Since the election, I’ve been wondering about what actually happened – how good a result was it really for Labour? and how bad a result for the Tories? Also, what about Mansfield? A certain kind of centre-left commentator has made hay out of Labour’s loss to the Tories of Mansfield and a few other seats (Derbyshire NE, Middlesbrough S, Stoke-on-Trent S and Walsall N); all of these, along with the by-election loss of Copeland, had been held by Labour for twenty years or more – considerably more in some cases. It’s all very well winning these places like Canterbury and Lincoln and Stroud, the thinking seems to run, but look what’s happening out there in the real Labour seats! Six losses plays 27 gains (in England), but look at the quality of those losses – if we can’t stem the drift of Labour’s core vote to the Tories, flukey wins in Sheffield Hallam and Kensington (majority: 0.03%) aren’t going to save us in the long run.
So what can we say about the 2017 result – and what is going on in places like Mansfield? I’ve been playing around with the figures, and (in the immortal words of Anya Christina Emmanuella Jenkins) I’ve got a theory. But first, let’s ask the real question about what happened in 2017, which is: what happened in 2015? What kind of status quo did that leave us with, and what kind of movement had there been to get us there?
Here goes. The dataset I’m working with consists of all seats in England that were held by either the Tories or Labour when the music stopped: every seat is either a hold (by Tory or Labour) or a gain (ditto). Here’s the overall picture for 2015, as compared with 2010:
X axis: change in the Tory share of the electorate since 2010; Y axis: change in the Labour share of the electorate. (These are not vote shares in the usual sense. I’ve done it this way because I’m interested in how changes in turnout affect the figures.)
Pink triangles: Labour holds; red squares: Labour gains; pale blue diamonds: Tory holds; blue squares: Tory gains.
All clear? As for the trendline, it’s for the Labour holds. I used a polynomial trendline because the curve makes it look like a better fit to the data; I’ve no idea whether there’s any mathematical justification for doing this with data like these.
A few things jump out at us from this chart. One is that 2015 was a substantially better election for the Tories than for Labour: the majority of seats fall in the range from -5% to +5% (Labour vote) and 0 to +5% (Tory vote). Another is that the different series occupy pretty much the same space. There’s some clustering – the seats where the Labour vote fell were mostly held or won by the Tories, and vice versa – but there’s also a lot of overlap: there are Labour holds where the Labour vote fell further than in any seat the Tories won. Oddly, almost all the Tory wins are seats where the Labour vote didn’t fall; they’re clustered in the 0 to +5%/o to +5% box. Labour wins are much more widely distributed. It’s also noticeable that a substantial minority of Labour seats – holds as well as wins – show a really large increase in the Labour vote, 10% and up.
But there’s no show without Punch, and there’s no telling the story of the 2015 election without UKIP. The following chart tells the same story about the same seats, but with the electoral shares for UKIP (and the BNP) added to the Tories for a single ‘Right’ share; I’ve also added the Greens’ share of the electorate to Labour’s. The result looks a bit different:
Now ‘Left’ votes are clustered in the 0 to +5% range, with smaller numbers in the +5% to +10% and +10% to +15% ranges – but ‘Right’ votes are almost entirely in the +5% to +10% range, with a scattering in the +10% to +15% range and above. It’s also noticeable that there are substantial numbers of Tory holds, and even wins, where the Left vote has risen by 5% and more. What we see here, I think, is the collapse of the Lib Dem vote – leading to increases in Left and Right votes – together with the UKIP surge, producing a substantial swing to the Right. This in turn leads both to Tory wins and to Tory holds, where UKIP put the lid on a rise in the Labour vote.
But it’s hard to say much more than that, from these data, about the seats that changed hands. Here are the Labour wins:
Labour won seats in 2015 with changes in the Left vote ranging from +4% to +24%, and in the Right vote from -7% to +8%; it’s hard to make out much of a pattern here, other than that it took a really substantial rise in the Left vote to counteract a rise in the Right vote. In the bottom right corner – Left vote +4%-+8%, Right vote +5%-+9% – the overlap is really substantial, with all four types of seat represented and some contradictory patterns: Chester gained by Labour (Left +5.5%, Right +6.3%); Lewes gained by the Tories (Left +6.5%, Right +5.3%).
Here are all the Tory gains:
Not many Tory gains are to the ‘northwest’ relative to a Labour gain, or above it on the trendline (i.e. showing a higher Left increase and a lower Right). But plenty of them are above a Labour hold, and every one of them is above at least one Tory hold. Two lessons for 2015: firstly, in terms of the swing to the Right, seats that the Tories could actually gain in 2015 looked very much like any other seat; secondly, there was a big swing to the Right. Another election fought by the same parties and on the same ideological battleground could have been very difficult for Labour.
So what happened this year? In terms of Labour and Tory, this happened:
A different box with different corners: still a substantial Tory increase (0-10%) but now the main Labour cluster lies between 5% and 10%. There’s a definite inverse relation between changes in the Labour and Tory votes, with falls in the Tory vote mostly corresponding to higher rises in the Labour vote and the lowest Labour rises corresponding to the higher Tory rises. There are only three seats in England where the Labour vote actually fell in 2017 – one Lib Dem gain (not shown here) and two Tory holds, Richmond Park and Waveney. The first of these was a Lib Dem target; in the second, the Labour vote fell between 2015 and 2017 by 0.5% of the electorate, or 268 votes.
But what about the Left-Right picture?
Ladies and gentlemen, I give you: the collapse of UKIP. Or, to look at it in a more positive light, a swing to the Left. If the Tory vote was mostly clustered between 0 and +10% relative to 2015, the Right vote as a whole is almost entirely contained between 0 and -10%. Labour gains are mostly within the -3% to -6% (Right) and +5% to +10% (Left) bracket, suggesting a combination of a direct swing to the Left and greater mobilisation of Left voters. This impression is confirmed by the clustering of the Labour and Tory holds; intriguingly, Labour holds, as compared to Tory holds, tend to have a higher increase (or lower decrease) in the Right vote as well as, more predictably, the Left vote.
And the Tory gains? I’m coming to them. (That weird one over on the left of the chart – Left up 6.4%, Right down 7.7% – is Clacton, gained from UKIP.) Here are the Labour gains, or most of them; there’s also another anomaly, which I’ll come back to.
That trendline is (still) the trendline for Labour holds; it’s interesting how many Labour gains are bang on it, not to mention how many are below it (i.e. how many seats were gained from the Tories despite the Left performing worse than they tended to do in seats Labour held). If I were a Tory this chart would worry me quite a lot; not only are Labour gains interspersed among Tory holds – as Tory gains were among Labour holds in 2015 – but most of them are below the trendline. This suggests that more focused mobilisation next time could really pay off. Turnout was up generally as against 2015, but the increase in turnout was highest in seats Labour held – and lowest in seats the Tories held.
But of course none of this answers the question we began with, the question Labour absolutely must answer if it’s ever to form a governmentget centre-left commentators off its back: what was going on in Mansfield? This. This is what was going on:
The highest blue square is Southport – a Tory gain from the Lib Dems on a three-way split, and hence not really part of the story I’m telling here (apart from noting that Labour pushed the Lib Dems into third place, where in 2015 the Lib Dems held the seat and Labour were not only third, but less than 2% ahead of UKIP). Moving down, the next blue square you come to is Stoke-on-Trent South – Left +6.9%, Right +0.004%[sic]. UKIP didn’t put up a candidate in Stoke-on-Trent South, having got the votes of 12.2% of the electorate in 2015; the Tory electoral share rose by 12.2% and they fluked a win.
The other five seats – Mansfield, Derbyshire NE, Middlesbrough S, Walsall N and Copeland – are best defined by their relation to the ‘Labour hold’ trendline: they’re a long way below it. Looking at the details, Middlesbrough S had no UKIP candidate and a rise in the Tory vote which didn’t quite fill the gap, as witness the drop of 0.9% in the Right vote; the other four all saw a collapse in the UKIP vote together with a rise in the Tory vote which more than compensated for it. Meanwhile the combined Labour and Green vote also went up, but only by between 2% and 3%. This, more than anything, is what singles out those five seats: there was a nationwide trend for Labour-held seats – involving the Left vote rising by between 6% and 12% while the Right vote fell by anything up to 8% – and they’re way below it. These are the outliers; they’re the ones that haven’t performed the way they should have done. It would take some intellectual contortions to argue that it’s the five underperforming losses – rather than the 27 gains or the 200 holds – that are typical of Corbyn’s Labour or crucial to its future. It’s hard not to feel that a bit more mobilisation could have made all the difference – Derbyshire NE has the biggest Tory majority of the five, and 1500 more Labour votes would have made it a Labour hold; 600 would have done the job for Mansfield. A few more Labour votes and they’d have been back in the main cluster – which is to say, a few more Labour votes and they’d look like all the other Labour holds.
As for why these five seats under-performed, different constituencies will have different stories, but it is striking that two of the five – Copeland and Middlesbrough S – were represented in 2015 by MPs who left Parliament rather than fight an election under Corbyn’s leadership (Jamie Reed and Tom Blenkinsop respectively). Of the three MPs who did stand in 2017, David Winnick (Walsall N) had forecast electoral disaster if Corbyn remained leader, while Natascha Engel (Derbyshire NE) is on record as being a fan of Maurice Glasman. Alan Meale (Mansfield) has a radical past but does not appear to have placed his views on Corbyn on the record – although the notorious 2016 ‘league table‘ placed him in the ‘Core group negative’ column, with Ben Bradshaw and Gloria de Piero. (Another ‘Core group negative’ was Rob Flello of Stoke-on-Trent South, who publicly called on Corbyn to resign after the EU referendum.) Even if these MPs strained every sinew to get Labour returned in 2017 – as I’m quite prepared to believe that they did – their opposition to Corbyn’s leadership was no secret; and in our current, quasi-presidential political culture, that was bound to cost the party votes (if they don’t support him, why should I?).
It’s noticeable, finally, just how unusual an area those five seats are in. They’re in the -1% to +5% (Right), 0 to +3% (Left) box, along with only eight other seats: two Tory holds, five Labour holds… and one Labour win. (For completeness, the Labour holds are Ashfield, Leigh, Hull W, West Bromwich W and Bolsover. The last of these was singled out by John Mann MP in his own “Labour heartland” polemic; Mann’s own seat, Bassetlaw, is just outside the box, on Right -1.1%, Left +2.5%. The lesson Mann draws, incidentally, is that “[t]he Labour Party is nothing if it does not represent the aspirations of the white working class in industrial areas”, therefore Corbyn must condemn the IRA, endorse shoot-to-kill and drop his opposition to nuclear weapons. I guess the workers of Bolsover mainly aspire to shoot terrorists and bomb North Korea.)
I think the main lesson of this corner of the chart is that, when you’re in a four-party system with differential levels of mobilisation, and when you haven’t got the momentum of a good chunky electoral mobilisation campaign behind you, electoral politics in a plurality-based system can be very chancy indeed. Looking at the four rightmost seats on that chart and reading from left to right: Right +2.2%, Left +2.5% gets you Ashfield; Right +3.1%, Left +2.5% gets you Copeland (Tory gain (relative to 2015)); Right +3.3%, Left +2.2% gets you Thornbury and Yate (Tory hold); and Right +4.4%, Left +1% gets you Jared O’Mara MP. Sheffield Hallam was also a seat where Labour was under-performing relative to the national trend, presumably because nobody had prioritised it as a potential target; Labour’s vote rising just enough, and the Tories taking just enough of a bite out of the Lib Dem vote, gave us a Labour majority of 2000 and a new MP whom nobody had expected, himself included.
What happened in Sheffield Hallam? Nick Clegg lost it. What happened in Mansfield? Alan Meale lost it. The Labour vote on June 8th? 40%; up 9.6% on 2015. (To put it another way, 27.5% of the electorate voted Labour in 2017, the highest share of the electorate the party has achieved since 1997; the comparable figure for 2015 was 20.2% (up from 18.9% in 2010).) Who won that? We did; the Labour party united behind its elected leader did it. The mean level of Labour support over the twenty opinion polls conducted since the beginning of September? 41.9%. Will it go higher? Yes.
Correction – in an earlier draft I misidentified the MP for Copeland, who left Parliament for a job in the nuclear power industry and triggered a by-election rather than continue to serve his constituents under the leadership of Jeremy Corbyn, as John Mann rather than Jamie Reed. John Mann has been MP for Bassetlaw since 2001; he was re-elected in 2017 with 52.6% of the vote on a 66.5% turnout.
Workers’ Playtime 