An inconvenient truth for messrs Liddle and Macintyre
Never forget the “Golden Polling Rule?
In the feeding frenzy that’s followed the Observer MORI poll we’ve seen so-called experts rushing in to pontificate because they thought they had a story. The election was not a foregone conclusion after all.
But there’s an inconvenient truth that you hardly hear mentioned – what PB regulars will know as “The Golden Polling Rule”.
This is that whenever polls have been tested against real election results it’s been the survey with Labour in the least favourable position that has been the most accurate.
This is not just a modern phenomenon – but has going going on in Westminster, Euro and London Mayoral elections since at least the 1980s. Just look at the following which I have rehearsed here before:-
June 2009: Euro Election Four pollsters did surveys and the worst share predicted for Labour was 16%. The party got a smaller share than even that.
July 2008: Glasgow East by-election Two polls showing Labour leads of 14% and 17%. Labour lost.
May 2008: Crewe & Nantwich by election Two pollsters did surveys showing that Labour would lose – but in each case the margin predicted was smaller than the outcome.
May 2008: London Mayoral Election. Four pollsters carried out surveys and three suggested that it was neck and neck between Labour’s Ken and Boris. The fourth had Boris ahead by what turned out to be the precise margin of victory.
May 2005: general election. All the pollsters’ final polls bar one had Labour with a bigger vote lead than was actually achieved. The one exception, NOP for the Independent, got it precisely right and then got dropped by the paper.
June 2004: London Mayoral race. Two pollsters did surveys – the one with Labour in the least favourable position got it almost precisely right.
June 2004: Euro elections. Two firms did polls both of them overstating Labour’s eventual position.
June 2001: general election. Labour won with a 9.3% lead on votes. None of the pollsters had this in single figures and one campaign poll had the party 30% ahead.
May 1997: general election. Labour won with a 14% lead on votes. The best pollster was ICM who had Labour in the least favourable position although their final poll did understate Tony Blair’s party.
April 1992: general election. The closest final poll had the Tories ahead by just 0.5%. John Major’s party won with a lead on votes of just under 8%.
I have not included the by elections where there was only one poll. For with the exception of Glenrothes in each by election where there has been polling there has seen the same pattern. In this parliament a survey suggested that Labour was heading for an easy recovery of Blaenau Gwent in 2006 – it failed. A couple of years earlier the one poll ahead of the Hartlepool by election had Labour holding on with a massive 33% majority – the actual margin was 7%.
So is it going to apply this time? Who knows? But surely the default position for punters, pundits and politicians should be to assume the worst for Labour.
We’ll see what happens at the general election. Maybe the recent changes that some pollsters have introduced will cause, perhaps, the second or third least favourable Labour position to be the most accurate? Maybe they won’t?
PaddyPower has just put up a market on when we will see a Labour lead from one of the main pollsters. It’s 10/1 Before the end of 2009; 6/1 Jan-Feb 2010; 9/2 March-April 2010; 7/2 May 2010; 2/5 Not at any time before June 2010. UPDATE: The dataset from last night’s PB Angus Reid Strategies poll is now available here