In the NH pre-election polls, as Steve mentioned, there was an very high percentage of undecideds. That alone should have been enough for the pollsters to have not made any definitive predictions regarding the election's outcome. Instead it looks as if they almost entirely disregarded those who said they were undecided, and simply assumed their votes would fall in line with those who had already decided.

Very foolish of the pollsters, but there is another consideration that persons who rely on opinion polls refuse to consider: there exists a segment of the population who intentionally lie on public opinion polls. It is unlikely that this small segment of the polling population would have told the pollster they were undecided. They would have picked a candidate that they had no intention of voting for. With a 30% undecided vote, these dishonest poll-takers suddenly have significance. There are many reasons why a person would be dishonest to a pollster. A very common reason is because they have had their dinner interrupted by a stranger calling into their home, and are angered by it.

Also, as far as I am aware of, random telephone polls still only use land-lines. There is a growing segment of the US population which do not use land-lines for their telephones. They are completely wireless. This tends to be more prevalent in younger age cohorts. If the telephone pollsters do not correct for this, their results will continue to be skewed from reality.

Last edited by a knight; 01/13/08 10:49 PM.