Reply #37: Interestingly enough, one of the reasons that Brit exit

[Internut]

polls are so accurate is because they take the "shy Tory" effect into account (the equivalent of "reluctant Bush responder" in the US):

http://pollingreport.co.uk/blog/index.php?m=20050205

The theory behind “Shy Tories” is that many people who tell pollsters they don’t know how they will vote or refuse to tell pollsters how they will vote, are actually just embarrassed to admit to voting in a socially unacceptable way. The evidence for this is largely drawn from panel studies before and after the 1992 & 1997 elections, which revealed that people who said “Don’t Know” before the election were disproportionately more likely to admit to having voted Conservative after the election. Attitudinal studies of those who refused to say how they voted both before and after the election, pointed to them too being disproportionately Conservative.

In an attempt to deal with this problem, ICM (and now Populus) reallocate a proportion of the people who say they don’t know how they’ll vote, or refuse to answer, according to how they say they voted at the previous election. When ICM first started doing this, they re-allocted 60% of don’t knows in this way - a decision that was broadly vindicated in 1997 when ICM were the best performing pollster and panel studies showed that between 55% and 59% of people saying don’t know ended up voting in the same way as they did in 1992. ICM’s system clearly worked for them again in 2001, when they were again the best performing pollster.