Earthquake In The Polls
Wow! Mild earthquake in the election polls. ICM has now brought this election to life in a poll for Channel 4, which I missed last night on the TV, being out to dinner with my family.
When I returned, I scoured the internet for details of the poll, starting with Channel 4's (no mention of it) and ICM's (nowhere in sight). Sky News's wonderful rolling polls update in the lower right hand corner of their screen gave me the first news of it last night and told me that the latest ICM results are Conservatives 31% (up three), Labour 43% (down four) and the Liberal Democrats 19% (up two), Others 7% (down one).
Still continuing to scour the press for it this morning, I see mention of it in only a few papers, the Indy, Guardian, Mail, FT and Times, all in passing and none except the Indy's John Curtice (who is writing for six papers in this election so far as we've noticed) doing any analysis.
On a uniform swing, this would lead to an overall majority of 171 in the MORI model, taking account of regional swing would be 169, almost the same, but inputting the differential turnout "certain to vote" factors would reduce the Labour majority to 87! This calculation shows just how important it is for the Labour Party to get out their vote on Thursday.
None of the papers or television give much detail on the ICM poll, and nothing beyond voting intention. No-one, not even John Curtice who doesn't miss much by accident, reports the continuing remarkable story of the British Election Survey's 'rolly-polly', with the fieldwork done on the telephone by Gallup, which has drifted downward somewhat but as of May 30, the last figures reported, still has Labour on 53%, 28%, 13%, which would project on a uniform swing to a Labour majority of 295!
The BES results of 295 would compare with the Gallup Poll for the Telegraph published from fieldwork dates overlapping that of the BES which on 31%, 47%, 16% would project to 209, 86 seats fewer.
By comparison, of the two ICM surveys published this week, the one for Channel 4 projects to 171 (see below) and the one published in the Guardian from earlier fieldwork to 229, 58 seats more.
My Internet "pen pal", Adrian Bailey, whom I did not know before and whom I have never met, has e-mailed his thoughts on what I've been writing in these essays (and pointing out the typos -- ed please note!).
He's been looking closely at the BES and tells me that his favourite measure, the "Who will win in your constituency?" question, has finally started to move in the last three days, and move significantly.
There's been a 4% swing from Labour to Conservative, mirroring, with a slight time delay, a similar change in the BES's baseline voting intention figures. The voters' prediction now (excluding nationalists) is 163/419/54 -- virtually the same as in '97, and simultaneously the best day for the Tories and the Lib Dems since the BES started publishing its figures.
The scale of the swing means that around 30% of people questioned yesterday [over the previous seven days] said they expect the Conservatives to win their seat, and by extension, expect a Labour majority of just 75. While commentators have been pooh-poohing the Tories' emphasis on Europe, their strategy seems to be working. There's been a doubling in the number of respondents who consider Europe to be the most important issue in the election, stealing a few percentage points from crime, the NHS and tax.
Although there's been a slight increase in the percentage of people who say they're certain to vote, my formula for calculating turnout (from the BES data) comes up with a figure of 73% whether I use this week's or last week's data.
The final BES tables (likelihood to vote for each party) are still among the most fascinating to me. First of all, if we take them at face value, they suggest that 60% of us are likely to vote Labour, 40% Tory and 35% Lib Dem.
If we try to make sense of the responses by comparing them with the likely real outcome of the election, then Labour and Lib Dems will get the votes of all their 8s, 9s and 10s while the Tories will get the votes of all their 7s, 8s, 9s and 10s. This suggests some reticence to admit to supporting the Tories.
Finally, William Hague's unpopularity has declined at last! His rock-solid -3.7 rating fell (or rose!) yesterday to -3.2. I don't really think he's become more attractive, I think it's just a consequence of a general softening of antipathy towards the Tories.'
Hartlepool
I do wish that someone would commission a poll in Hartlepool to give guidance to those who above all would like to see the end of Peter Mandelson (PM), Arthur Scargill or John Booth. When the NOP poll in the Mail in 1983 gave Simon Hughes a point lead over the Labour Leader of the Southwark Council John O'Grady, he went on the defeat Peter Tatchell, the gay, draft-avoiding Australian whom the left-wing Labour selectors in Bermondsey decided that working-class, Catholic, dock-yard constituency wanted to have as their Member of Parliament.
NOP/Daily Mail, where are you now that we need you?
Never did I think I might wish to a) give up my American citizenship, b) move to Hartlepool, and c) vote for Arthur Scargill!
I also see some whinging in the Telegraph on the part of the BBC Radio 4 Today team that they are having difficulty getting Labour politicians to appear on their programme "because they are frightened of John Humphrey", or so they allege. I have another explanation: yesterday I got several telephone calls from one of their researchers asking me to take part in a "light" (so it was described, but hardly "light" to me) item with some academic psychologist they'd dug up who was "sure" that people lie to pollsters, but who they told me had no empirical evidence to support her allegation.
I agreed with misgivings, and spent half an hour preparing myself to appear, with further misgivings that it would all be for nought. Sure enough, as so often happens with the Today programme and more than all other broadcast media put together, on the next morning I received the usual phone call from the morning team standing me down. It is probably half a dozen times this has happened to me from the team at Today. Every politician and pundit I know has the same story, the day team stand you up, and the night team stand you down again.
So what did I find when I began to prepare my evidence?
First, that the average error in the percentage share of all the final week's polls in all the 15 general elections since 1945, has been precisely -0.13%, and in 1997 was precisely zero.
So much for the "shy Tory" theory we've heard so much about over these past four years. To be sure, MORI and NOP both underestimated the Tory share by three points last time, but were balanced by Gallup and ICM, both of whom overestimated the Tory share by two points.
Harris, in the Independent, said 31% for the Tories, and they got 31%.
Sadly, Harris is not in the game this time. The exit polls underestimated the Tories by two points (NOP) and by one (MORI), although the MORI exit poll was carried out in the marginals, and our estimate of the national result was a psephological projection of the exit poll projected to the national share from the swing. The average error in measuring the vote in the 100 marginals in which MORI did their exit poll for ITN was 0.5%.
Second, when projecting from the final polls to Labour's overall majority in seats, when the actual result was a Labour victory of 179 seats over all other parties, NOP projected to 221, MORI to 183, Harris to 173, the poll average 159, Gallup 137 and ICM, nearly 100 seats under, at 81.
Third, the average results found by the different polling companies during the 1997 campaign was as I so often say was nowhere near the final result, as "polls don't predict, they are a snapshot at a point in time and that is at the time the fieldwork is taken".
Even ICM, with all its adjustments, over shot both the Tory vote and the Labour vote by two points each, and underestimated the Liberal Democrats by one and the others by two points as well.
Of course, I lose the argument with the media over this question of reporting the "gap".
Close readers of my essays will recall that time and time again I have urged watching the share, not the gap, and especially the Conservative share, as that will best indicate the state of the parties at the time the poll in question was taken.
5 days and counting.
