Closing up shop

Anyone who used to regularly read UKPollingReport has presumably noticed it has been many months since I have updated it. You’ve probably already assumed that it is being wound down, but the time has come to make it official. Over the next few weeks I’ll be archiving some posts I want to keep for another day, and then shutting the site down.

I started writing UKPollingReport in 2004, back when blogs were the new exciting thing on the internet. It was when I was an amateur observer of polls, rather than someone who actually carried them out. Over time my hobby became what I actually did for a living, which is nice at first (after all, who wouldn’t like to get paid for something they are interested in doing anyway). As the years go by it switched over though, and there comes a point when the last thing you want to do when you log off from work is to write about your work. UKPollingReport had long since become a chore rather than a pleasure, so it’s time to move on.

I hope I achieved something along the way. Opinion polling has become less opaque since I started writing (though that is largely because of the advent of the British Polling Council, rather than anything I have done), but it still remains something of a mystery to many people. Any poll that people don’t like still generates a thousand replies asking how come they weren’t polled, or everyone they know thinks different, or how polling companies are all run by lizards (plus, of course, the Peter Hitchens meme, which even Peter Hitchens appears sick of). Making polling methodology more open only helps if people want to understand how it works, and naturally enough, many people have better things to do, but I hope I provided a useful resource for those who do have an interest.

More realistic is spreading the message to journalists and political observers (and, indeed, to my own industry. Being called out by your peers should always be a good deterrent). What polls can and, more importantly, really can’t do. Why a question like that won’t work, or doesn’t actually tell you what it appears to on the surface. What that question is terribly biased and should be quietly ignored. Why that poll using sampling that is probably skewed and one sided.

I think newspaper reporting of polling has genuinely improved over the last twenty years. There is a habit of being extremely cynical about journalists and their honesty or ethics – “one cannot hope to bribe or twist, etc, etc”. I don’t subscribe to that point of view. Most of the political journalists I’ve worked with over the years have been my favourite clients. Pressure groups, campaigning groups, etc, inevitably have an answer they want to get (and we often have to spend long hours saying no to various ways they try to bias questions). Journalists don’t – they tend to be the clients who just want to ask a fair question and report the results.

Overall, I think the press treat voting intention polls far better than they once did – putting it in the context of other recent polls, not massively over-reacting to small movements. It is very easy to point to places where they have fallen over, but these tend to be polls that have been commissioned and spun by pressure groups, rather than ones where the journalist talks to the pollster (or worse, things masquerading as “polls” which have never been near a professional research company, like the Observer twitter poll last month!).

Nevertheless, I’ll keep on rolling the boulder up the hill in other places. I am hoping that not having the weight of expectations of maintaining a blog will mean I feel actually able to write more often, rather than less. I’m going to continue grumbling about polls on twitter (@anthonyjwells, if you don’t follow me already). I’m also planning to put the occasional longer form piece as a (free) newsletter on Substack for when I want to rant about something that’s too long for a Twitter thread. I’ll put some of the pieces I keep referring back to up there as well in due course, so I’ve still something to link to when I want to explain why a Twitter poll of 100,000 people is still worthless, and why agree-disagree statements are almost always a terrible way to ask a question. Subscribe (for free) to that at https://anthonyjwells.substack.com/. I hope to see you there.


With a huge quantity of elections on Thursday, we also have a huge amount of opinion polling ahead of it. Here is a summary of the polling on this week’s elections and what we can predict about this week’s results from it.

SCOTLAND

The Scottish election race has been heavily polled and with fairly consistent results across polling companies. Everyone has the SNP clearly ahead in both votes, and we can be confident as confident can be that the SNP will win. The broad questions are:

a) How close the SNP will get to an overall majority in their own right, rather than with the Greens
b) Whether Labour or the Conservatives will be in second place
c) Whether Alex Salmond’s Alba party will win any seats

At least six companies have released Scottish polls in the last week or so (and we may well get more tomorrow). Most have shown the SNP around 50% in the constituency vote (SavantaComRes in the Scotsman were lower at 45%, but was conducted a week ago now, so I don’t know if they have another to come). In the regional vote most companies tend to show the SNP at around 36%-41%, with everyone showing the Conservatives in second place on around 22% and Labour third.

It is hard to translate votes into seats with confidence between so many of the final regional seats end up being won by extremely small margins, but it is probably the best way of understanding the interaction between the two votes. The most recent projections have tended to suggest that the SNP may just scrape a majority.

Sky projects 67 seats for the SNP based on the Opinium poll, John Curtice in the Times suggests 68 SNP seats based on the YouGov data. Panelbase in the Sunday Times at the weekend was projected to deliver them 65 seats by Curtice. The Herald’s BMG poll was projected to deliver 68 SNP seats. If the SNP do fall short then they will easily have a pro-independence majority with the Greens anyway, but an outright majority may give them a stronger moral case in the inevitable argument with the UK government over a referendum.

Finally, there is Alba – Alex Salmond’s new party. There has been some difference between polling companies on their projected level of support. For a while Panelbase were giving them around 6% while other companies were giving them derisory support. The final YouGov & Opinium polls did at least see them climbing to 3% (possibly enough to get a seat somewhere depending how unevenly their vote is distributed), while the most recent Panelbase poll had them dropping to 4%.

WALES

In the past Wales has not tended to attract much polling – there have been elections when it was YouGov and no one else. This year has been more interesting, with polls from SavantaComRes, Opinium and ICM. There are final Welsh polls from YouGov and Survation due out out tomorrow (and possibly SavantaComRes too) so we’re not done here yet.

The polling show far has also shown Labour ahead, but by differing amounts. Back in February and March YouGov produced a couple of polls showing the Conservatives within touching distance of Labour. Since then Labour have pulled back ahead, with a lead of 9-11 points on the constituency vote, 7-10 points on the list vote. We will see what the final polls show tomorrow.

LONDON MAYOR

Perhaps the most foregone conclusion that we have polling for – there has never really been the slightest doubt that Sadiq Khan was going to be re-elected as London mayor once Rory Stewart dropped out (even when Stewart was in the race polling showed an easy win for Khan, but at least Stewart had the potential to shake things). Given the race hasn’t been competitive there hasn’t been nearly as much polling as in past years, but we have had two final calls today (and perhaps more to come tomorrow). Both show Khan winning easily.

Opinium’s final call has Khan winning the first round by 48% to Bailey’s 29%, with Porrit in third on 8% and Berry on 7%. Khan wins easily on round two.
YouGov’s final call has Khan slightly lower on the first round, winning by 43% to Bailey’s 31%, with Berry on 10% and Porrit on 5%. Again, Khan wins easily on round two.

MAYORAL ELECTIONS

As well as London, there are elections for seven combined authority mayors (Cambridgeshire, Greater Manchester, Liverpool City Region, Tees Valley, West Midlands, West of England, West Yorkshire). We have polling for two of those, and in the case of the more recent polls, both show incumbents who were initially elected on a knife-edge now cruising to re-election.

The West Midlands mayoralty is being defended by the Conservative Andy Street. We have two decent sized polls there. Redfield & Wilton polled about a fortnight ago and found a 9 point lead for Street on the first round, but only a 2 point lead once second preferences were redistributed. Opinium released a more recent poll on Tuesday morning, commissioned by the Northern Powerhouse Partnership which projected a clearer win for Street – a 17 point lead and victory on the first round.

The only other poll I’m aware of is for the Tees Valley race. This was won by the Conservative incumbent Ben Houchen by an extremely narrow margin in 2017. The only poll is by Opinium. It has a sample size of only 387 (243 once you take out don’t knows and won’t votes) and the size was such that Opinium didn’t have the space to filter for likelihood to vote. For what it’s worth though, the shares were CON 63%, LAB 37% – suggesting a very easy hold for the Conservatives.

HARTLEPOOL

Survation have produced two opinion polls of Hartlepool. The first for the CWU early in the campaign, the second for Good Morning Britain, conducted at the end of April. Constituency polling is a difficult challenge that has a somewhat patchy record, and small sample sizes mean they are large margins of error. That said Survation’s latest poll showed a very solid 17 point lead for the Conservatives, well outside the margin of error. The poll would have to be very wrong indeed for this not to be a Conservative gain.

LOCAL ELECTIONS

Local elections rarely get much polling because of the nature of the contest. There is very uneven pattern of contestation, so many people don’t get a choice between all parties. In many places people get more than one vote so can vote for different parties. From a pollsters point of view, it’s also difficult to know what you are measuring – the “Projected National Share” that the BBC calculate on election night is not the actual shares of the vote – its a projection based on the votes in some key wards – so even if pollsters did ask about local elections, the numbers wouldn’t match the numbers the BBC announced on the night!

Normally the only attempt we get to predict likely gains and losses in local elections is therefore the Rallings & Thrasher modelling based on how people have voted in local council by-elections. This time round the coronavirus lockdown means we’ve had hardly any local council by-elections, so even the Rallings & Thrasher model lies dormant.

Unusually though we have got a couple of predictions using polling data. YouGov released a poll last week of councils that cover Northern & Midlands seats that the Tories won at the last elections (the so-called “Red wall” seats) that was used to project gains and losses here – that predicted Tory councillor gains in those areas.

Meanwhile Electoral Calculus have make predictions of district and unitary councillor gains and losses based on data from newcomers “Find Out Now” (a company that polls people visiting the PickMyPostcode website). Not sure why the county elections were excluded, but there goes. That predicts gains of about 300 seats for both Labour and Conservative, at the expense of Lib Dems and Others (note the overall number of seats rises because council changes). Note that the YouGov fieldwork was mid-April, the Electoral Calculus data last week, so quite possible that it pre-empted some of the recent poor coverage for the Government.

POLICE COMMISSIONER ELECTIONS

The final set of elections on Thursday, and very much the poor relation, are the Police Commissioner elections. Given the lack of public profile these – rather unsurprisingly – don’t appear to have been polled at all (indeed, given the uneven pattern of contestation and the fact I expect many people have no idea they are happening and won’t make their mind up how they’ll vote before being presented with a surprise additional ballot paper on Thursday, they would probably be rather hard to poll if we tried).


-->

This morning the Communication Workers’ Union released a Survation constituency poll of Hartlepool, the first one we’ve seen (earlier in the campaign figures were released from a Focaldata MRP of the North, but you can’t really use MRP for a by-election – it doesn’t pick up the unique circumstances). Topline figures with changes from the last election were are CON 49%(+20), LAB 42%(+4), NIP 2%(+2), GRN 1%(+1), LD 1%(-3), ReformUK 1%(-25).

I should start by saying that constituency polling is difficult. It is mostly done by telephone and often has small sample sizes (in this case, the sample was 500, but the actual voting figures are based only 302 who gave a response). Its track record has sometimes been patchy. Nevertheless, it’s the best evidence of where the race stands that we are going to get. What can we tell?

The Conservatives are ahead (though the two main parties are within the margin of error for a sample of 302). Compared to the general election the poll suggests an 8 point lead from Lab to Con, significantly better than how the Conservatives are doing in national polls.

It would be extremely unusual for a governing party to gain a seat in a by-election. There have been only two instances in the last fifty years (Copeland in 2017 and Mitcham & Morden in 1982). Few governments poll ahead of their last election performance mid-term anyway, and if anything they do worse than that in by-elections.

The reason the Tories are doing better in Hartlepool than nationwide appears fairly straightforward, and doesn’t offer any obviously transferrable lessons. In Britain as a whole the Brexit party got 2% at the 2019 election. In Hartlepool they got a very healthy 26%. That vote has almost completely vanished, presumably to the benefit of the Conservatives.

As ever, by-elections are extremely unusual beasts that do not necessarily tell much about national politics. Maybe if the actual by-election turns out like this it will be a steer on how other seats with a high level of Brexit party support in 2019 may go… but then, come the actual by-election we’ll have a glut of other data from the local, Scottish, Welsh, Mayoral and London elections due to be held on the same day, so hopefully we won’t be trying to desperately read too much into one single by-election.

Also worth noting that – given this poll was commissioned by the CWU – it also asked about some of the issues that they are concerned about like broadband, Royal Mail privatisation, nurses pay. The answers in Hartlepool were as you’d expect from national polling (people like free stuff & nurses. They don’t like privatisations). It doesn’t tell us anything particularly useful about why Labour aren’t doing better. Don’t assume because the CWU chose to ask about those issues that they are necessarily ones that are driving support in Hartlepool. Maybe people in Hartlepool care more about Corona, or crime, or Brexit, or economic regeneration, or taxes…

Finally, before this poll there was also significant social media buzz about the Northern Independence party having an impact, not least because their candidate is Thelma Walker, a former Labour MP who resigned over the party’s refusal to re-admit Jeremy Corbyn. Realistically a party that hasn’t even been registered yet may be very pleased indeed if they manage to get third place, but nevertheless, the poll suggests they are not significant players here.

UPDATE: The tables for the Survation poll have appeared, and worth adding a further caveat. At the last election the Brexit party got 26% of the vote. Among people who took part in the poll, only 3% recalled voting for the Brexit party. This does not *necessarily* mean its a duff sample – there will undoubtedly be issues of false recall, of people re-aligning their past vote to match with present circumstances (especially since the Brexit party has rebranded itself into ReformUK and no longer exists in its old form), but it should be an extra reason for caution.

Scotland

There were two Scottish Parliament voting intention over the weekend, one from Panelbase, one from Survation. Topline figures are that both show the SNP continuing to cruise towards victory and on the edge of winning a majority. Both show a tight race for second place between the Conservatives and Labour.

However, these were also the first two to measure support for Alex Salmond’s new list only party, Alba. The Panelbase poll showed them at 6%, the Survation poll showed them at 3%. To understand the significance of these we need to explore the nuances of the Scottish Parliament electoral system.

The Scottish Parliament elects members using an additional member system. 73 MSPs are elected in constituencies using first past the post, a further 56 are elected on a proportional regional list system. The regional list seats effectively operate as a “top-up” to the constituency seats already won, so that overall the seats won should be proportional to the list vote. For example, if party A won 6 constituency seats, but got 10% of the list vote, they’d be awarded another 7 list seats so they had 10% of the total seats. It’s more complicated than that because it’s done by region, meaning there is an effective threshold to get any seats at all, but we’ll come to that.

Crucially people cast two votes – you don’t have to cast your constituency vote in the same way as your list vote, you can vote for different parties.

The SNP did extremely well at winning constituencies at the last election (59 out of 73). This meant that that despite winning 42% of the list vote, they didn’t receive many list seats, because they had already won almost their fair share through constituency seats. Compare this to the Scottish Greens – they don’t win any constituency seats (they barely stand), so there is nothing to set against their list vote and their list vote of 7% translates into 6 seats.

Therefore, the Alba argument goes, SNP votes on the regional list are “wasted” votes, that are unlikely to return MSPs. If a significant chunk of SNP voters voted Alba instead, it would return more pro-independence MSPs.

So far, so good. However, because the Scottish system uses regional lists, there’s an effective threshold to get any seats at all (about 5-6%). There is also already a second pro-independence party, the Scottish Greens. That means in practice Alba could have a positive or negative impact on the number of pro-Independence MSPs elected. If they get over 5% in a substantial number of regions, and do so by taking SNP second preferences, rather than taking votes who would otherwise back the Greens, they will increase the next number of pro-independence MSPs. If they get under 6% in most regions, they are unlikely to win any MSPs at all. If they get under 6%, but in doing so, take votes from the Scottish Greens, they could even reduce the the number of pro-independence MSPs.

Hence, in judging the impact of Alba, the thing to look at is the level of Alba and the Scottish Greens in the list vote, and whether each is above or below that threshold of around 5-6%. The two polls so far paint contrasting pictures – in the Survation poll, Alba were at 3% and the Scottish Greens were unchanged at 11%. In the Panelbase poll Alba were at 6%, the Scottish Greens at 8%, again comparable to their showing in previous Panelbase polls.

So in neither case was there any evidence that Alba were cannibalising the pro-independence list vote by taking support from the Greens, but the evidence on whether they’ll actually win seats of their own is unclear. On the Panelbase figures they may well do (John Curtice tentatively projects 6 Alba seats, with a total of 79 pro-Independence MSPs). On the Survation figures they probably wouldn’t, but the SNP and Greens would get 77 pro-Independence MSPs between them anyway.

And that, in itself, maybe underlines the extent to which this matters. As things stand most polls show the SNP getting a majority or getting close to one. Taking the SNP & Scottish Greens together, there will very likely be a majority of pro-Independence MSPs anyway. Whether Alba manage to scramble over the threshold to win some seats or not doesn’t look likely to change that given their present level of support.


Post-budget polling

There were three immediate post-budget polls on the day of the budget from Opinium, SavantaComRes and YouGov. They all showed very high levels of approval – YouGov found 46% support, 11% opposed; Opinium found 52% approve, 12% disapprove; SavantaComRes found 60% satisfied and 11% dissatisfied. The following day YouGov also put out new voting intention figures, which showed a Conservative lead of 13 points. While the Conservative lead has been steadily growing over recent weeks anyway, at least part of that looks likely to be a bounce from a positively received budget. Worth remembering that it was conducted immediately after the budget, before the press coverage turned rather less positive…

NHS Salaries

This would not be the first initially popular budget to come unstuck in the following days as people look in more detail at the details. In this case the downside doesn’t appear to be in the details of the budget itself, but in the proposed NHS pay settlement that was announced the following day. The first actual polling on this has come from Opinium in the Observer. This found 24% thought the proposed pay rise was too high or about right, 72% thought it was too low.

While high, this is by no means new, or different, or surprising. All past polling on nurses salaries has always shown the vast majority of people support increases – look, for example, here or here or here or here or here or here. That’s about a minute of googling. The public have very consistently and very widely supported increasing nurses’ pay for many years. It has not led to large rises, or prevented the Conservative party winning elections in their absence. A better question is perhaps to what degree it is a more salient issue given the coronavirus, and whether it actually damages Conservative support. Time will tell if that’s the case.

All past polling has shown that nurses are held in extremely high public regard (look, for example, at the Ipsos MORI veracity index). In terms of public sympathy, this is a fight that is probably unwinnable for the Government. The question is more weather they knuckle down and accept any hit to public support (because something is unpopular it does not necessarily damage support at the ballot box), or whether they U-turn on it.

Scotland

There are two Scottish polls in the Sunday papers, one from SavantaComRes for the Scotsman on Sunday, one from Panelbase for the Sunday Times.

The Scotsman on Sunday one makes great play of No being ahead in their SavantaComRes poll – 46% No, 43% Yes, Don’t know 10%.

This has been rather complicated however by methodological issues. Previous ComRes polls on Scottish Independence have been weighted by likelihood to vote. This one was not, meaning it can’t be easily compared to the previous figures. More to the point, the impact of weighting by likelihood to vote has sometimes been quite large in previous ComRes polls, so it’s quite hard to work out what the impact would have been.

In ComRes’s last Scottish poll, conducted in late February for ITN, applying turnout weighting had minimal effect. However, in their last poll for the Scotsman at the start of Feb 2021 it made a huge impact, transforming a 3 point No lead to a 5 point Yes lead. In other words, it’s quite possible that if the same turnout filter had been used this wouldn’t have shown any change in support for Independence at all.

The Scotsman on Sunday article also reports people claiming that the Salmond Inquiry makes them less likely to support Independence. I’d take that with a bucket of salt. I’ve written before about the perils of “does X make you more or less likely to vote Y” questions. They are tricky to ask, as committed supporters of something tend to pick “more likely” as the most supportive option (even if their opinion is completely unchanged), and the same for committed opponents. This means you get odd results like 16% saying Alex Salmond & Nicola Surgeon calling each other liars makes them support independence more*. Really, it is just committed supporters of independence saying they still support independence, and the claimed percentage saying it makes them less supportive of independence will include a lot of people who were concrete opponents of independence anyway.

The way to see if an event has actually changed the level of support for independence is to ask about support for independence and compare it to a poll conducted before the event. It does require you to ask it in a comparable manner though, rather than buggering about with the turnout filter.

The Panelbase poll does at least do that. They found topline figures of YES 46%(-3), NO 47%(+3). Changes are from their previous poll in January. The No lead is pretty much an artefact of the rounding – according to John Curtice who has written them up in the Sunday Times, without don’t knows it ends up as a 50/50 split. It is, as ever, just one poll, but does suggest some damage from the Salmond-Sturgeon row. One should again consider the context of the polling – just as the YouGov voting intention poll at the top of this post was conducted in the midst of extremely positive coverage about the budget, this poll was conducted in the midst of the SNP having a huge row between its current and former leaders. In both cases one should probably wait for things to calm down before assuming any long term impact.

(*As an aside, the best example of the useless of more/less likely to vote questions I’ve ever come across is this one from the US Alabama Senate election in 2017. The Republican candidate faced several accusations of sexually assaulting underage girls. A question asked in that same “more or less likely to vote” format found 36% of people were less likely to vote for him because of these accusations, and 28% were MORE likely. Either 28% of people were genuinely more likely to vote for a candidate because he’d been accused of molesting schoolgirls… or perhaps questions like that don’t really show what they appear to show.)


Yesterday Channel 4 news and JLPartners released an interesting poll of so-called “red wall” seats. A lot of things get written about “Red Wall seats” that don’t necessarily have much thought behind them. It is the Essex man or Worcester woman of the 2019 election, an easy buzzword that is too often a substitute for proper understanding. They are important, but you need to look careful at the nuances.

Let us start by going back to where the term originated, with James Kanagasooriam. It wasn’t just a generic word for northern marginals or Tory targets – James coined it when talking about as seats that demographically should have been Conservative, but which consistently voted Labour. James actually identified several groups of these seats – some in Wales, some in County Durham, and a big swathe of them across urban Lancashire and Yorkshire that looked like a red wall. Hence the name.

The whole point of James’ argument was that there were seats that in terms of their make up (class, economy, education, age structure and so on) you would expect to vote Conservative, but that they actually voted Labour because of a cultural, historical and social hostility towards the Tories. These weren’t seats full of horny-handed sons of toil, they were seats that were or had become more affluent but yet not become Tory. Sefton, for example, is largely affluent suburbia, with some of the highest home-ownership rates in the country. Yet it votes Labour, because it’s Liverpool’s suburbs, and people in Liverpool are not inclined to go about voting Conservative.

One can debate the reasons for this dislike, but the most obvious explanations are historical: the identity as former mining communities, the legacy and memory of Thatcherism and the dismantling of industry in the North in the 1980s. The point was, this was an obstacle to the Tories – how to appeal to these people who “should” demographically be their target audience, but for whatever reason were not interested.

Skipping ahead, we know that the Conservatives did manage to do this in many areas in 2017 and 2019. In fact in many of these areas there has been an incredible sea-change in voting behavior. Across the two elections the Conservatives have made gains there that would have looked unbelievable ten years ago. In 2015, the Conservatives won the national share of the vote by 7%. In 2019 they won by 12%, that is, a 2.5% swing across those two elections. Compare that to the Lab=>Con swing in some of the “red wall” seats. Sedgefield experienced a 14% swing across the two elections, Blyth Valley 13%, Bolsover 19%, Leigh 18%.

While there are some areas that did not follow this tide (Merseyside in particular is still extremely unforgiving territory for the Conservative party), among other areas in Lancashire, Country Durham, and Derbyshire mining areas the “red wall” decisively crumbled.

There are different explanations one can come up with for what happened. Part of it was probably the disruptive effect Brexit had upon traditional party ties, part of it perhaps a general change to the way the Conservative party has presented itself and its message. Much will simply be to the passage of time – those old mining identities can only sustain for so long once the mines have closed, the miners have passed on, the old sites regenerated and replaced by new build housing estates.

However, the Tory advances of 2017 and 2019 were not just in James’s red wall seats. Here is where it gets complicated, and why one should be cautious about throwing all those 2019 gains in together. The Conservatives gained other seats as well, some of which don’t match this description at all. Lewis Baston has written about this well previously. Some of them were in perennial marginals – places like Darlington, Stockton South, Keighley or Lincoln that have been competitive for decades and just happen to be in the North or the Midlands. If you are looking at opinion in the “red wall seats”, you have to be careful how you define it, and what you are actually looking at.

All that brings me round to the actual JLpartners/Channel 4 polling. This polled 500 people in the seats that the Conservatives won from Labour in the North & Midlands in 2019. The write up and full tables are here (do go and have a read, as there is lots of detail I have not explored below).

Overall the poll shows Labour at 47%, the Conservatives at 41%. In comparison, in the same seats the vote share in 2019 was Conservative 48%, Labour 39%. That translates into a swing of 7.5% from Conservative to Labour. In comparison the national polls conducted over the same period showed on average a Conservative lead of 1 point, a swing of 5.5% from Conservative to Labour.

On the face of it, that suggests the Conservatives are doing marginally worse in these seats than in the country as a whole. If that was to happen at an election it would be unusual – parties actually tend to do a bit better than average in seats they gained at the previous election because they have gained the incumbency advantage (the MP’s “personal vote”), and their opponents have lost it, so this would be a particularly poor performance. However, I should add the caveat that it’s just one poll of 500 people, so there is a margin of error of 4% on there. We should not put too much confidence on whether the Conservatives are doing a couple of percentage points better or worse in an area based on a single poll.

More interesting of course would be to be able to look under the bonnet at the different types of seat within those we’ve lumped together as “red wall” seats. Are there different patterns at work in those traditional marginal seats to those former mining and industrial seats that have been part of the bigger red-wall sea-change. There is no particular reason to think that seats like Lincoln or Blackpool South or Gedling would behave any differently to marginal seats elsewhere – but for seats like Sedgefield, Bassetlaw or Bolsover there’s a question of whether the political re-alignment we’ve seen over the last few elections has come to a halt or is still ongoing.

That’s not to say the JLPartners/Channel 4 poll isn’t good stuff – it is – it’s more than it’s only a starting point.

The question people tend to ask on the back of polls like this is whether the Tories need to worry unduly about keeping these seats in their column, and whether Labour can win them back. In that context, it is probably too simplistic to look at them as a single lump. In one sense, obviously these seats will be part of the battleground – but that’s just a truism. These are marginal seats, elections will be always be won and lost in the marginal seats. The more important question is whether these marginal seats are the ones that are most likely to change at the next election, or whether by looking at the fashionable “red wall” seats we miss looking at potentially more vulnerable seats elsewhere?

It may be that the political re-alignment in the true “red wall” seats is so seismic that they actually become safer Tory seats than some of the more traditional marginals. It may be that the more vulnerable Tory seats next time round are actually some more affluent seats with high proportions of graduates. The pattern of key marginals next time round could be those that are similar to North West Bristol or Canterbury, rather than winning back old mining seats.

Northern Tory gains last time weren’t monolithic – it isn’t one single “red wall” – they mix up some traditional marginals, as well as some sea that have seen truly transformational change. We shouldn’t assume they’ll behave as one block, or in the same way in the future either. Equally, we shouldn’t necessarily assume that all the interesting changes at the next election will happen in the same place as the last one. There are risks and opportunities elsewhere too.