Sunday 10 February 2019

UK polling: what's going on beneath the surface?


So one thing Polling Club would tell you, if there were such a thing, would be to look beyond the headline Voting Intention numbers that the hard-of-thinking throw around all the time. They’d also tell you not to talk about Polling Club, so it might exist after all, but that’s another story. Anyway. Since we took a perhaps ill-advised look at those very topline figures last time, this month we thought ‘Public Policy and the Past’ would take a look below the surface – at the numbers that might really determined who commands the House of Commons after the next election. Hopefully, it’s imminent, and thousands of numbers will soon pour across screens that for now remain sad and empty. So this is not a – shall we say – completely academic question.

The don’t knows – where are they now?

As you’ll also know if you were with us in January, one key to any election is the Don’t Knows, the Won’t Says and the Refused. Say one in five survey respondents tells you that they’re not sure who they’re going to vote for – even though they’ve clicked on the invite. Now imagine that they are not a balanced or normal sample of voters – that they all actually favour one ‘side’ or the other. That means that your nice clean 50/50 split from the poll might actually blow up into a lead of twenty points on polling day. We’ve actually got a good recent example of this, because in the lead up to the 2017 General Election a huge slice of ex-Labour voters had gone over to Don’t Know. And hey presto, by polling day most of them were back, boosting Labour’s score higher than even its surge in the polls guessed at.

Where are we at with this lot now? Well, the most recent evidence shows that they are now more likely to be voters who chose the Conservatives in the recent past, not Labourites disillusioned with the Party’s recent, well, troubles. Take the most recent Ipsos-Mori Political Monitor. Only five per cent of 2017 Labour voters are now completely unsure how they would vote. Only three per cent of ex-Conservatives say the same – though the numbers of ‘won’t vote’ pretty much even up the score of those who’ve shed supporters to ‘unsure’ or ‘won’t turn up’. In the last Survation poll, which returned a slight Labour lead, Conservative ‘not sures’ were 10.5 per cent to nine per cent. ComRes’ last poll showed that Don’t Know and Won’t Vote were tied at five per cent of their past supporters. And so on. One of the causes of what polling fail there was in 2017 came from citizens who weren’t sure which way they’d jump – who turned up to be Labour voters in disguise. That doesn’t seem to be holding this time.

Leaders… and the lack of them

Leadership ratings are usually a good indication of voting intention. When Neil Kinnock and John Major fought it out in 1992, one indication of how well the Conservatives were actually going to do was the Prime Minister’s leader on ‘best Prime Minister’. When David Cameron was more popular than Ed Miliband, the suspicion stuck that there was something ‘wrong’ with the headline voting figures. And so on. So we need to look at these numbers too, to test visceral reactions to the parties’ main personalities as figureheads and lightning rods. And what we find here is very interesting – that Theresa May is unpopular in a normal way, at about the level one would expect for a Prime Minister who’s now nearly three years in power, while Jeremy Corbyn is very, very, very unpopular – indeed spectacularly so, and probably more unpopular than he has ever been.

Let’s take a look at this historically. After a long period of slowly deflating, recent rows over Brexit and antisemitism appear to have done further damage to Mr Corbyn’s already-tarnished brand. Ipsos-Mori’s Political Monitor for January has just given him the highest Unfavourable rating that any Leader of the Opposition since 1977 has ever recorded – bar none – and the second-worse net result (of -55) after Michael Foot in August 1982. Now in a world of increasingly fluid political loyalties, it might not be particularly surprising that these scores are less ‘sticky’ as it were, but that’s still an absolutely dreadful result. Theresa May, on the other hand, has a net score of-25 (with 33 per cent satisfied). This is obviously pretty bad too: for comparison, Donald Trump given some rather different questions has an overall rating of -15 (itself the second-worst in history). But Mrs May’s rating is about the same as Gordon Brown’s and David Cameron’s at this point in their Premierships, and better than John Major’s numbers at that point. So it’s not particularly remarkable. Mr Corbyn’s favourables do now on the other hand appear to be at an all-time low, since they have also reached the same trough with YouGov. Those statistics might recover – they did, after all, surge very rapidly during the 2017 General Election campaign – but for now what we can say here is the Prime Minister is unpopular, and the Leader of the Opposition is very, very, very unpopular.

The extraordinary longevity of the Scottish National Party

One of the main battlefields next time will be Scotland. Which is one of the reasons why the next General is so unpredictable. Labour must make progress here to govern with an overall majority. The Conservatives must try to hang on to their impressive 2017 gains if they are to get anywhere near an absolute advantage in the House of Commons. At the moment? They’re both falling back - Labour slightly more than the Conservatives - in the face of a small but noticeable bump in support for the Scottish National Party. Partly we suspect because they have that vital political quality of clarity when they talk about Brexit, and partly because the Conservatives’ main star Ruth Davidson has been absent on maternity leave, the SNP have been clocking up some pretty impressive poll leaders – which is extraordinary when you think they have been in power in Edinburgh for over a decade. An average of the last two Scottish polls puts them on 38.5 per cent, up from the 36.9 per cent they gained last time. Labour has fallen back rather, from 27.1 per cent to 23.5 per cent.

This matters a lot. Seven out of Labour’s top twenty targets are held by the SNP, and all of them have majorities under 1,000 voters which will fall on a swing of less than one per cent. Right now, polling says the SNP will hold them all. Yes, Labour can govern without making a single gain in Scotland. But the more they win there, the less they will have to rely on and listen to SNP leader Nichola Sturgeon, and the less in government will they risk English voters’ ire by appearing to rest on Scottish voters and Scottish MPs. Looking at the other side of the equation, the Conservatives have eight seats vulnerable to the SNP on a swing of less than five per cent. These are much less vulnerable on the whole than Labour’s vulnerable Scottish outposts, but the small size of Scottish seats (and a fall in turnout between 2015 and 2017) means that we’re not talking very many actual votes here – perhaps three or four thousand at most. Lose just a few of those, or worse face SNP voters who stayed home in 2017 coming back to the polls, and the Tories could lose a scattering of absolutely vital seats. At the moment, the SNP has advanced enough to put all but one of Labour’s seats in danger, but not far enough to expose more than the Tories’ Stirling seat. None of this changes the size of likely Commons coalitions. But if the SNP push forward any more, the balance might start to change again as Conservative seats come within their range.

The unbearable lightness of council by-elections

Last but very much not least, we really should take a look at local council by-elections – contests that go on round the country week in, week out – and for the most part with very little fanfare. Sure, these are low-turnout affairs, they often throw up eccentric results, and in various parts of the country they are contested by local or regional parties that have very little chance of winning a Parliamentary seat. Last Thursday night, a Tower Hamlets-only party won a ward off Labour, while an excellent and surprising Labour win in deepest Buckinghamshire was rather marred by the fact that their candidate had been suspended from the Party before the polls had even opened. So you can’t put a vast store by these results. You can, however, use them as a broad-brush overall guide to exactly where the parties are. If one of them was absolutely tanking or surging, you would expect to see it show up here.

Except that’s not what we see at all. In fact, the overall stasis that we observe from the national Voting Intention headlines is borne out here too. There has been a small swing from the Conservatives to Labour since the 2017 General Election, though one that seems to have become a little less powerful over time (for now disregarding the small number of results we’ve had in 2019). The swing is a modest 2.3 per cent since the last General Election, and since the 2018 local elections it has been running at just under the two per cent mark. It’s not exactly a King’s ransom for the Opposition, especially when we note exactly when these wards were last fought. They were last up for election in the 2015-18 cycle, four local elections in which Labour certainly did not do very well. They lost the National Equivalent Share of the Vote in three out of the four, and only in one of them (2016) did they squeak ahead, that time by a single point – 33 per cent to 32 per cent. So this two per cent swing means that Labour might overall be on average there or there about with the Tory score – exactly what we would expect from the opinion polls. It’s not much of an insight, but as with ever confirming what you think you already know isn’t nothing.

Inner mechanics and outward appearance

So there we have it. Below the big-ticket numbers, there are a number of very interesting things that we can say about any upcoming General Election. What we know about the Don’t Knows tells us that they are not so maldistributed as they were in 2016 and 2017. Labour is unlikely to be able to draw on such a ready-made pool of sympathetic voters again. Another surge might happen, but it will need a different source this time. Leadership ratings are a bit trickier, because Mrs May has made clear that she will not fight any election held in the medium- to long-term. If there’s an election soon, though, her numbers will matter. And they’re really bad – though not so bad as Mr Corbyn’s, who is about as popular as video rental. Again, that might change, just like it did in 2017, though to come back from these lows twice might be a harder ask again. For now, amazingly, the Prime Minister has the edge.

In Scotland, Labour face a greater challenge than do the Conservatives, not that it will matter in terms of who will sit in Downing Street – the SNP could not possibly do any sort of deal with the Conservatives. Even so, what really matters there will be whether the Conservatives lose more than one or two seats. If they don’t, so much the better are their prospects of retaining power. Local by-elections tell us that we are probably are reading this all aright – that the two major parties in England and Wales do seem to be locked together in a macabre political struggle to bestride the realm of the unpopular. This time, the inner mechanics of what we know seem to confirm the words up in lights. It might not always be so: we’ll keep tracking the detailed churning of the data, and get right back to you if things change.

Update, 12 February: YouGov tells us more

No sooner had the blog above been posted, but pollsters over at YouGov published an update of results from their famous Multilevel Regression and Post-stratification model (or MRP for short). Now although this got things pretty much spot on at the last election, we shouldn’t deify MRP as anything more than an interesting new piece in the puzzle. But it is pretty much the closest thing we have to the state of the art right now, and it’s where most polling is going. Take an absolutely massive sample, and map it onto all sorts of turnout and demographic data, and you can much more accurately project a seat-by-seat analysis. And lo and behold, what YouGov have found is pretty much in the ballpark as all the straws in the wind above told us. The Conservatives do indeed have their noses in front, as their continuing hold over their 2017 voters, their leadership ratings and council byelection results have been indicating they might. Labour are as we suggested going backwards in Scotland, with five of their seven seats there in deep danger, and none of the Conservatives’ Scottish seats looking likely to fall to the SNP. So – it all fits. We’ve got a good (if blurry) picture. Whether that scene withstands the Brexit hurricane or an actual election campaign is another matter.