UK Politics

Headline numbers: The problem with unusual elections

Pen marking a cross in a box Image copyright Thinkstock

Opinion polls so far this year have suggested that about a quarter of people plan to vote for a party other than the Conservatives, Labour and the Liberal Democrats.

If they actually do, that will be remarkably high. In 2010, fewer that 12% of voters put their cross next to anyone other than those three parties.

So it's an unusual election, which is a problem for people conducting opinion polls, as I discussed during the Scottish Independence campaign.

Why is it so much of a problem? Well, I used to think that if you were conducting an opinion poll you'd go and ask a bunch of people which party they planned to vote for and then publish their responses, but it turns out it's not as simple as that.

You do go and ask a random group of people, or alternatively you go and ask a panel of people you have selected because you think they are representative of the whole population.

Playing with numbers

But it is at that point that the tricky bit starts.

First of all, you might want to adjust your findings because you have not asked enough women, or enough young people, or enough people in the Midlands.

Image copyright PA
Image caption Trying to gauge which way people will vote is a complicated business

But your adjustment will have to be based on previous polling you have done that gives you an idea of how those people would have responded if you'd asked them.

The next question is how likely the people you have asked are to actually vote. If they're not going to vote then their opinion is irrelevant to the poll.

So you ask them how likely they are to vote and you can also ask them whether they voted in the last election, and based on their answers you decide how much weight to give their party support.

'Shy' voters

You might even weight your responses based on their other answers - perhaps you have found from previous polling that people who say they are going to support a smaller party are less likely to turn out and vote, for example.

Finally, you have to decide whether you believe the responses you have been given. There is a phenomenon known as shy voters or the spiral of silence, which is that poll respondents may not be comfortable telling you whom they plan to vote for.

The classic example of this was at the 1992 election, when Conservative voters were "shy", so the polls made support for Labour look stronger than it really was.

Professor John Curtice, the polling guru from University of Strathclyde, says that the polling company ICM has decided Liberal Democrat voters are shy ahead of this election.

It may be right. Or it may be wrong. The adjustments that pollsters make are based largely on their experiences from previous elections.

And that is why having an election that is significantly different to previous ones is a problem for opinion polls.

Around the BBC