Thursday, 31 January 2019

What are the polls telling us about a snap election?

There is quite a lot of talk at the moment about the possibility of a snap election in the UK – of Prime Minister Theresa May trying to break the Parliamentary stalemate over Brexit by changing the composition of a Parliament that seems too confused about what it wants to make any decision at all. So this month, we thought we’d take a look at the polling and ask: what does it tell us about the standing of the parties right now?

First things first: overall Voting Intention numbers. If we take an average of each pollster in the field’s January report, we get vote totals that look rather like: Conservatives 38%, Labour 37.8%, Liberal Democrats 9.4%. So the Tories would be down 5.5% on their 2017 performance; Labour will have shed 3.2%; and the Liberal Democrats have advanced 1.8% since the last national contest: but the swing from Conservative to Labour over this Parliament would be only 1.15%: and that means the Government has actually made a very small amount of progress over the last twelve months. A year ago, those numbers actually reported a slender Labour advantage, rather than the tiny Conservative lead that we see now: Labour were on 41.5% as against the Conservatives 40.3% and the Liberal Democrats’ 7.3%.

So the ratings of both major parties have taken a hit over the last year, both drifting downwards a tad as the Liberal Democrats’ polling looks a couple of points healthier. But really, for a number of reasons, there’s even less to that change than meets the eye. You win seats under the First Past the Post, and how many votes you get regionally or nationally don’t come into it. Here, the picture has moved perhaps even more glacially than the overall numbers. So many seats are on a knife-edge that a really decisive break in one direction or the other would pile up the numbers in the Commons for whoever leads the charge; but at the moment, that just isn’t happening.

If we take those overall averages and also factor in sub-national polling from Scotland, Wales and London, in January 2018 we would have expected a House of Commons that looks rather like this: Conservatives, 293; Labour, 283; Scottish National Party, 37; Liberal Democrats, 14; Plaid Cymru, 4; Greens, 1. Right now, we’re looking at something more like Conservatives, 301; Labour, 268; Scottish National Party, 42; Liberal Democrats, 17; Plaid Cymru, 3; Greens, 1. So in a whole year of sound and fury, the Conservatives have moved forward by less than ten seats, while Labour have moved backwards a bit more, partly because of what looks like a deteriorating situation in Scotland: they’ve gone backwards in our virtual election by fifteen MPs, and forward on their unexpectedly good 2017 showing by just six seats.

That means even less when we look at what those numbers would mean in terms of forming a new government. In both situations, those seat totals add up to a Hung Parliament in which Labour would likely try to form a minority government relying on the support of the SNP and the Liberal Democrats. The only difference is the security and stability delivered by a deal between those two parties. Labour plus SNP in January 2018 looked like 325 seats; just about enough on their own to govern for quite a while, with vote-by-vote support from the minor parties. Now that picture looks slightly less rosy for those two parties: they only add up to 310, not enough without the Liberal Democrats (and probably Plaid and some liberal Tories) to do anything much at all.

This is a strange kind of stasis. We appear to be living in a period when the parties might fracture, or even break up. Their discipline in the House of Commons seems almost shot. Frontbenchers on the Opposition side are even allowed to rebel against a three-line whip and keep their jobs. There has been a chemical weapons attack on British soil, killing a British citizen. Brexit is casting a pall over everything. Public services look increasingly threadbare, and in some cases (for instance if we look at homelessness and rough sleeping) appear to be failing altogether.

But the public are just not moving. The vote totals are stuck. It is our contention that voters are so fed up – a fact that you can see in almost all the qualitative and quantitative evidence – that they’ve got into a 'plague on all your houses' mindset, promising to vote for one team or the other just because they hate the alternative. With both parties retiring into their own increasingly bizarre and fantastical comfort zones, these wide but thin coalitions might last for quite a long time.

Ah, Labour partisans say – but look at what happened in the last election, where we surged from the mid-20s to more than 40% in just a couple of months. Once reporting restrictions are introduced to give due balance to the Opposition, and once Labour gets out in sunny rallies and canvassing with its huge membership, then these numbers will be transformed. There’s probably something in that case, too: strategically, the Conservatives have very little to the public except ‘delivering Brexit’, which is likely to consolidate their base without reaching out to any new voters. And Labour’s domestic policies – nationalisation, higher taxes on the rich, more public spending – are undoubtedly pretty popular.

It is however possible to doubt that any Labour gains during the campaign will be on anything like the scale that they managed last time. For one thing, there is just much less of the vote to bite into. The Liberal Democrats were polling higher than they are now at the start of the 2017 race, and the United Kingdom Independence Party have gone from polling in the teens then to mid-single figures now. For another, the number of don’t knows that were once Labour is far smaller than when Mrs May made her ill-fated break for it in 2017. As the Bristol University academic Paula Surridge explains here on her blog (using YouGov figures) there were many more ex-Labour undecideds from 2015 at the start of the 2017 campaign than there were uncertain Conservatives. That allowed Opposition Leader Jeremy Corbyn to squeeze voters who preferred Labour, but who weren’t sure about him and his agenda, by urging to come home and ‘stop the Tories’. But that gap has now been reversed or closed altogether. There aren’t the don’t knows to make the same sort of progress.

Other pollsters confirm this picture. Take the first ComRes poll of the 2017 campaign: 11% of 2015 Labour voters said ‘don’t know’at that point, as against six per cent of Conservatives. The last ComRes poll we have in 2018 records 15% of Conservatives as uncertain, as against 13% of previous Labour supporters. Labour might indeed do quite well in a campaign. But they seem unlikely to jump upwards quite so vertiginously as they did in 2017.

What does this all mean in practice? Well, the exact polling numbers should be taken with a pinch of salt. For one thing, UK polling has a mediocre rather than a bullseye record of predicting General Election results even just before the poll, let alone months or even years before it happens. As US polling guru Nate Silver noted in 2017, the average error on UK polls fifty days out from a vote is exactly six points on the margin between either the two sides in a referendum or the first place and runner-up slots in a ‘normal’ election. That means that our tiny Conservative lead of 0.2% could actually end up meaning anything between a 6.2% lead and a 5.8% deficit on any polling day in March. Even poll numbers from the week before have a five-point error margin. The biggest miss from the week before is 9.5% (at the 1992 General Election). That means that it should not completely and utterly shock us if the Conservatives ended up either winning a snap election by 9.7% or losing by 9.5% - not much of a guide to anything, you might think.

Secondly, the whole country is entering the white-water phase of any political crisis, where the thrills and spills make you feel sick rather than excited: it’s perfectly possible that one or both main parties will splinter, with breakaway groups forming and recrimination spreading. There’s even potential for a really unplanned or disorderly Brexit, which will likely bring about more disruption and therefore political upheaval than any emergency since the miners’ strikes, the three-day week and the winter of discontent in the 1970s. It is likely that if we do crash out of the European Union, the Government will get much of the blame; but even that is unpredictable, were the Conservatives to be led into a snap election by a Eurosceptical leader who blamed everything on Eurocrats and foreigners.

Third and last, it’s important to note that everything we’ve written above is based on averages – the safest, but by no means foolproof, way of going about measuring any group of indicators churning so dynamically. And the polling differs. One pollster, YouGov, is showing a consistent Conservative lead – of five points at the last count. The others aren’t, and the last reports from pollsters Opinium and Kantar actually gave Labour three-point leads. Relying on YouGov alone would give us a House of Commons would look like: 331 Conservatives, 241 Labour, 37 SNP, 19 Liberal Democrat, 3 Plaid Cymru, 1 Green. The Tories would gain a small absolute majority that looked very much like David Cameron’s in 2015, though with a rather different geographical spread of where their seats actually were. But if we take YouGov out of the picture entirely, each party’s total of MPs would likely look much more like the average, with perhaps a handful of seats going over to Labour from Conservative.

Even so, the polling does tell us something, and that’s why it’s better than nothing – just as it told us a lot during the 2017 General Election campaign, during which both main campaigns whispered to anyone listening that they didn’t believe the surveys showing a Labour surge upwards. Well, the polls were right about that, and the canvassing data and human intelligence was wrong. So it might prove again. The polling tells us that the parties are running pretty much neck-and-neck, that the remarkable longevity of the SNP’s popularity in Scotland looks to be still holding up, and that Labour looks unlikely to make the same gains during the next campaign as it did during the last one. In short: any General Election would be a massive gamble for all concerned. Right in the middle of the bell curve of probability is a Hung Parliament that produces nothing very useful except the chaos that populists love so much: but at one end (with YouGov) the Conservatives win a majority, and at another (with Opinium or Kantar) Labour have enough MPs to govern, albeit without much comfort. That is the heat map of where we would land if there were a General Election tomorrow – which, of course, there won’t be.

Now, that level of precision (or lack of it) might be a poor return for all the work everyone is putting in. It isn’t a Rosetta Stone. But it isn’t nothing either. Such is the world of known knowns and known unknowns – and of statistical art and practice, rather than science. Experts, in this case pollsters, do know things: it’s just that they don’t know everything. They can map out the landing zone. The rest is up to you.