On Opinion Polls

There has been much exitement in the Blogosphere about opinion polls out tomorrow that supposedly show Labour with a 7% lead. It appears that these "two" polls are actually different presentations of the same polling data.

There are two equal and opposite mistake that people interested in politics can make with opinion polls.

One mistake is to assume that, when poll after poll gives the same message, they're all wrong if it's not what you want to hear.

The opposite error is to panic - or be too pleased - as a result of a single poll. Some polls are incompetently run, or slanted, or happen to get wrong results - even if polls are perfectly conducted one in twenty will produce results outside the margin of error, which means quite a way off.

Let's be clear about what this means. All polls have a margin of error. For a voting intention survey with a sample size of a thousand voters, what statisticians call the 95% confidence limit is about 3 percentage points for each party. And what that means is that there are 19 chances in 20 that the true figure is within 3 percentage points of what the survey shows.

So, for example, if supposing we had a perfectly conducted poll which shows Labour on 38%, Conservatives on 34%, and the Lib/Dems on 20%, then nineteen times out of 20,

Labour support will actually be between 35% and 41%
Conservative support will actually be between 31% and 37%
Lib/Dem support will actually be between 17% and 23%.

In this case, the higher Conservative vote share is above the lower Labour vote share. The range of real support consistent with our hypothetical poll is from a 10% Labour lead to a 2% Tory one.

In other words, leads of up to 6% are within the statistical margin of error for a single poll with a sample of about 1,000 voters. And one poll in 20 will be even less accurate. So if one poll produces results which are very different from all the others, it is almost certainly a rogue.

But this does not mean that it is wise to ignore all opinion polls.

Most pollsters are not incompetent, most of them try to learn from their mistakes, and most of them get it more or less right most of the time. So if a whole string of polls suggest that someone's position has improved or got worse, it probably has.

Hence this poll showing Brown 7% ahead is not good news for the Conservatives, and if we get a whole series of polls like that it will indicate that he is getting a honeymoon with the votesr.

But it is much, much too early for Conservatives to panic or Labour people to assume they are coasting to a fourth term.

The next election is not going to be a walkover for either side. It never was. And there is still everything to play for.

Comments

Anonymous said…
Chris - are you the same Chris Whiteside whom worked for Concert/BT in Holborn?
Chris Whiteside said…
Yes.

There are about 12 people in BT called Whiteside but to the best of my knowledge I'm the only one called Chris: I was based in Morley House by Holborn Viaduct for a couple of years starting when I joined Concert in 1999.

Popular posts from this blog

Nick Herbert on his visit to flood hit areas of Cumbria

Quotes of the day 19th August 2020

Quote of the day 24th July 2020