Polls Apart!

Never did I ever think I'd write the most trite poll story headline, but I'm tempted tonight, writing on the eve on the publication of three opinion polls in the Sunday newspapers: MORI in the Sunday Telegraph with 1,010 interviews nationwide conducted over the last couple of days, NOP in the Sunday Times with 1,000+ interviews, and ICM in the Observer also with just over 1,000, all done over the telephone.

This spread contrasts with the same three pollsters' findings on Wednesday and Thursday, when as I said at the time it appeared that you couldn't get a cigarette paper between them. I went on to say in that essay however, that the 'top line' hid a number of adjustments, weightings, and what in the US would be called 'Kentucky windage' by some pollsters, which in one case made the difference between an 18% lead for Labour and a 12% lead that is to be published, by weighting for the respondents' recall of their past voting behaviour four years ago, which the BES panel following after the last general election I had thought showed how unreliable a reed this was upon which to rest.

On Sky at 10:00 Saturday night, their projection would have MORI on 233, NOP on 217 and ICM on 170 overall seat majority for the Labour Party, all within near 200 +/- 30 seats, but seen as 'polls apart', and giving licence to the poll pickers to point out the vast discrepancy between us. Of course if it were a close election this could be the difference between success and failure, if one were on one side, the winning side, and another the loser.

We'll all know on Thursday night, and MORI and NOP will have early warning, as MORI are doing the Exit Poll for ITN and NOP will be doing the same for the BBC. This gives us the benefit of interviewing only those who have voted, and reduces probably the single most important factor causing the differences in projection to seats. With only three percentage points separating the 'certain to vote' score between Labour supporters at 50% and the Tory supporters at 27% (astonishingly), 72% to 69% respectively 'certain to vote'. This projected to 257 seats; if the turnout had been reversed and Labour supporters were 69% and the Tories at 72%, then a Labour majority of 241, a difference of 16 seats just on three points turnout reversed.

It is even more important for the Liberal Democrats. On 61% turnout, they'd lose 19 seats; on 71% turnout they'd save 9 of them.

The most interesting finding of the work we've done for the Sunday Telegraph is in Indexing the 'Like him/dislike him; like his policies/dislike his policies' 2x2 ratings of Blair and Hague, against our measure on the same bipolar scales last January. Then Blair's MORI Index was +14 and Hague's was -23. Blair has moved his up to +23, mostly improving his policy image, while Hague is now -22, effectively no change.

One might wonder about the embittered attack on MORI and me personally in the Guardian on Friday. My CEO Brian Gosschalk wrote a mild letter to correct the errors of fact and gently chided Alan Travers for writing such a poisonous article in a reputable newspaper without having a word with us to enquire why we did what we have done, which replicates what we did in our final poll in 1997, and Robert Waller reminds me he did years before when doing the Harris polls. Described as 'irascible' in his piece, I'll put that along side the 'amiable' which Frank Johnson used to describe me in a piece in the Spectator some years ago, and lose no sleep over it.

Roger Mortimore has put his finger on what is the difference in philosophy between the approach that ICM takes, attempting to always 'forecast' the general election, year in and year out, while MORI takes the 'snapshot at a point in time' approach. As roger put it to an enquiring email on Saturday:

"The figures are measuring two different things. Between elections, the question on voting intention is effectively a hypothetical one. We try to measure the mood of the nation as a whole regarding the parties. We want to avoid adding any elements that might cause our respondents to be unrepresentative or different from the rest of the public; therefore we don't artificially remind them of the possibility of supporting other parties if such ideas are not in their minds already, because our respondents represent the wider public who do not have such thoughts either.

The function of the final polls in an election, though, is rather different. Our last poll has to be designed to provide as much information as possible to predict the result of the election; here we are interested in finding out what the voters will actually do in the polling booth, and need to confront them with the real choice that they will face so as to judge their reactions. Even in an election-eve poll, we are not really concerned with what the nation as a whole thinks on election-eve but with what those who get to the polling booths will do on Election Day. Therefore a different measuring tool is needed, and the results may be different.

Between the moment where the election is first called and the final poll, it is more of a grey area; we are no longer asking about an entirely hypothetical election, but nor are we predicting the future. We took the view that the most appropriate point to switch methods was once the final lists of candidates were available, at which point the various local factors can be properly brought into the equation. In particular, electors can begin to make their calculations about tactical voting and to realise which other parties apart from the big two are standing in their constituency.

Incidentally, it would be easy to misunderstand the implications of the difference in the figures by thinking in terms of the Labour lead over the Conservatives. Simply because the Labour lead is lower on the new poll figures, this does not imply better news for the Conservatives. In fact the Conservative share of the vote has not been affected; but there is a shift from Labour to the Liberal Democrats and smaller parties. If anything, this implies a worse position for the Tories as the Liberal Democrat increase will be greatest in the seats where the Conservatives and Liberal Democrats stand first and second, where Labour votes would be wasted in any case -- as I wrote at the start of the campaign: www.ipsos-mori.com/election2001/ec0508.shtml

I think that about sums it up.

4 days and counting.