Over on Labour Uncut Atual Hatwal worries that Labour’s lead may only be 3 points, not 10.

He rightly thinks back to those towering 20 point leads that Labour was enjoying in early 1990, before going on to lose the election in 1992 and warns Labour supporters not to be too complacent – sound advice. However, people can be somewhat more confident in the figures now than in 1990. Two things led to those 20 point leads in early 1990 translating into a Conservative victory in 1992. The first was a genuine swing back to the Conservatives in popular support (largely through removing Mrs Thatcher), the second is that the polls were getting it wrong, and Labour support was probably never that high to begin with. Right up to the election most polls were showing Labour with a small lead, when in reality the Conservatives went on to win with an eight point lead.

In his article Atul ponders whether the polls could be overestimating Labour support again, pointing out that Labour’s lead in the London Assembly elections was not as large as in London polls of Westminster voting intention. The problem with this is that people vote differently in different sorts of elections. This is something I’ve already addressed this year in terms of the Liberal Democrat performance at local elections, which some people thought was a sign that the Lib Dems were actually doing much better than the polls suggested, when in fact it is perfectly normal for the Lib Dems to perform seven percentage points higher at local elections than their national level of support in Westminster polls.

The final YouGov poll of the mayoral election campaign is here. As you can see, London’s Westminster voting intentions in the poll were CON 31%, LAB 49%, LDEM 9%, Others 11%. However, voting intentions for the London Assembly list were somewhat different – they stood at CON 32%, LAB 42%, LDEM 9%, Others 18%. Part of the difference is likelihood to vote (YouGov don’t normally weight by likelihood to vote outside election campaigns), part of it is people simply voting differently – particularly, in the case of the London Assembly list, for minor parties.

Anyway, if you are looking for evidence that the polls may be overestimating Labour’s lead, the only polling figures you can legitimately compare a result to are polls asking about that particular contest. In this instance, the results in the London assembly constituency London Assembly list vote were CON 32%, LAB 41%, LDEM 7%, Others 20%. YouGov’s final poll showed London assembly list intentions of CON 32%, LAB 42%, LDEM 9%, Others 18%. YouGov showed a 10 point lead, compared to the 9 point lead Labour actually acheived.

Needless to say, this does not indicate that the polls are overestimating Labour’s lead.

So, more generally what is Labour’s lead in the polls? Here are what each polling company’s recent polls have been showing

Angus Reid 16
YouGov 12
TNS BMRB 11
Opinium 11
Ipsos MORI 10
ComRes 9
Survation 8
Populus 8
ICM 5

There are several reasons for the variation. Part of it is likelihood to vote – the two companies showing the highest Labour leads, Angus Reid and YouGov, do not weight or filter respondents by their likelihood to vote which tends to reduce Labour’s lead in other companies’ figures. Another factor is that online pollsters are tending to show higher levels of support for UKIP than telephone polls, which in some cases may be impacting on their reported levels of Tory support. There is also how pollsters treat don’t knows – the three pollsters showing the lowest Labour leads all reallocate a proportion of people who say don’t know to the party they voted for last time, which again tends to reduce the level of Labour support. Finally there is the normal random variation within the margin of error that we expect to see in all polls. Taking it all into account though, Labour’s lead does tend to be around about the 10 point mark.

At the last couple of elections the overestimating of Labour support seems to have been solved, however all companies overestimated Lib Dem support at the last election, so it would be hubris to deny the possibility that the polls could get it wrong again in the future. Nevertheless, there is no evidence to suggest it. Atul’s other point – whether a ten point Labour lead now translates into one in three years time, however, is an entirely different question. All polls can do is measure what people think now, not what they’ll think in three years time.

Labour DO have a lead of somewhere in the region of 10 points now. Whether they’ll have one in 2015 is an open question.


59 Responses to “A polling question to which the answer is no”

1 2
  1. Well,I think the polls tonight could be a bit odd.A lot of lack
    of focus,and lets face it,a lot of people disengaging with
    politics.A lot of people away or whatever.This time next
    week perhaps a more valid poll.Does anyone know when
    parliament is ever going to sit again?

  2. This sort of thing drives me crazy: the repeated conjuring up of the 1983 and 1992 examples like they point to some sort of psephological inevitability, whilst wilfully ignoring the events that produced radical shifts in polling: the SDP split off (and the Falklands) in the case of the early eighties, and the removal of Margaret Thatcher in the case of the early nineties.

  3. COLIN………….Elegant ! :-)

  4. We’ll have a 32 / 43 / 8 tonight, or I’m a potato.

  5. Good Evening All.
    i, Wonderful work, Anthony, very useful for my own lessons.

    ii. CRAIG: You mentioned me. Have I said things about the Lib Dem polling figures being high?

    I think we are at a high water mark of Labour leads.

    COLIN: not an infallible statement, there have only been two of them in my neck of the woods. unlike on UKPR

  6. Latest YouGov/The Sun results 6th June CON 34%, LAB 43%, LD 8%, UKIP 7%; APP -36

  7. MICHAEL ELLIOT.
    Nearly a potato then.

    CRAIG.
    The LD figure may be a high outlier, imo/credo.

    In any case UKIP will come home in 2015, I think

  8. Anthony: All polls can do is measure what people think now, not what they’ll think in three years time.
    Not quite, surely, since they are a indicator. which can be validly seen in relation, for example, to the comparative VI of 2010 voters by party support, or demographic trends which will see specific factors in changing age group VI, such as attitudes to job and educational prospects and changing entry and characteristics of the youngest age group, or to pension and working prospects of the growth, changing entry and characteristics of the oldest group.

  9. I think that considering evidence from actual elections constantly points to the phone companies being better than the internet ones and they consistently give a lower Labour lead then we should tend towards the lower end of the range.

    That would only mean I would say the Labour lead was 8% rather than 10%, but still. It’s definitely not 3% and any coalition member (as opposed to Labour person attempting to scare his party out of complacency) who thinks otherwise is deluding themselves.

1 2