Guest Pollster | April 22, 2010
Topics: election results , Interpreting polls , UK elections
Anthony Wells is the editor of the UK Polling Report and an associate director at YouGov. He spoke with Emily Swanson on Tuesday about polling and the UK elections. Below is part 1 on the state of the race and interpreting polling data. Tomorrow we will post part 2 on polling methodology.
First of all, could you tell me a little bit about UK Polling Report and about yourself?
UK polling report I started about 2005, really just a similar thing to what Mark does. About me, I'm associate director at YouGov, which is with Polimetrix in the US and in the UK now as well, so it's the parent company for pollster.
Could you tell me a little bit about what the state of the race is in the UK right now?
Until last week, it was a Conservative lead of about 6 or 7 points, before our first televised election debate between the leaders, and after that , the third party, the Liberal Democrats just sort of rocketed in support, so they're up to just about 30 percent or so, and realistically it's about neck and neck between them and the Conservatives as the polls bounce back and forth. [Note: The second debate in this race took place earlier today. Wells has posted instant reaction polls here and here.]
So what does that mean for the Liberal Democrats? If they're neck and neck in the polls, could they get the most seats in parliament?
Assuming a uniform swing, then no, they'll still be miles behind. On the latest polls, we had the Liberal Democrats on top, neck and neck with the Conservatives, then Labour, but in terms of seats, it would equal out to Liberal Democrats having the fewest seats, then the Conservatives, then Labour having the most, despite having the fewest votes.
How is it that that could happen?
It's actually different reasons. The Conservative and Labour disparity is largely out of demographics. Labour seats tend to be smaller because the demographic movements in the UK population is people moving from the inner cities, which tend to be Labour seats, out into the suburbs, which tend to be Conservative seats. Boundary distributions normally lag about 10 years behind. So that helps Labour. You also get very low turnout in a lot of Labour seats, and you also get lots of tactical voting against the Conservatives. So they all mean, on an equal vote [between Conservatives and Labour] Labour do much better.
The Liberal Democrats, they do much worse for different reasons, it's basically because they almost broke through in the early '80s, then they were third place in the 1983 election, but it was a fraction of a percentage point behind Labour. Then they didn't get any seats, they got about 20 seats, despite being a quarter of the vote, because their vote was very evenly spread across the country. Since then, they've become very, very good at targeting - very, very good at focusing their campaign on winner-take-all seats. Which means now they win more seats on a lower percentage of the vote than they had back in 1983. So they've suddenly got this great big grapefruit. They've got 60 or so seats they hold, then maybe 40 or 50 more marginals, and then in the rest of the country their vote is very very low. Over half the seats, they've got under 20 percent, so there's just a huge gap for them to climb over before they start getting a large number of seats.
Labour vote is very efficiently distributed, the Liberal Democrats one is very efficiently distributed for a party that has 20%, but if you suddenly get up to 30%, their vote is atrociously distributed.
You talked a little bit about the assumption of a uniform swing. Is that usually how analysts predict the results?
Yes. And right now, we're all being very cautious (laughs) and hedging a bit. Typically, yes, everyone uses uniform swing. All the media, all the newspapers, all the broadcasters will talk about uniform swing. And it's not that bad. Labour, in 1997 in the big landslide, they outperformed it by a lot because there was tactical voting against the Conservatives. And that stayed around a bit in 2001, but apart from that, uniform national swing has been a pretty good predictor. Now you get lots of pundits and pollsters on television saying, "If there's a uniform swing, this would happen," but if one of the parties suddenly has gone from 20% to 33%, overtaking the other two, then we really don't know!
So when you say uniform swing, that's from when?
From the last election.
Right now you're predicting [at UK polling report] a hung parliament. Could you talk a little bit about what that means?
It's no overall majority, so the largest party is fewer than 326 seats. What happens in practice is there will either be some form of coalition, so two parties both taking seats in the cabinet, or there will be some form of informal pact where one party will opt to support another one without taking cabinet seats, but will vote in favor of their broad program and their budget resolutions, but take other legislation on a case by case basis.
What role, if any, do any regional differences in party support play in interpreting poll results?
The main one would be Scotland. In the past there haven't been big regional differences in swing apart from Scotland, which has quite often gone in sort of an opposite direction from the rest of the country. 1992 is sort of the obvious example - the rest of the country swung towards Labour, but actually the Conservatives gained strength in Scotland. Elsewhere in the country there's just no history of big differences, so no one pays it too much attention.
So is Scotland treated differently? I noticed there was maybe at least one pollster who was polling specifically in Scotland.
Quite a lot do - it depends who commissions it. That's probably why more separate Scottish polls exist, because there's more separate Scottish media. There isn't really a separate media for the East Midlands, or separate polls. In Scotland, there are separate newspapers so they do commission separate polls. In terms of projections, most of the time the media just do go for a simple uniform national swing, even though actually factoring in Scotland separately would make it a bit more accurate. The reason is probably part simplicity and not being bothered to do that bit, and partially because there aren't a huge number of marginal seats in Scotland, so while it would make it more accurate it's not going to make a great difference in the headline figure.
Note: Check in tomorrow for part 2 of this interview