An ICM poll
in The Guardian
this morning gives the Conservatives a 13-point lead, roughly the same as the 14-point lead in our ComRes poll
for The Independent on Sunday
at the weekend, and another 14-point lead from YouGov in The Sunday Times
. ICM actually shows a narrowing of the gap from 17 points last month, although The Guardian
's report emphasises evidence from other questions that David Cameron is "closing the deal" with the voters.
Despite the big mid-year shift from Conservative and Labour to "Others" at the time of the MPs' expenses hoo-ha and the European elections, the average gap between the two main parties has been around 14 points all year. It is enough for the Tories to win, but does not leave them much margin of error, as they need to be 10-11 points ahead in share of the voteto win a majority in the House of Commons.
Just as Nick Robinson's list of reasons
for Labour MPs to fight on under Gordon Brown could just as well be read as reasons to get rid of him, because they are so much more likely to be able to exploit them under David Miliband, the apparent "closeness" of the opinion polls is another reason for changing leader.
But how close is the horse race?
There used to be a good rule, now named the Smithson Rule after Mike Smithson at Political Betting, that the poll that gave Labour the lowest share of the vote was most likely to be right. But the pollsters have spent a lot of time tweaking their numbers to reduce the Labour figure, and now we do not know whether the process has gone too far. In the last election campaign, the average of all the polls overestimated the Labour lead by 2.2 points
- but this was down from a 7.8 point error in 2001.
The evidence of other elections is scant. The European and local elections are so different from general elections that their predictive power is weak. In the Scottish parliament election in 2007, System Three polls on average underestimated Labour's share of seats by about five points - see page 36 of this Scottish Parliament briefing
. But Scotland is another country, and System Three is another pollster.
The most competitive recent English election was the London mayoral contest
in May 2008. YouGov got the margin of Boris Johnson's victory exactly right at six points, while MORI gave Ken Livingstone a four-point lead. The only other polling company in the field, ICM, gave Johnson a two-point lead a month before polling day. (YouGov's average of five polls during the campaign was an eight-point Johnson lead; MORI's average of three was a one-point Livingstone lead.)
London was, of course, a two-horse race under a preferential voting system but, taking all three polling companies together, it suggests that if there is still an overall bias, it flatters Labour slightly.
As a service to readers, I have updated my calculations of the averages reported by each polling company this year:
Con Lab LD Con lead
Ipsos MORI 41.2 25.2 18.8 +16.0
ComRes 39.2 24.9 18.8 +14.3
ICM 41.0 27.0 20.1 +14.0
YouGov 40.8 27.5 17.3 +13.3
Populus 40.2 27.2 18.1 +13.0
(Average, last poll each month, February to November 2009.)
An adapted Smithson Rule, which focused on the gap between the two main parties rather than just the Labour share, would choose Ipsos MORI as the most likely to be accurate, although I am happy with ComRes's position as the second-least flattering to Labour.
Even if the Tory lead is really 16 points, it is worth Labour fighting to close the gap by five points to secure a hung parliament. But, as I argued on Sunday
, it would take a new leader to do it.