Dominic Lawson: Polls do much more for the pollsters than for the public

If you see someone with a smug grin on his face the chances are he's an opinion pollster. Over the next month-and-a-half he will be raking in enough money to make a banker jealous. Some newspapers are promising to publish an opinion poll every day from the moment the date of the general election is confirmed: at a charging rate of thousands of pounds per poll, that is seriously good business. No wonder Bob Worcester, founder of the polling organisation MORI, has been able to afford an estate on the Caribbean island of Mustique, more widely known as the favoured holiday home of the world's richest rock stars and fashion moguls.

It's understandable why opinion polls are so addictive, especially at times such as these. Everyone wants to know the future; and while the pollsters dutifully put out the health warning that their output is not predictive, but only paints a picture of the mood of the moment, they are studied in the same spirit as Etruscan soothsayers gazing at the entrails of sacrificed animals.

In a close general election, however, the ancient art of hepatoscopy would be about as useful as modern opinion polls at divining the identity of the winner. Most pollsters like to change the topic of conversation if you mention the 1992 General Election. On average, their final polls were out by over 8 percentage points, underestimating the Conservative vote by 4 per cent, overestimating Labour support by a similar amount, and thus failing to forecast the Tory victory.

There is in any case an acknowledged "sampling error" of 3 per cent when the usual random collection of 1,000 people is used as a substitute for the entire electorate – a standard deviation which can itself be of dramatic significance in a close contest. Because the last three general elections have been very one-sided affairs, the erratic performance of the pollsters has not further damaged their reputation. In fact, they gained kudos by predicting that the 1997 general election would be a Labour landslide – as in fact it proved to be despite the insistence by a number of journalistic political pundits that no such avalanche would occur.

Yet as Nick Sparrow of the polling group ICM later pointed out, the 1997 general election was in reality an equally dismal performance at statistical prediction by his industry. While the actual percentage lead of Blair's New Labour over John Major's Conservatives (after the real votes were counted) was 13 per cent, Gallup, Harris, MORI, and NOP had all, in their final polls, stated that Labour's vote would exceed the Tories' by between 19 and 21 per cent. Only ICM came close to the final result, which perhaps helps to explain Mr Sparrow's willingness to publish a dispassionate assessment of his industry's lamentable performance. Yet because the pollsters hadn't actually failed to identify the winning party in 1997, they did not experience the public scorn which followed their 1992 debacle – although their performance was actually no better.

Since then all the polling organisations have followed ICM by introducing various "adjustments" which seek to make allowances for the fact that, for one reason or another, many of those who ultimately vote Conservative disguise this fact from people who ask them about their intentions. These are the so-called "silent Tories". Yet this form of pro-Tory "adjustment" – based on the idea that people have been reluctant to admit (even anonymously) to supporting "the nasty party" – might well not be appropriate in the current election. After all, it is now Gordon Brown's Labour Party which it is most fashionable to denigrate as hopeless and incompetent: what if the pollsters are busily allowing for "silent" Tories in their methodology for polling in the 2010 campaign, only to discover too late that they should instead have been concerned that those intending to vote Labour were reluctant to admit it?

It is true, however, that there has been a long history of polls underestimating the Tory vote, even before the New Labour era. Ted Heath's victory over Harold Wilson in 1970 came as much as a shock to the opinion pollsters as John Major's over Neil Kinnock did 22 years later. Among the pollsters' somewhat self-serving explanations at the time was the claim that many Labour voters, convinced by the accuracy of the polls, thought that they would not need to turn out to vote in order for Harold Wilson to be returned to Number 10.

A similar argument was put by embarrassed American opinion pollsters in January 2008, after Hillary Clinton beat Barack Obama in the New Hampshire Democrat Primary, in stark contrast to their allegedly scientific predictions; but if this defence by the pollsters has any merit, it would suggest that polls predict best when the public have little confidence in them – hardly a recommendation for their business.

Some critics, in fact, make an entirely opposite criticism of the opinion polls and their effect on the democratic process: that they have a dangerously self-fulfilling property. This is the so-called "bandwagon effect", which proposes that the sheep-like tendency of much of humanity would make them more likely to vote for a party which seems to be very popular with other people.

Some psephological studies have purported to show that there is indeed some force in the "bandwagon effect". For example, when people are asked after a general election how they did vote, the numbers never add up: more people say they voted for the winning party, than actually did.

It is because polls actually have their own effect on voting intentions and outcomes, that it is sometimes argued that they should be banned during election campaigns; indeed in such European countries as Italy, Greece (and until recently, France) it has been against the law to publish opinion polls during the last week or two of an election campaign.

My instincts are resolutely against any such infringement on what amounts to the freedom of the press – especially as there is no sense that the public's sense of decency is outraged by the publication of opinion polls, however lurid the accompanying headlines and however distressing the sight of all those pie-charts. On the other hand, there would be something marvellous about an election campaign free of opinion polls. First of all, it would be good for the politicians: they would resemble less the sort of businessmen who are so obsessed with tracking their share price daily that they forget about the long-term needs of their company.

It would also be good for political journalism: we would spend less time pontificating about the prospects of a hung parliament and the various putative deals which the parties would or would not make with each other in such an eventuality, and more time thinking about the actual issues which are at stake in the election – and which, after all, are the matters which will actually decide how the public eventually cast their votes. Besides, if all newspapers were able to agree a mutual poll non-publication pact, imagine how much money they would save, to be spent instead on genuinely creative journalism rather than the mere purchase of rows of numbers.

Above all, think of the benefit of the likely increase in the uncertainty about the outcome: it would surely lead to a sharp rise in voter turn-out, something which all parties purport to desire. On the downside, there might be fewer pollsters with Caribbean estates. Somehow, we would all just have to live with that.

d.lawson@independent.co.uk