Why bother with opinion polls?

DESPITE the humiliation, despite the ridicule, the opinion pollsters, who without exception got the 1992 general election result wrong, are flourishing.

It is almost as if nothing had happened. Figures released last week showed that marketing managers, newspapers and political parties are paying hundreds of millions of pounds for polls which, if the Conservatives' victory is a guide, are at best inaccurate and at worst worthless.

In the election year itself, members of the Association of Market Survey Organisations, which represents the big polling companies, saw their business grow by 11.3 per cent. In 1993, the take rose by a further 10 per cent to pounds 329m. Derek Martin, the association's chairman, estimated that the total market for polling and surveys is about pounds 500m.

After the election there was a great deal of criticism of newspapers for filling their front pages with poll results, which turned out to be spurious, instead of discussing issues affecting the electorate. But two years on, newspapers are still paying between pounds 5,000 and pounds 10,000 a time for polls during the local and European election campaigns, putting them on the front page and treating the results as if they were facts. (This newspaper and its daily sister have dropped regular polls but still use them occasionally.) Advertisers and manufacturers, meanwhile, who provide the vast bulk of opinion companies' business, are commissioning more and more surveys despite the recession. 'What's the alternative? There is no other way to find out about public opinion,' said Professor Ivor Crewe, of Essex University, echoing the general view of academics and pollsters.

But beneath the surface, the opinion poll industry is nervous. In the summer, a Market Research Society inquiry will issue a 125-page report on what went wrong in the 1992 election. It is already clear, however, that the society has found no agreement among opinion poll companies on how they should change their procedures to prevent another mistake. More serious for the industry, some of the most respected voices in social research are now suggesting that all opinion poll companies' methods are fundamentally flawed.

Next month, The 1992 British General Election Study - the most authoritative account to date of what happened in the campaign - will be published by Social and Community Planning Research (SCPR) and Nuffield College, Oxford. Its findings will make disturbing reading for the assorted pundits, journalists and party managers who still believe that polls are accurate.

There was no late swing to the Conservatives which could explain how they got 42 per cent of the vote and Labour 35 per cent when all the polls had put Labour on 40 and the Tories on 38, the study found. The acres of newsprint devoted to polls turned out to be a waste of space. 'The Conservatives were ahead throughout and nothing changed very much during the campaign,' said Roger Jowell, the director of SCPR. 'They were always going to win.'

Mr Jowell and his colleagues explain the fiasco by highlighting a key error in the pollsters' method. They have pointed out that the companies failed to take account of the huge numbers of people - sometimes 45 per cent of a sample - who more or less politely tell market researchers to go away when they are stopped in the streets.

The 1992 British General Election Study shows that these 'refusers' are more likely to be Conservatives than Labour or Liberal Democrat supporters. Opinion poll companies do include 'don't knows' in their results. These people co-operate with the pollsters but answer 'don't know' or 'won't say' to specific questions. But the poll companies just ignore the 'refusers' who will not co-operate at all. If a middle-class, middle-aged man declines to answer, they go off and find another middle-class, middle-aged man. In this way, they hope to produce a representative sample of the public by asking quotas of people from different classes, ages and sexes their opinion.

The 1992 British General Election Study, however, did not use quotas but selected 2,800 people at random - far more than in a normal opinion poll - to find out what had happened in the campaign. If people on the list were not in or said they were too busy to talk, the pollsters went back again and again until they got answers.

The study, described by academics as 'the bible of election research', found a 6 per cent bias in favour of the Conservatives among the people the newspaper opinion polls missed during the 1992 campaign - those who would not or could not answer first time. No one knows why. But the numbers of hidden Conservative voters missed by the opinion poll companies has been growing since the 1983 election. There is every reason to believe that there will be even more Tories hiding in the 'spiral of silence' at the next election, the researchers concluded.

'I don't know why it is happening,' said Mr Jowell. 'Maybe Labour and Liberal Democrat voters are less likely to say 'go away' or words to that effect. Maybe they are less egocentric and more polite; more willing to take part in a communal exercise and less likely to get uppity. But, whatever the reason, my fear is that unless the opinion poll companies do something before the next election the whole ship will go down.' The findings of the study mean that the Tories may not be as far behind as the current spate of opinion polls claim. Meanwhile, results showing a one or two-point gap between the parties - such as the recent poll claiming that Labour is now slightly more trusted to run the economy than the Tories - are pretty meaningless.

There is, however, no agreement on what should be done. The Oxford and SCPR researchers advocate random sampling with pollsters going back time and again to get answers. But opinion poll companies say the method does not work and fear, rightly, that newspapers, political parties and marketing managers would not pay the far higher costs of random polls. One company, ICM, now hands out secret ballot boxes in the hope that it will persuade more people to reveal how they will vote.

But NOP, which does this newspaper's polls, says that a ballot box makes no difference to participation rates.

Nick Moon, of NOP, said: 'What worries me is that there is a false sense of security. People are saying: 'We got the results of the Christchurch and Newbury by-elections right, so there's no problem'.' Mr Jowell is more scathing: 'There's a terrible conspiracy by the media to pretend that rough-and-ready opinion polls done cheaply and quickly are really decent research. I wonder how long they will get away with it.'