Forget the party spin. What does history suggest success really looks like for the parties at the 2023 Local Elections?
As far as the next General Election is concerned, what really matters is the 'projected national share' - and the relationship this has to polling and General Election outcomes.
It’s pretty much impossible to look at any coverage of the Local Elections without coming across a great deal of expectations management.
The Conservatives have been briefing that they expect to lose 1,000 seats. It has been suggested that 500 Conservative losses would be ‘just midterm blues’, even though they had already lost over 1,300 last time these seats were fought in 2019. Meanwhile, Labour sources, including Keir Starmer himself, have reportedly said that Labour achieving just 400 gains would represent ‘good progress’, despite Labour’s large poll lead, and the fact that they are starting out at a low base, with potentially huge scope for gains. It has been suggested that only by making 150+ gains can the Liberal Democrats demonstrate a ‘promising advance’, and that they would be ‘struggling to pose a threat’ if their gains are below 50 - even though holding their 700 gains from May 2019 would still nonetheless see them very competitive in many Conservative-held marginals at the next General Election.
Oh dear. I’m not sure about you, but all this spinning is making me dizzy. It seems very difficult to get a clear understanding of what really does represent a good outcome for any party, and what implications - if any - the results could have for the next General Election, to be held sometime in 2024.
So let’s forgo all of that, and let’s take a data-driven approach. In this piece, using historical polling data, and Local Election and General Election results, we’ll seek to understand:
What today’s polls tell us about how we should expect the parties to perform;
What the data suggests these Local Election results can tell us about the next General Election - indeed, if anything;
Based on the above, some (heavily caveated) data-informed benchmarks for what a good result might look like for the parties.
Scroll down to the bottom if you want the sugar rush of numbers and benchmarks – but they will make much more sense having read the rest of the piece!
Net seat losses and gains are difficult to interpret
The most simple and crucial point to understand upfront is that, as far as any implications for General Elections are concerned, seat losses and gains at Local Elections actually not very helpful benchmarks for assessing who is doing well and who is doing badly. This is because seat losses/gains tell you more about the baseline - i.e. what happened at the Locals 4 years previously - than they do about the state of public opinion today.
For instance, in the May 2021 Local Elections, the Conservatives stormed home with a big lead, making significant gains off the back of their ‘vaccine bounce’ (including winning the Hartlepool by-election). When some of these seats are fought again in May 2024 (the 2020 set, which were postponed due to the pandemic), the Conservatives will face significant losses - even if Sunak does close the 15-point polling lead that Labour currently enjoy. But he would be doing very well to have closed that gap, and would be on course to win the next GE if he had, despite the losses.
Similarly, the baseline the Conservatives begin from in this year’s Local Elections is indeed quite poor - the Conservatives incurred 1,330 losses in 2019. But there is some nuance here: these 2019 losses were incurred from the Conservatives’ high-water mark of 2015. So although their 2019 performance was indeed poor, it was not dreadful in seat terms - they still won 43% of the total seats up in 2019 (down from about 59% in 2015), more than Labour and the Lib Dems combined (40%). So the Conservative performance in 2019 was certainly not great, but perhaps not quite as dreadful as some opposition politicians may like to suggest.
However, in terms of votes, it’s perhaps quite a different picture. The Conservatives did do very badly in this respect in 2019 - estimates of their projected national vote share vary from 28-31%, the same as Labour’s at this election. In actual votes cast, the Conservatives achieved 31.4%, i.e. significantly lower than the 43% of seats they won, despite these councils being more Conservative-leaning than the country as a whole. By comparison, Labour and the Lib Dems combined won 43.4% of votes, slightly higher than the 40% of seats they won. I haven’t drilled down enough into the detail, so I could be very wrong, but this might suggest that the Conservative vote in May 2019 was still relatively ‘efficient’, winning lots of seats by relatively small margins, despite their anaemic vote share and big seat losses. If so, even a small further swing against them in May 2023 could mean council seats fall like dominoes, especially if anti-Conservative tactical voting is at play. That is to say - many seat losses might just be expected, to some degree.
More broadly, there are also wider reasons why seat losses and gains are an unhelpful metric: e.g. progressive parties often tend to step on each others’ toes in Local Elections – and the Liberal Democrats and Greens are standing in a greater proportion of seats this time. This can often allow the Conservatives to retain seats due to split votes, perhaps contributing to the efficiency with which they convert votes into seats relative to Labour / Lib Dems / Greens1. Meanwhile, Reform UK are only standing 6% of a full slate of candidates across England - meaning the Conservatives are barely challenged on the right (something that may not be true at a General Election, if the Reform UK leader is to be believed). This should help them to retain more seats.
I’ve dwelled on the above as it should hopefully be useful context for cutting through the ‘expectations management’ we’re seeing from all parties. But the overarching point is this - it can be difficult to interpret what Local Election seat gains and losses really mean for the wider electoral landscape. It’s no surprise that spin thrives.
Some indicators can be useful pointers as to the direction of travel, e.g. Labour overtaking the Conservatives as the biggest party of Local Government is something that should happen if they are really headed for Government, and it is probably a pre-requisite for Labour to be able to claim a strong result this year. But in general, seat losses and gains largely just tell you about what happened 4 years ago.
To highlight this - nobody would argue that the Labour Party did badly in the 2001 General Election because it lost 6 seats. It was, of course, a landslide, just as it was 4 years previously. If we only consider gains and losses and not the absolute position of parties too, we can end up with silly narratives that often seem to exist for the sake of existing - such as (in my view) the idea that the Liberal Democrats would be ‘struggling to pose a threat’ to the Conservatives if they held their relative high-mark in the 2019 Local Elections, without further significant gains.
Nobody would argue that the Labour Party did badly in the 2001 General Election because it lost 6 seats.
For these reasons, to assess performance, I’m actually going to forget about seat losses and gains entirely - at least for the purpose of this piece. (I have made a silly little forecast for fun, but I’m not taking that too seriously.)
National vote share is most important, not seat changes
If we are thinking about the wider electoral landscape, what matters most is how the parties are performing in terms of votes.
For this, we will use national measures of vote share, which include the Projected National Share (PNS), produced by the BBC, and the National Equivalent Vote (NEV), produced by Colin Rallings and Michael Thrasher. These figures estimate what the parties’ vote share would have been had the Locals been a GB-wide general election, taking place across the whole country. For more information about the methodology for calculating the Projected National Share, see this excellent article from Stephen Fisher and John Curtice. Here you can also see a chart showing how the parties have performed in Local Elections since the 1980s.
These projections are estimates and are by no means perfect (and there are often slight differences between PNS and NEV), but they are nonetheless the best indicators we have of the state of the parties.
For the purpose of this post, I will primarily be using Rallings and Thrasher’s NEV estimates, as NEV has a slightly longer time series, stretching back to 1976. The NEV data is sourced from Mark Pack’s excellent LocalBase, and the polling data is sourced from the equally excellent PollBase too.
Hold on to your hats for the next section, as I’m afraid there’s going to be a lot of statistics. Feel free to skip straight on to section 2 if you don’t want to endure the maths.
1. What do current polls suggest the 2023 Local Election results will look like?
To understand what we expect the NEV to look like this year, we can analyse the relationship between Labour polling prior at the time of past Local Elections, and the resulting Labour / Conservative NEV lead - see below (Note I’ve excluded Local Elections years that coincided with General Elections, for consistency).
As you can see, a consistent trend since 1976 is that Labour’s lead in NEV is generally lower than the polling lead that they hold at the time of the Local Election. This is a long-running, consistent trend, which needs to be understood and accounted for to accurately interpret Local Election results, as this appears to be a structural feature. Indeed, Local Elections expert Michael Thrasher has often remarked at how Labour tends to perform better at General Elections than Local Elections.
For the stats nerds, the regression formula is:
y = 0.63x - 0.4
To explain what this means - to work out the typical expected NEV lead for Labour at a Local Election (while in opposition), we work out 0.63 multiplied by the poll lead Labour holds at the time, minus 0.4. To simplify it even further, we essentially expect the NEV Labour lead at Locals to be about 60%, or three-fifths, of the Labour poll lead at the time of the Local Elections. So a 10-point Labour polling lead would be expected to translate into roughly a 6 point NEV lead, for example.
Looking to this year’s Locals, the 15% poll lead Labour currently hold would imply roughly a ~9 point Labour lead in NEV according to the regression (perhaps, say, 37% Labour, 28% Conservative). This would be my central, data-informed estimate of the likely Labour lead at the Locals this year. Higher than that is above expectation; lower than that is below expectation. This is, of course, notably lower than Labour’s current polling lead of ~15 points.
A 9-point Labour lead would represent a 4.5 percentage point swing from Conservative to Labour compared to the 2019 Local Elections, when the two parties were tied at 31% on this measure. For those wondering what this means for seat counts, the rough estimate of 1,000 Conservative losses was based on a 6 point swing. So we might crudely expect a 4.5 point swing to indicate something like ~750 losses (although, the eventual figure depends on how the Lib/Con fights go, too). But this is real finger-in-the-air stuff, so don’t take those figures too seriously.
A 15% poll lead implies a ~9 point Labour lead in National Equivalent Vote. This would be my central estimate of the likely Labour lead at the Locals this year.
Local Elections taking place under Conservative Governments are very different to under Labour ones
In the above regression, I’ve only included Local Elections that took place under Conservative Governments (1980-1996; 2011-2022). This is for a very good reason: the relationship is notably very different under Labour Governments. The ‘Labour under-performance’ effect is much more dramatic.
For instance, in the 2000 Local Elections a year prior to the 2001 GE, the Conservatives had a NEV lead of 8 points, despite an impressive 13 pt Labour lead in the polls at the time. Similarly, when Labour held a 25-point poll lead in 1999, it only won the Local Elections by 2 points.
You can see the relationship under Labour Governments below:
The relationship is noticeably different under Labour Governments (data included is 1976-1978; 1998-2009, excluding General Election years). The regression formula is:
y = 0.54x - 10.5
This means that while Labour is in Government and the Conservatives in opposition, even if both parties are tied in the polls (an okay position to be in during the midterm), we would still expect the Conservatives to win NEV by 10.5 points. But if the polls are tied at the time of a Local Election while Labour are in opposition, we would expect the parties to be (roughly) tied in NEV. This is quite a dramatic difference.
Others will have a much better explanation for this seemingly inherent, consistent Conservative over-performance. I imagine it’s a combination of factors, but I would speculate they include:
Differential turnout, with Conservative voters historically being more likely to turn out at Locals than Labour voters (especially when Labour are in power);
Perhaps would-be Labour General Election voters disproportionately vote Lib Dem at Locals (again - especially when Labour are in power) relative to Conservative voters, which acts to disproportionately reduce the Labour share.
But more widely, this really matters for interpretation of Local Election results - because what it tells you is: if Labour achieved a (say) 11 point lead in NEV this week, this would be far more significant than if the Conservatives achieved this.
The Conservatives achieved a 11 point NEV lead in 2004 with just a 1 point polling lead over Labour, in line with our regression.
Labour achieving an 11 point NEV lead in opposition, on the other hand, on average requires nearly a 20 point lead in polls.
If Labour achieves a result similar to an 11 point NEV lead this week, we may well hear spin that this is a mediocre result, because the Conservatives achieved this result in 2004 too and still went on to lose in 2005.
However, this would be quite misleading - this kind of result would be in a similar ballpark to those achieved by the Labour Party in the mid-90s.
In a sense, ordinary GE polling - in general - probably explains far more about the parties’ true positions than NEV does. NEV simply acts as a real-world confirmation of that polling lead - but for it to do even that, it has to be understood through the nuanced statistical relationship that it has with polling.
When we take that relationship into account, it’s not difficult to see why the Conservatives went on to lose the 2005 General Election, and why an opposition Labour Party that achieved similar results in NEV a year prior to a GE would probably be expected to win it.
2. What can these Local Election results can tell us about the next General Election?
So - if we are thinking about what a ‘good’ performance looks like at the 2023 Locals, and what this could mean for the next General Election, the ‘Conservative advantage’ needs to be borne in mind - and we probably want to ignore periods where Labour was in Government, sticking to periods where Labour is in opposition as a point of comparison - as others who write on this topic have also done.
The other point of reference for this question, relevant to this week’s Local Elections, would be - how have Labour generally performed in Local Elections in the year prior to a General Election?
Sticking with periods where Labour was in opposition, the NEV results, and the national polling immediately prior to the Local Elections, look like this:
1982: 11 pt Conservative lead (Cons won GE 1983)
Pre-LE polling: 11 pt Conservative lead
1986: 3 pt Labour lead (Cons won GE 1987)
Pre-LE polling: 7 pt Labour lead
1991: 3 pt Labour lead (Cons won GE 1992)
Pre-LE polling: 1 pt Conservative lead
1996: 14 pt Labour lead (Lab won GE 1997)
Pre-LE polling: 25 pt Labour lead
2014: 1 pt Labour lead (Cons won GE 2015)
Pre-LE polling: 3 pt Conservative lead
2016: 1 pt Labour lead (Cons won GE 2017)
Pre-LE polling: 2 pt Conservative lead
2018: 1 pt Conservative lead (Cons won GE 2019)
Pre-LE polling: 2 pt Conservative lead
The trend is quite striking. You can pretty much tell that Labour is going to lose the General Election 12-18 months or so out - because every time it lost a GE, its NEV the year prior was a 3 point lead, or worse, at the Local Elections. That said, the past is not always a guide to the future, and these trends can always be broken - the May 2017 elections certainly did not imply a hung parliament one month later!
1991 is of particular relevance here - that was a very similar situation in which a new, more popular PM replaced a deeply unpopular one, seeking to reverse the party’s fortunes. It paid off handsomely for the Conservatives in 1992, as we all know.
At the moment however, examining the parallels of Sunak’s situation with Major only reinforces why 2024 is not (yet) looking like 1992. 5 months into Major’s premiership (April 1991), Major had reversed Labour’s 12-point poll lead that he had inherited from Thatcher in November 1990, holding a slender 1 point polling lead over Labour according to PollBase. His approval ratings were also stratospheric (+33), miles above his opponent Kinnock, and the Government’s net satisfaction (Ipsos) rating had recovered to a reasonable -26.
Sunak’s situation is rather different at the moment. The Labour poll lead (15 pts at the time of writing) has narrowed by about 5 points since November, but the lead remains worse than where Major had to start from. And Government satisfaction remains stuck at -61, every bit as bad as it was in the last 18 months of John Major’s Government.
Sunak’s personal ratings are so-so - they are well above his party’s ratings, but are nothing like Major’s, who never went negative - not once - between becoming PM and fighting the 1992 GE. And although Starmer is not especially popular either, Sunak’s ratings are nonetheless worse than Starmer’s overall. Sunak is in touching distance of Starmer’s ratings, and this does provide some genuine hope of recovery for the Conservatives. But given the state of the fundamentals (and the wider Tory brand), Sunak would need to be doing some very heavy lifting. Better get to the gym.
3. Benchmark setting
All of which is to say - if the Conservatives want to establish momentum behind the narrative that the next GE is going to be 1992 (or indeed, 2015) all over again - a surprise narrow victory - we will need to see that his party is performing as well at these Local Elections as John Major was 6 months after he took over - which was the 1991 Local Elections, when Labour only won a slender 3 point NEV lead.
So for the ‘1992’ narrative to have any substance to it, this means that means Labour’s NEV lead definitely needs to be no higher than 5 points, and ideally 3 points or less. Anything higher than 5 points, and that narrative does not stand up. Not achieving this doesn’t mean the Conservative can’t win the GE, of course, it just means they’re not on track to do so at present based on where we are in the electoral cycle - just as Labour certainly was not on track back in May 2021.
But equally, the challenge for the Labour Party is that - if it wants to look like it’s on track for another 1997 - the landslide win some are arguing that we are on course for - it needs to win big at these Local Elections too.
What further complicates this is that - whilst we have lots of examples of what Labour’s performance at Locals looks like prior to a GE loss - and so we very much know what a losing NEV for Labour looks like, for the purpose of benchmark setting - we actually don’t know what the Locals performance of a Labour Party in opposition that is on course for an ordinary, non-landslide win looks like.2 The data doesn’t stretch back to 1974 - the last time this happened.
Probably the closest point of comparison that we have for this is during John Smith's leadership of the Labour Party - pollsters suspect Smith was on course to win the 1997 General Election had he not died; but that Blair turned it from a likely ‘healthy Labour majority’ to a landslide. For this, we can reference the 1993 and 1994 Local Elections, but these aren’t great points of reference as they were 3-4 years before a GE, not 12-18 months. NB: like Starmer, Smith’s personal ratings were so-so, and tended to be float around neutral or slightly negative, with high numbers ‘don’t knows’. Had he fought the 1997 GE, we might have had a much better point of comparison for Starmer than Blair.
We don’t really have a great idea of what the Locals performance of a Labour Party in opposition that is on course for an ordinary, non-landslide win looks like.
It would be foolish not to also mention here that there was one Local Election where Labour did gain a double digit NEV lead and fail to win the subsequent general election - May 1990, where Labour won NEV by 11 points in the aftermath of the Poll Tax riots. There are, however, obvious reasons why this situation is not particularly relevant or predictive in this instance. Thatcher may well have been on course to lose the subsequent GE - we don’t know, because we did not get to find out; she was of course replaced by the very popular John Major that Autumn. Insofar that 1990 is a relevant parallel to the current Parliament, the closest thing would have to be if these Local Elections were in fact being fought by Truss, not Sunak - thankfully for the Conservatives, they aren’t.
Anyway, back to benchmarks. Overall, I take the same view as John Curtice and Chris Hopkins (Savanta) that around a 10 point NEV lead for Labour is probably a fair threshold for a successful night. This would easily be Labour’s best midterm NEV lead since 1996. A 10 point lead would also be broadly in-line with other results from the mid-90s (if slightly below), except for the extreme event of the 1995 Locals, where Labour won by 22 points while holding an eye-watering 33 point average lead in the polls. Perhaps similar to what the Conservatives would have faced if Truss was leading them into these Locals.
For reference, I’ve given the midterm Local Election NEV figures from the 92-97 parliament below:
1993: 8 pt Labour NEV lead
Pre-LE polling: 15 pt Labour lead
1994: 12 pt Labour NEV lead
Pre-LE polling: 21 pt Labour lead
1995: 22 pt Labour NEV lead
Pre-LE polling: 33 pt Labour lead(!)
1996: 14 pt Labour NEV lead
Pre-LE polling: 25 pt Labour lead
A 10-point NEV lead would fit neatly in this grouping (albeit on the lower side) - especially if the bar is not ‘on track for a 1997-style landslide’ but merely ‘on course to win the GE’, as it seemed like they were in May 1993 and May 1994. But again, even achieving this guarantees nothing about what will happen 12-18 months later. It’s merely indicative.
Wider reflections
The nuances of the above might well lead one to conclude that we shouldn’t overthink this - that as far as the next GE is concerned, Local Elections might simply amount to a really rubbish opinion poll?
In many ways, this would not be a wholly unfair judgement to make. Locals certainly have some predictive value but really only as one indicator. I still think the other indicators – national vote intention polling; party brands; which parties lead on the key issues of the day (especially the economy); and leadership ratings – are more useful variables. If polls change dramatically, it doesn’t matter how well you did in the Locals last year, of course.
But where Locals are useful is for allowing us to identify the extent to which what we’re seeing in polling is translating into reality. But one can only understand how polling is translating into reality through the lens of these complex relationships that exist between polling and Local Election results, hence the purpose of this piece.
In light of all of this, below I’ve created a fuller set of scenarios and benchmarks based on the analysis and discussion above.
Overarching benchmarks for the 2023 Local Elections
Labour win by 5 points or less: Possible 1992 redux as the Tories surge
Labour winning the Locals by 5 points or fewer would be a huge disappointment for the party, who would have performed much worse than their polling implies. A result like this would be broadly in-line with Major’s performance in 1991, giving the Conservatives a significant boost, and providing genuine hope they may be able to repeat their 1992 shock victory - especially if the Labour lead is 3 points or lower.
That said, whilst the Conservatives may become favourites to win in 2024 in this scenario, they would not be out of the woods yet - their recovery would need to continue over the next year.
Labour win by 6-8 points: Labour under-performance - uncertainty
This kind of result would still suggest an under-performance for Labour compared to what we would expect from polling. Labour’s worries about the Tory recovery will certainly be exacerbated by this performance, which would still be a something of a boost for Number 10. However, this should be tempered by the fact that such a result 12-18 months out from a General Election would not typically point towards a Conservative majority, either.
As far as it has predictive value, such a result may point towards a hung parliament at the next election.
Labour win by 9-11 points: Labour probably still on course for a General Election victory, but a landslide looks unlikely
I consider this kind of result to be essentially ‘par’ for Labour, in line with what polls would historically predict based on the current polling averages, and probably roughly what the parties themselves are anticipating. This would easily be Labour’s best set of Locals since 1997, and I suspect the party would be satisfied with these results, but may have hoped for better.
Such a result would be an improvement for the Conservatives compared to what January and February’s polling would have implied, but would still point towards Labour broadly being on course to win the next General Election. That said, there’s not an enormous amount of padding in these figures, and certainly still time for Labour to mess it up. Nor would it suggest a 1997-type result was on the cards.
This is the kind of result that (in my view) we are likely headed for, and which may be erroneously seen by some as ‘not as bad as current polls suggest’ because it’s smaller than the current poll lead - even though it’s actually exactly in-line with what polls would suggest the result should be, historically.
Labour win by 12-14 points: Labour comfortably on course to win the next General Election, matching its mid-90s performances.
Getting well into double digits would be a very strong result for Labour, and one that is entirely in-line with the party’s performances in the 1994 and 1996 Local Elections, when the party was very much on course for a majority. These kinds of numbers have been described by a pollster as “great for Starmer” and “terrible for the Tories”. Such a result would also likely come with substantial Lib Dem and Green gains.
This kind of lead would certainly disappoint Conservatives as it would suggest their recent recovery in the polls has amounted to little. It would make any ‘Conservative recovery’ narrative difficult to sustain, and would mean Labour’s position was relatively robust heading into 2024 - although nothing is guaranteed, of course.
Labour win by 15+ points: Things Can Only Get Better
A true landslide result for the Labour Party, which would likely also involve a rout by the Liberal Democrats in many Conservative heartlands in this scenario. The Conservatives would likely be distraught by such a result - with the situation being significantly worse than they feared, certainly losing well over 1,000 seats. The ‘narrow path’ to a Conservative election victory in 2024 would feel very thin indeed.
This result might suggest that the real Labour lead larger than polls suggest at present, and that Labour could be course for a landslide majority at the next General Election after all. Labour campaigners would spend the day of the King’s Coronation playing D:Ream on repeat.
There is nuance here of course - many Lib Dem or Green votes are not ones that would go to Labour in a straight Lab/Con contest; not all Lib Dem votes would go to Greens in a straight Con/Green contest, and vice versa. This is especially true at Local Elections, where people who are not typical Lib Dem or Green voters may be more likely to consider voting for these parties than at General Elections.
There is presumably also significant geographical variation in e.g. where the Lib Dem/Green vote is more Labour-leaning. But the point here is that in aggregate, across GB, voters who would consider voting these parties have significantly greater overlap between them with the Conservatives. Bracknell Forest could be an interesting test of this - though it is just one council, and especially with new boundaries, there’s no easy counterfactual to assess against.
This also points to a wider issue around commentary of Labour’s prospects of winning Government. The only precedent that most commentators have witnessed of a successful Labour LOTO was Blair, arguably the most electorally successful postwar politician, who achieved the largest Labour majorities ever - twice. The ‘threshold’ for Labour success is inevitably lower than his electoral achievements is, perhaps significantly so. But as we lack recent precedents for what a Labour Party heading for an ordinary, or narrow victory looks like, many commentators - including those sympathetic as well as hostile to Labour -possibly set the bar of Labour success far too high as a result.
What's the R squared for the Lab poll lead vs NEV regression?
Fascinating stuff. When will the NEV be calculated? Something to distract us from the coronation...