« Spread markets moves mean that the Tories could come top in England
Flying blind in London… »
h1

Are ICM’s Guardian polls including too many non-voters?

April 30th, 2004

make votes count
A big issue for political gamblers in the UK is how you do distinguish between the two polling organisations that were most accurate with the 2001 General Election and who both claim to be “Britain’s most accurate pollster” ICM and YouGov?

    Is ICM, the Guardian’s pollster, overstating what Labour will get at the General Election because it is giving too much weight to people who say they are not certain whether they will vote?

This is a critical question for political gamblers for currently ICM is showing a split of LAB-37,CON-32: and LibD 22%. This is by far the biggest Labour lead and contrasts sharply with YouGov which is predicting a Conservative lead which, as we showed last week would make a dramatic difference to the way the next House of Commons looks.

At the moment, at least, the spread betting markets seem to lean more towards the YouGov view than ICM.

Understanding how ICM produces its figures is important in order for political gamblers to determine how much significance to attach to the firm’s monthly findings.

Each ICM interviewee is asked to rate on a scale of 1 to 10 their likelihood of voting with 10/10 being “absolutely certain”. This was the detailed response in the April survey:-

  • 235 said LAB: 10-52%; 9-6%; 8-13%; 7- 6%; 6-4%; 5-10%; 4-1%; 3-4%; 2-1%; 1-0%
  • 196 said CON: 10-65%; 9-11%; 8-8%; 7-5%; 6-1%; 5-6%; 4-1%; 3-2%; 2-0%; 1- 0%
  • 136 said LD: 10- 55%; 9-%; 8-11%; 7-8%; 6-2%; 5-7%; 4-2%; 3-3%; 2-2%; 1-0%

    • Looking at the detailed breakdown it is clear that Conservatives say they are much more likely to vote than Labour supporters. The critical thing is the weighting to attach to these answers. The Mori polling company asks a similar question for its political monitor and only includes those responses from people who say they are “certain”.

    With the April ICM survey a totally different picture emerges if you just include those who are “absolutely certain” or are 90% sure. Politicalbetting.com has done that recalculation and this is how it would look:-

    LAB 33.5%
    CON 36.8%
    LIB D 21.5%

    What is remarkable at the moment is that whichever pollster is carrying out the survey or however you crunch the numbers the result for the Lib Dems is, within a very small margin, the same.

    We have raised thepoints in this post directly with ICM and will publish its reply.

    Illustration - www.makevotescount.org.uk



    MessageSpace Advertising

    4 comments to “Are ICM’s Guardian polls including too many non-voters?”

    1. The proof will be in the general election result, but on the face of it the ICM method seems reasonable enough - nearly all of those who say they are certain to vote probably will vote, but clearly some of those who rate themselves a 5 will vote. I think I read somewhere that the ICM method gives you a turnout of about 65%.

      Basically this accounts for the difference, the higher the turnout the higher the Labour lead- I did a chart but I can\’t find it, but Chris Lightfoot did a very similar one (see http://ex-parrot.com/~chris/tmp/votes-vs-turnout-2.png). ICM and YouGov basically differ in their (implied) forecast of turnout.

      In terms of seats we decided it would make less difference — the bias to Labour in the electoral system is in part because it wins seats with smaller majorities. On a higher turnout it would just waste more votes.

           by Matt May 1st, 2004 at 12:14 am

    2. Sorry I meant Mori, not YouGov

           by Matt May 1st, 2004 at 11:49 am

    3. vans shoes Heh. How it goes? Buy it all. ASAP. Last discount in your live (AAAAA!!!!!). Take a rest.

           by vans shoes February 14th, 2006 at 10:04 pm

    4. aztmhsf…

      pbpoivxo risqbemi zqsvgags…

           by ykjpvpvxdoi September 7th, 2006 at 11:28 am