Hit and Myth

The Daily Telegraph's lead editorial on 14 February, argued on the basis of recent Conservative successes in local government by-elections that the Conservatives are in a much stronger position than current opinion polls suggest.

Some of their arguments were valid: no poll a year and more before an election can make any realistic measure of how likely voters will be to turn out at the next general election or how they will vote. Recent trends in other local/regional/European/by-elections have suggested that fewer of Labour's supporters have been likely to vote than Conservatives, but this is no predictor of what will happen at a general election. Opinion polls do not predict the future - when the time comes, we will hope to diagnose accurately as we have in the past any differential turnout and get an accurate reading of the final vote shares, but at this stage our job is to measure what is happening now, not to make projections for an election which is probably more than a year away.

The Telegraph's primary argument was based on a fallacy - that the polls are underestimating the present strength of the Conservatives - and the editorial went on to back up this argument with a series of myths and misunderstandings about opinion polls. Although we wrote a letter to the editor to point these out for the benefit of Telegraph readers, the letter has not appeared.

The Telegraph claimed :-

• "opinion surveys have tended to underestimate the actual level of Conservative support" - no, they have not. Apart from the exception of the 1992 general election (after which all the polling companies extensively reviewed and reformed their methodologies), the polls have measured Conservative strength remarkably well; in fact in the four most recent other general elections (1979, 1983, 1987 and 1997) the average performance of the final polls has been a slight overestimate of the Conservative share. (Voting Intention Trends)

More recently, the polls have measured Tory support perfectly well in both the European elections and by-elections. MORI's last poll before the European election found just over 8% of adults "certain to vote" and supporting the Conservatives, the measure we would have used had we produced a prediction (which we didn't, as the poll was more than a fortnight before the election) - that would project to a Conservative vote of 3.54 million, and the actual Tory vote was 3.6 million. Our poll in the Eddisbury by-election (again, more than a week before the contest and therefore not published as a prediction) measured the Conservative share of the vote within 1% and the margin of victory within 2%. So where were those 'shy Tories'?

• "face-to-face interviews ... tended to be biased because Tory voters spend more time at work". This is nonsense. Any competent pollster can allow for such factors in the quotas and/or weighting used in the poll design, so that the sample accurately represents all work patterns. (If it were a problem, polling by telephone would be little better, since different groups of voters are at home and available to be telephoned for different fractions of time).

Furthermore, regardless of methodological differences, the pollsters agree on the current standing of the Tories: MORI's last two published face-to-face polls (in The Times) have put the Conservative share at 28% and 30%, and our last telephone poll (in the Daily Mail) on 29%; ICM's last two telephone polls (in The Guardian) put them at 29% and 30%, and Gallup's (in the Telegraph itself) at 30% and 28%. In other words, 29% plus or minus 1% in seven polls, two in person and five over the telephone.

• "most of them [the pollsters] were wrong again in 1997" - not so. Four of the five final polls in 1997 were within the accepted 3% sampling tolerance for the Conservative share - two overestimates (Gallup and ICM, both by telephone), two underestimate (MORI in the Times and Evening Standard, one face-to-face and one by telephone), and one spot-on to the nearest whole number (Harris, face-to-face).

The Conservatives may well do better in the next election than their strength would suggest, if there is large-scale Labour abstention in marginal seats and an unwinding of tactical voting; although most commentators seem to feel that disillusionment and apathy among Labour supporters is greatest in their 'heartlands', where most of the seats are safe and where they could afford a dramatic fall in turnout which would cost them few if any seats to the Tories. (It might, however, mean gains for the Scottish National Party and Plaid Cymru, and perhaps also the Liberal Democrats.) At any rate, it will not help the Conservatives to continue relying for their salvation on the non-existent millions of 'shy Tories' who have already once, in 1997, failed to gallop to the rescue.