Are the polls getting it right?

'We have a clear message for the undecided voter. If you have not made up your mind how to vote on the basis of Europe or the economy, think about guns. If it is important to you that we make this country safer and ban all handguns, then don't vote Tory.' Chair of the Gun Control Network, Gill Marshall-Andrews. 23 rd April 1997.

May Day could lead to cries for help from gurus

By Alan Cochrane in the Scotsman 24 th April 1997

IF TONY BLAIR, John Major, Paddy Ashdown, Alex Salmond et al regard 1 May with a mixture of hope, fear and trepidation, there is another group of men and women for whom polling day is an even greater date with destiny.

They are the representatives of the polling organisations, those modern-day soothsayers who make a very nice living, thank you very much, from predicting the outcome of the general election. Except for one thing... they didn't last time.

In 1992 all of the major pollsters got the result wrong. On 27 April, 1992 the five polls published that morning confidently predicted a lead for Neil Kinnock's Labour Party of 0.8 per cent. This would have produced a hung parliament, with Labour as the largest party. The exit polls conducted as people left the polling booths have traditionally been the most accurate estimate of voters' intention. They suggested a Conservative lead of 4 per cent - which would still have produced a hung parliament.

The truth was somewhat different - the Conservatives ended up with a 7.6 per cent lead over Labour, easily enough to form a government with an overall majority of 21. For the polling organisations it was pretty humiliating stuff.

The portents this time are not especially good for our friends with the clipboards as was demonstrated only yesterday, with ICM in the Guardian causing Tory hearts to beat faster by reporting that the Labour lead had been slashed from 14 per cent to only 5 per cent in the space of a week. With only eight days to go, it suggested that the Tories, for so long in the doldrums, might actually snatch a sensational victory.

Unfortunately this theory was more than tempered by Gallup's survey in the Daily Telegraph. Far from reporting a Tory revival, it suggested that Labour's lead had actually increased by 6 per cent to a whopping 21 per cent.

So, how come they can predict different results? The answer may be in the methods used. The Conservative Party, as well as this newspaper and the Guardian, use ICM. This company has responded to past errors by identifying where it went wrong in 1992 and trying to correct it.

One of the interesting theories adopted, now, by ICM is the so-called 'spiral of silence' approach to polls, developed by the German psephologist Elizabeth Noelle-Neumann of the Allensbach Institute.

In essence it suggests that people are reluctant to take part in polls if their first-choice party is unfashionable or unpopular. If they do take part, these people consistently make up the bulk of the 'don't knows'.

She first noticed this trend in watching the right-wing Christian Social Union of Franz Joseph Strauss in Bavaria which consistently did better in elections than it showed up in opinion polls.

Now, Nick Sparrow at ICM believes that this spiral of silence is a feature of British opinion polls, with Conservative voters either reluctant to take part or refusing to state a preference. "When the climate of general public opinion is anti- Conservative, as it has been in this country in recent years, then natural Conservative supporters would be less willing to take part in polls or answer questions about their voting intentions," he says.

As a result, ICM, when delivering its forecasts, assumes that as many as Ireland consistently do better in elections than their showing in opinion polls would suggest.

Labour's polling organisation, NOP, has a different approach. When confronted by 'don't knows', NOP allocates them according to which party they say they identify most closely with or which party the voters say has the best policies on the economy - generally the most important election issue in most voters' minds.

The most important service the polling organisations provide for the politicians - or at least have persuaded the politicians that it is the most important - is their detailed surveys of voters' reactions to particular policies.

They provide the main parties with daily, hourly sometimes, snapshots of what issue is 'playing' particularly well in Perth or Peterborough and with what class of voter. For another important aspect of polling is the class reaction to specific policies. It was by targeting the C2, working-class, vote which helped Mrs Thatcher win in 1979.

This reliance on polling results has caused many a conflict within political parties. So-called 'gut' politicians, who believe they have a feel for the electorate's view, say that too much attention is paid to the pollsters, to the extent that they are in danger of dictating policy.

To be fair to them, until the 1992 general election, Britain's polling organisations had a fairly good record. Since general election polls first began - in 1945 - the pollsters have forecast fairly accurately the results, to within plus or minus 2 per cent. In 1951and 1970 the polls were wrong by 5 per cent and 6 per cent respectively.

1992 was the worst ever - nearly nine per cent out. As embarrassments go, it was on a huge scale and this time the pollsters are determined to get it right. But with the wide variations now showing up, they can't all be correct.

A few people are set to have very red faces come 2 May ... and they won't all be politicians.