Watch The Share, Not The Gap

Parliamentary Monitor 'VOX POP'

Sir Robert Worcester

How many times do I have to tell you, watch the share, not the gap!

Every week I seem to see another political pundit misread the polls in this heated run-up to the General Election on May 3rd, because they watch the gap, and not the share. Naturally enough, the politicians pick up what they read in the press or hear on radio and television, and don't really study the figures. I sometimes wonder if the backroom boys and girls do either, from what I read about what the parties are finding, leaked from Millbank or from Central Office. Of course they have their own agenda, but that is so transparent it ought to be printed with a health warning, not as if it were gospel.

Take the run of polls since Labour got over its little local difficulty last September. Let's start from the first of November, and take all the national polls done by the major polling organisations. In all there have been 19, 10 by MORI, who poll regularly for the Times, c. 2,000 each month, face-to-face, in home, using CAPI (Computer-Aided Personal Interviewing). In addition, MORI does telephone polls occasionally for the Mail on Sunday and sometimes for other daily and Sunday newspapers. Four were done by ICM (which does the Conservatives' private polling), for the Guardian, and four by Gallup, for the Telegraph, all by telephone. Only one was done by NOP during this period, for the Sunday Times, again, by telephone. NOP does the private polling for the Labour Party. The telephone polls are usually c. 1,000 interviews, the minimum agreed by the Association of Professional Opinion Polling Organisations.

Since the beginning of November, the 19 polls have interviewed more than 23,000 people, and have found that Labour, with little variation, has averaged 47%, the Conservatives 32.5%, the Liberal Democrats just under 15% and the other parties, including the Nationalists, 6%, for an average Labour lead of 14.5%, which if projected into seats in the House of Commons, using a uniform swing from the 1997 election, would 'suggest' (using the BBC's tortured language reserved for those few poll reports it broadcasts -- other than those they commission, of course) a majority after the election for Labour of some 195 seats over all other parties, around 175 for the Tories, around 425 for Labour, and 35 for the Liberal Democrats. Of course the swing won't be uniform, and projections of these averages to individual seats is bound to be problematic, as we pollsters are reminded by every MP who bucked the trend in any election in the past.

And let me remind you that these polls are not forecasts of what will happen on May 3rd, but snapshots at the point they are taken. Why people think that polls can produce the result of the election before the election has been called always baffles me, but there are many politicians, pundits and others who still believe they can. Let me tell you, polls don't forecast (but sometimes pollsters do, or at least I do). Polls are not fortunetellers, or gaze into teacups, or read palms, and they cannot foretell the future.

Much is made, recently by the BBC's political analyst on their web site, of the 'shy Tory' theory spread on every occasion by spokesmen for the Conservatives. Let's look at the evidence.

ICM employs a secret 'weighting' process which is supposed to adjust for these 'shy Tories, yet the difference they find in the Labour lead, around 10% rather than the 15% found by MORI, Gallup and NOP, is only accounted for by a 1%-2% difference in the share they find for the Conservatives, as ICM finds the Tories at 33.5%, a point above MORI (32.5%), two points above Gallup (31.5%) and half a point over NOP's share for the Conservatives (32%). Of course the 'gap' doubles that.

So what else accounts for their seemingly large difference in the gap? Mostly, their question wording, which asks people who they are intending to vote for, mentioning the names of the parties, including the Liberal Democrats. The other pollsters ask an open ended question, such as "How would you vote if there were a General Election tomorrow?", then asking those who say they are undecided or refuse to answer: "Which party are you most inclined to support?", combining the answers to the two questions which over many years have proven to be the most accurate way of asking the public how they will vote. After all, if the Liberal Democrats do poorly, and the Tories pick up seats from them, the overall majority Labour would have in the House of Commons would be unchanged, as it will be when the Tories regain Tatton from the Independent who now holds the seat, Martin Bell.

Next month: the potential effect of differential turnout.

[50] days and counting.