The rise and fall of the Brexit realignment

Marching up the Brexit hill, then marching down again

Last week’s local elections were bad for the Conservative government on any measure, with over 1,000 council seats lost. When we dig into the detail of the results, there is evidence of further bad news for an incumbent government due to face the voters with 18 months: the electoral coalition which delivered an 80 seat majority in 2019 is unravelling.

Labour’s successes last week in many strongly Leave voting areas, from the “red wall” councils of the North East to the town halls of strongly Leave Kent, provided strong hints that Labour is bouncing back in the Brexit heartlands. But the overall picture was muddied by the unusual baseline for these elections. Nearly all the seats in the 2023 local elections were previously fought in May 2019 - during the chaotic final weeks of the May administration. A “plague on both your houses” mood was in the air then, and both Labour and the Conservatives did poorly. It was only six months later that Boris Johnson united Leave voters behind the Tory banner with his pledge to “Get Brexit Done”.

A comparison of May 2019 with May 2023 therefore isn’t a good test for Brexit unwind - the 2019 locals took place before “peak Brexit” had occurred. Yet thanks to the unusual nature of English local elections, with many councils electing in “thirds” and holding contests in three out of every four years, we can split the past four years up to bring things into sharper focus. COVID prevented local elections from happening at all in 2020, so the first round of local contests fought in the wake of the 2019 general election came in May 2021. These were the “peak Brexit” local elections - the Conservatives’ general election advance in Leave areas was now replicated in local contests, with the Johnson government winning a swathe of red wall councils and taking Hartlepool from Labour in a Westminster by-election.

The 2021 locals were thus the high point for Boris and Brexit in the Conservatives, and Labour’s lowest ebb under Keir Starmer. By splitting the past four years in two, and looking at change from 2019-2021 and then from 2021-2023, we can answer two questions. First, how high did the Brexit tide rise in 2021? Second, how far has it fallen back since? The change in party fortunes since 2021 likely provides a better proxy for Labour’s progress in Leave voting Westminster constituencies, as the 2019 General Election, like the 2021 locals, were a “peak Brexit” election.

To investigate these questions I use data from the BBC elections unit’s “key wards” dataset, a sample of all wards from 45 local authorities who elect councillors annually in “thirds”, where ward boundaries are unchanged since 2019 (and usually since 2015 too) - this is a total of 794 wards.¹ I focus on the wards the Conservatives, Labour, the Liberal Democrats and at least one other party fought in both elections, to limit the effect of churn in candidates on the results. To test the strength of Brexit alignments, I make use of statistical estimates of ward level referendum voting patterns specially compiled for the BBC key wards analysis.²

Up the Brexit Hill: Changes in Conservative and Labour voting, and Conservative to
Labour swing, 2019-2021

Source: BBC key wards dataset. Wards with Conservative, Labour, Liberal Democrat and at least one other candidate in both elections

As the above chart makes clear, Leave voters marched a long way up the Brexit hill with Boris Johnson in 2021. The average increase in Conservative support was ten points plus in all the wards with above average Brexit support, and close to 16 points in wards where the estimated Leave vote was above 63 per cent. Meanwhile, there is no evidence that Labour made any progress in unwinding their own Brexit divides in 2019-21. The party made modest gains in vote share across the board, around two points on average, and a little less in the most Brexity wards. The result was modest swings to the Conservatives in Remain areas, but whopping Labour to Tory swings in the Leave heartlands. No wonder Johnson and his allies were so jubilant in the summer of 2021.

…And Down Again: Changes in Conservative and Labour voting, and Conservative to Labour swing, 2021-2023

Source: BBC key wards dataset. Wards with Conservative, Labour, Liberal Democrat and at least one other candidate in both elections

The Leave voters who marched up the Brexit hill have marched a long way down since, as my second chart reveals. The Conservatives have lost substantial support everywhere since 2021, but the average Con decline in the most strongly Leave wards is, at nearly 13 points, almost twice the 7 point fall registered in the most Remain wards. And while there was no clear Brexit pattern in Labour’s 2019-2021 performance, there is a very clear Brexit effect in this second period. Labour gains steadily rise as wards become more Leave leaning - from a 2 point gain in the most remain wards, to 5 points in evenly divided wards, and 7.5 points in the Leave heartlands.

This is clear “Brexit unwind” - the more Leave a ward is, the more the Tory tide goes out, and the more the Labour tide comes in. The 2021-23 Conservative to Labour swing in Remain areas is a healthy 5 points, but in the strongest Leave wards it is twice as high. The ten point swing to Labour in the most Leave voting wards since 2021 is equivalent to the swing Tony Blair achieved in the 1997 general election.

The evidence from the 2021-2023 local elections thus suggests the Brexit divide is now rapidly fading. But does this mean a return to the pre-referendum status quo ante? As part of the key wards exercise, we also gathered data on voting patterns in the May 2015 local elections, conducted on the same day as the general election when David Cameron achieved a surprise majority. This gives us a picture of the political landscape before voters had even heard the term “Brexit” (and also before Jeremy Corbyn was known for anything other than serial backbench rebellion).

No going back: Changes in Conservative and Labour voting, and Conservative to Labour swing, 2015-2023

Source: BBC key wards dataset. Wards with Conservative, Labour, Liberal Democrat and at least one other candidate in both elections

This picture looks rather different. When we use a peak Brexit baseline, the Conservatives are down most, and Labour up most, in Leave areas - Brexit unwind on both fronts. But when we use a pre-Brexit baseline, we find Brexit no longer matters for Labour performance but it still matters for the Conservatives. Labour’s gains on 2015 are evenly spread - they are up by between 6 and 10 points everywhere, with no clear Brexit pattern. The opposition are advancing on a broad front compared with 2015, their final pre-Brexit defeat.

But for the Conservatives Brexit has left a clear, and clearly negative, legacy: the 2023 Tory share in more Remain wards is far below what the party achieved under David Cameron. Conservative shares in 2023 were down almost 14 points on their showings in 2015 in the most Remain wards, while in the strongest Leave areas they are down only one point on their 2015 share.

The Conservatives achieved their large 2019 majority by advancing strongly in Leave areas while limiting their losses in Remain areas, helped by hostility to Corbyn in the latter. Local election patterns suggest trouble on both fronts. The party’s advance in Leave areas is unwinding, but its retreat in Remain areas is still gathering steam. While we have to be cautious when projecting from local contests to general elections, these patterns do suggest that the voter coalition Boris Johnson assembled may be coming apart at both ends. That could presage an explosive final act to the long Brexit electoral drama in next year’s general election.

1. The key wards exercise is a team effort, involving myself, Sir John Curtice, Patrick English, Stephen Fisher, and the BBC elections team headed by Julia Walker. This analysis would not be possible without their efforts. Any errors of fact and interpretation are however mine alone.

2. I am grateful to Patrick English and Jonathan Mellon for generating these estimates.