Some commentators on Twitter and in the media are saying that Labour lost the 2019 General Election because it lost too many votes to remain parties, and that failure to retain support from remainers was the problem. Angry Labour activists on Twitter have been listing off the remain seats that were lost, and saying that a strong remain strategy would have saved the party.
This is completely wrong, and I will show this using data from the 2019 election and the 2016 referendum.
Methods
First I used the dataset of constituency-level results I assembled over the weekend, which contains results for 339 constituencies, semi-randomly sampled from the list of all constituencies on the BBC election site and linked to leave voting data from the 2016 EU referendum. The detailed methodology for assembling this dataset is given here. I then assembled a separate data set of only the seats Labour lost, using this handy (but not quite alphabetical) guide from the Metro newspaper. I merged these with EU referendum data.
Using the full constituency data set, I created a logistic regression model of probability of retaining a seat against constituency leave vote, for all the seats that were held by Labour at the 2017 election. I plotted the predicted probability of losing a seat against the proportion of the population in that seat. Then, I conducted a crosstabs and chi-squared test for the seats held by Labour in 2017, showing the probability of losing a seat in 2019 by whether or not it was a leave-voting constituency. I defined a “leave-voting constituency” as any constituency voting above the median leave vote (which was 53.55).
Next, using the data set of the 59 constituencies Labour lost, I calculated the mean vote in this set of constituencies, and the proportion of constituencies that were leave-voting constituencies. I compared this with data for all Labour held seats that were not lost in the 2019 election.
Results
In my constituency data set there were 142 seats held by Labour in the 2017 election, of which 30 (21%) were lost in the 2019 election. Figure 1 shows the cross tabulation of leave seats with seats Labour held in 2019[1].
As can be seen, 92% of remain seats were held, compared to 66% of leave seats. This is extremely statistically significant (chi-squared statistic 14.35, p<0.001). That’s a nasty sign that the main risk of losing a seat was that it was a leave seat, not a remain seat.
We can show this explicitly using logistic regression. Figure 2 shows the predicted probability of a seat being held by Labour in 2019, plotted against the proportion of the seat that voted to leave in the EU referendum. The red dots on this figure indicate whether it was held by Labour in 2019: red dots on the top of the figure are seats retained, plotted at the value of their leave vote; red dots at the bottom are seats that were lost, plotted at the value of their leave vote.
This model was highly significant, and showed that every 1% point in the leave vote reduced the odds of Labour holding the seat by 7%. Note that this figure includes Scotland, so the results might be slightly different if only England were considered, but even the strongest remain-voting seat that was lost – even were it in Scotland – is well above the remain vote of some seats that were held. This model shows that at the extreme end of the leave spectrum, up above 60% of the electorate voting for leave, the probability that Labour retained the seat dropped to around 50%. That’s terrible!
My constituency data set contains only 142 Labour seats, and 30 seats that were lost, but actually 59 seats were lost. Since my data set is semi-random, there is a small chance that it will misrepresent the results. So I checked with the dataset of all seats that were lost. This data set contains 59 seats. Here are some basic facts about this data set, and comparisons with the constituency data set and the full list of Labour-held seats:
- Labour lost 14 remain-voting seats (24% of all seats lost) and 45 leave-voting seats (76%). This is very similar to my crosstabs, where 24 of 30 seats lost (80%) were leave
- The average leave vote in the 59 seats that were lost was 57.7%, slightly above the median, ranging from 31.2 – 71.4%.
- In contrast, the average leave vote in the 112 seats in my constituency data set that Labour held was 48.8%, ranging from 20.5 – 72.8%
- The average leave vote in all seats held by Labour going into this election was 51.1%, ranging from 20.5 – 72.8%
This is clear statistical evidence that Labour went into this election having a slightly remain-leaning set of constituencies, primarily lost leave-voting constituencies, and emerged from the election even more remain-focused than when it went in.
Conclusion
Labour did not lose this election because of a large swing in votes to the remain parties. It did not lose a large number of remain-voting seats, but was decimated in the leave-voting areas. Labour held on to all of its most heavily remain-focused seats. In attempting to appeal to both leavers and remainers, Labour managed to retain most of the remainers and lose a lot more leavers. Labour emerged from this election even more remain-focused than it was when it went in[2]. There are some very simple reasons for this:
- The swing to the Tories and away from Labour was much bigger in leave-voting seats
- The Brexit party was only active in Labour-held seats, and got its largest vote share in the strongly leave seats
- The swing to the Lib Dems was much less closely related to the leave vote than was the swing away from Labour (see my last blog post, Figure 4)
- The intensity of the relationship between leave voting and swinging to Lib Dems was lower in Labour-held seats than Tory-held seats (see my last blog post, Figure 4)
In trying to please both sides of the Brexit divide, Labour failed to satisfy the leavers. Pro-brexit Labour voters were simply much, much more committed to Brexit than pro-remain Labour voters were to remain, and so Labour lost the leave areas. There are lots of remainers out there who want to claim that remain is wot did it, but they are simply wrong. I’m super pro-remain myself, but the data makes it very clear: British Labour voters want to leave, and they were willing to pack in their allegiance to the Labour movement to get that done. Whatever you might think of their politics, that is the simple hard fact of the electorates Labour represented.
It’s worth noting that in 2017 Corbyn campaigned on Brexit. The Labour manifesto explicitly accepted Brexit and said Labour would negotiate and leave. At that election Labour won a historically high share for a party in opposition, a higher share of the vote in fact than Blair won in 2005 (when he retained government). In that election they came within a bees’ dick of winning government, and in that period before Corbyn accepted the compromise of a second referendum two Tory PMs left, and Johnson only held onto government by the skin of his teeth (recall there was talk of a unity government). Blair and Cameron have both shown it’s possible to hold government with 35% of the vote, so it’s perfectly possible that had Corbyn gone into this election on a leave platform he would have seen a much smaller swing against him, and could have won it. We don’t know, but on the basis of all the evidence here it seems like the second referendum policy was a disaster for Labour.
This gives two clear lessons for Labour to take in over the next few years and as they choose a new leader:
- Labour’s policies and Corbyn were not the primary problem, and dropping them is not going to help. Obviously Corbyn is going to go, it’s traditional, but the manifesto’s policies were not the problem. The Labour right is going to push for the party to throw the Corbyn years down the memory hole (in today’s Guardian we have Suzanne Moore begging for a vet to “sedate” the Corbyn supporters!), because they are and have always been intent on fighting these genuinely left wing policies. Ignore them, and stick to the real Labour platform that will really help the country as it recovers from the horrors of this Tory leadership
- Labour – and the British left generally – have to get over Brexit. There is no option left to remain, and no chance it will ever happen now. The Labour right want to claim that Corbyn doesn’t understand working class voters but his original policy – of full-throated Lexit – was much more in tune with what ordinary working class Labour supporters want than anything that the Blairist rump have to say. The debate now for Labour has to be about the type of Brexit, and how to make it work. This means fighting Johnson’s bullshit deal, but on the basis that they can make a better one – obviously this doesn’t matter now but it is the job of the opposition to hold the government to account, and they should do so from the clear perspective of their voters, that Brexit has to happen. This is going to be hard for some of the urban remainers from the south and east, but that’s life if you’re a politician. Further talk of remain just has to end
For 20 years the EU was a thorn in the Tory side, constantly causing them trouble. Cameron ripped that thorn out with this referendum, and although May spent some time botching the healing process Johnson has patched up the damage and squeezed out the last remainer pus from the Tory body politic. If Labour don’t face the reality of Brexit and what it did to this party at the 2019 election, then the issue will fester for them – as it did over so many years for the Tories – and hold them back just as it did the Tories. It is time for Britain to move on from Brexit, and for the Labour movement to accept the reality of the disaster that is coming. Once people realize how Johnson’s Brexit has screwed them, they will turn to Labour – and Labour needs to be ready with a transformative, genuinely left wing agenda in order to recapture its heartland and do what is right for working people. Corbyn was right about Brexit, right about the policies Britain needs, and after he is gone he will still be right about what has to be done. Don’t repudiate those lessons, and in the process destroy the movement.
fn1: My apologies for pasting this as a picture directly copied from Stata, instead of making a nice pretty table – I hate it when people do this but it’s late and I hate making tables in html. Stata offers an option to copy as html but it doesn’t work. Sorry!
fn2: This final conclusion is shakey because it depends on my constituency data set, and I don’t know if it would still be true once all the remaining Labour-held seats are entered into the dataset. I think it will, but there’s a chance the final data set will end up the same level of leaviness as the 2017 constituencies, statistically speaking. But this conclusion is not a very important one anyway, so it doesn’t matter if it isn’t held up by the full dataset.