Why the pundits could be wrong again

Here is the most widely held interpretation of what the polls tell us will happen on 5 May. Labour is enjoying a respectable lead in the polls. The size of that lead varies, partly because of the pollsters' different methodologies, but with rare exceptions most research in recent months has put Labour on roughly 37 per cent and the Tories on roughly 33 per cent (in both cases, plus or minus a sampling error of

3 per cent). This should be more than enough for a comfortable

Labour victory. On these figures, assuming a uniform swing, Labour would win 373 seats and a very comfortable majority of 100. The Liberal Democrats would climb to 63; the Tories, stuck on 179 seats, would have made a net gain of just 14 seats since 1997. Though some disgruntled Labour voters would have defected - to the Lib Dems or some other non-Conservative party, or just by staying at home - they would not have voted Conservative in sizeable numbers.

But here is a second - equally valid - interpretation of the current situation. On average, non-internet polls in February 2001 put Labour 19 percentage points ahead of the Conservatives; the equivalent figure for 2005 was just seven. Moreover, four years ago the polls seriously overestimated Labour's share of the vote, just as they had done in 1992 and 1997. The average Labour lead in the polls during the 2001 campaign was a touch over 17 percentage points, but Labour's lead in the ballot was 9.3 per cent, lower than in every single poll published during the campaign. Because the result was a landslide Labour victory - exactly as the polls had predicted - only anoraks noticed how wrong many of the polls had been. In a tighter election, errors of the same magnitude would make a real difference.

Ah, say the optimists (or should that be pessimists?), even with the Tories on level pegging, Labour will get a comfortable majority, because of the pro-Labour bias that has developed in the electoral system. And it is true that, in 2001, Labour gained 64 per cent of the constituencies in Great Britain on just 42 per cent of the vote - the largest gap between the seat share and vote share of any winning party since 1945. It is also true that, if we assume a uniform swing away from 2001, Labour should still gain a Commons majority in seats even if the Tories gain parity in numbers of votes. A dead heat on 36 per cent, for example, should give Labour a majority of 52.

Yet this is to assume that the bias in the electoral system will work its magic in the same way again, and there is absolutely no reason to think that it will. For one thing, the reduction in the number of Scottish seats hurts Labour disproportionately. For another, the Tories appear to have started to target properly, focusing on key voters in swing constituencies. And third, there is a sign that the pro-Labour tactical voting of the past couple of elections has worn off. A Populus poll last month found that Labour supporters who were willing to vote tactically still split heavily in favour of voting Lib Dem to stop the Tories rather than vice versa (by a ratio of almost 8:1). But Lib Dems now split almost evenly between those who would vote Tory to stop Labour and those who would vote Labour to stop a Tory.

It is the Conservatives who will benefit from this. Following the Scottish boundary changes, Labour would lose its Commons majority if it lost 79 seats. The Conservatives are the main challengers in 85 per cent of these 79 constituencies. You don't have to be a statistical whizz to work out what will happen in these seats as the anti-Conservative tactical voting of previous elections begins to disappear. As Labour's vote falls, the Conservatives would start to gain the key marginal seats, even without any sizeable increase in their own vote. As the electoral bias that benefited them so much in previous elections starts to unravel, Labour could find itself not doing any better in its marginal seats than it does elsewhere. What the electoral system gave, the electoral system can take away.

Since 2001, the pollsters have made adjustments to their methodology. So they are unlikely to get the result as wrong this time as they did last time, you may argue. But the pollsters all changed their methodology after 1992, and they were out in 1997; they changed it further after 1997, and they were still wrong in 2001. The performance in the recent London and Euro elections was not terribly impressive, either.

A handful of recent polls have put Labour's lead in double digits and they may be a more accurate guide than those that give a more narrow gap. Labour may continue to do disproportionately better in its marginal seats, perhaps because of its MPs' efforts over recent years. Perhaps, perhaps, perhaps. Labour could still be on course for a comfortable victory - but there are just as many signs of a much closer race.

Philip Cowley is reader in parliamentary government at the University of Nottingham who also runs www.revolts.co.uk