies is a but too much!
6. bit, not but in the lasr line
He’s quite right Andrea. If MPs had no skeletons to hide then the Government would never be able to get its legislation through the Commons.
I’m not sure if that’s what he meant though!
7. Alex, yes
After the whip tried to seduce him, he became a loyalist. Do you think Hilary Amrstrong should seduce Jeremy Corbyn? 
In fact, if you read Giles Brandreth’s book, you get the impression that the only time the Tory whips get worried is when they don’t have something on an MP. And I’m sure they haven’t been averse to going out of their way to rectify the situation. I don’t think we should draw any inferences about the whip in question
It is not surprising that it is much harder for Labour to control its backbenchers. Especially the rich ones.
They key man at the next GE will not be DC, or anyone in the LDs. It will be GB.
There has been much analysis—here and elsewhere—on his record as Chancellor. Very little on his political abilty. He conceded the leadership of his party to a much better polician.
TB has always been able to talk complete bollocks, and everybody likes him. Remember cool Brittania? When GB tries to tap into the same vein–a ‘Britishness day’–it was more thought through, and better constructed. But it struck the wrong note. Alistair Campbell could have told GB that for nothing, in 5 seconds. But GB has no AC. Even when TB was at the peak of considerable powers, he worked hard at making sure he struck the right tone. TB could be told when he was wrong, listen, and act on it. Who tells GB when he’s wrong?
Look at GB when he is on an overseas tour, in an open-necked shirt talking to African children. He is stiff and unnatural. Listen to the interview. His tone is the same as if he were explaining tax-credits to a dim sixth-form.
GB’s supporters will welcome complacency, and point to his record. On Europe in particular, his views are more mainstream than TB’s. You could certainly argue that GB’s legacy will be seen to be more significant than TB’s.
GB is respected rather than liked even within his own party. Can such a cold man win a GE, in this televisual age?
Only if DC seriously cocks it up.
Local elections this week pointed more to YouGov than ICM.
The leadership web site polls still have the same trend. Will give figures tomorrow after todays live TV cross examinations.
Though the D’ancona article was interesting only a Tory supporter could agree with a word he says (perhaps agreeing with his thoughts would be a good way to flush out shy Tories?). For instance;
“…….only three weeks after his resignation, 39 per cent of Lib Dems and 37 per cent of all voters already want him back. It is a measure of the panic ……..”
If the same question had been asked of Tory voters after the dumping of Thatch I guess 80% of Tory voters would have been against. Even IDS would probably have got Charlies 39%!
But without being too churlish he can be witty. I loved this;
“….It was surely Mr Oaten’s great misfortune to have picked the only rent boy in Britain who was familiar with the Liberal Democrat front bench……”
At first sight the gap between the Yougov and ICM polls is odd and does need some explaining. When I was a pollster (for Harris, 1986-94) the Alliance/SLD/LD figures were usually much more stable between the survey companies than those for the other parties, partly because the ‘Liberal’ vote is spread more evenly, among classes for example, which meant that sampling variations were less important.
However, it should always be remembered that each figure in a poll is subject to a statistical error based on sample size (as well as any other sources of ‘error’ in the case of a poll which is not truly random, which none of them are). Both 13% and 18% are perfctly consonant with a figure around mid way between. In addition either might be a statistcial rogue - about 1 in 20 polls should be. All this shows why it is foolish to place weight on any one or two sets of figures, and why the much-maligned ‘plethora’ of polls is a good thing.
However, I will throw in one other idea. I have always felt that the kind of respondents who are included in on-line polls may be different from the general public for one specific reason. From whatever age or class they are, they will be people who are hooked up regularly to the internet: not just possessing internet access, but using it on a very frequent basis (otherwise they would not be able to respond to these surveys, one of whose selling points is rapid turn-round). Perhaps such folk are more aware of the latest news, e.g. the Hughes “revelations” and the Lib Dems’ strife in general, which frankly will largely have passed by more ‘normal’ voters. We should always remember that most people have much better things to do than pick over the bones of topical political matters in the way that those who use this site do!
Generally, voters will have noticed that the LDs are in the news, which usually benefits a party in poll ratings, and that they seem to be having some troubles and divisions, which counteracts that effect. If I had to make a guess, then, I’d say that the next set of polls would show figures closer to ICM’s latest than Yougov’s. But unless we’re betting on that itself - wait and see!
Aren’t we missing the most striking thing about the 8 months since the election? We’ve had a new Tory leader with an orgy of favourable coverage, we’ve had senior LibDems convulsed by controversy, we’ve had huge rows over pensions, education and several smaller issues like smoking, and…well, the Tories are up a few per cent, the LibDems down a few per cent, and er that’s it. In particular, the two ICM polls in the last week (always risky to compare polls from rival institutes, for the reasons Mike sets out) suggest that the entire controversy has shifted opinion by 1%.
This is not entirely normal for British politics. It’s common for parties to open up big leads which then fall back and for polls to oscillate wildly. My sense is that most people are largely disengaged at the moment. They’re bemused by the LibDem stuff and, as Mike says, it’s increasing the number who normally vote LD and say they’re not sure, but that looks pretty transient to me. Many are a bit annoyed with the government for one reason or another without really hating them, and they still think they’re reasonably good at the economy. Many think Cameron sounds quite a nice chap who might be changing the Tories positively. But they feel it’s all a bit peripheral (exemplified by the fact that over 50% have no opinion on Cameron at all) and mostly when asked how they’ll vote they say the same as last time. My largest postbag at the moment is from cyclists who want changes to speed and insurance regulations.
So what? So all parties have plenty to play for, and the LDs in particular shouldn’t worry that they’re melting down.
On the reshuffle, by the way, it’s generally thought that it will be very small, with a larger one possible later in the year. The stuff about it being held back as a weapon in the education negotiations is correspondingly wide of the mark, since few expect to be affected either way this time round.
“…….the only time the Tory whips get worried is when they don’t have something on an MP. And I’m sure they haven’t been averse to going out of their way to rectify the situation”.
Alex, was this in answer to Andrea’s post about the whip having oral sex with Michael Brown in Douglas hogg’s office? Sounds too similar to the well known ‘producer-director joke’ to be believable
12 - I think it’s a pity that we don’t still get Harris polls. Am I right in recalling they were pretty accurate in 1992 - and at one point gave an 8% Tory lead, when all the others put Labour ahead?
13 - You’ve misread Andrea’s post.
I love the story in the Independent today about poor whip Andrew Stunnel- every Lib Dem MP is disclosing to him their most intimate secrets that could be revealed- they said on the paper review yesterday one was terrified about discussions of gay relationships, some about finance and some about marriage problems. I am sure John Hemming may have taken some time to get through his list!
Very interesting post Robert and an excellent point about the quick turnaround of the internet polls. By far the most persuasive reason for some of the differences between the internet polls and the rest
Both 13% and 18% are perfctly consonant with a figure around mid way between
That’s not really true, is it? My (limited) understanding was that for any individual poll, the most likely figure was the result achieved, with the probability of the true level of support being some other % declining the further you moved from the given figure. Until you reach the point where the poll becomes a rogue and doesn’t tell you anything.
ie. if a poll gives a figure of 35% the most likely level of support is just that - 35%. 34% is less likely than 35% and 33% is less likely than 34%. The point about rogues (ie. this example if the true level was greater than 38% or less than 32%) is that having a poll of around 3-4% from the true level, is about as likely as having a poll of 10% from the true level.
So summarising, if one poll shows 13% and one shows 18% it is very likely (assuming the same methodology, which is not true in this case) that one is a rogue.
Apologies if that made no sense at all. Or is just wrong (which wouldn’t be surprising :()
13. Roger, it’s a journalist who had oral sex with Michael Brown in Hogg’s office. Then if you don’t believe it, write Michael Brown and tell him you think he’s a liar.
17 - an alternative explanation in the internet pollsters favour is that internet polls will pick up trends much quicker than others.
The idea that internet polls are inherently more unstable is undermined somewhat by YouGov’s record over the past 3 years.
17, thanks, Roger, it’s just an idea …
14, SBS: yes, Harris has a series of polls for the Daily Express for which we did fieldwork at weekends rather than midweek: they all showed a Tory lead. I clearly remember being accused of massaging the figures to fit the client - which no reputable pollster ever does! - and there was such a fuss about the difference from what most other polls we saying that we were persuaded to change to all-midweek for our final polls .. and ultimately suffered the same fate as everyone else.
The final figures in Harris-ITN exit poll (5% lead and no adjustment for ’shy Tories’) also showed that a Conservative overall majority a very plausible result, but by then this seemed so unlikely that the story placed around it was that of ‘hung parliament’.
I myself had previously placed bets on John Major’s return (though I refrained from betting during the campaign for ethical reasons)so at least there was some financial compensation for the ‘Waterloo of the polls’ !
12 - Robert an interesting point - but how do you then explain YouGov’s consistency prior to the General Election (unlike every other pollster), its accuracy in the run up to the GE and its accuracy in for example the Conseravtive leadership election (sampling an electorate whom everyone seems to claim is geriatric and therefore less likely to be online constantly)?
See UKIP has performed a U-turn with Nigel Farage doing his best for European relations - see NOTW.
“The idea that internet polls are inherently more unstable is undermined somewhat by YouGov’s record over the past 3 years”.
Though they are consistant is there any evidence that they don’t react more quickly to a major scandal than the other pollsters? And anyway I always suspect that Lib Dem votes tend to alternate with ‘don’t knows’ rather than any other party so if there is volatility their vote would be the first to be affected.
Andrea. Good heavens no! Nothing Tory MP’s do in the privacy of a whips office surprises me!
18 - perfect sense! Also bear in mind ‘3% MOE’ only applies at a figure of 50%, it would be much lower below 20%. So the two polls aren’t within two margins of error even if as alex says that was a likely explanation anyway. Rogue poll or flawed methodology by one or the other very likely. Not sure which one though - I’ll go for ICM being deadly accurate, but may Be a bit biased!
25 - can’t they both be slightly rogue?
18, Alex, you do indeed have a point - but I was saying that I would still counsel against treating polls as if their figures are exact. 15 or 16% are still ‘consonant’ with these results.
And 20 - yes, quick turnround polls do have that advantage, but I thought in this case that the ICM fieldwork was the later?
22, Rik: yes, Yougov have a good record in practice whatever my theoretical sampling doubts. As I said, just an idea. I would say re the C leadership election though I wonder if the sample response may have been less likely to vary according to exactly what kind of person was easily contactable.
24 - Reacting quickly to major scandals is not incompatible with being consistent. In fact i specifically made the point that they do react quickly to changing events - and are therefore the first to pick up significant changes in public mood. What they don’t seem to do is have as much “natural variation” as the other pollsters, which might suggest that changes in support detected by them should be treated far more seriously.
21. Very interesting about getting these starkly different results by weekend polling. Presumably the point here is that at the weekend, your sample bias might be lower - good chance of getting more people who work for a living for example. Do any of the current telephone pollsters poll at the weekend, and if so is there any evidence of different results from midweek polls?
On another subject - any markets open yet on the future of Britain’s (or the Lib Dems’) most popular policeman, ‘Sir’ Ian Blair? I reckon with his sensitive speech about the Soham girls plus today’s intriguing cover-up story in the indy, early retirement beckons. A labour peerage, too - if he can avoid total disgrace and keep the knighthood he most certainly does not deserve.
25,27 - That’s a relief! That means i can continue to protest whenever people attempt to explain away polls they don’t like by talking about “margins of error” and “statistical ties”
29 - Weekends are good, bank holiday weekends are bad! (Just ask Populus ;))
As I have continually said, the points Robert makes support the contention that we should never take a single poll as being triumph or disaster. Several polls together over a cycle, or the trends of a single pollster across several polls are far more revealing.
Thus I would suggest that the position is:
Cons 36-39%
Lab 36-40%
LD 13-19%
The trends are:
Cons Drifting up slightly
Lab Stable
LD
Ooops hit wrong button!
The trends are:
Cons Drifting up slightly
Lab Stable
LD Drifting down
32. Rik. LD what?
OVerall therefore the Tories and Labour are neck and neck in the upper 30’s and the Lib Dems are trending down to the mid teens at the moment.
Cameron’s challenge is to get us over the 40% hurdle and clearly ahead of Labour!
29 - I’m not sure you’ll find many Lib Dem fans of Ian Blair, effectively after his intercessions on behalf of the Labour whips in the vote on 90-day detention.
36 - I thought fred might have been referring to his (in vain) efforts to keep the LibDems off the front pages.
37. Haha actually I hadn’t thought of that….! Book Value have a look back a few months ago and you will find one or two of your yellow peril brethren (Tabman in particular)hotly defending ‘Sir’ Ian against my last assault. He may have change his mind in the meantime. I certainly hope so.
fred @ 29: ICM and Populus’s monthly polls are both carried out at the weekend. I’m not sure there is any particular difference - after all, this ICM poll (conducted midweek) and the last ICM poll (conducted over a weekend) are almost identical.
Rik at 32. I thought you were being subtle and asking us to fill in the blank!
Just saw Huhne on ITV. Most fluent speaker of the three candidates by far. Needs to have a bit more humour. What he says is also very sensible and plausible. Absolutely no rhetoric like “great tradition of liberalism / radicalism” “we are at our best when we at at our boldest” that you hear all the time from the other Lib Dems. Good! Let’s get some PR gurus in to help him project his personality more. Didn’t Mark Oaten used to work in PR?
39. Thanks Anthony. So no reason then to think that the weekend polling effect Robert noted earlier has any significance now. I wonder what has changed to make this so? Are the pollsters just better at ‘adjusting’ their samples?
Rik’s blank at 32. could be filled in ‘RIP’
o/t I am sure Andrea will be interested to know that Adam Rickett is scheduled to be on Question Time this week, according to today’s Telegraph.
41. yup, he was group director and Managing Director of Westminster Public Relations.
43. Andy, I’ll try my best to see him next thursday, but I would stress that I’m not a AR’s fan!
44 - Not Westminster council Public Relations?
I was campaigning for the Lib Dems in Dunfermline yesterday. We’ve been pleasantly surpised at how little negative feedback we’ve received over the leadership “issues”. Local issues - Lexwell; Forth Bridge tolls; 2nd Forth Bridge; state of Dunfermline town centre etc. - will decide the result.
30 - no, Alex, as someone who made my living by conducting survey research for a decade I prefer the approach adopted at post 32 above.
Any one poll has so many sources of variation, or ‘error’, not just that based on sampling size (and remember’margin of error’ is a term which relates to statistically random surveys, which are not done and cannot be done)that to look at or extrapolate from differences of one or two points in single polls really is misleading.
Polling methodology has indeed improved vastly since my day; but remember that even now even that margin of error on a single poll does not really apply; and also you may be talking to the wrong people, asking the wrong questions, or making the wrong adjustments.
Of course those involved in polling will want to stress the unique value of each poll. Partisan observers will want to spin polls seen as favourable and decry those which are not.
Taken together, polls are very useful indicators of opinion - and please do not forget the other questions, often much more interesting and revealing than hypothetical voting intentions.
re Anthony 39. I seem to recall that at the General Election there was concern about distorted samples in the surveys carried out during the May bank holiday weekend. More Tories than Labour people, it was suggested, were away on short breaks thus producing inflated Labour shares. In a cold January everybody stays at home.
47 - yes, sorry Robert. I didn’t mean to imply that one poll should ever be taken as gospel.
47 The polls ought to be able to produce a confidence interval such that 95% of the time the “true” percentage support lies inside this interval. Assuming a totally random sample size of about 1000 a 4%-wide confidence interval is possible. By adding in further variables and calculating the variables, theoretically a narrower bound would be possible, but due to modelling issues +/- 2% should be as accurate as any pollster claims. On this basis, a 5% difference between the polls suggests at least one is rogue, or at least one has an unusually flawed methodology.
Hughes has now moved into the lead in terms of supporters listed on each candidate’s website:
Hughes 953
Campbell 782
Huhne 420
I know this is self-selecting but all candidates are asking people to declare their support so I would have thought this must mean something.
Hughes and Campbell are fairly close but Huhne is well behind. Surely this would suggest Hughes would be a good bet at his current odds of 8.0 on Betfair (ie 7-1).
I can’t remember the exact numbers but in the Tory contest Cameron had about twice as many supporters as Davis had listed on their websites.
http://news.bbc.co.uk/1/hi/uk_politics/4659596.stm
Do we believe David Blunkett?
52. I suppose they mean 100 Labour MPs, not 10 not being happy about the Education White Paper.
With regard to internet polling, do the pollsters attempt to make any adjustment for the following possible factors:-
The internet is used more by the under 50’s than the over 50’s - probably working against the reported level of support for the Tories
Computers are still nowhere nearly in every home - probably working against the level of Labour support
Those who are unemployed or retired have more time to reply to pollsters on the internet - net impact: ???
If the “Shy Tory” syndrome still exists, such respondents are more likely to admit their true allegiance on the internet than face to face or on a telephone line - therby increasing their party’s poll rating.
54 - they should weight the sample so it matches the electorate in distribution of ages. However you still have the question of whether over-50s with internet access are representative of over-50s in general.
In addition they profile, and presumably weight, on things like newspaper readership. And you have to say that whether or not it should work for YouGov, the results are generally pretty good.
Bank holidays are notoriously tricky. The reason it damaged Populus’s tracker poll so severely was because it was also a one day snapshot poll. A normal 3-day poll has problems over a bank holiday weekend, but you do at least have 3 days to get hold of people. No pollster in their right minds would recommend a 1 day snapshot poll taken on a bank holiday to a client, unfortunately because Populus was doing a daily tracker poll they couldn’t avoid the bank holiday.
I think the best way to establish voting trends is to compare polls by the same pollster.
E.g. Ipsos/MORI
November Con 32 Lab 42 LD 19
December Con 40 Lab 31 LD 21
January Con 39 Lab 39 LD 15.
I think this clearly shows that Cameron’s election produced a Lab to Con swing and the LD’s leadership problems produced a LD to Lab swing.
46. sp. Lexwell/Lexmark
51 - yes, numbers of supporters is interesting. I wonder if the ugly media treatment of Simon Hughes as sparked a flurry of sympathetic supporters to him? Maybe it would have happened anyway - he’s still very popular among a range of acticists.
I think the LD leadership betting markets are probably pretty sound. Here’s why:
Huhne is almost certain to come 3rd, albeit with a pretty respectable showing. Most of Huhne’s 2nd preferences will move to Ming. Which means Ming wins.
However, if (and I think it’s a huge if) Huhne were to pip Hughes, it’s anyone’s guess where Hughes’s 2nd preferences will stack up. Probably Ming would still win - name recognition will count for a lot - but it’s not certain.
In short, it’s very hard to see Hughes winning. It is just about possible to imagine Huhne squeezing through. But by far the most likely result is a convincing Ming win. Which is more or less what Betfair is telling us.
(FWIW, I think Ming’s campaign should try a variant on the ‘Cheer Churchill, vote Labour’ slogan to Huhne supporters, many of whom want their guy to do well, but not win.)
55 - Of the relatively few over 70’s I know, but of whom there are millions in this country, very very few own a computer or have even ever used one - surely this could lead to a significant discrepancy if this sector’s opinions are effectively being ignored by the internet pollsters?
Any movement on the Blair departure date odds after Blunkett’s comments?
As BV says, all the things Peter raises can and are weighted for (specifically age, work status, social class & political ID). The potential problem is if internet access itself changes attitudes or reflects different attitudes (if a Labour voting 50 year retired man from Gateshead with 1 car who takes 1 foreign holiday a year with internet access has the same views as a Labour voting 50 year retired man from Gateshead with 1 car who takes 1 foreign holiday a year without internet access). Obviously in some specific cases they don’t (attitudes to internet usage are almost certainly different!) but given YouGov’s accuracy to date their views on politics seem to be the same.
59 If a section of society is underepresented then the views of those in the sector expressed can be used, however also on categories such as age extrapolation is valid. So if 60 year olds are on average more Labourite than 50 year olds and 40 years less so again, then you draw a regression line to express Labour tendencies, check it against what data you have collected and then make a best guess.
Peter - it would, if they were, which they aren’t. A panel just drawn randomly from the internet would hideously under-represent the elderly. YouGov’s panel though is pro-actively recruited to include groups under-represented on the internet so does include enough elderly people.
Re 51 and 58. That is not what the web site polls are saying.
57 - Alan J - I disagree. Some pollsters can fluctuate wildly and MORI seems to be guilty of that in the last year or so. I always remember that this seemed to be a characteristic of Gallup when they used to poll for the D Tel.
The only reliable way of looking at polls IMHO is to take the full cycle of each pollster and take an average.
Having said that where you have a pollster that has proven over time not to be erratic and to be on the money with results you could (with caveats) draw some limited conclusions from its findings (I am of course thinking of YouGov at present - although my fave used to be ICM). However all pollsters can have the occasion rogue poll - and MORI seems more prone to that than most!
I will always remember Bob WOrcester declaring the last US election for Gore just before it became clear Bush had won!
62 - Reading Mike’s opening commentary and subsequent related posts here, some of the pollsters appear to be bending over backwards to find some reason to discount the apparent Tory revival,in the belief that this will inevitably reverse once the DC honeymoon period is over - I believe this could be a dangerous strategy since every one,I recall,overstated Labour/understated Tory support at the last GE by 2-3 % points or more, which they have not yet adequately explained.If they are now compounding this error factor,then their findings could be some considerable way off the mark.
“I will always remember Bob WOrcester declaring the last US election for Gore just before it became clear Bush had won!”
… Or indeed for Kerry.
66 I agree the random discounting of Tory votes or creating of them under the shy Tory hypothesis doesn’t seem very scientific. I’m doing a degree in statistics and I’ll admit it has its artistic side, but some pollsters need to check their methodologies.
50, Lisa, margins of error: you say “assuming a totally random sample” - but no poll ever has anything like a random sample (i.e. a controlled and achieved sampling interval of the universe of voters).
People are not available, some refuse to answer; some may not tell the truth (whatever that is). Obviously surveys which only include those with internet access are not random, but that does not make them particularly un-random. Surveys are very often subject to quotas, which = not random. Also you have question wording error, then all the adjustments e.g. by voting recall which may or may not (usually may)improve accuracy of results for prediction purposes.
All these other sources of variation or ‘error’ are unquantifiable in statistical terms. Sample size itself is only validly subject to margins of error in random surveys. My experience was that sample size was hardly ever the source of problems or variations between polls.
I was never happy ‘claiming’ any margin of error - it is at best a minimum figure, but clients wanted it.
There are all kinds of things you can do with polls, and sequences with the same methodology help, for example, but I always myself believed that the plus or minus three per cent etc is based on a series of fallacies.
People are not beans in a barrel and polling just cannot be robust enough to sustain such theoretical claims. In practice we can see that often it does provide effective results.
67 - Indeed - there is one joy about this site and that is that any error in debate WILL be corrected
69 “Sample size itself is only validly subject to margins of error in random surveys.”
What on Earth do you mean? The pollsters managed to collect a sample size of 1026 with a 5% margin for error?
FYI a “random” sample is not weighted, a weighted sample should provide more accurate data.
Also, the variation of error attributed to “other factors” is quantifiable. Compare with election results and you can see that it is around 2-3% on current methodologies. In addition the margin of error given out by pollsters should not be a minimumu amount that they think they are out, but an indication of the confidence interval i.e. a maximum predicted error.
65. But as Gordon Brown would say, Rik, it all depends on when you start the cycle!
69 Sorry to reply again but…
“I always myself believed that the plus or minus three per cent etc is based on a series of fallacies.”
Have you heard of the Central limit Theorem which inform use that the distribution of the maximum likelihood estimate is approximately normally distributed mean the parameter, variance Fischer’s Information inversed for large sample size? The +/- 2/3% on a sample of 1000 from the UK population is based on this theorem. All adjustments made by the pollsters to the raw data should be reducing the chance of error, not increasing it. Hence the stated MoEs.
It looks like Huhne is putting the boot into Hughes now:
http://www.epolitix.com/EN/News/200601/15920184-ca34-46df-95cc-6d778c94d330.htm
He really is sticking the knife into Hughes and reminding everyone about Oaten. What a git!
74. Has he already produced a “2 horses race” barchart?
64. Fair enough - can you direct us to these website polls?
74 - Hmm - ‘Huhne winning here’. He doesn’t deserve to get anywhere near the leadership, I hope that members treat him accordingly.
52 Anna “Do we believe David Blunkett?”
Is the Pope a Protestant?
75 One of Huhne’s skeletons made then student newspaper and is in the MoS today…
http://www.cherwell.org/ousu_worth_fighting_for
I suggest you all await the web site figures!!
79. Bit like the suffragettes really then? - seems Huhne & co were pushing at a partly-open door.
74 - he has to get into the top two. This is politics.
81. zebidee, I’ve just imagined Huhne dressed as a suffragette. Not a very nice picture!
80 - what “website figures”?
A bit of student radicalism getting into the papers should be good to improve his profile.
84 - Probably Huhne supporters rigging an internet ‘voodoo poll’.
79 - the MoS is slow then, this was well known a couple of weeks ago.
73 - All adjustments made by the pollsters to the raw data should be reducing the chance of error, not increasing it. The adjustments are corrections for inaccuracy, not imprecision. The MoE as determined the way you describe is a fair measure of precision (i.e. the variance you would get if you repeated the survey under identical conditions). But it is not a fair measure of accuracy, i.e. how well the poll serves its stated purpose of predicting population-level voting intentions. The accuracy is affected by non-randomness in the sample, and this is what the adjustments aim to counteract.
87 My point being that having shifted the results up down and around the houses by a methodology that accounts for modelling inaccuracies, the raw underlying data has 95% confidence intervals of 5% and these should not increase under transitional shifts these significantly.
88 apologies don’t know why I put a second “these” in the second line.
88 - sure. I think Robert W’s point was that the layperson looking at a poll tends to see the MoE as a measure of accuracy - which it isn’t, even if the way it is calculated as a measure of precision is perfectly sound.
Being a Yougov panel member myself, my theory is that because panel members are so regularly ask for their opinions, they are more likely to give an opinion based on recent events than just on long-term beliefs. That is, it’s a kind of “protest polling”.
rE 86. One of the sites is a declared Campbell supporter.
91 - Again, Adrian, that is not borne out by the evidence of remarkable consistency in the year leading up to the election. If anything the fact that the contributors have the time to consider their answers carefully, without the pressure of blurting out the first thing that comes into their head (I did a phone poll recently, and my answers were probably consistent with a regular problem on the telephone that i find it difficult to think when on it) could lead them to filter out current events. If they have done polls before they know they will probably be asked their views on those later on.
80, 84, 86. Surely the official candidates website lists of supporters are more credible.
If hypothetically, Huhne has more supporters than Hughes, why have only 420 Huhne supporters registered whilst 953 Hughes supporters have registered?
It could be argued that Hughes has fewer, more enthusiastic supporters. But there is a huge difference between 953 and 420.
All I can say is that as more votes have gone on the web sites the same pattern/trend has been steady since Wednesday/Thursday.
I think it is worth recalling that there has in the past been a difference, probably in all parties, between what activists may feel and vote and what ordinary members do. At present it is probably the activists who are speaking. No one can predict until the non activist members have considered. In 1976 history says the activists wanted Pardoe but he was heavily beaten by Steele.
The Labour activists I knew in 1994 were at the time declaring for Prescott not Blair, look what happened there. People will make up the own minds individually and it is most unlikely they will log onto political betting , Betfair, look at any of the candidates web sites or any voting web sites.
Huhne is proving to be far more resilient that anybody had thought. It makes you wonder what Clegg could have achieved.
Has Clegg actually clarified whether he didn’t stand for the leadership because he was the wrong man for leader, or because he didn’t think he would win?
Re 96, now there is a question? Another for the history books.
97. He has said he thought the party needed an experienced leader when he came out for Campbell. Whether that was just what he said, or actaually what he believed is unclear.
71, 73 et passim; Anna I do apolgise for calling you Lisa .. that’s someone on another forum I’m looking at (it’s on football, if anyone was wondering!).
Apart from that, I’ll leave you to those who know what I’m talking about, as Book Value seems to. Suffice it to say that polling samples are not random representative samples of the voting public, no academic has ever before suggested to me that quotas are as accurate from their point of view as pure random, and I don’t think you can use observed empirical results of elections to put a figure on error limits - for a start, the voting process is at a different time from the survey.
Meanwhile, if anyone is still reading, they might find the following useful:
http://en.wikipedia.org/wiki/Category:Lists_of_UK_Parliamentary_constituencies
It contains outline maps of the new constituency boundaries, for example.
I wonder if there will be a falling out between Huhne and Clegg. Remember when Major stepped down saying “put up or shut up”. Portillo dithered and then chose not to run; Redwood decided to run and fell out with Portillo.
Polling Adjustments.
My doctorate in Mathematical Statistics is a little rusty, but I feel that I may still have something to contribute to this debate.
I am surprised that anybody could suggest that the adjustments applied by the polling companies should affect the margin of error of the results.
What the adjustments will affect is the bias in the results. That will in turn make the results more accurate, but it will not affect the randomness in them. Thus the results will still dart around, with occassional outlyers. Hopefully, though, the means of the distributions of the estimators derived from the polls will be centred on the actual means.
N.B. Removing bias from estimators sometimes INCREASES the MoE.
96/98/99 can’t believe we’re condemning a guy for not running for leader of his party after 8 months as an MP! He has a long career ahead of him, why risk doing a Hague and going for the leadership too early? I look forward to seeing both himself and Chris Huhne proving themselves on the front benches in the next months (possibly CH at the very front of course)
Just noticed that the archived poll is the first link I pointed to. The updated poll is here -
http://www.colin-ross.org.uk/polls/2.html
Look at the question again.
In essence, if there was a general election tomorrow, and you voted, who would you vote for.
The Liberal Democrats do not have a leader. Why therefore would you vote Liberal Democrat? (PB.com users may think such a conclusion absurd as you can see the bigger picture, this though is a question of the general populace and I would not be surprised if people thought that way).
Analysis of polls mid leadership contest is crazy, volativity is off the chart as each good or bad news story breaks (in the Lib Dem case, bad plus bad).
If the Lib Dems are down in the low teens come May then they have everything to worry about.
105 - Chris there is some truth in what you say but the polls during the Conservative contest broadly pointed to an improved performance for us. I dont think you can dismiss them completely. None of the three LD candidates are particularly inspiring and I cant see any great bounce coming from any one of them being elected.
“Though the D’ancona article was interesting only a Tory supporter could agree with a word he says ”
This Tory supporter agrees with very little D’Ancona says.
talking to student tories is often great.
on friday night at the student union, i was treated to some fantastic statements.
upon asking them what they thought of mr cameron, the reply i got was ” well we’re not f****** liberal democrats are we”
upon me mocking them about being fellow comrades, “it takes the p*ss, we are to the left of you now, even you dont talk about re distribution” - i dont agree with them btw
So you wish you hadnt voted for him “cameron could only win if the country was f******, and its not (!!!!) so he will lose, he will never be pm, labour will win the next election with a 60 seat majority, people respect brown, cameron is a popinjay”
good to see the tories have conceded 4 years in advance lol
and the absolute killer, do you think you will win the next election- “well, i certainly hope not”
10 - good grief, what sort of students use words like “popinjay”. In fact, what sort of anybody uses words like “popinjay” (Galloway aside)? Are you sure they weren’t taking the rise?
11 - Respect supporting students use the word popinyay of course, especially when referring to the ridiculous figure of Christopher Hitchens.
‘Popinjay’ rather.
110. I’ve heard some more upbeat view than that from student Tories. Yet i would agree they are a little confused about direction of the party (lots of nice things about Cameron’s attitude and image, blank faces about policy). It seems to be more about getting rid of Labour than actually having an agenda they believe in.
Also many of student Tories i have spoken to still wish that Ken Clarke had won.
111. “peasant” is a word i frequently hear from Tory supporting students here in Durham. But then there students here who are bred from good stock, so you would expect such words!
no they were serious, albeit i think one of them was banking on the mosing in 09 so he could elected in the following election.
typical tories hey
Re 106. suprised you have given details, anyone, from anywhere could vote. Probably lost all its future interest.
Figures are:
Huhne 43
Campbell 39
Hughes 18
2nd Poll set up for STV
Huhne 43.7
Campbell 34.1
Hughes 22.1
with Hughes dropping out becomes Huhne 56.9 Campbell 43.1
It maybe all the Huhne people have voted or it may be indicative
of what is going on at the moment. There is a month to go.
110 - Yes, I’ll agree that on some issues Cameron is a bit cuddly, and Libidemmy now. But come election time I fully expect normal dog-whistle service to be resumed - right to shoot burglars in the back, immigration issues, gypsy issues, lock all criminals up, bungs to the middle classes to leave the NHS etc, blame France and Germany for everything. We are already seeing a very illiberal campaign against Cadbury using French sugar run by some daft backbencher.
The trouble is that the Tory instincts are not liberal. If they’re still so nice and touchy feely at the next election, I’ll eat my ballot paper.
AOL Poll
Who would make best Lib Dem leader?
Menzies Campbell 177 13%
Simon Hughes 86 6%
Chris Huhne 79 6%
Charles Kennedy 1015 75%
Total Votes: 1357
17 SBS Why do you have a problem locking up criminals ?
18 Heard the Kennedy rating @ 39% this morning.
Troubled is probably the best word to describe the LDs at present !
“But come election time I fully expect normal dog-whistle service to be resumed - right to shoot burglars in the back, immigration issues, gypsy issues, lock all criminals up, bungs to the middle classes to leave the NHS etc, blame France and Germany for everything. We are already seeing a very illiberal campaign against Cadbury using French sugar run by some daft backbencher. ”
Sounds okay to me, apart from the last issue, which is absurd.
Interesting discussion as ever.
(a) wouldn’t take the lists of supporters to seriously. Simon has had a wave of sympathy sign ups of late and you must not confuse activists with members. If this was an activists election then simon would win and i suspect activists are disprortionately represented among those that sign upto websites.
(b) have had my first two mailings from candidates. bizzarely they aren’t given access to the membership list (though cowley street will relay one email for each of the candidates) so they seem to be targeting councillors at this early stage. ie people they can identify. so far one email from mike storey, former leader of liverpool city council (no tory jibes please! hes enormously respected in the party) backing ming. However i thought it sent a rather confused message. the tone seemed to reassure me that ming understoof the parties grass roots but then banged on about his national standing and gravitas. it all rather seemed to sign post the alledged weakness that it was designed to rebutt.
had a snail mailing from Huhne. two focus style leaflets which to be honest were very good. bit irritated that his wife and 5 children were top of his “factfile” however this more successfully got across his messages. good pictures of him on the anti war march, weapons decommisioning in cambodia say ” I have gravitas” while loads of stuff about his work in his euro constituency say ” I have a grass roots record) all you can really conclude from this is that his campign team know what its doing and has got hard cash in the bank and off to a flying start. one to watch.
119 - No problem with it. Perhaps I could have been clearer…
It’s the tone of the rhetoric though. Not every criminal needs to be banged up in perpetuity. The tradition view for the Tories is prison works. Prison has three functions: firstly, to stop the banged up criminals re-offending while banged up (yes, that works fairly well!); secondly, as a deterrent - may work for some, but to a drug addled thief, I don’t think the fear of being caught enters into their decision on whether or not to commit a crime; thirdly, rehabilitation - utter failure.
Building more prisons as part of the existing system is empty headed, crowd pleasing, right wing, sensationalistic, tabloid propaganda.
119. I take it you are comfortable with the rest of the issues SBS raised there!
And he said “lock all criminals up”, which isnt that great for the taxpayer (not to mention the fact that it doesnt really work).
123 - apologies, if I sounded like a Trot in the last paragraph. I am not. But if there’s one thing that makes me rant like an idiot, it’s politicians who rant like idiots…
If the police were more successful at catching criminals, one would have little alternative but to create more prison places.
If there were fewer laws there would be fewer crimes to commit.
Anyway, the question is has Cameroon really defected to liberal democracy (and will he bring the rest of his party over to us)?
128. I think the question will be more whether he can convince the average voter that he cares about liberal democratic issues. His actual beliefs on liberal issues are, in the short term, irrelevant (unless he angers a large proportion of his support/party by sounding too liberal, but that seems unlikely).
They might become more relevant should he win, and then be shown up not to give a sh*t about liberal issues.
He doesn’t need to attract committed Liberal Democrats anyway. Just people who, having deserted the Tories since 1994, have ended up in the LibDem camp for reasons more due to dissatisfaction with the two main parties than any positive feeling towards Liberal Democratic principles and policy.
23 Seem to recall a certain Mr Oaten lecturing that Prison “never, ever works”
Happy to fight as many elections as you fancy opposing that concept.
Prior to recent LD hiccups, along with Tax ( the gap over which has now blurred significantly between us - (our fault !)) and Europe your ’soft’ policy on law and order was one of the key reasons you were unable to advance further in May 05.
Until you look at these 3 issues (IMHO) you are not likely to make the advances you need (certainly in London & the SE)
Quite happy to reaffirm that I believe if we need more prisons we should build more.
25 Fully agree with your comments about ranters (of all hues !)
Kind regards
24 No, but I’m going through rendition and its painful !!
28 Hmmmm….. “No comment”
Perhaps if Charlie gets himsulf sorted out over the summer, runs a decent time in the Great North Run (on behalf of a suitable medical charity of course), he could offer himself back to the LD front bench?
They’d be glad to have him back.
130,
Lucky ` he doesn`t need to attract committed Liberal Democrats`.
If he did it sure would be a close election, with the dedicated few making all the difference.
However I do think the next British Election, could be a re-run of Gore v Bush.
With the compassionate Conservative coming out on top just, with the help of the Ralph Nader , third party.
Then once elected shoring up the right wing base.
In light of Blunkett’s comments today, will there be a Labour leadership election when Blair resigns?.
I notice the guardian did a very favourable piece on simons campaign launch next to a very undignified picture of ming squashing into a racing car. on the same day the indie did a “saint simon” front page and article. perhaps have helped contribute to the jitters reported intodays observer.
Not sure what Charlie would do. If Ming wins, Clegg will get foreign affairs. If he loses, Ming will stay at foreign affairs. Cable won’t leave the treasury brief (and if he does Laws will get it), but there is a permanent vacancy at home affairs. Simon Hughes may want this, but probably prefers being president. Can’t see Charlie going for this; we would need somebody reliable who would turn up to debates and know what he’s talking about. I was no fan of Oaten, but he was pretty good at his brief, and was making the running. So, I can’t see a swift return… perhaps he could shadow the new chancellor of the duchy of lancaster - when we have one.
130 - Cameron may not need to attracted committed Liberal Democrats, but I feel he does need to attract committed liberal democrats (small “l” small “d”).
sorry, but this EDM is probably one of the most hysteric EDM I read recently:
http://edmi.parliament.uk/EDMi/EDMDetails.aspx?EDMID=29909&SESSION=875
136. “Then once elected shoring up the right wing base. ”
Bush was way down in the polls prior to 9/11. Frankly without that i doubt he would have stood a chance of re-election. It sounds harsh, but it was 9/11 and the popular support for the Afghan and Iraq wars which ensured he got re-elected, not him shoring up support.
138 - I found the Indy piece (front page something like “I just want to be remembered as a human rights activist”) utterly nauseating, self-righteous rubbish.
SBS @128 - I am here to apply balm to your troubled brow - I’m that sort of chap :P. Neither Cameron nor the Conservative Party is, or will ever ‘defect’ to liberal democracy (whatever that is). What will emerge in the weeks, months and years ahead will be rooted firmly in the mainstream progressive Conservative tradition.
You can relax, though if you’re prudent, be just a tad apprehensive about your own party’s fortunes
38. Have you really found the coverage of Hughes affair sympathetic?
I got the opposite impression.
The only person they tried to present as a saint was Peter Tatchell. Even when the Indy had the self written article by Hughes, they had in the same edition the article about Tatchell (presented as a poor saint).
143 - I am partly resassured; this mainstream progressive Conservative tradition is presumably different from the ugly regressive tradition that formed the basis of the Tory campaigns of 2001 and 2005.
145 - Oh, please follow your own admirable counsel, and make an effort to stop ranting