What’s happening with the polls?
I’m glad you asked. This. This is happening with the polls:
That’s a bit of a ski-slope on the right, wouldn’t you say? Mind you, it started high. Only a couple of months ago it was up in the… low 20s? Can that be right?
OK, back up a bit. This chart covers the period from the beginning of December 2021 to the 20th of April 2023. The red line, plotted against the left-hand axis, represents the Labour Party’s polling average, with each point calculated up to and including a poll taken on the date on the X axis (or in a period ending on that date). The red mountain, plotted against the right-hand axis, represents Labour’s lead over the Tories in each of those polls; if you squint you can see a submerged blue ice floe at the very start of the chart, representing a brief period when Labour’s average poll lead was negative (i.e. the Tories were ahead). Both of these series are averaged over 14 days – which is to say, the figure against any given date is the average of any and all polls taken on (or ending on) that day and the 13 previous days: the figures for the 17th of April, for example, are averages of all polls from the 4th to 17th. The left and right axes are scaled 1:1; 0% on the right axis is matched with 40% on the left axis, for no reason other than convenience.
So, what are we looking at? First of all, yes, that is a bit of a ski-slope on the right. But let’s do it in order. The first thing you see here, in the leftmost ‘peak’ (circa the end of January 2022), is the discrediting of Boris Johnson. As the ‘Partygate’ scandal unfolded, Labour’s support rose from the high 30%s to the very high 30%s or even 40% – an undramatic improvement which nevertheless resulted in 5-6% leads over the Tories. This is worth pausing over, as it’s the result of the current leadership’s determination to take votes directly from the Tories. In 2017, a campaigning approach focused on mobilising non-voters and hegemonising the centre-left meant that Labour’s share of the (intended) vote could rise from 25% to 40% without the Tory share going below 42%; the current approach doesn’t run that risk. It does, however, run the risk of amplifying the volatility of the system, with every 1% movement towards Labour improving Labour’s lead by 2%. (And poll leads, like share prices, can go down as well as up.)
Anyway, after that first peak things settled down a bit; Labour polled a bit below 40% or a bit above, with round about a 5% lead over the Tories, from February through to July. In July, Boris Johnson resigned, Labour’s polling moved decisively above 40% and the poll lead started hitting 10%. (I don’t know what was behind that dip at the beginning of August.)
Then Liz Truss was elected leader of the Conservative Party, and whoa, Nelly: Labour’s polling climbs to 53% (up 11%) and their lead over the Tories, following the same pattern as before, climbs to 31% (up 21% – which isn’t twice 11, but it’s close). But all good things must come to an end: by November 2022, Rishi Sunak is Prime Minister, and Labour’s polling average has settled back down to 47-8%, with a lead over the Tories of 20-22%.
Yes, we’re in the sunlit uplands of the Twenty Point Lead – and there we stay, give or take a point or two, from mid-November 2022 through to the end of February ’23. After that, though, comes another phase, and that phase we are still very much in: not to put too fine a point on it, it’s the ski-slope phase. In the month and a half since the end of February, Labour’s average support has fallen by 3% (from 48% to 45%), and their average poll lead has (predictably) fallen by 6%, from 22% to 16%. Update: on 20th April those two averages dipped below 45% and below 16% respectively, for the first time since Liz Truss was PM.
Like all the other movements on this chart, this drift back to the Tories can be explained fairly exhaustively by two factors. One is the public perception of the Prime Minister, which in Sunak’s case was initially fairly unfavourable – after the debacle of the Truss premiership and Sunak’s anointment by default – but has improved over time, and (at least for now) is continuing to improve. The other is electoral gravity, a.k.a. reversion to the mean: it’s not a frequent occurrence for Labour to be polling above 45% or the Tories below 30%, and when something like that does happen the smart money is on it not happening for much longer.
What the factors driving Labour’s polling don’t seem to include is anything contributed by Labour. The Labour leadership isn’t making the weather when it comes to the party’s popularity; moreover, they haven’t done so for some time. Quite some time…
[wobbly dissolve effect]
What’s this? It’s the same kind of chart, with the same kind of scale (1:1 between polling average and average lead, zero matched to 40%), for the period September 2015 to March 2018.
There are three main periods. In the first, Labour’s polling under new leader Jeremy Corbyn starts low, then gets lower – 32% in September 2015, touching 25% in April 2017. The Tories’ lead over Labour starts at 8% and grows at a bit less than the 2x rate we’d expect, peaking at 20%. (The area with the bite taken out of it represents the period of the Brexit referendum, when some Tory support went to UKIP, and the Tory lead over Labour fell without Labour doing any better.) In April 2017 Theresa May called an election; between then and the election itself, in an extraordinary and memorable short campaign, Labour’s support (as registered in opinion polls) grew from 25% to 36% – a remarkable gain in support, which actually understated Labour’s support in the ballot box. (While Labour’s polling rose by 11%, incidentally, the Tories’ lead over Labour shrank by… 12%: almost none of that gain in support corresponded directly to a loss of support by the Tories.)
After the election, of course, the polls turned round for Labour, although in retrospect low 40%s and a 5% lead don’t look like that much of a honeymoon. It also wasn’t that long a honeymoon. The attacks on the leadership from inside the party may have quietened down for a while after the 2017 election, but the people attacking from outside never stopped – and in the Skripal poisoning they finally found a way of making Corbyn look ridiculous (or dangerous) that actually stuck.
This chart takes us through to December 2019. Between March 2018 and March 2019, Labour’s polling slides from around 40% to 33%, and the Tory lead over Labour grows from 3% to 7%; so far from gaining 2% poll lead for every 1% of support lost, the Tories aren’t even gaining from Labour’s lost support at a rate of 1 for 1. (The Greens and Lib Dems between them put on about 5% in those twelve months, possibly because the Brexit storm clouds were already gathering.) The European elections of 2019 see polling briefly go completely crazy, with support for both major parties collapsing; the Tories’ support collapses further and faster than Labour’s, leading to the bizarre sight of Labour taking a lead over the Tories while polling in the low 20%s. Labour’s polling settles down around 25% and, following the defenestration of Theresa May in July, the Tories’ recovers, giving them a solid 10% lead. In November an election is called. Labour’s polling recovers during a slightly less extraordinary (and very cold) short campaign, but not by enough: up 8% instead of 11%, and without the eventual understatement of votes cast. Strikingly, even less of the support recovered is pulled back from the Tories than in 2017: Labour begins the campaign polling an average of 25% with a 12% Tory lead, and finishes it polling 33%… with a 10% Tory lead. (The Lib Dems and Greens began the campaign polling around 20% combined, and ended it on 10-15% (14% on the day).)
But what is this? The bad times ended in December 2019, surely? Well, not quite. Trivia question for future politics quizzes: under which Labour leader did the Tories poll a 23-point lead in April 2020? The answer is, of course, Keir Starmer. To be fair, from that moment on Labour’s support rises steadily from 28% to 38% (10%), and the Tories’ lead falls from 23% to 5% (18%, very nearly double). In other words, the very first thing Keir Starmer did as leader was pull back support from the Tories, in the way that Corbyn had never done even when Labour were polling in the 40%s. (Given the lack of any correlation under Corbyn between rises in Labour support and loss of support by the Tories, those 8-10% of voters who came back to Labour for Starmer are a bit of a puzzle. I mean, where had they been hiding? One can’t help wondering how many of them had been supporting the Tories in 2019 and even in 2017 – and how differently those elections might have gone if they’d been won over earlier.) After that, though, the chart shows Labour’s support bimbling around in the mid- to high 30%s with a Tory lead of 5-10%, barring a period of neck-and-neck polling in December 2020 (which you may remember as Disastrous Pandemic Mismanagement Month). The leadership seem unable to land a glove on Johnson’s Tory government – or even present themselves convincingly as a more desirable alternative – until November 2021. At which point Partygate erupts, the British press turn on the government, and shenanigans ensue.
And that’s where we came in.
What these charts tell us is that there have only been three periods in the last five and a half years when there’s been a sustained rise in Labour’s support both in absolute terms and relative to the Tories: the 2017 election, the election of Keir Starmer and the election of Liz Truss. (Borderline cases are the 2019 election (which improved Labour’s polling in absolute terms but not relative to the Tories) and the removal of Boris Johnson (which improved Labour’s polling by 7% and their relative position by 15%, but did so over a period of eight months); including them wouldn’t change the argument, though.) Simply listing these events demonstrates how little – under the current leadership – Labour’s popularity has depended on anything the party’s representatives have done: instead, events in the Tory Party and its associated media seem to exert a tidal pull on levels of support for the Tories, with Labour support following as the default alternative. The election of Keir Starmer as leader might seem like an exception, but it really isn’t: the event that made the difference there wasn’t anything Starmer did or said but how he was received by the political-media establishment, and in particular how he was hailed as a decisive break from Corbyn.
As for where we are now, the most recent of these vote shifts, after an unsustainable peak (and the removal of an unsustainable Liz Truss) took Labour from averaging 42% and a 10-point lead to averaging 47% and a 20-point lead. (47-42=5, 20-10=10 – same ratio as ever.) It’s that lead which is now unwinding, with the current averages standing at or slightly below 45% and 16% respectively.
Will the Labour vote share (and lead) go lower? Almost certainly. Will it go a lot lower? Depends what you mean by ‘a lot’, but I can’t see anything militating against Labour support falling back to somewhere around 40%. Which might be OK on the day of an election, but could be disastrous going into an election – particularly for a party which has limited capacity to build its own support and has pinned its hopes on the Tories self-destructing.
On that topic, here’s one final chart.
Crowded and unreadable, I admit, but it’s not as if you’re interested in individual data points. All you really need to know is that the dotted lines mark election campaigns in the course of which Labour’s polling was pretty much flat (up a bit in 2005, down a bit in 2015); the dashed lines are campaigns during which Labour’s polling either went down a little (3-4% in 2010) or went down a lot (8-10% in both 1997 and 2001); and the solid lines are the ones where I went out canvassing with some friends. (Hey, what can I tell you? We knocked on a lot of doors.) The serious points here are that Labour’s polling went up by 6-8% during the 2019 campaign and 8-10% in 2017; that this was very unusual, and can reasonably be related to other things that were unusual about the party in those campaigns; and that, if we’re not going there again, past form says that we’re very unlikely to put on support during the short campaign, and pretty likely to lose it.
Now, you might say that another unusual factor is that there was plenty of room for improvement in the polls on both those occasions, and that this could also be related to the then leadership; you might also say that, when it comes to winning elections, Labour is better off starting in the mid-50%s and losing 8% than starting below 30% and gaining the same amount. All I’d say to that is that polling in the mid-50%s isn’t to be had for the asking – and we don’t want to go into the next election campaign on 42% (say) and lose 8% out of that.
At the moment, glancing back at that asymmetrical red mountain, I’d say that when the music stops and the ski-slope levels out, Labour are likely to be polling in the 38-42% region, with at best a low-single-figure lead over the Tories. And if that’s how we go into the short campaign (whenever it is), I don’t hold out much hope of a Labour majority.
[Update: after this post was published Labour’s polling average and their average lead continued to fall, hitting a low of 43.5% and 14% respectively around the end of April. They then recovered steadily, and are currently (20th May) standing at 45% and 17%. Maybe 45% – and, say, a 15% lead over the Tories – is the new floor. We shall see.]