Where the polls went wrong

[carlton43]

That is interesting. Thank you. I was in error in assuming a LD melt-down would predominantly aid Labour, so I got the LD seats in right area but really over-estimated Labour gains. I thought the weighting model as applied to UKIP was consistently wrong and underestimated our support and I still do. I think there was a significant swing from UKIP and LD to Conservative very late in final few days (or on the day)because of the SNP perceived threat to English affairs and still do.

[Andrew_S]

I think another major factor was that they overestimated the turnout of young people who are slanted towards Labour.

[Robert]

Matt's analysis is good. There is however a question which seems to have gone unasked let alone unanswered and that is 'why did the parties (particularly Labour and Lib Dem) get it so wrong?'.

There was plenty of information around, other than the polls. Did the 4 million conversations give uniformly the same message, or the door to door canvassing, or telephone polling, or postal vote openings etc. etc..

It is far too easy just to review what the polls did. The party structures have a lot to answer for. Because both parties moved on to leadership elections so quickly, there has been little attention paid to how/why the parties didn't see the result coming.

Just as the polls need a review, if the parties don't know what mistakes they made in 2015, they may make the same mistakes again in 2016, 2017 or 2020.

[Deleted]

I do seem to be alone in suggesting that national opinion polling cannot detect the often considerable power of first time incumbency. The vast majority of Conservative MPs who originally gained their seats in 2010 (and thus were in the front line of defence) increased their majorities, despite the overall swing to Labour of just under half of one per cent. We saw a similar phenomenon in 1987 and in reverse for Labour MPs in 2001 (if you will, in the calm elections that followed more dramatic ones).

That said, I'm not convinced there was a late swing at all and suspect the Conservatives were heading towards an overall majority, or at least approaching one, for about a year or so before the election. In retrospect, the British public were not ready to re-elect a party they had dismissed only five years earlier.

A final point: perhaps national opinion polls have become obsolete, certainly much more obsolete than they were, say 50 years ago when there was much more uniformity in election results.

[J.G.Harston]

Nov 20, 2015 15:25:59 GMT carlton43 said:

That is interesting. Thank you. I was in error in assuming a LD melt-down would predominantly aid Labour, so I got the LD seats in right area but really over-estimated Labour gains.

I kept pointing out to people (with admittedly first-approximation figures) that a LibDem meltdown with voters switching to Labour would result in LibDem seats tumbling to the Conservatives, and quickly give a Conservative majority government. To my own disbelief I even got the actual numbers right, and 7 of the remaining 8 correct.

[carlton43]

Nov 20, 2015 18:03:04 GMT J.G.Harston said:

Nov 20, 2015 15:25:59 GMT carlton43 said:

That is interesting. Thank you. I was in error in assuming a LD melt-down would predominantly aid Labour, so I got the LD seats in right area but really over-estimated Labour gains.

I kept pointing out to people (with admittedly first-approximation figures) that a LibDem meltdown with voters switching to Labour would result in LibDem seats tumbling to the Conservatives, and quickly give a Conservative majority government. To my own disbelief I even got the actual numbers right, and 7 of the remaining 8 correct.

That is most impressive JG and must have resulted in less of a shock for you on the night.

[gwynthegriff]

Nov 20, 2015 16:41:39 GMT Robert said:

Matt's analysis is good. There is however a question which seems to have gone unasked let alone unanswered and that is 'why did the parties (particularly Labour and Lib Dem) get it so wrong?'.

There was plenty of information around, other than the polls. Did the 4 million conversations give uniformly the same message, or the door to door canvassing, or telephone polling, or postal vote openings etc. etc..

It is far too easy just to review what the polls did. The party structures have a lot to answer for. Because both parties moved on to leadership elections so quickly, there has been little attention paid to how/why the parties didn't see the result coming.

My gut feeling from the two LD-held seats I visited (both of which fell to the Conservatives) was that we knew what was coming.

But didn't want to admit it. (And I mean that in two senses - psychologically and practically)

If you anticipate getting smashed you can do one of two things - give up and go home to watch the telly, or fight on.

Fighting on didn't help where I was, but probably saved Southport (for example).

[justin124]

Nov 20, 2015 17:20:21 GMT swanarcadian said:

I do seem to be alone in suggesting that national opinion polling cannot detect the often considerable power of first time incumbency. The vast majority of Conservative MPs who originally gained their seats in 2010 (and thus were in the front line of defence) increased their majorities, despite the overall swing to Labour of just under half of one per cent. We saw a similar phenomenon in 1987 and in reverse for Labour MPs in 2001 (if you will, in the calm elections that followed more dramatic ones).

That said, I'm not convinced there was a late swing at all and suspect the Conservatives were heading towards an overall majority, or at least approaching one, for about a year or so before the election. In retrospect, the British public were not ready to re-elect a party they had dismissed only five years earlier.

A final point: perhaps national opinion polls have become obsolete, certainly much more obsolete than they were, say 50 years ago when there was much more uniformity in election results.

I believe that the incumbency bonus actually goes back quite a long way. In 1964 the Tories - despite a national swing of 3.5% against them - managed to hang on to a fair number of seats gained in 1959 - Uxbridge - Reading - Rugby - Lowestoft - Brierley Hill. Others such as Wellingborough , Meriden and Rochester & Chatham were only narrowly lost. Similarly in 1970 Labour held on to a significant number of seats only gained in 1966 despite an adverse swing of 4.75% - Brentford & Chiswick - Norwood - Rugby - Berwick & East Lothian - Plymouth Sutton - Portsmouth West.- York -Eton & Slough. Again other seats such as Hampstead and Conway were only narrowly lost.

[carlton43]

Are you both suggesting that the actual electors have a different attitude to first incumbency members standing again as against subsequent incumbency members they know better, or yet again to newbies they don't know at all? I can't see that could possibly be true. Why would the electorate react like that? My view is that most of the electorate have only a dim idea of which party holds the seat let alone the name of the incumbent. Are you suggesting that there really are sentient beings reflecting in their heads along the lines of 'Well I would have changed to Labour in most normal circumstances, but our member is a first incumbency Conservative, so I shall give him a bit more of a chance'? I don't think so.

What about considering that a constituency that got a change at the last election by good organization is likely to still have elements of that team and that spirit and an investment in that gain that they fight hard to protect? Whereas after say 4-terms and some increases of majority the party has become more laid back, lost its edge, change of officers, all too easy, gives more mutual aid to other marginals, gets less help from Party HQ, etc?

[justin124]

I think it is perhaps a case of a personal vote for a new member revealing itself after 4 or 5 years. Thereafter, its magnitude remains pretty fixed so that subsequent results for the same seat are likely to reflect the national swing. In the 2015 election the factor was probably more apparent because the Con to Lab swing in England was barely 1%. Had the swing been 3 or 4% far fewer seats would have gone against the trend - though the swing would still have been well below average.

[The Bishop]

Nov 20, 2015 16:31:47 GMT Andrew_S said:

I think another major factor was that they overestimated the turnout of young people who are slanted towards Labour.

They overestimated turnout more generally (70-75% range rather than the actual 66%) and there is little doubt this involved younger voters disproportionately.

[carlton43]

Nov 21, 2015 1:23:27 GMT justin124 said:

I think it is perhaps a case of a personal vote for a new member revealing itself after 4 or 5 years. Thereafter, its magnitude remains pretty fixed so that subsequent results for the same seat are likely to reflect the national swing. In the 2015 election the factor was probably more apparent because the Con to Lab swing in England was barely 1%. Had the swing been 3 or 4% far fewer seats would have gone against the trend - though the swing would still have been well below average.

If you think about that it makes no sense at all. How can a personal vote build up that fast and why? If it does, then surely it would get better and better, term by term, as more electors get helped and come to see him at work? My theory of it being the organizational effort inside the party makes far more sense.

[marksenior]

The personal first time incumbency effect is mostly driven psychologically . I have experienced it at first hand . I voted Labour for the first and only time in 2001 . It was a personal vote for Des Turner In Brighton Kemptown as a personal thank you for the substantial help he had given myself and my wife . Even so it was not easy to go and put a cross on the ballot paper against someone who was not a Lib Dem .
Four years later although my gratitude was still there it would not have been strong enough to override my desire to vote Lib Dem . In fact I did not vote at all in that GE as I had just moved from Brighton to Worthing and it was too inconvenient to go back to Brighton just to vote when with FPTP one vote makes no difference to who wins anyway .

[Deleted]

The other important thing to bear in mind is that any personal vote that a defeated MP had built up by 2010 will have disappeared by 2015 (probably even if they are attempting a comeback), while the new MP is able to build up a personal vote of their own.
We're talking about a relatively small proportion of voters here, but enough to make a crucial difference.

[YL]

The Tories also overperformed in some 2010 gains where the MP elected in 2010 stood down: Cannock Chase, Dudley S, N Warwickshire.

It has occurred to me that, although in practice I voted Lib Dem in 2010 and Labour in 2015, I'd probably have voted Labour in both elections had I been living in a Lab/Con marginal. Now, I don't make any claim to being typical of LD to Lab shifters, but is it possible that there were fewer potential Labour voters among the 2010 Lib Dems in Lab/Con marginals because many of the ones who might have existed tactically voted Labour in 2010 anyway?

Or are LD/Lab swing voters underrepresented in most Lab/Con marginals for demographic reasons?

[Deleted]

Nov 22, 2015 9:49:49 GMT YL said:

The Tories also overperformed in some 2010 gains where the MP elected in 2010 stood down: Cannock Chase, Dudley S, N Warwickshire.

It has occurred to me that, although in practice I voted Lib Dem in 2010 and Labour in 2015, I'd probably have voted Labour in both elections had I been living in a Lab/Con marginal. Now, I don't make any claim to being typical of LD to Lab shifters, but is it possible that there were fewer potential Labour voters among the 2010 Lib Dems in Lab/Con marginals because many of the ones who might have existed tactically voted Labour in 2010 anyway?

Or are LD/Lab swing voters underrepresented in most Lab/Con marginals for demographic reasons?

Using that argument, surely LD/Con switchers would also be more likely to vote C in C/Lab marginals, cancelling out the LibLab switchers.

[YL]

Nov 22, 2015 9:56:45 GMT joe said:

Nov 22, 2015 9:49:49 GMT YL said:

The Tories also overperformed in some 2010 gains where the MP elected in 2010 stood down: Cannock Chase, Dudley S, N Warwickshire.

It has occurred to me that, although in practice I voted Lib Dem in 2010 and Labour in 2015, I'd probably have voted Labour in both elections had I been living in a Lab/Con marginal. Now, I don't make any claim to being typical of LD to Lab shifters, but is it possible that there were fewer potential Labour voters among the 2010 Lib Dems in Lab/Con marginals because many of the ones who might have existed tactically voted Labour in 2010 anyway?

Or are LD/Lab swing voters underrepresented in most Lab/Con marginals for demographic reasons?

Using that argument, surely LD/Con switchers would also be more likely to vote C in C/Lab marginals, cancelling out the LibLab switchers.

Possibly. It depends on the numbers of the two types of potential switchers and the rates of 2010 tactical voting in the two groups, and the latter is very hard to estimate.

[gwynthegriff]

Nov 21, 2015 14:09:46 GMT carlton43 said:

Nov 21, 2015 1:23:27 GMT justin124 said:

I think it is perhaps a case of a personal vote for a new member revealing itself after 4 or 5 years. Thereafter, its magnitude remains pretty fixed so that subsequent results for the same seat are likely to reflect the national swing. In the 2015 election the factor was probably more apparent because the Con to Lab swing in England was barely 1%. Had the swing been 3 or 4% far fewer seats would have gone against the trend - though the swing would still have been well below average.

If you think about that it makes no sense at all. How can a personal vote build up that fast and why? If it does, then surely it would get better and better, term by term, as more electors get helped and come to see him at work? My theory of it being the organizational effort inside the party makes far more sense.

I agree with Carlton

I would point out two things which (generally) happen when a seat changes hands.

1. The new MP receives funding to set up a constituency office, employ staff etc etc

2. The losing party loses the funding to ...... etc etc

[No Offence Alan]

Nov 22, 2015 9:28:39 GMT swanarcadian said:

The other important thing to bear in mind is that any personal vote that a defeated MP had built up by 2010 will have disappeared by 2015 (probably even if they are attempting a comeback), while the new MP is able to build up a personal vote of their own.
We're talking about a relatively small proportion of voters here, but enough to make a crucial difference.

Though the incumbency effect will be reduced by any boundary changes.

[Deleted]

Nov 22, 2015 23:39:54 GMT No Offence Alan said:

Nov 22, 2015 9:28:39 GMT swanarcadian said:

The other important thing to bear in mind is that any personal vote that a defeated MP had built up by 2010 will have disappeared by 2015 (probably even if they are attempting a comeback), while the new MP is able to build up a personal vote of their own.
We're talking about a relatively small proportion of voters here, but enough to make a crucial difference.

Though the incumbency effect will be reduced by any boundary changes.

You're quite right, I am of course speaking of this year's general election.