Belfast Telegraph poll: read with large pinch of salt

Mark Devenport’s blog at the BBC has the results of the latest opinion poll published in the Belfast Telegraph today (I cannot find it online yet). It seems a larger poll that the “Slugger O’Toole one” Mick mentioned.

Currently the Belfast Telegraph has:

DUP 26%
SF 25%
SDLP 17%
UCUNF 13%
Alliance 7%
Independent 6%
TUV 5%

In terms of the seat tally they have all the sitting MPs reelected, though FST is impossible to call being a 1% margin between Gildernew and Rodney Connor. Paisley junior appears ahead in North Antrim and Jim Shannon in Strangford.

The problem with opinion polls in Northern Ireland is that they have never been particularly reliable due to the tendency for voters not to reveal their true intentions. In addition although the margin for error is 3% (for statistical geeks I presume on 95% confidence limits though I do not see it mentioned), the error margin for individual constituencies is not stated and sub group analysis is notoriously problematic. Finally of course the Belfast Telegraph itself has an excellent track record of producing opinion poll results only fractionally more accurate than the NIO’s. Lest anyone forget shortly before the 2005 general election some polls predicted the DUP being destroyed by the UUP. Still I suppose we can now all argue about this poll. I meanwhile am going to do a blog about werewolves.

  • snowstorm

    Anyone have the poll the Tele ran prior to the 2001 and 2005 elections? In 2001 they had the SDLP as the largest party, and look what happened.

    The Belfast Telegraph polls are always wrong – and thats before they brought in a DUP linked PR company to carry them out.

  • fitzjameshorse1745

    Presumably the headline in a Belfast Telegraph edition announcing a new poll will generate enough extra sales to justify the expenditure on polls which usually turn out to worthless.

  • Cynic

    Why a new thread on this?

  • snowstorm

    Because the new Slugger is too akward to find old threads unless they remain on the home page?

  • Greenflag

    Good point snowstorm . I think it it may be adversely affecting debate and contributions although I can’t prove it . Perhaps the timing of the changeover right in the middle of the election was not a good idea ?

  • unionistvoter

    How was it that Inform provided the DUP with the poll results earlier this week to use. Did any other party get the results early?

  • Cynic

    What debate?

  • Greenflag

    cynic ,

    you are living up/down to your name ;)?

  • Pickled Badger

    I would not dismiss the poll are garbage it was a masterpiece, accurate no, but it got the messages across that it wanted to. All polls must be taken with a pince of salt, especially those done by party activists, in this case a DUP man. Add this to the fact that the telegraph is owned by the former owner of that bastion of ant-Toryism the Independent then this result from this poll is hardly surprising. I do take my hat off though to however made it up.
    In East Belfast it portrays an almost evenly divided anti-Robbo vote to minimise the chance that voters would swing behind Ringland there. In North Antrim and South Antrim it deters tactical voting by the scale of the margins between the DUP and its chief rivals. The piece de resistance though is Upper Bann. Portaying what is in actuality a DUP-UCUNF marginal as a DUP-Sinn Fein marginal is an inspired move to rally votes behind Simpson and full credit should go to the author.

  • Over on the BBC website in advance of the Leader’s Debate this evening the poll is across 3,200 voters, which is about 180 per seat (178.777777777777777). So the margin of error in each constituency is very large up to 5/6% but snapshot and broadly relative strengths probably OK. As usual, BT making out that things are much more certain than they are on the ground. Though on overall numbers the size of the poll makes the Province wide number more accurate, however relevant that is?

  • the old Manxman

    The sampling is even worse than that. 31% of that 3200 said they weren’t going to vote and 20% said they didn’t know (whether if they were voting or who for it doesn’t say).

    So it’s less than 90 per constituency.

    Add to that it’s face to face interviewing, so some people won’t be telling the truth (“whatever you say, say nothing “) and frankly you’d get a more reliable answer gutting a chicken.

    “Details” of the poll here:

    http://www.informcommunications.com/opinion-polls.aspx

  • Lionel Hutz

    Is this not true of any political poll. I was a bit taken aback by the unionist result but whilst no-one should pay any attention to the individual poll, is an NI wide poll with 3200 sample not quite reliable. Certainly bigger than the average ipsos-mori.

  • Gerry Lynch

    For what my views are worth. In terms of the NI wide poll, the sample size and breadth of sampling points are genuinely impressive. This is good. Literally every other poll I’ve ever seen here has understated the nationalist vote – this doesn’t. That is better. Whether it avoids the other classic boob of NI polling, massively understating the Sinn Féin vote, we shall see on Thursday.

    I have to say, I would feel more confident if I saw the methodology and full crosstabs.

    As far as the constituency results go, THEY ARE ABSOLUTELY MEANINGLESS. The Telegraph should not have published them, and the polling company ought to have avoided reputational damage by trying to ensure they weren’t published.

    Assuming they have been able to get a genuinely demographically representative sample in each constituency, the margin of error just from random error in a sample of ~90 is over 10%. BUT it is virtually impossible to get a representative sample from 90 people in six sampling points. You might as well just ask 90 of your mates. They’ll have the right number of left- and right-footers in each constituency (itself an improvement on most polling here), and that’s what keeps the results looking half-way credible, but that’s as far as it goes.

  • Sounds like the BT standard. Still the 1600 is still a sizeable overall picture. On those numbers however, the local projections are worthless.

  • Marion Smith

    Show me whwn I already made those comments before, WHERE and WHEN show me dates and times.

  • Picture This

    UK polling report is a highly respected website that analysis political opinion polls. It is run by Anthony Wells from You Gov.

    This is what Anthony has to say about yesterday’s Belfast Telegraph poll.

    The Belfast Telegraph have the only poll of Northern Ireland in the campaign so far, carried out by Inform Communications. Polling in Northern Ireland does not have a stellar record of success, but for the record their topline figures are DUP 26%, Sinn Fein 25%, SDLP 17%, UCUNF 13%, Alliance 7%, TUV 5%.

    They have also published breakdowns by constituency – with a total sample size of 3200, this suggests about 180 or so respondents per seat which is, frankly, nigh on useless for predicting a winner in all but safe seats.