Greater Manchester 2026 · Ward-Level Structural Analysis
Local elections, 7 May 2026 · 213 of 214 wards declared · 10 boroughs

Reform won 104 of
213 GM wards. The
structural divide is stark.

Across the 213 declared wards in the 10 Greater Manchester boroughs, Reform vs % apprenticeship sits at r = +0.64 — the single strongest correlation in the dataset. Reform's mean ward has 6.8% apprenticeship-trained residents vs 3.2% in Green wards. Reform also tracks negatively with degree-holding (−0.50), working-from-home (−0.45) and density (−0.36), and positively with UK-born share (+0.46). Conservative and LibDem wards share the highest-degree, highest-WFH profiles — Conservatives in lower-density rural-fringe seats (Bowdon, Hale, Bamford, Norden), LibDems in dense suburban knowledge-worker belts (Stockport, Timperley). Greens win the dense, young, privately-rented inner cores. Reform sweeps the post-industrial periphery.

Sources: All 10 GM borough councils' published 2026 results · ONS Census 2021 ward-level data (UK Data Service XLSX + NOMIS CSV): TS006 (density), TS007 (age), TS063 (occupation), TS067 (qualifications), TS054 (tenure), TS004 (country of birth), TS061 (method of travel to work), TS008 (sex), TS022 (ethnic group). ONS small-area income estimates (FY ending March 2023, "Net income before housing costs") at MSOA level, aggregated to ward via Census-2021 OA-count weighting using the ONS Open Geography Portal OA→MSOA + OA→Ward lookups — see scripts/03b_aggregate_income.py. Council Tax Support claimants by Manchester ward, June 2025 (Manchester only — other boroughs run separate schemes). Bolton, Stockport, Trafford had ward boundary changes between Census 2021 and the 2026 election; ~30 of those wards use approximate Census mappings flagged in the methodology. Manchester's Miles Platting & Newton Heath ward is still pending. Bury's Moorside ward election was cancelled (Reform candidate died; rerun expected June).
§ 1 Top correlations · all variables Pearson r · 213 declared GM wards
Strongest
+0.64
Reform shows a strong positive correlation with % with apprenticeships: wards Reform won average 6.8% apprenticeship-trained residents vs 3.2% in Green-won wards. The single strongest correlation across 213 wards.
Reform
−0.50
Reform shows a strong negative correlation with % with a Level 4+ qualification (degree): Reform-won wards average 26% degree-holders, against Green 41%, LibDem 40%, Conservative 46%.
Reform
+0.46
Reform shows a strong positive correlation with % UK-born: Reform-won wards average 90% UK-born residents vs 69% in Green-won wards — a 21-point gap.
Reform
−0.45
Reform shows a strong negative correlation with % working from home: Reform 24%, Green 34%, LibDem 37%, Conservative 40%. WFH is a graduate-occupation marker.
Greens
+0.63
Greens show a strong positive correlation with % aged 18-29: Green-won wards average 29% young adults vs 14% in Reform-won wards.
Greens
+0.62
Greens show a strong positive correlation with % private renting: Greens win the dense, renter-heavy inner cores (37% private rent vs Reform's 17%).
Greens · density
+0.55
Greens show a strong positive correlation with population density: Green mean 6,229/km², Reform 2,664/km², Conservative 1,849/km². The urban-rural axis.
Conservative
+0.35
Conservatives show a moderate positive correlation with % in SOC 1-3 (graduate-level jobs): the Tory ward profile (n=11) carries the oldest median age (44.3), highest L4+ (46%), and highest SOC 1-3 (63%) — the affluent rural fringe.
§ 2 Top correlations per party Top 5 positive · top 5 negative

For each main party, the five strongest positive correlations and the five strongest negative correlations between a 0/1 party-win indicator and the ward's structural profile. To avoid showing the same age signal three times, only the strongest age-related variable (median age, % under 18, % 18-29, % 30-49, % 50-64, % 65+) is shown per direction. Conservative (n=11) and LibDem (n=19) have smaller samples so their r values are noisier — read them as direction-of-pattern rather than precise effect size.

§ 3 Full correlation matrix All 213 declared GM wards · Pearson r

Each row is a structural variable; each column is the party-win indicator (1 = ward won by that party, 0 otherwise). Reform now has the largest sample (n=104) so its correlations are tightest. LibDem (n=19) and Conservative (n=11) signals are real but noisier. Independent (n=9) and Other (n=6, mostly The Oldham Group + Workers Party + Radcliffe First + community parties) are too varied to show clear structural patterns. Bolton, Stockport, and Trafford had ward boundary changes between Census 2021 and the 2026 election; ~30 wards there use approximate Census mappings (full list in methodology).

Ward variable Green Labour Reform Lib Dem Combined
Green / Reform
Sort by
Where each party wins, structurally — Greater Manchester 2026 wards Pearson r between a 0/1 party-win indicator and each ward's Census 2021 structural profile, across 213 declared wards. Showing the ten variables most correlated with Reform's wins. PARTY · WARDS WON Reform n=104 Green n=30 Labour n=34 LibDem n=19 Conservative n=11 −0.6 −0.3 0 +0.3 +0.6 % Apprenticeship % UK-born % White % Aged 50–64 % SOC 1–3 % Asian % WFH % Mixed % Degree (L4+) % Other ethnic +0.64 −0.50 Where each party wins, structurally — Greater Manchester 2026 wards Pearson r between a 0/1 party-win indicator and each ward's Census 2021 structural profile, across 213 declared wards. Showing the ten variables most correlated with Green's wins. PARTY · WARDS WON Reform n=104 Green n=30 Labour n=34 LibDem n=19 Conservative n=11 −0.6 −0.3 0 +0.3 +0.6 % Mixed % Aged 18–29 % Private rented % Other ethnic Pop. density Median age % UK-born % Aged 65+ % Aged 50–64 % Apprenticeship +0.64 −0.59 Where each party wins, structurally — Greater Manchester 2026 wards Pearson r between a 0/1 party-win indicator and each ward's Census 2021 structural profile, across 213 declared wards. Showing the ten variables most correlated with Labour's wins. PARTY · WARDS WON Reform n=104 Green n=30 Labour n=34 LibDem n=19 Conservative n=11 −0.6 −0.3 0 +0.3 +0.6 % Aged 0–17 % Other ethnic % Aged 30–49 % Asian % Mixed Pop. density % Aged 65+ % UK-born % White % Apprenticeship +0.22 −0.19 Where each party wins, structurally — Greater Manchester 2026 wards Pearson r between a 0/1 party-win indicator and each ward's Census 2021 structural profile, across 213 declared wards. Showing the ten variables most correlated with Lib Dem's wins. PARTY · WARDS WON Reform n=104 Green n=30 Labour n=34 LibDem n=19 Conservative n=11 −0.6 −0.3 0 +0.3 +0.6 Income £/yr % SOC 1–3 % WFH % Aged 65+ % Owner-occupied Median age % Degree (L4+) % Private rented % Social rented % No qualifications +0.29 −0.26 Where each party wins, structurally — Greater Manchester 2026 wards Pearson r between a 0/1 party-win indicator and each ward's Census 2021 structural profile, across 213 declared wards. Showing the ten variables most correlated with Conservative's wins. PARTY · WARDS WON Reform n=104 Green n=30 Labour n=34 LibDem n=19 Conservative n=11 −0.6 −0.3 0 +0.3 +0.6 % SOC 1–3 Income £/yr % Degree (L4+) % WFH % Owner-occupied Median age % Aged 65+ Pop. density % Social rented % No qualifications +0.35 −0.28

Each view shows the ten Census variables most correlated with the selected party's wins, arranged left-to-right from most-positive r to most-negative — so the selected party's line descends by construction. Where another line climbs against that descent (clearest in the Reform / Green pair), the inversion is real signal, not an artifact of variable choice.

§ 4 Per-borough breakdown 214 wards · 213 declared

Reform won 104 of 213 declared wards across Greater Manchester — sweeping Wigan (24/25), Tameside (18/19), and dominating Oldham (13/20), Rochdale (13/20), Salford (12/20). Greens broke through Manchester (18/32) but won only sparsely outside it. Stockport stayed Lib Dem (13/21). Trafford stayed Labour (8/21) with significant Conservative gains in the Altrincham area. The full borough-by-borough picture:

Borough · ward count by winning party

Manchester's Miles Platting & Newton Heath ward had not declared at the time of writing. Bury's Moorside ward election was cancelled in April after Reform candidate Victor Hagan died (rerun expected June). Salford's Cadishead & Lower Irlam had two seats up but is counted once here.

§ 5 Average ward profile All 213 declared GM wards · means
§ 6 All wards · 10 boroughs 214 wards · sorted by borough → party → density
Method. All 10 Greater Manchester borough councils held local elections on 7 May 2026; ward-level results were collected from each council's official results page or the most authoritative local source (bolton.gov.uk, oldham.gov.uk, stockport.gov.uk, trafford.gov.uk, electionresults.wigan.gov.uk for council-direct; Bury Times via Yahoo, Tameside Correspondent, Rochdale Times, Salford Now, salfordmedia.co.uk, Northern Quota for the rest). For each ward, the structural profile is built from ONS Census 2021 ward-level data: TS006 (density), TS007 (single-year age), TS063 (10-category occupation), TS067 (8-category qualifications, ages 16+), TS054 (9-category tenure), TS004 (12-category country of birth), TS061 (12-category travel to work), TS008 (sex), TS022 (ethnic group, detailed). Aggregations: % apprenticeship and % L4+ are share of residents 16+ (TS067 codes 3 and 5 respectively); % owned = TS054 codes 0+1+2; % social rent = codes 3+4; % private rent = codes 5+6+7 (rent-free bundled with private); % UK-born = TS004 code 1 / total; % WFH = TS061 code 1 / sum of codes 1-11 (employed denominator); % female = TS008 Female / (Female + Male); the five TS022 shares (% White, Asian, Black, Mixed, Other ethnic) are each top-level group / TS022 "All usual residents", with % Black combining the African and Caribbean sub-groups per the standard ONS 5-category presentation. Correlations are Pearson r vs a 0/1 party-win indicator across the full 213-ward declared sample.

Boundary mismatches. Bolton (2023 review), Stockport (2023 review), and Trafford (2023 review) all had ward boundary changes between Census 2021 and the 2026 election. Where 2026 ward names match 2021 ward boundaries, an exact GSS code lookup is used (104 of 161 new wards). Where names changed but boundaries are substantially the same, a fuzzy mapping is applied (33 wards) — e.g. Bolton's "Heaton, Lostock & Chew Moor" maps to 2021's "Heaton and Lostock". Where 2026 wards are splits or significant re-bounds of 2021 wards, an approximate proxy is used (e.g. Bolton's "Farnworth North" and "Farnworth South" both inherit the old "Farnworth" Census profile). This introduces some imprecision for these three boroughs, particularly Trafford, but doesn't materially affect the borough-level patterns. Wigan also had a 2024 boundary change handled the same way.

Coverage notes. 213 of 214 wards are declared. Manchester's Miles Platting & Newton Heath was still pending at the time of writing. Bury's Moorside ward election was cancelled in April 2026 after Reform candidate Victor Hagan died; rerun expected June 2026. Salford's Cadishead & Lower Irlam had two seats up (both Reform); counted once. Bury count assumes a 16-of-17 declared sample. Per-ward turnout is published for Tameside, Trafford, Oldham, Rochdale, Bury (45% borough-wide), Stockport (44% borough-wide), and one Salford ward (Swinton Park 39.45%); other boroughs published only borough-level turnouts. Council Tax Support data is Manchester-only — other boroughs run separate Council Tax Reduction schemes that aren't directly comparable.

Caveats. Independent and Other categories combine heterogeneous local parties (Horwich & Blackrod First, Radcliffe First, The Oldham Group Party, Edgeley Community Association, Workers Party of Britain, etc.) so their correlation profiles are weak by construction. The Conservative sample (n=11) and LibDem sample (n=19) are smaller than Reform (n=104) so their correlations are wider; the means in §4 are more reliable than the per-variable r values. Reform's structural correlation (vs % apprenticeship +0.64, vs L4+ −0.50) weakened slightly from the 48-ward Manchester+Salford sample (+0.75 / −0.50) but remains the strongest single signal in the dataset.