Follow @PoliticsIE
 
 
 
Page 1 of 4 1234 LastLast
Results 1 to 10 of 40

Thread: Analysis of Lansdowne & RedC polls

  1. #1

    Default Analysis of Lansdowne & RedC polls

    Hi,

    For anyone interested, I've finished (over 4 posts) the constituency breakdown/analysis of the two polls last week.

    Part one is at 48 Hours « Irish Polling Report and links to the other 3 parts are at the end of that post.

    D

  2. #2
    Politics.ie Member
    Join Date
    Aug 2010
    Posts
    285
    Mentioned
    0 Post(s)

    Default

    Excellent analysis, however surely you are mistaken in Louth are saying that Arthur Morgan would lose his seat based on figures in the Red C poll? (20% first preferences by your prediction)

  3. #3
    Politics.ie Member gijoe's Avatar
    Join Date
    Jul 2010
    Posts
    15,301
    Mentioned
    0 Post(s)

    Default

    It was not a Lansdowne poll on TV3, it was Millard Brown.

  4. #4

    Default

    I think you're making the same mistake that I've been trying to avoid, that is assuming hat if Labour increase their vote by 100% nationally (not impossible), that that will mean a 100% increase in each constituency.

    Their vote in areas where they are 5% or less, will see a much higher proportional gain, perhaps increasing three of four times what they got in 2007, but it still woon't come near to a seat.

    By contrast, where they are already strong they'll go up by a lot less. They could turn ~15% to 25% in several urban constituencies.

    There is also one factor that no polls at this stage in the election cycle can project and that's name recognition. I wouldn't be surprised if you did a poll in Lmerick, the people who would say they'd vote FF would be ~25%. However put Willie O'Dea's name on the ballot and watch that shoot up.

    FF and to a lesser extent FG benefit from name recognition. Labour lose on it. I still think that the odds being offered at the bookmakers on seat ranges are the best indiction of the result.
    Thank you for the six thousand likes.

  5. #5

    Default

    Quote Originally Posted by tullahogue View Post
    Excellent analysis, however surely you are mistaken in Louth are saying that Arthur Morgan would lose his seat based on figures in the Red C poll? (20% first preferences by your prediction)

    You're quite right I'd omitted the SF seat- a typo, it's 2-2-1, not 2-2 (ta, changing that now)

  6. #6

    Default

    Quote Originally Posted by Keith-M View Post
    I think you're making the same mistake that I've been trying to avoid, that is assuming hat if Labour increase their vote by 100% nationally (not impossible), that that will mean a 100% increase in each constituency.

    Their vote in areas where they are 5% or less, will see a much higher proportional gain, perhaps increasing three of four times what they got in 2007, but it still woon't come near to a seat.

    By contrast, where they are already strong they'll go up by a lot less. They could turn ~15% to 25% in several urban constituencies.
    Nope, I've factored that in -you're right though in that I used to, knowing that it would result in incorrect results locally, but that these would cancel out. the new method is outlined in the first post, and the effect of all these computations does indeed 'flatten' the swing, but on a mathematical and historical basis, rather than 'gut feeling' one.

    48 Hours « Irish Polling Report
    Before reading this, please bear in mind that local results always vary somewhat from the uniform swing. This gives you an idea of what that uniform swing is that they’d be deviating from, rather than is some sort of portal into the future. However, if your party over-performs the national swing somewhere, it will under-perform elsewhere. Effectively, unless you’re a small party, the differences will cancel out.

    The spreadsheet takes the 2007 results as a base for the distribution of the vote, and makes some small changes, e.g. Regarding PD votes, personal votes, effects of boundary changes etc.

    It then computes uniform swings in (a) the vote each party received, and (b) the vote each party didn’t receive (e.g. Non-LP vote down from 90% to 65% is X% etc). The composite of these effects is combined with a further swing which take into account the proportion of the national swing in ’92 experienced in each area (the idea of this is that while LP got big gains then nationally, it was quite low in some areas such as Kerry). The composite of all these votes is them calculated. In each constituency, the total will now be above/below 100%, and each party total is then apportioned up/down to adjust to 100%. National totals are then calculated, and where they are deviating from the input figures, there is an adjustment across the board to bring them back in line. You then have a distribution of votes that adds up, and is remarkably similar in pattern to what has happened in the past when there was significant swings.

    Sounds like a lot, but it takes less than a second, running on OpenOffice on a pretty old laptop.

    Next, it feeds into 43 constituency tabs. Each of these assumes a percentage of each parties vote will go to particular candidates, and is set up to auto-compute counts when you enter your estimation of how the transfers go. This is a manual job, and is the most subjective part. I’m assuming good transfers between FG and LP, except when there’s a good left-wing candidate present (although FG still get a decent share there also). Better transfers too to SF this time, and also more SF transfer to FG than in the past (but more to LP, about 35-40%). GP / FF about 35-40% each way. A typical LP vote split is 2:1, although late 3rd additions are assumed to do pretty badly (3.5/2/1).

    When I’ve finished the final count, I enter the party seats in each tab, and that feeds back to the ‘front’ tab, although if it’s very obvious from the headline figures, I may not bother with the constituency tab, and just enter the seat totals on the front tab.

    So that’s how I do it. Disagree with the assumptions? Fine. Do your own, and share the results!

  7. #7
    Politics.ie Member
    Join Date
    Mar 2005
    Location
    Dublin
    Posts
    12,331
    Mentioned
    0 Post(s)

    Default

    Composite figures are interesting:

    Party MRBI RedC
    Fianna Fail 39 43
    Fine Gael 54 59
    Labour 67 50
    SF 1 11
    GP 1 1

    I don't really find the MRBI result there believable. Gilmore is doing well, certainly, but I don't see something greater than the Spring Tide happening. That, in turn, throws off the other figures a fair bit.
    Never let the best be the enemy of the good.

  8. #8

    Default

    Quote Originally Posted by ibis View Post
    Composite figures are interesting:

    Party MRBI RedC
    Fianna Fail 39 43
    Fine Gael 54 59
    Labour 67 50
    SF 1 11
    GP 1 1

    I don't really find the MRBI result there believable. Gilmore is doing well, certainly, but I don't see something greater than the Spring Tide happening. That, in turn, throws off the other figures a fair bit.
    Well it's MRBI, and Lansdowne/Millward Brown (and Quantum, although I discount them). RedC are the only company at present putting LP in the 20s.

    They do of course have to deliver, but as I say, it's only a few months since we were all arguing if they were at 17% or 24%. Red C have since caught up with where LP were with the other companies, and the argument is now whether they are at that level, that of the others. I do think LP being in the mix for top 1 or 2 parties is a game changer, as it makes the GE a choice between a LP/FG coalition led by Gilmore, or a FG/LP govt led by Kenny.

    Regardless of what they believe, you can see why FG in particular would be anxious to rubbish this, as they know the effect it would have on the GE.

  9. #9

    Default

    Quote Originally Posted by Dolores Ibárruri View Post
    Great work, Dotski.

    Are those composite figures above correct? We haven't had the MRBI poll yet.
    Ta.

    Just had a quick look (am running out on school run) I think ibis may be assuming L/MB is MRBI, rather than Lansdown/Millwood (I was a bit lazy in just using initials)

  10. #10
    Politics.ie Member
    Join Date
    Mar 2005
    Location
    Dublin
    Posts
    12,331
    Mentioned
    0 Post(s)

    Default

    Quote Originally Posted by dotski_w_ View Post
    Ta.

    Just had a quick look (am running out on school run) I think ibis may be assuming L/MB is MRBI, rather than Lansdown/Millwood (I was a bit lazy in just using initials)
    Mea culpa, I did assume that! Also, I didn't include the Independents in the count, although you can assume they're the remaining figures.
    Last edited by ibis; 29th September 2010 at 10:09 AM.
    Never let the best be the enemy of the good.

Page 1 of 4 1234 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •