19 Jan 2015

How opinion polls under-estimated Tories and over-estimated Labour

Conservative hopes of a winning the election will be buoyed by analysis from a polling expert which shows that last year opinion polls regularly underestimated the Tory vote, while  over-estimating support for Labour.


The analysis, published today,  examined opinion polls before the three main sets of elections in 2014 – the local and European elections in May, and the five parliamentary by-elections which took place at various times during the year.

The results are striking.  The Conservatives got around 2 per cent more in real elections than the polls suggested they would achieve, yet equally Labour performed around 2 per cent worse in reality than polls forecast.  The Greens also performed better when people actually voted than they had in polls beforehand.

The results were compiled by Rob Hayward, a former Conservative MP who occasionally advises the party on election matters, but who has earned a reputation among journalists as an independent polling expert. Hayward presented his findings to political journalists at Westminster this lunchtime.

Hayward found that in the six polls conducted ahead of May’s European contest, asking how people would vote in that election, five opinion polls understated both the Conservative and Green vote, and five overstated the Labour vote.  The Tories did 2.2 per cent better on average than the polls predicted, and Labour 2 per cent worse.

Ten separate polls were conducted before the five different parliamentary  by-elections.  Eight of these understated actual Conservative results, while two overstated them.  One poll understated a Labour result, and eight overstated Labour results.  Overall the by-election polls understated the real Tory vote by 1.8 per cent, and overstated Labour’s by 3.7 per cent.

Polls conducted before May’s local elections, in contrast, actually overstated the Conservative vote by 2 per cent, but Labour was overstated by a lot more – by 5 per cent.

Taken together, these polls roughly overstate Labour by 2 per cent and understate the Tories by 2 per cent – a net figure of 4 per cent.  The trends suggest there may have been a fundamental flaw in opinion polling in 2014, which consistently suggests Labour will do better than they actually do, and which leads the Tories to perform better than expected.

Similar flaws have been found in polls in the past, but the pollsters seemed to have corrected them.  Hayward’s analysis suggests that the problem may have re-emerged.

Current polls show Labour and Conservatives almost neck-and-neck, with a slight Labour lead since 1 January of around 1 per cent on average.  But if current polls are making the same wrong estimates now, ahead of the 2015 election, as they did before people voted on various occasions in 2014, then there could instead be a real Conservative lead of around 3 per cent right now.

Quite why the polls got their results wrong in a fairly consistent way is not clear.  It may be that some people are unwilling to admit that they are going to vote Conservative, or feel they ought to say they will vote Labour.

The analysis is also good news for the Greens, whose figures were also regularly underestimated in 2014.  The party is becoming one of the political stories of 2015, having been buoyed by last week’s figures suggesting that their membership figures have now overtaken both Ukip and the Liberal Democrats.

Follow @MichaelLCrick on Twitter

6 reader comments

  1. Rich says:

    The ‘social desirability’ effect of not wanting others to know you vote Tory is interesting and entirely plausible. Surely though, if that is what’s leading to underestimating the Tories support there would be a similar, if not even more dramatic, effect for UKIP. As the article does not mention this I presume there is no such underestimation of UKIP’s support?

  2. Alan says:

    Opinion polls purpose is to create opinion, hence the results are understandable.

  3. Andrew Dundas says:

    More people tell pollsters that they definitely will vote than the actuality. Moreover, many more people claim they have voted in previous elections than the aggregate number of votes cast.

    In short, it’s turnout that matters. And Conservative voters – being older than average – are more likely to vote than the Labour voters. And women are a little more reliable than men; especially more than young men.

    All of which helps to explain why Osborne is so anxious to reward pensioners, whilst punishing the low-voting types. Isn’t that a rational response from our Chancellor? He wants to keep his job, and focuses on the people most likely to actually vote in local and national elections. QED?

  4. Ant says:

    We saw this with the Scottish independence vote where the No vote ended up being much higher than the polls suggested. The Yes’s were loud and proud to tell the polls they were voting yes while many No’s kept quiet and clearly told the polls they would vote yes when in fact they voted no.

  5. CWH says:

    When you say the ‘Greens’ are you referring to all three Green parties in the UK? Is it the total membership of the three Green parties that is greater than that of UKIP and the LibDems or is it just the membership of the Green Party of England and Wales? Is this membership, whether of one of the Green parties or all three, anywhere close to the membership of the SNP which is the 3rd largest party in the UK?

    Everyone talks about the ‘Greens’ as if the were a single UK wide party but in fact Scotland and Northern Ireland each have their own Green Party. They may cooperate with the Green Party of England and Wales but they are distinct parties with their own leaders.

  6. Nigel Wilson says:

    Most people just say anything to get rid of the pollster.

    I will never forget the 1970 election and the general expectation – polls and all – that Wilson would win. He lost and we had the joys of Mr Heath, for whatever that was worth!

Comments are closed.