As you’ve probably noticed, Labour is doing well in the polls at the moment. One polling result that got a lot of exposure recently was this one:
The pollsters – J L Partners – hail from Downing Street, no less; James Johnson was previously a SpAd to Theresa May and Rory Stewart.
What I found particularly interesting about this was the reference to “45 Red Wall seats” – the constituencies in which the polling had been carried out, presumably. Could this be a definitive answer to the old question, what is the Red Wall?
Well, (a) it’s not that old a question, and (b) yes it could, sort of – although this is, to my knowledge, the fourth distinct version of the “Red Wall”, so it could all change again. (Update 15th February: it turns out that this was actually the sixth version; see below for details.)
Let’s go back a bit. (NB Some overlap with my earlier series of posts, but at least this way it’s all in one place.)
Red Wall v0: to August 2019
The Red Wall as we know it is the creation of a right-wing think-tanker, an FT writer who previously worked for the Telegraph and the Spectator and an FT dataviz specialist, with additional contributions by Downing Street advisors. And the Red Wall is something we didn’t know – at all – until relatively recently. Up to the middle of August 2019 – less than four months before the election where the Red Wall would feature so prominently – the Red Wall as a political concept didn’t exist; the only people who talked about a Red Wall on a regular basis were Wales football supporters (not shown here).
And then there was
Red Wall v1 (August 2019)
In August 2019, James Kanagasooriam of right-wing think tank Onward identified four groups of seats where the Conservative Party tended to under-perform relative to what the demographics of the area would lead one to expect. One of the four was
a huge “red wall” stretching from N Wales into Merseyside, Warrington, Wigan, Manchester, Oldham, Barnsley, Nottingham and Doncaster. When you talk about cultural barriers to voting Tory – this is where it is. This entire stretch shouldn’t be all Labour but is
This Tweet was accompanied by a map showing 46 constituencies. Removing one Lib Dem seat and a number of seats that had either changed hands multiple times or only been formed relatively recently – and where, either way, we can’t presume those “cultural barriers to voting Tory” applied – gave 39 seats. Here’s what happened to them in December 2019:

Five of the 39 went Conservative, one of them (Leigh, in paler blue) quite narrowly – and one of the remaining four was Bassetlaw (chequered), whose sitting MP had left the Labour Party and was actively campaigning against it at the time of the election. Cultural barriers one, demographics nil.
But the really odd thing about this, first version of the Red Wall, at least in retrospect, is how little traction it got: nobody really picked up on it at all.
Red Wall v2 (October 2019)
Not, that is, until the end of October, by which time the December election had already been called (and was less than six weeks away). It was then that Kanagasooriam – in the context of a report about something else entirely – revived the Red Wall; now it referred to
a belt of sixty seats in the North and Midlands which the Conservatives have never won. They include places like Wakefield, Great Grimsby and Penistone and Stockbridge. Termed elsewhere as the ‘Red Wall’ by the framework’s author James Kanagsooriam, it is made up of a mixture of constituencies which for demographic reasons have always been quite marginal but have consistently remained Labour; constituencies where the Conservatives significantly increased their vote share in 2017 but didn’t win; and a scattering of seats with five figure majorities but which could be become marginal because of voters’ strong pro-Brexit views.
This more expansively defined group was itself said to be one of three groups of seats which would be determinant of the election result, totalling 109 battleground seats – 60 ‘Red Wall’, 38 ‘Uniform National Swing’ and a cluster of eleven seats in Wales. It’s not clear which were seen as the Red Wall seats, though; while the JRF report listed all 109, it didn’t break them down into the three sub-groups. Sebastian Payne’s book Broken Heartlands does include a table supplied by Kanagasooriam and itemising the ‘Red Wall’ and ‘Uniform National Swing’ seats; however, the table only lists 43 ‘Red Wall’ seats, and several of those listed are not named in the JRF report. In search of a definitive list, I put the two lists together and took out any seat listed under ‘Uniform National Swing’ and anywhere south of the Midlands, then did a bit more tidying-up. It seemed to me that if we were going to talk about seats that the Conservatives have never won, “never” ought to mean something; strictly speaking a seat that was created in 2010 and won by the Tories in 2017 had never been won by the Tories up to that point, but it’s not the impression that word gives. So I removed any seat that had come into existence since 1983, and any seat that had been held by the Tories or Lib Dems at any time between 1983 and 2017.
At the end of all that I didn’t have a list of sixty seats, but I did have 45; and here they are. I give you the Red Wall, version 2, late October 2019. The dark blue seats (14 of them) are big Tory wins; the mid-blues (11) are narrow wins; the chequered area is Bassetlaw, whose sitting Labour MP was campaigning against the party by the time of the election; and the remaining 19 are bricks in the Red Wall that unsportingly stayed red.

It’s… not that much of a wall, really, is it? It’s an awfully long way from Birmingham Northfield to Blyth Valley – 230 miles, in fact – and neither of them has much in common with Blackpool South, Don Valley or Great Grimsby, or Workington for that matter. Apart from being (a) Labour seats up to 2017 and (b) Up North, that is.
Shortly afterwards it was decided – by Kanagasooriam, James Burn-Murdoch of the FT or both – that anything called a ‘wall’ really ought to look a bit like a continuous series of blocks leading from A to B. The FT duly publicised a third iteration of the Red Wall, which is partly a Lancashire/Yorkshire/Midlands slice out of the map above and partly… not. As you’re about to see.
Red Wall v3 (November 2019)

This is what the FT described as “a near-contiguous span of 50 Labour-held seats stretching from the Vale of Clwyd in North Wales to Great Grimsby on the East Coast”. The 43 English seats shown on the FT‘s accompanying map are above, colour-coded according to what would happen in December. Again, grey chequers indicate a seat whose former Labour MP was campaigning against the party; again, deep blue is a big Tory win (9 constituencies), the mid-blue is a narrow win (7) and the red are Labour holds (15). The remaining 12, in the pale blue, are seats that went Tory in 2019 but hadn’t consistently been Labour since 1983 – and consequently didn’t feature in previous versions of the Red Wall. The point of the Red Wall rhetoric, let’s not forget, was that these were Labour strongholds which were now tumbling due to the waning of tribal loyalties. The mid-blue seats – many of which are constituencies where the Tories squeaked a win, as parties having a good election campaign often do – are already a poor fit with this model; the pale blue seats depart from it altogether. In this iteration, “Red Wall” didn’t mean much more than “Conservative targets north of the Wash and south of Morecambe Bay”.
Red Wall v4 (December 2019)
(Updated February 2022) Up till now I’ve somehow missed this morning-after Telegraph story – variously headlined “The 24 Labour heartland seats lost to the Tories for the first time in decades” and “The 24 Labour heartland seats lost to the Tories for the first time” tout court (as we’ll see, the latter is actually more accurate). The seats (22 in England, two in Wales) are listed in a table headed “Fall of the Labour wall”, so I think this listing deserves its place in the genealogy of the Red Wall.

22 wins, right enough; but they weren’t all big wins, they weren’t all seats that had been Labour for longer than a decade or so, and some of them were won with assistance from the former MP. Also, with the exception of Stoke-on-Trent (whose politics have been decidedly troubled for some time), these constituencies look less like “Labour heartlands” than rural and semi-rural seats where Labour supporters had been in the majority for historical reasons. (Which, to be fair, was more or less what James Kanagasooriam was getting at to begin with, even if he later helped bend the concept out of shape.)
Red Wall v5 (The Definitive Red Wall) (September 2021)
(Updated February 2022) It’s been brought to my attention that James Kanagasooriam has not only identified the seats making up the Red Wall but explained how they were selected, in an article in Political Insight co-written with Elizabeth Simon and modestly entitled “Red Wall: The Definitive Description”. All right! Let’s get some political science on this thing!
The Red Wall, in this telling, began with a demographic model predicting the level of the Conservative vote in a given constituency – factors such as deprivation (negatively correlated), higher education (also negatively correlated but less strongly) and the proportion of residents in managerial positions (positive correlation). Constituencies not held by the Tories in 2017 were assessed according to whether they had an anomalously low score on this model, as well as three other factors: a Conservative vote share over 25% in 2017, a swing of over 5% to the Conservatives between 2010[sic] and 2017, and a greater than 55% Leave vote.
70 of the 269 eligible seats hit all four criteria. This group was then winnowed down by excluding seats outside England (and perhaps making other unspecified “geographic exclusions”), as well as excluding seats “deemed too unlikely to switch allegiance”; this gave 28 seats. From the pool of constituencies meeting only three of four factors, another 11 were “designated part of the Red Wall through qualitative selection” (Kanagasooriam doesn’t mention how many were in this pool); finally, “a further three seats, which met two or less of the criteria, were also included based on geographical proximity to other Red Wall seats”.
The 42 included the Speaker’s seat of Chorley; excluding Chorley gives 41 seats, as follows.

That’s 30 out of 41, although only 13 of the 30 are big (dark blue) wins – and again, the East Midlands excepted you’d be looking at that map a long time before you thought you were looking at a ‘wall’ of any kind.
In the discussion section of the paper, Kanagasooriam suggests that his results would have been even better if he’d trusted the data more: none of the three seats that were added on “geographical” grounds (despite only ticking one or two boxes) went to the Tories in 2017, while two of those that hit all four marks but were excluded as “unlikely to switch allegiance” – Leigh and Redcar – did. It’s nice to see a researcher own up to fudging the data, but Kanagasooriam’s suggestion that an unfudged version would have been more accurate isn’t borne out by the data he presents.
The problem is that, as soon as any judgment calls were made on inclusion or exclusion, the whole sample was fudged (to put it euphemistically). What we really need to know is the content (and hence the hit-rate) of all the subsamples – the 28 constituencies that were judged to be ‘true’ Red Wall seats; the 42 that weren’t despite hitting all four marks; the 11 that qualified on three criteria and were added to the sample; the unknown number that qualified on three criteria and weren’t added to the sample; and, of course, the three erroneous ‘geographical’ choices. If all the judgment calls had been omitted, Leigh and Redcar would certainly have been on the list, but none of the 11 added at the third stage would have been. In any case, the list would have numbered 70 constituencies – which, given that the Tories only made 48 gains from Labour in the whole of England, would be bound to bring down the Red Wall’s hit rate and hence its predictive accuracy.
Red Wall v6 (the pollsters’ Red Wall) (January 2022 and doubtless earlier)
So far the December election result has seen the Tories win 5 out of 39 Red Wall seats, 26 out of 45, 28 out of 43, 22 out of, er, 22, and 30 out of 41. The changing meaning of the concept is clearly closing in on the actual result – although the big, eye-popping, “dude where’s my core vote?” victories account for 3 of the 39, 14 of the 45, 8 of the 43, 12 of the 22 and 13 of the 41. (Needless to say, there have been varying degrees of overlap between the 39, the 45, the 43 and the 41.)
You may well be wondering how it can be that opinion polling shows the Tories potentially losing all but three of their 45 Red Wall seats. The three they’re projected to hang on to are familiar enough – Dudley North, Bassetlaw and Great Grimsby, or one former deep blue seat and two chequered in defectors’ grey (interesting in itself) – but where had J L Partners found another 42 Conservative gains? Particularly since, as just noted, the Tories only made 48 gains from Labour in the whole of England (and one loss)…
Hold on to that thought. Here’s the pollsters’ Red Wall.

The list of 45 seats published by the pollsters includes one in Wales (Delyn) which I’m ignoring. The other 44 are shown here, with the usual colour coding. And, wouldn’t you know it, the Tories won all 44! Anyone wondering if there was perhaps a touch of the Texas Sharpshooter about one of the earlier versions can relax – that’s all this is. Red Wall = Tory gain, Tory gain = Red Wall, with a handful of exceptions – in fact the only Tory gains in England not forming part of the Red Wall are Kensington, Stroud, Ipswich and Peterborough, which presumably weren’t considered “Northern” enough. (Scare quotes used because Peterborough is actually on this map – by latitude it’s slightly North of Birmingham.)
The phrase “Red Wall” now means nothing more than “one of the seats the Tories won from Labour in 2019” – which is to say, it means nothing.
Postscript: Is there a real Red Wall?
No. No, there isn’t. Stop it now. Put the psephological buzz-phrase down.
What there is – and what is quite interesting – is a relatively small group of seats which had genuinely been Labour for a long time, and which genuinely went Tory in a big way in 2019. The ‘deep blue’ seats in all the above maps are defined as long-term Labour seats where the 2019 Tory majority was 5% or more and the Labour vote had fallen by 10% or more relative to 2017 and the Labour vote was lower than at any time since (and including) 2001. These three criteria do seem to identify a real phenomenon, setting these seats apart from the ‘mid-blue’ seats (long-term Labour seats won by the Tories in 2019 but where one or more of those factors don’t apply). The highest Labour vote in 2019 in a ‘deep blue’ seat was 39.8%; the lowest Labour vote in a ‘mid-blue’ seat was 39.3%.
On investigating the deep-blue seats more closely I found that almost all of them showed a similar pattern over the previous three elections, with Labour’s margin over the Tories going down in both 2010 and – most unusually – 2017. There are, in point of fact, only 50 Labour constituencies (of 232) where Labour’s margin fell in 2017 – anyone who looked at that election with a degree of objectivity would have to say that 2017 was a good result in lots of ways (as long as they can silence the nagging voice saying yeah but we didn’t win did we…).
Here are those 50 seats.

Key: as before, except that seats with defectors aren’t singled out any more. There’s also one seat – Scunthorpe – in deep purple; this was a ‘deep blue’ where Labour’s margin over the Tories didn’t go down between 2015 and 2017 (it went up by 0.05%).
What does this tell us, though? I think it tells us that, while substantial numbers of people in the East Midlands and the North East (and a couple of other areas) didn’t really feel the love for Corbyn, this only created the opportunity for a major (local) political upset when other factors were present. In a previous post I suggested that the distinctive characteristics of the ‘deep blue’ constituencies were the presence of an sizeable anti-system, “none of the above” protest vote, together with the legitimation of the far Right as a vehicle for protest votes. Both of these localised trends – established in those seats since the New Labour years – made it possible for substantial numbers of voters to switch from the Lib Dems to UKIP in 2015, and for the Conservatives to attract a majority of those voters in 2017. This led to a reduced Labour majority in those seats – and a lot of publicity for some of the affected MPs, who blamed the result on the new leader (although, ironically, under a more ‘establishment’ leader than Corbyn the minority of anti-system voters who returned to Labour might have been much smaller). In 2019, Labour’s Brexit positioning left it looking like the ‘establishment’ party, standing in the way of the Tories’ endorsement of the revolt of 2016; as a result it lost whatever ‘protest vote’ credibility it still had, and lost ground to both the Tories and the Brexit Party. The result, in those 17 constituencies, was a dramatic collapse in the Labour vote.
But note: in those 17 constituencies. The Tories made 47 net gains in England; if they’d made 30 in the entire country they would still have come out with a solid majority. The map immediately above tells us where in England Labour’s margin over the Tories fell – against the current – in 2017, whether or not the Tories won those seats two years later and if so, how well; the map above it tells us where in England the Tories won a Labour seat in 2019, and what kind of a win it was. There’s a lot of overlap, but there are also an awful lot of differences; and lots of the wins – enough wins to make an election victory – were narrow, chancy, unpredictable wins. (They still count, that’s the thing.)
What this whole exercise tells us is that the Red Wall is useful as a concept if you define it tightly enough, but that what it’s useful for is telling us why Labour lost some of the seats it did in 2019 and, perhaps, where similar factors might apply in future. What it definitely doesn’t tell us is “why Labour lost”; this is a small, untypical group of seats, meaning that any reorientation of the party based on the idea of “winning back the Red Wall” would be disastrous.
We can also see that, in practice, the concept of “Red Wall” has steadily converged on that of “Conservative target seat”, to the point where it’s now more or less synonymous with “Conservative gains from Labour in 2019 in England”. (It’s precisely synonymous with “Conservative gains from Labour in 2019 in England, north of 52.4 degrees N, with the exception of Peterborough”, but that’s not quite as snappy.)
Once more with feeling: there Is No Red Wall. If you mean “seats the Conservatives gained from Labour”, say “seats the Conservatives gained from Labour”. If you mean “longstanding Labour seats in the North and the Midlands that the Conservatives gained from Labour by unexpectedly large margins”, say that – but be aware that:
- The North is a very big place; it’s 140 miles from Manchester to Newcastle, 115 to Grimsby, 90 to Birmingham. What do you suppose all those places have in common? (Similar distances from London would get you to Southampton, Hereford and, well, Birmingham.)
- Our electoral system paints a constituency in the colour of the largest single sub-group of voters, no more and no less than that. If a 38%/42% split between Conservative and Labour at one election turns into 42%/38% at the next, the constituency has certainly gone to the Tories, but the Labour voters haven’t – or rather, only some of them have, and they’re not necessarily the most representative ones.
- The story of those longstanding… gained… unexpectedly large seats is an interesting one, but it doesn’t really tell us about anywhere else (e.g. The North, or even The East Midlands), or about anything else (e.g. the overall election result).
- Those places have their own political history, which you can dip a toe in by looking at previous election results. You may be surprised by what you see; you may not like what you see. (“Cultural conservatism” doesn’t just mean you aren’t a fan of RuPaul’s Drag Race.)
- Not everyone who offers Labour bad news and hard truths is doing so because they want Labour to win. The party that doesn’t have any bad news to deal with – the party whose common sense is the other party’s hard truths – starts at an advantage.
- Generalising about Labour strategy on the basis of an imaginary version of the ‘Red Wall’, what it stands for and why it supposedly fell into the arms of the Tories (suit and tie, support our boys and God save the Queen) would be idiotic: basing strategy on fantasy can’t possibly work.
- Generalising about Labour strategy on the basis of an accurate understanding of the ‘deep blue’ seats, what voters there believe in and why they actually switched to the Tories would be cynical and unprincipled. And – again – it couldn’t possibly work, outside those seats; it wouldn’t even be a recipe for winning back the other 31 losses or holding Labour’s existing seats, let alone making the additional gains the party needs.
What Labour needs above all is to set its own direction, without looking over its collective shoulder at how policy X might play with demographic Y. A good start would be to stop listening to people offering to feed this habit – and to remember that those people aren’t always friends of the party.