|
To those of you who know me, I'm a British ex-pat, grew up in the Thatcher years and had my formation of politics around the time of John Major and the rise of Tony Blair. I remember the 1992 General Election and the BBC coverage of the night.
According to exit polls, and previous opinion polls, the BBC announced minutes after polling closed, Neil Kinnock and the Labour Party had won enough seats so that if he tied up with the Liberal Democrats that he could form a government, based on an average 0.8% opinion poll lead. The end result was the opposite: John Major was able to form a majority parliament with 21 seats - no coalition needed - based on an actual 7.5% of popular vote lead. The BBC was updating its prediction reguarly and I saw it creep from hung Labour to hung Tory to Tory majority. In other words the polls were wrong by 8.3% - quite a big margin. The 1992 poll inaccuracies were apparently due to undersampling conservative voters and underassigning the "don't knows" to Labour rather than the Conservatives. Apparently Conservative voters didn't want to admit they're going to vote Conservative.
Subsequent to that, the polling organisations re-checked their methodologies, tried out different things and basically got back on track. Most polls afterwards got the election right within a percentage point or two
How does this play out here Stateside? Well, polling organisations want to be accurate. They devise schemes and statistical modelling to ensure that any biases in their samples are ironed out. The question is each time have they got it right? Are the polls skewed in our favour? Are they skewed the other way? We'll only be able to tell after the poll that actually matters: November 4th, 2008.
So the take away message here is: 1. Even if the polls show us 5-8 points ahead, treat it as if we're 5-10 points behind and work accordingly. 2. Are there hidden Republicans, not admitting who they're going to vote for? 3. Are the polls right? If the polls are wrong then why worry?
Mark.
|