I like comparing election forecasts from politicians, people who are considering voting but don't watch the political programmes on TV, opinion pollsters, market traders and bookies. The people I never ask, except for entertainment are policy analysts, because they seem to be the only people more clueless than politicians at recognising when they're about to take a good kicking in an election.
One of my funnest moments in politics was sitting cross-legged in a Bayswater flat calling half of Kent's parliamentary constituencies as Labour gains on the evening of May 1, 1997 at the same moment as a buffoon (employed as a special adivsor to a worthless government minister) ostentatiously waved his brick-like mobile phone to ring up someone called "Charles", (he pronounced it "Chaaarlz" in a strangled yet booming Hooray-Henry voice) to wail "Whaaatz haaaapuning?"
I think he wanted to know why the voters weren't magically coming out to keep him in free lunches for another four years. Only a few minutes before, he'd been plotting with other "well-connected, inside-track" political "experts" on how his Tory Party was going to do deals to keep Mr Blair out of Downing Street.
The fact that I won three bets that night, a £75 sweepstake, a £5 where I'd offered 10 to 1 odds against John Major winning 17 months beforehand and a £20 personal bet that Michael Portillo would fall, also made for my only truly successful gambling experience.
The more people were connected to "the centre" of British politics, the less they really believed what actually happened. If they'd had a chat with people like the number one cashier at a high-street bank in Cricklewood about her husband's business, or thought for a minute about how many mini-cab drivers at the time had previously been self-employed, upwardly-mobile supporters of Thatcherism, alarm bells might have rung.
To be fair, the problem was not all on the government side: the opposition Labour did not dare believe the scale of its victory. Five years previously, in April 1992, one life-long Labour supporter with a relative in Parliament was so convinced, as were all her colleagues, that Neil Kinnock would win, especially in the Westminster North constituency, that champagne was being drunk as soon as the polls closed.
The AEI/Brooking study of indicators, titled Partisan Impacts on the Economy: Evidence from Prediction Markets and Close Elections, by Erik Snowberg, Justin Wolfers, Eric Zitzewitz, finds that markets are good at predicting the effects of elections and acting swiftly on information as it emerges.
The Wisdom of Crowds, an excellent little book by James Surowiecki, examined the Iowa Electronic Markets and other indicators of aggregated knowledge and judgement. I recently gave a talk on the subject, an audio recording is available here.
The key message I take from this, is that all the election polling in the world is wrong, because the wrong question is being asked. Instead of trying to find a demographic sample and asking "Who do you WANT to win?" or "Who WILL you vote for?" my impression is that the correct method is to take a random sample, not weighted for demographic representativeness (so it should be quite a bit cheaper).
The question to ask is: "Leaving aside who you WANT to win the next election, who do you think WILL win"?