John Curtice: The lesson of 2010? Ignore all the forecasts except for the exit polls

Analysis

This election marks something of a watershed for the polling industry. At each of the last five elections, the final polls conducted just before polling day on average have overestimated Labour's vote as compared with the Conservatives.

That unenviable track record has finally been brought to an end. On average eight polls conducted in the final two or three days of the campaign and published on polling day suggested the Conservatives would win 36 per cent, Labour 28 per cent and the Liberal Democrats 27 per cent.

This means that, in practice, the Conservative performance was still slightly underestimated; David Cameron's party managed to win 37 per cent, one point above the final average of the polls. But the Labour vote was also underestimated – and in this case by two points. So the polls actually overestimated the Conservative lead over Labour by a point.

Yet this is only a limited consolation for the polling industry. For collectively the final polls overestimated the Liberal Democrat vote by no less than three points. Even the company that was closest to the Liberal Democrats, ICM, still overestimated Nick Clegg's party by two points. The Liberal Democrat surge the polls claim to have detected after the first leaders' debate apparently disappeared like the winter snow in the spring sunshine that bathed much of the nation on polling day.

Questions are bound to be asked about whether this means much of the reporting of the campaign was based on a false premise. The alleged surge transformed the campaign and dominated media coverage of much of the ensuing campaign. Now some might wonder whether the surge ever really existed at all.

The polling industry will be hoping it can demonstrate there was a late swing away from the Liberal Democrats in the final 24 hours of the campaign, after most of the interviews in the final polls were conducted. Perhaps too this might have been augmented by a tendency for Liberal Democrat supporters to stay at home.

Given the possibility that another election might be held sooner rather than later, ascertaining exactly what happened may also need to be pursued with a degree of urgency. In the meantime, exactly what happened to the Liberal Democrat surge remains the great mystery of the 2010 election.

Yet there was one poll whose forecast proved to be prescient. At 10pm an exit poll commissioned by the BBC, ITV News and Sky, based on interviews conducted by GfKNOP and Ipsos-MORI and analysed by a team of voting experts and statisticians, forecast that the eventual outcome would be Conservative 307, Labour 255, Liberal Democrat 59 and others 29.

It was greeted with disbelief when initially forecast, not least because its statement of the Liberal Democrat performance was well below everybody's expectations.

Yet, once Thirsk finally elects its MP in three weeks time, the forecast should prove spot-on in its estimate of the number of Tory seats, while it was just three seats out in its statement of Labour's tally and two in the case of the Liberal Democrats.

Anyone who went to sleep immediately after the exit poll was forecast was well-informed about what would transpire in the many long hours that subsequently ensued. It follows a similar exercise in 2005 that accurately forecast Labour's overall majority of 66.

For those who did stay up, the polls also provided a good guide to some of the potentially bewildering variation in the pattern of results. In particular they identified the Conservatives' difficulties in making an advance in Scotland and in constituencies with high ethnic-minority populations.

John Curtice is professor of politics, Strathclyde University