Welcome to FiveThirtyEight’s U.K. election coverage. It’s 59 days until the vote on May 7 and we’re just a few short weeks away from the start of the short campaign. We thought we’d take a moment to explain how we’ll be covering the election and why.
Mona: So Nate, aside from the fact that I’ve obviously convinced you that all British people are fascinating, why do you want to cover the U.K. election?
Nate: Well Mona, as we’ve discussed, it’s a fun election to think about from a forecasting standpoint. But also: It’s the bloody United Kingdom! Sixth-largest economy in the world! Twenty-sixth-best footballing nation! Third-largest source of FiveThirtyEight readers!
Mona: But it’s not the same U.K. as it was the last time we voted, in 2010. Britain’s two-party system, deeply rooted in history, has been disrupted. And it happened fast — the polling we’re looking at now would have been unthinkable as recently as a decade ago; support for two dominant parties has slowly eroded and smaller parties, once considered inconsequential, have made big gains. No longer able to count on simply switching power between themselves, the Labour and Conservative parties are now closely watching the fate of the Liberal Democrats, the UK Independence Party (UKIP), the Scottish National Party (SNP) and the Green Party in constituencies (that’s what we call our electoral districts) across the country. If you don’t know who all of those parties are, don’t worry too much — I’ll write you a little cheat sheet. You’re not used to dealing with so many parties are you, Nate?
Nate: No, one advantage we have here in the U.S. is that it’s mostly a two-party system. Two-party systems make forecasting really easy. When voters don’t feel another choice is viable, that leaves them only the option of shifting their vote from Democrat to Republican, or vice versa.
But the U.K. has become less of a two-party system, as you mention. Labour and Conservatives may collectively win only 60 to 65 percent of the vote on May 7. That compares with 98 percent combined for Democrats and Republicans in the presidential election in 2012.
If voters are choosing between four parties — say, UKIP and Lib Dems in addition to Labour and Tories — the complexity increases a great deal. What’s more, some voters will have the option of voting for regional parties, which in the UK are far from irrelevant. SNP, in particular, is set to win a large number of seats in Scotland, and that could make it hard for either Labour or the Conservatives to win a majority nationally.
Mona: Right, and Labour or the Conservatives probably won’t win a majority according to just about every forecast, including ours. If that’s the case, then the parties will need to choose whom they want to be in government with. Last time, the party with the biggest share managed to team up with the third-biggest party (we’ve had a Conservative-Liberal Democrat coalition for the past five years) but things could have gone another way — there’s no constitutional basis for forming those coalitions. Predicting voters behavior is one thing. Predicting political alliances is quite another!
Nate: Yes, and we should be careful to distinguish forecasts of the election outcome from forecasts of the coalition that might emerge. We’re really trying to forecast the former rather than the latter.
Mona: But even predicting the choices of British voters has become more complicated. In a winner-takes-all system, there can be big majorities, so lots of voters make seemingly strange strategic choices so their voice isn’t lost in a “wasted” vote. This so-called “tactical voting” has been a feature of British politics for ages, but it’s probably going to become even more important now.
In Scotland, there are concerns that Conservative voters will “hold their noses” and vote for the Labour party just to keep the Scottish National Party out of power. Elsewhere in the country, as support for the Liberal Democrats (that’s the U.K.’s third-largest political party) has waned, voters are reconsidering their strategies. The two most important tactics that could affect votes in May would be Labour voters turning Conservative to keep UKIP out in their districts and Liberal Democrat voters turning to Labour to stop the Conservative party from gaining a seat.
Those kind of tactical strategies can be hard to predict at the best of times, but it’s even trickier when we have so little constituency-level polling. All in all, this is going to be seriously hard to forecast. And the last time you did it, things didn’t work out exactly as planned, did they?
Nate: Our forecast in 2010 overestimated how well Conservatives and Lib Dems would do, and underestimated Labour. And there were a couple of reasons for that. One was that it’s not so straightforward to go from the national vote total to the seat count. Unlike in the U.S., as you said, there’s not much in the way of local (constituency-level) polling. It would be a bit like trying to forecast the U.S. presidential election from national polls only. You might get somewhere, but the data isn’t nearly as rich, and that makes models more assumption-driven. (We had a big argument in 2010 about whether the shift in the vote would be uniform or proportional; it wound up being somewhere in between.)
But what perhaps proved more problematic in 2010 is that the “bounce” Lib Dems experienced in the polls faded just before Election Day. That hurt not only Lib Dems but also Conservatives, since some ostensible Lib Dem voters turned out to be Labour voters instead. We hadn’t been very focused on tracking movement in the polls — we just thought we had a clever way to go from the vote count to the seat count — when that proved to be the more important part. U.K. election campaigns are considerably shorter than American ones and the vote can shift more at the end. Even the Election Day polls aren’t so hot.
Mona: But this time, we’re doing things differently.
Nate: Yeah, this time we’re partnering with Chris Hanretty, Benjamin Lauderdale and Nick Vivyan, three professors who together run the site electionforecast.co.uk. They’ve developed a model that’s very much in the spirit of FiveThirtyEight’s U.S. election forecasts — it’s empirical, it’s probabilistic, it emphasizes uncertainty. But they’ve had much more time to study these questions thoroughly instead of taking the (let’s be honest) half-assed approach we did in 2010.
For instance, they’ve figured out that the vote for parties often reverts toward where it was in the past election. So their model will be more skeptical of a “surge” for one party that goes against historic norms, especially early in the race. (UKIP’s numbers have already faded some, for instance.) They’re adjusting polls for house effects. And they’re taking better advantage of constituency-level polling, which has become a little more plentiful.
We’re also going to supplement our coverage with reporting on the ground. Particularly from you, Mona, since all Americans find Brits endlessly fascinating. But also from Carl Bialik, who’s London-based, and Harry Enten, who’s so desperate for a fix before the Iowa caucus that he’s considering developing a model to predict the student council elections at Bronx Science. Apparently I’ll even be visiting something called Skegness myself.
Mona: But our editorial plan isn’t set in stone yet, is it? We want to hear from readers, British and American alike (and anyone in between), about what they want us to cover.
Nate: For sure. One reason we’re planning to devote a fair amount of coverage to the U.K. election is because we heard from our readers — they were pissed off that we didn’t do more to cover the referendum in Scotland last year. So we’re aiming to please, to experiment, and to have some fun.
Have questions, comments or concerns about our U.K. election coverage? Send us an email or, better yet, join the discussion in the comments.