How much credence should we give to polls two years before the day?

There’s a great facility on the ICM website for those who want to check back on polling history. Just restricted to those commissioned by the Guardian, which is by far and away the longest media relationship in the UK industry, you can sort results how you like and, of course, check back on each monthly survey and compare it with the following general election.

Just before Christmas Andy Cooke wrote a guest column here which got quoted in the national press examining the widely held notion that “Governments always recover in the polls”. His conclusion was that this was more myth than fact.

The challenge here is the lack of consistent polling data over a long period on which to work. So yesterday during siesta time here in southern Spain where I am still on holiday, I played around with a quarter of a century of Guardian ICM polls to see whether this would shed further light on the issue. Comparing what the pollster was reporting for the paper two years before a general election and what happened on the day and you get this:-

June 1985 – ICM poll CON 31%: LAB 36%: LD 32% (LAB +5)
General Election June 11 1987 CON 43%: LAB 32%: LD 23% (LAB -11)
Net difference in the LAB-CON margin: minus 16%

April 1990 – ICM poll CON 32%: LAB 56%: LD 6% (LAB +24)
General Election April 9 1992 CON 43%: LAB 35%: LD 18% (LAB -8)
Net difference in the LAB-CON margin: minus 16%

May 1995 – ICM poll CON 29%: LAB 48%: LD 19% (LAB +19)
General Election May 1 1997 CON 31%: LAB 44%: LD 17% (LAB +13)
Net difference in the LAB-CON margin: minus 6%

June 1999 – ICM poll CON 29%: LAB 46%: LD 19% (LAB +17)
General Election June 7 2001 CON 33%: LAB 42%: LD 19% (LAB +9)
Net difference in the LAB-CON margin: minus 8%

May 2003 – ICM poll CON 29%: LAB 42%: LD 21% (LAB +13)
General Election May 5 2005 CON 33%: LAB 36%: LD 23% (LAB +3)
Net difference in the LAB-CON margin: minus 10%

May 2008 – ICM poll CON 41%: LAB 27%: LD 22% (LAB -14)
General Election May 2010??? CON ??%: LAB ??%: LD ??% (??)
Net difference in the LAB-CON margin: Who knows?

So whether in government or opposition the record shows that Labour’s polling position in relation to the Conservative deteriorates markedly from two years out and what happens on election day. If this is repeated again then my suggestion that we could see a Tory landslide in 2010 might not be wide of the mark.

Betting – I have emailed Ladbrokes and William Hill to ask what odds they would give me on a Tory majority of 100 or more next time.

Mike Smithson