OPEN LETTER TO DANNY FINKELSTEIN
Dear Danny
Your open letter to Roger Alton got me thinking about polling and its place within the news agenda.
You have implied that we at ComRes have in some way been at fault or even irresponsible in our recent work. Your suggestion that The Independent needs to hold an inquisition into our voting intention methodology given that our result stood out from other polls conducted over the past few weeks has, happily, been rendered unnecessary by the advent of the British Polling Council which requires us to publish methodology details and full tables on our website.
Media organisations often rely on polls to generate news and their contribution can often stimulate great debate. However this does not mean that, as a polling company, we are either under pressure or inclined to fudge results or let standards drop.
Last week’s poll was interesting and I will not deny that the results surprised me – indeed we scrutinised the figures and methodology, as we always do, to be sure quality control had been maintained. It seems uncontroversial to assume that the Tories suffered at least some of the electorate’s anger in the wake of the expenses scandal.
Interestingly, in the final national voting intention poll before the previous European election in 2004, Populus for The Times had the Tories on 29%. They went on to get 27%. No other pollster put the Tories on less than 30% in each of the preceding 12 months (and a month earlier YouGov had the Tories on 40%).
You suggest two alternative explanations. Either that after the worst sleaze scandal of our lifetimes, Conservative support dipped by 10% and then quickly recovered as the Labour Party almost imploded. Or we produced one of the 1 in 20 polls which, statistically speaking, is bound to be wrong. Either is equally plausible. You acknowledge that rogue polls are as inevitable as tube strikes and happen to everyone, including therefore Populus whom you favour at The Times. Yet you alight on a third possibility which is that we are consistently wrong which is a claim I strongly dispute.
Interestingly there hasn’t been a great wail of complaint about this week’s voting intention figures which show the Conservatives back nearer 40%. Which reinforces the old adage that that a rogue poll is usually just one with which the reader disagrees.
Yours ever
Andrew
I love spats about polling. I remember going to a post election conference after the 2001 election when there was a massive falling out between three of the country's leading pollsters. I have rarely seen such viciousness and complete loathing between political professionals. Well, maybe I have. Witness Hazel Blears and Gordon Brown.
14 comments:
I've long held the view that ComRes were from another planet - given their strange findings - and the only thing which surprised me was that newspapers continued to hire them. It's a small outfit but I know that their continually anomalous findings have worried even the ComRes team in the past. Either they sort it out or the doubts of newspaper editors will see the back of this nascent pollster.
I love the arguments about how 1000 people can be a representative sample
They may well be but the fudging that goes on to minimise the sampling error is the stuff that needs to be published
Of course all polls use different methodologies - even the real election.
Polling is a business and the market will sort out poor methodologies. ComRes will go out of business is they get it wrong too often.
The only person being sneaky here is Gordon Brown who tries to pick his polling methodology to suit himself rather than mirror the intentions on the population - which a decent and honest person would do ...
... and the best methodology of all is knowing that the population consists of decent honest people who will vote Brown out for trying to fix the next election.
Surely this has more to do with competition between newspapers than anything else? Fink comes across as if he resents the Indie for having two big stories in a week. Perhaps the Times realises its cock-up in missing out on the expenses data
All polls are educated guesswork and merely give an indication of likelihoods. Both the Fink and the pollsters take themselves far too seriously. Pollsters should all have a notice on their desks saying "Remember 1992", and the Fink needs an office junior to crouch by his desk repeating the words, "Remember Daniel, Thou art Human". In fact, Iain, you and Danny could share one to cut costs! :)
...Conservative support dipped by 10%...
I think he means 10 percentage points, not 10%.
A bit worrying if a pollster doesn't understand the difference.
There's only one poll that matters. It's the one Gordon won't let us have.
Interestingly, a lot of use of the word interesting. Or not, as the case may be.
Over at politicalbetting.com, we know that ComRes is short for "Comedy Results"...
I have a distinct feeling that Fink is on to something here. In the real world it is childs play to warp results and the pompous defensive tone of this letter confirms it
Two statisticians in a room will have three different viewpoints.
Mike Smithson,the High Priest of Polling has plenty of articles on PB.com with regards to ComRes and their lack of vote weighting.
They are a joke outfit as far as I'm concerned and seem to produce Polls that help the cause of the Labour Party and no one else!!!!
Does anyone still take ComRes seriously? Or should I say did anyone ever take ComRes seriously?
I know I didn't - it was that little "Being wrong in a truly epic fashion every single time" thing that did it for me.
ComRes = ComedyResults
Post a Comment