How accurate can these various opinion polls be?

[Harty]

Sorry if fixtures.

But does the hung scenario look odds on, or could we end up with an emphatic win for one of the big two, unlike what they are predicting?

And the poll companies having to hold their hands up and say they got it wrong.

 

[supaseagull]

Opinion polls mean nothing...You've only got to look at Major's election win in '92.

 

[Lord Bracknell]

"A week is a long time in politics"

(c) H Wislon

 

[Pavilionaire]

But does the hung scenario look odds on.

I'm assuming we're talking about the election, and not what you've got planned for tonight?

It's looking like a 'hung' Parliament. Barring some sort of leadership personal skeleton "dynamite" emerging the only party who might get a majority are the Tories. To do this Cameron will have to win the last 2 debates, Clegg will have to commit a major league gaffe and senior Labour politicians will need to be seen distancing themselves from Brown before polling day.

 

[Tricky Dicky]

The alledged polls are changing depending on what is currently headlining the news on any given day. I doubt it reflects very much on how people will actually vote - at least I hope people are not that fickle.

Of course, for tomorrows debate, a lot is now expected of Clegg, so he can only fail in comparison. Cameron will be trying hard not to flounder again. El Gordo will continue to "agree" with Nicholas.

 

[Giraffe]

You only have to look at the wide disparity between each poll that comes out to realise that they are no more than a small indication as to what is happening, and only by referral to the previous poll of the same type.

Worse case scenario for me would be a hung parliament with Gordon Brown in charge. This would undoubtely lead to a general election in the next two years and further uncertainty at a time when the economy needs strong leadership and no faffing about. If Gordon is constantly trying to keep all factions happy it would lead to no decisions.

 

[Pavilionaire]

That poll on the Sky News footer that shows Lib Dems on 21% is an absolute joke (and it's not often in the past we could say that!).

The graphic shows 4 or 5 different polls on a scrolling basis and the Lib Dems are 26%+ on all the others.

 

[Wozza]

You should check the QUESTION being asked in the polls - they vary wildly, and so do the results.

 

[Tyrone Biggums]

I've always found with polls people say one thing one day and do another thing on the day of the vote.

Once they have the voting sheet in front of them people who often said they would vote one way will at the last minute just go back to who they always have voted for.

 

[Giraffe]

28% of NSC users have been with a hooker. Really?

 

[Giraffe]

This sums it up pretty well:

"What I'm interested in is the big poll on 6 May, when people really have to choose five more years of Gordon Brown - the uncertainty, the bickering, the haggling of a hung parliament - or a decisive clean break with the Conservatives." Well said Big Dave!

 

[Gwylan]

Opinion polls mean nothing...You've only got to look at Major's election win in '92.

Well, the opinion polls did show a gradual narrowing of Labour's lead until the two parties were neck-and-neck (which of course was a couple of days before the election). But the pollsters changed their methodology after that as the Tories' vote had been underestimated - they've been pretty spot on in the last couple of elections.

Wozza's right, the questions do matter. There was a Sun/YouGov poll a couple of days that had the LibDems at much lower than others. It turned out there was a leading question about the LDs in it.

Apart from that, I don't think there is too much disparity in the polls.

 

[Kumquat]

Think i've said before, but some advice on opinion polls i was given which is pretty sound. Instead of looking at all the different polls, pick one company and follow it day-by-day to the election. That will give you an accurate trend of how people's views are changing (because while each company measures it differently, each company uses the same sample for it's own polls) if not necessarily the correct figures. The trend will tell you more than anything in this election because it will tell you which way the swing voters are swinging. In 1992 on the last day the swing to Tories was confirmed if you did this and people could have picked up a fortune.

 

[Giraffe]

Think i've said before, but some advice on opinion polls i was given which is pretty sound. Instead of looking at all the different polls, pick one company and follow it day-by-day to the election. That will give you an accurate trend of how people's views are changing (because while each company measures it differently, each company uses the same sample for it's own polls) if not necessarily the correct figures. The trend will tell you more than anything in this election because it will tell you which way the swing voters are swinging. In 1992 on the last day the swing to Tories was confirmed if you did this and people could have picked up a fortune.

Yep, exactly.

 

[beorhthelm]

yes the trend is by far the only important thing in these polls. and even then the Clegg effect has thrown that out, i cant seriously believe people will be so swayed simply by his performance being better than two poor performances. the answer to "did you like Clegg in the TV debate" does not translate to "do you want Clegg and the liberals to run the country", which is the question asked on 6th May.

 

[Kalimantan Gull]

Think i've said before, but some advice on opinion polls i was given which is pretty sound. Instead of looking at all the different polls, pick one company and follow it day-by-day to the election. That will give you an accurate trend of how people's views are changing (because while each company measures it differently, each company uses the same sample for it's own polls) if not necessarily the correct figures. The trend will tell you more than anything in this election because it will tell you which way the swing voters are swinging. In 1992 on the last day the swing to Tories was confirmed if you did this and people could have picked up a fortune.

Yes, and in addition to that, if you want to know what the numbers actually are, take an average of all the polls (the poll-of-polls), that gets pretty close.

 

[Was not Was]

Yes, and in addition to that, if you want to know what the numbers actually are, take an average of all the polls (the poll-of-polls), that gets pretty close.

A poll-of-polls will never be more accurate than the most accurate - it'll just lessen the effect of the occasional (due to statistical significance) inaccurate outlier.