TL;DR Sian Astley’s Reform campaign in Greater Manchester fronted a grooming-gangs pledge explicitly built around protecting women and girls, while Labour’s first-preference vote fell by close to a third against the last comparable contest and turnout fell by 130,000. Neither shift shows up as a swing of women toward Astley. One pollster’s crosstab has Reform running nine points stronger with men than women, and the largest gap in the whole survey is between men and women who don’t know how they’ll vote at all. The right’s safety pitch and Labour’s collapse are both real. Neither has yet been decided on the terrain this series exists to track. A postscript addresses the strongest version of the counter-argument directly, Sir John Curtice’s own account of the result as a ‘convincing’ win, and why it doesn’t settle the question either way.
It started with a comparison, not a table. Reading Dave Kellaway’s piece on 1 August, the word ‘bounce’ sat wrong against something I’d watched six weeks earlier in Makerfield, when Burnham returned to Westminster on a result that looked, at 54.8% to 34.5%, like vindication. It wasn’t quite that. Labour’s raw vote in Makerfield rose by about a third on its 2024 baseline, but almost all of that rise traced back to the Conservative, Green and Liberal Democrat vote collapsing to a combined 3.3% around him, tactical squeeze rather than enthusiasm. Reform’s own raw vote rose too, by just under 3,000, even as its share fell against the swollen turnout behind Labour. Tactical voting did what it is built to do: it kept the far right out of the seat. It did nothing to shrink the far right’s actual vote, which grew regardless, and nothing but stronger, substantive policy, not the tepid technocratic management Labour offers in its place, is going to do that. If Makerfield was a bounce, it was one built on other parties disappearing, not on Labour convincing anyone new and not on Reform’s own base going anywhere. Watching the same word land on Manchester five weeks later meant going looking for what was actually moving underneath it. What turned up was not where the framing assumed it would be.
Sian Astley launched her campaign for Greater Manchester mayor with a promise to publish ‘all files’ relating to grooming gangs ‘going back at least 40 years.’ It was the centrepiece of her manifesto, repeated in launch interviews. Asked what should happen to the people she blames for the region’s failures, she was specific: ‘Personally I’d just ensure the Green voters, uni-party MPs (present or past) and Cllrs house the illegal criminals, rapists and economic chancers.’ Marlon West, Restore Britain’s candidate, carried a version of the same appeal from a different angle, a former nurse whose own daughter was a victim, running explicitly as a campaigner against child sexual exploitation rather than as a party functionary. Two candidates, two parties competing for the same flank, both fronting the same claim to be the true defenders of women and girls against a Labour establishment they say spent decades looking away.
This is the mechanism the Terrain of Survival series has traced elsewhere: the far right does not win working-class women by accident. It wins them, where it wins them, by occupying ground the left has vacated, the ground of safety, of material precarity, of who is actually present when the immediate danger is a person rather than a statistic. Astley’s campaign was built to occupy exactly that ground in Greater Manchester. What the numbers from this contest show, and it is the sort of finding a rigorous series has to sit with rather than write around, is a region where two things are true at once and neither is what the loudest readings claim: the right’s safety pitch has not converted into a female vote, and Labour’s own coalition did not hold either. Both campaigns lost the argument to the same thing, an electorate that simply did not show up.
Start with the scale of what didn’t show up, and start with the right baseline. The 63.4% Burnham won in 2024 gets quoted more often, but 2024 was the only Greater Manchester mayoral contest run under first-past-the-post, imposed by the outgoing government’s Elections Act 2022 and reversed again before this one. The properly matched figure is 2021, the last time Greater Manchester used the same supplementary vote system Craig faced: Burnham took 67.3% of first preferences that year. Against that baseline, Craig’s 47.1% is a fall of twenty points, close to a third of Burnham’s vote, a larger collapse than the more commonly cited comparison to 2024 suggests, not a smaller one. Burnham set a second, harder floor seven weeks earlier at Makerfield, 54.8%, a result already propped up by tactical squeeze rather than Labour enthusiasm, as the previous paragraph sets out; that comparison holds cleanly, since both Makerfield contests were fought under first-past-the-post. Craig’s first-round vote came in under that floor too, before a single second preference was counted. Turnout dropped seven points on the 2024 baseline, roughly 130,000 fewer people voting than last time despite the electoral register growing by 60,000. Labour’s own count lost around 170,000 votes. The combined Reform, Restore and Conservative vote rose by only 30,000 over the same period, a fifth the size of Labour’s loss. Craig’s eventual second-round margin, 66 to 34, looks like resilience taken on its own, but it is manufactured by transfers, not earned by turnout: roughly nine in ten Green voters broke for Labour, while Reform picked up only about half the Conservative and Restore transfers on offer. A voting system built to reward broad coalitions rewarded Labour’s for having one. It is not evidence that Labour’s own vote held.
None of that demobilisation shows up as a story about women moving right, whatever the framing of Astley’s own campaign implies. YouGov’s internal crosstab, fielded 8 to 15 July among 2,200 adults across the city-region, models Astley at 44% among men against 33% among women in the projected second round. Restore’s first-round figure splits the same way in miniature, 8% among men against 6% among women. If the grooming-gangs pledge is functioning as a gendered recruitment tool in this contest, it is not showing up in a female skew toward either candidate built around it. Reform’s vote, for now, looks like a fairly conventional male-skewed populist coalition, not a coalition of women rallying to a self-styled protector.
Editorial note: this is one pollster’s snapshot from three weeks before the vote, not verified post-election behaviour. UK returns don’t record voter gender, so there is no way to check the model against the actual result directly, and the actual second round (66-34) ran four points more favourable to Labour than this poll projected (62-38). Something moved in the final fortnight that this data can’t see. Treat the gender figures as indicative of the state of the race in mid-July, not as a settled account of who voted for whom.
What the same table shows instead is more useful than a simple negative finding, and it is where the series should actually be looking. In the raw, unmodelled first-vote intention, before YouGov’s likelihood-to-vote adjustment smooths the picture into a headline figure, 27% of women say don’t know against 14% of men, and a further 18% of women say they will not vote against 13% of men. Add refusals and the gap widens further. Women are not choosing Astley over Labour in any numbers that show up in this poll. They are, in far greater proportion than men, choosing neither. That is the same demobilised bloc that swallowed Labour’s own vote, and within it the gender gap is the widest single split in the entire table, wider than the gap between any two parties.
There is a further tension in Astley’s own positioning worth naming. She is a landlord who owns seventeen flats across three properties in south Manchester, campaigning in part on a promise to audit the quality of the region’s social and rented housing stock, having concluded, in her own words, that too much of it ‘is not up to scratch.’ A landlady proposing to inspect the standard of a housing market she profits from is not disqualifying in itself. It is, though, precisely the kind of gap between rhetoric and material position that the terrain-of-survival framework exists to expose, and it sits alongside a grooming-gangs pledge that promises exposure and accountability without connecting either to a single concrete demand: a refuge, a childcare place, a transport route that feels safe after dark. Protection as spectacle, files to be published, someone else’s failure to be exposed, is not the same offer as protection as material change, and the distinction matters more to a demobilised electorate than to a committed one.
None of this should be read as reassurance. A movement can occupy a demobilised bloc slowly and then all at once, and Astley has not been in politics since May. Rupert Lowe’s Restore operation is younger still and appears to be drawing votes from Reform’s own right flank rather than losing the argument entirely, which is its own kind of warning about the durability of any comfort taken from this month’s numbers. Farage himself heads into the contest’s aftermath with his personal ratings among the worst recorded in a decade of tracking, amid the unresolved Cottrell donations affair and his decision to resign his seat and refight Clacton, but a wounded leader is not the same thing as a beaten movement, and Restore exists precisely to catch whatever Reform drops.
What the Manchester data complicates is a version of this series’ own thesis that was becoming too comfortable to repeat: that the far right recruits women principally through safety rhetoric, and that the numbers will show it doing so wherever that rhetoric appears. They do not, not here, not on this evidence. What they show is a contest decided, for now, by who does not turn out rather than who does, a female electorate more unattached than the male one, and two campaigns, Astley’s and Labour’s both, that lost ground to the same withdrawal rather than to each other. Whether that withdrawal is available to be organised, by anyone, on either side of the argument, or whether it simply persists as a permanent feature of a region that no longer expects much from a mayoralty whose main job is keeping the buses paid for, is not a question this poll answers.
Postscript: The Right Baseline, and the Wrong Inference
Three things worth adding after publication: a correction; an observation and a distinction.
The electoral systems
The correction: this piece originally measured Labour’s Manchester collapse against Burnham’s 63.4% in 2024. That figure is real, but it is the wrong comparison. 2024 was the only Greater Manchester mayoral contest run under first-past-the-post, a temporary change under the Elections Act 2022 that was reversed before this contest. The properly matched figure is 2021, the last time Greater Manchester used the same supplementary vote system Craig faced: 67.3%. Measured against that baseline, Craig’s 47.1% is a fall of twenty points, close to a third of Burnham’s vote, not the quarter first reported here. The piece above has been corrected accordingly. The finding gets stronger, not weaker, once the systems are matched properly.
The Burnham bounce
The distinction: Sir John Curtice’s own account of the result, published in BBC InDepth the day of the count, calls Labour’s win ‘convincing’. His case: 47.1% on first preferences was ‘a near doubling’ of Labour’s 24% tally in May’s local elections and four points up on Labour’s local performance in the 2024 general election. Curtice attributes the national recovery mainly to Labour clawing back voters from the Greens and Liberal Democrats: the share of 2024 Labour voters who had defected to the Greens fell from 14% to 9% after Burnham took over, and 67% of 2024 Labour voters now say they would vote Labour again, against 57% when Starmer resigned. Those figures are real, and they describe precisely the mechanism this piece has been calling a Green-to-Labour swing rather than new territory won.
What Curtice’s piece does not do is measure Craig’s result against the last time Greater Manchester used the same voting system she stood under. Every comparison in his analysis, the 2024 general election, May’s locals, the 2024 mayoral race, either uses a different contest entirely or the one Greater Manchester mayoral election run under first-past-the-post. Measured against 2021, the last supplementary-vote contest, Craig’s 47.1% is still a fall of twenty points on Burnham’s 67.3%. A convincing win by the standard Curtice is using is compatible with a real collapse in Labour’s own vote by the standard this piece is using, because the two are answering different questions: whether Labour beat Reform convincingly, and whether Labour’s own coalition actually turned out. The recovery Curtice describes is, separately, concentrated in the North West and Yorkshire and Humber, the two English regions where Labour has clawed back the most ground, according to the nations-and-regions analysis addressed below rather than to Curtice’s own piece, which doesn’t break the national figure down by nation or region. Greater Manchester sits inside the region where that recovery is strongest anywhere in the UK state, since it is Burnham’s own home ground. If the conversion Curtice describes, Green voters returning to Labour, is not reaching the 27% of women in Manchester who say they don’t know how they will vote, it is hard to see where else it would.
Four nations, not one, in the UK state
It is also worth keeping the word ‘national’ out of this, since the recovery Curtice tracks is a Great Britain figure, not a description of all four nations moving together. Separate nations-and-regions analysis, cited independently of Curtice’s piece, shows Labour’s recovery landing unevenly: strongest in the North West and Yorkshire and Humber, minimal in Scotland and Wales, where Labour had just lost heavily to the SNP and Plaid Cymru in May. Curtice’s own numbers describe the mechanism. The nations-and-regions data describes where it is and is not working (and, of course, outlook the UK statelet that divides Ireland). Put together, they say the same thing this piece has been arguing from a different angle: a recovery built on winning back people who never stopped voting is not evidence about people who stopped voting at all, wherever in the UK state you look for it.
Addendum: May Was the Exception, Not July
One more thing worth adding, because it changes which number in this piece should be doing the most work. Everything above measures the by-election result against May’s local elections, since May is the most recent prior contest. But treating May as the baseline gets the comparison backwards. The Mayoral elections in 2021 and 2024, two separate elections years apart, run under two different voting systems, both landed Labour and Green on a combined 70 to 72% against a combined right of 18 to 22%, a gap of roughly 49 to 52 points. That consistency across two unrelated systems and two different years is what a structural baseline looks like. The vote in May 2026 is the outlier against that baseline, not July.
On this reading, May’s near-parity, a two-point gap between the blocs, was the emergency. July’s 22-point gap is a partial retreat from that emergency, not a recovery in any strong sense. It recovers only two in five of the votes lost between the historic baseline and May’s low point, and it leaves the result still 29 points short of where 2021 and 2024 both independently sat, well over half the historic gap still open. Curtice’s ‘near doubling’ of the May tally is measuring the right change against the wrong anchor. Against the only anchor that reflects a stable, unremarkable political moment for the region, this by-election result remains a landslide-scale departure from normal, only a smaller departure than the one two months before it.
None of this settles the harder question this piece opened with, and it sharpens rather than answers it. There is no gender breakdown available for May’s collapse, so it cannot be shown directly that the recovery since then reached men and women unevenly. What can be shown is that whatever partial recovery took place in those eight weeks, it left behind a pool of undecided and non-voting women in July twice the size of the equivalent male pool, the widest single gap in the entire crosstab. A general emergency and a partial general recovery would be expected to move roughly evenly across the electorate unless something else was also at work. The gender skew in who was still undecided by the time of the by-election is that something else, and it is not explained by treating May as the baseline any more than July is. Whatever pulled the region back from a two-point gap to a twenty-two-point one over the summer, it evidently pulled men back into the electorate faster, or further, than it pulled women, and that asymmetry, not the size of the recovery itself, is the question actually worth answering.