Don't know the answer Graham but during the run up to the last Federal Election, a friend swore she was going to vote Labor for the first time in her life - even though it was going to "kill her" to do so (her words) but in the end voted for the Liberals.

Disappointment doesn't always influence vote in some cases but clearly in the case of the 2007 Federal Election, there was enough of it to go around to change government. Like Ludwig said, there is very little to distinguish the two majors, so it comes down to maybe one or two major issues and perceptions of a government having lost touch with the grass roots and the perceived 'viability' of the opposition.

We saw the seriously flawed NSW State Government elected not through popularity but through lack of a reasonable and viable alternative.

Like any survey, the accuracy of polls would probably depend on various factors in the sample taken (demographic, education, income) on any given week. There might be a tendency for some participants to provide a preference during the survey/poll even if their intentions are really undecided, and this percentage is reflected in the final results.

The Nationals have held the seat of Gipplsand for 86 years approx, so this result does not have any real meaning for the current Rudd government.

I think the answer lies in the question - not so much mood as opinion. Hugh Mackay talks about the mood of the nation which always strikes me as a fuzzy way of describing it.

Your mood is something within you, your opinion about something outside yourself can change with your mood, yes, but it mostly changes according to what that thing outside yourself does. So people can have a good opinion of, for example, Kevin Rudd while he was apologising and signing Kyoto, but their opinion changes when his deeds fail to match his words.

The mood of the nation does change, but much slower than opinions, which can change in an instant. Look what the handshake did for Mark Latham.

Opinion polls are measures of people's responses to whatever is happening at the time. In my opinion that's pretty important.

I think there are too many variables to make polls useful, Graham. I don't take any notice of them.

As the politicians say - the only poll that matters is the election poll.

Yes, polls are at best, a guide.

But before we criticise their accuracy too much, remember how many straight polls predicted a Labor victory at the last election. I seem to recall hearing that there were 40 or 50 consecutive newspolls that predicted a Labor win, without a single poll to the opposite.

That's quite persuasive. Not even a blip the other way.

I think polls have a very brief use-by date, and you need to factor in a certain 'mood' percentage. I define this mood percentage as an answer on the poll that is contrary to how you would really vote, because you may be annoyed about a certain issue adopted by your favored party, that isn't really enough to make you change your vote, but is enough for a spiteful entry on a poll.
These 'mood' percentages could be minimised by assessing when the poll was taken. Was it in the wake of a particularly controversial party stance? If not, then it's more likely to be accurate, I think.

Actually the question is not if they are 'accurate' but whether they are representative.

Being 'representative' depends on the questions and the questioned.

It does not matter how many polls are taken, if the questions or the questioned are skewed, the outcome will be skewed.

Col's right, of course - as is TRTL, of course :) I'd bet that's because they're both OLO members who take notice of public opinion polls, regardless of their opinions about their validity and reliability - and of course that's what this topic's about.

It all depends on the question, particularly if the sample favours the hoi polloi.