Westminster By-Elections

Results and model signal — 2024–present Parliament

5 by-elections held since July 2024. 2 Labour losses.

By-election forecastFrozen · final estimate

Makerfield

Lab–Reform marginal · Thursday 18 June 2026 — triggered by Josh Simons' resignation to clear Andy Burnham's path back to Westminster. Fourteen candidates; effectively a two-horse race.

85%Burnham · Labour
15%Kenyon · Reform UK
Probability of winning the seatLikely Lab · 200,000 simulations
Projected vote share · mean & 90% interval
Burnham (Lab)
45.941.250.5
Kenyon (Reform)
39.934.944.9
Shepherd (Restore)
7.03.710.8
Austin (LD)
1.20.81.7
Green (Green)
3.22.14.3
Winstanley (Con)
1.40.91.9
Other (Other)
1.00.71.3
03060%
Lab–Reform +5.9pp, 90% interval -3.4 to +15.1. The vote margin sits within polling error — the favourite is not safe.
The pivot · right-wing consolidation

The left is consolidated behind Burnham; the right is split. Whether Reform squeezes Restore’s vote decides the seat.

Restore holds 5–10%86%
Restore surges ≥10%93%
Restore late-squeeze <5%78%
Primary input: Recency-weighted average of 5 polls / Survation ×2, More in Common, Opinium, Convergent, 18 May - 12 Jun 2026Blended with: 5 constituency polls (combined N≈2,605): Survation ×2, More in Common/UCL, Opinium, Convergent/Times · Burnham lead spans +5 (Opinium) to +12 (Convergent); Restore 5–8% across the set · cautious recency-weighted anchor: Lab 45.7 / Ref 39.5 (margin +6.2) — Convergent's +12 treated as a high outlier, not over-weighted · 2024 GE baselineUpdated: 15 June 2026Status: frozen
By-election forecastFrozen · final estimate

Aberdeen South

SNP–Con three-way marginal · Thursday 18 June 2026 — triggered by Stephen Flynn's resignation. A genuine three-way at the top: the SNP starts narrow favourite, but a by-election protest against the SNP government and unionist consolidation behind the Conservatives make this genuinely competitive — the Conservatives are a live threat.

65%SNP · Scottish National Party
35%Conservative · Conservative
Probability of winning the seatTilts SNP · 200,000 simulations
Projected vote share · mean & 90% interval
SNP (SNP)
33.126.939.4
Labour (Lab)
6.00.511.1
Conservative (Con)
30.120.240.7
Reform (Reform)
18.210.825.6
Liberal Democrats (LD)
6.31.111.8
Green (Green)
3.60.28.9
02550%
SNP–Con +3.0pp, 90% interval -10.5 to +15.9. The vote margin sits within polling error — the favourite is not safe.
What could change it

The unionist vote is split between Conservative and Reform. Whether Reform's vote consolidates behind the Conservatives — strongest in the Deeside wards — decides whether the SNP holds.

Conservative More weight on the close Holyrood read · stronger unionist consolidation · oil-salience live48%
Central Blended Holyrood + Westminster reads65%
Aggressive Right stays split (Reform holds) · SNP firm · low consolidation79%
Primary input: Holyrood→Westminster transfer / 7 May 2026 Holyrood result + Scotland-wide Westminster VI, Holyrood 7 May 2026; Norstat 27–29 May 2026Blended with: Holyrood 7 May 2026 (weighted Aberdeen Deeside and North Kincardine 86%, Aberdeen Central 11%, Aberdeenshire West 3%) · GE2024 Westminster baseline (live from election_results) · Scotland-wide Westminster movement since GE2024 (Norstat 27–29 May 2026)Updated: 7 June 2026Status: frozen
By-election forecastFrozen · final estimate

Arbroath and Broughty Ferry

Safe SNP hold; Reform–Labour fight for second · Thursday 18 June 2026 — triggered by Stephen Gethins' resignation. A safe SNP hold on every read; the live questions are the size of the majority and whether Reform or a collapsed Labour finishes second.

>99%SNP · Scottish National Party
Probability of winning the seatLikely SNP · 200,000 simulations
Projected vote share · mean & 90% interval
SNP (SNP)
39.934.445.8
Labour (Lab)
14.48.720.1
Conservative (Con)
14.79.320.6
Reform (Reform)
19.913.926.0
Liberal Democrats (LD)
6.32.011.1
Green (Green)
1.90.25.6
02855%
SNP lead over nearest +19.0pp, 90% interval +10.9 to +27.1. A clear lead, comfortably outside the modelled uncertainty.
What could change it

The SNP holds comfortably on every read. The real contest is for second place: Reform (71.8%) versus a collapsed Labour (14.8%).

Conservative Holyrood-weighted · softer SNP · stronger Reform>99%
Central Blended Holyrood + Westminster reads>99%
Aggressive Firmer SNP · weaker Reform>99%
Race for second
Reform UK72%
Labour15%
Conservative13%

Probability of finishing second (the seat itself is not in doubt).

Primary input: Holyrood→Westminster transfer / 7 May 2026 Holyrood result + Scotland-wide Westminster VI, Holyrood 7 May 2026; Norstat 27–29 May 2026Blended with: Holyrood 7 May 2026 (weighted Angus South 60%, Dundee City East 40%) · GE2024 Westminster baseline (live from election_results) · Scotland-wide Westminster movement since GE2024 (Norstat 27–29 May 2026)Updated: 7 June 2026Status: frozen

What the by-elections tell us

Five by-elections have been held since the 2024 general election, and they no longer tell one story. The first two — Runcorn & Helsby (Reform gain from Labour by 6 votes) and Gorton & Denton (Green gain from Labour) — followed abnormal triggers and showed Labour squeezed from both flanks at once. The three contests on 18 June 2026 followed routine vacancies and broke the pattern: Labour held Makerfield comfortably (54.8% to Reform’s 34.5%), the SNP held Arbroath & Broughty Ferry, and the Conservatives — under 8% in every North-West England by-election this parliament — took Aberdeen South from the SNP on 49.5% of the vote.

We froze forecasts for the June contests in advance and scored them against the declared results. Two winners were called correctly; Aberdeen South was missed outright — we had the SNP at 65% to hold a seat the Conservatives won by 20.9 points. The June errors share a direction: we understated Conservative support in all three contests and overstated Reform in all three.

Scored forecasts: Makerfield ✓ (Lab 85%) · Arbroath ✓ (SNP 99%) · Aberdeen South ✗ (SNP 65% — Con gain)

No direct by-election adjustment is applied to the GE forecast — the 0.08-weight swing signal was retired at v2.4. Results now enter the model as out-of-sample tests, frozen in advance and scored in our calibration ledger.

Makerfield

18 June 2026 · North West

Labour hold — majority 9,241 (20.3%)

Result
Turnout: 58.8%
PartyBy-electionVotes
Lab54.8%24,937
Reform34.5%15,696
Restore6.8%3,111
Con2.2%997
Green0.7%308
Lib Dem0.4%163
Lab
Reform
Restore
Con
Green
Lib Dem
Context
Trigger: vacancy

Lab hold. Per-party detail in by_election_result_parties. Source: Makerfield count declaration 2026-06-18.

Aberdeen South

18 June 2026 · Scotland

Conservative gain from SNP — majority 6,050 (20.9%)

Result
Turnout: 37.4%
PartyBy-electionVotes
Con49.5%14,308
SNP28.6%8,258
Reform8.6%2,478
Lab5.4%1,550
Lib Dem4.4%1,270
Green3.4%974
Con
SNP
Reform
Lab
Lib Dem
Green
Context
Trigger: vacancy

Con gain from SNP. Per-party detail in by_election_result_parties. Source: Aberdeen City Council declaration 2026-06-18.

Arbroath and Broughty Ferry

18 June 2026 · Scotland

Scottish National Party hold — majority 5,278 (22.1%)

Result
Turnout: 31.4%
PartyBy-electionVotes
SNP41.1%9,802
Con19.0%4,524
Reform18.2%4,341
Lab15.3%3,651
Lib Dem6.1%1,452
SNP
Con
Reform
Lab
Lib Dem
Context
Trigger: vacancy

SNP hold. Majority 5278 = SNP 9802 minus Con 4524 (council); supplied 5178 rejected as derived from bad Con=4624 (Con-Reform gap of 183 confirms 4524). Source: Angus Council declaration 2026-06-18.

Gorton and Denton

26 February 2026 · North West · metropolitan

Green Party gain from Labour — majority 4,402 (12.0%)

Result
Turnout: 47.5%Electorate: 77,501Votes cast: 36,814
PartyBy-election2024 GESwing
Green40.7%13.2%+27.5pp
Reform28.7%14.0%+14.7pp
Labour25.4%50.8%-25.4pp
Conservative1.9%7.9%-6.0pp
Lib Dem1.8%3.9%-2.1pp
Green
Reform
Labour
Conservative
Lib Dem
Context
Trigger: resignation

Andrew Gwynne (sitting as Independent after Lab suspension for antisemitic WhatsApp messages) resigned on health grounds 22 January 2026

Green Party's first Westminster by-election win and first MP in the North. Labour collapsed -25pp. Three-way tactical voting: Forward Democracy, TacticalVote.co.uk all endorsed Greens to stop Reform. Reform strong second on 28.7% (+14.7pp). Con wiped out (7.9% to 1.9%). Highly abnormal seat: large Muslim community, Manchester urban, student areas -- not representative of Lab marginals. Trigger was Labour-suspended MP, further reducing signal quality. Polanski Green leadership (elected Sep 2025) appears to be driving membership and vote surge. Weight: 0.5.

Model Signal
Signal weight:
0.5

Why discounted: Abnormal trigger (resignation). Result inflated by protest effects specific to the by-election context.

Historical contribution (signal retired at v2.4 — not applied):
Reform
+0.59pp
Lab
-1.02pp
Green
+1.10pp
Con
-0.24pp
LD
-0.08pp

This swing signal was retired in the v2.4 reconciliation — it fed only one of the model’s two calculation paths, creating an artificial gap between them — and contributes nothing to the current GE forecast. Shown for transparency.

Runcorn and Helsby

1 May 2025 · North West · unitary

Reform UK gain from Labour — majority 6 (0.0%)

Result
Turnout: 46.2%Electorate: 70,666Votes cast: 32,655
PartyBy-election2024 GESwing
Reform38.7%18.1%+20.6pp
Labour38.7%53.0%-14.3pp
Conservative7.2%16.0%-8.8pp
Green7.1%6.4%+0.7pp
Lib Dem2.9%5.1%-2.2pp
Reform
Labour
Conservative
Green
Lib Dem
Context
Trigger: misconduct

Mike Amesbury (Lab) convicted of assault, sentenced to 10 weeks prison (suspended), resigned 17 March 2025

Reform's first Westminster by-election win. Six-vote majority. Con collapsed from 16% to 7%, split between Reform and potential Lab tactical voters. Lab vote held better than feared (~53% to 38%). Seat won by scandal effect (Labour MP convicted) + national anti-government swing. Abnormal trigger: weight reduced to 0.6.

Model Signal
Signal weight:
0.6

Why discounted: Abnormal trigger (misconduct). Result inflated by protest effects specific to the by-election context.

Historical contribution (signal retired at v2.4 — not applied):
Reform
+0.99pp
Lab
-0.69pp
Green
+0.03pp
Con
-0.42pp
LD
-0.11pp

This swing signal was retired in the v2.4 reconciliation — it fed only one of the model’s two calculation paths, creating an artificial gap between them — and contributes nothing to the current GE forecast. Shown for transparency.

Historical Context

By-elections are an imperfect signal for general elections. Turnout is lower (31–59% here vs ~60% at GEs), protest voting is stronger, and tactical voting is often explicitly coordinated. An earlier version of the model nudged national vote shares using by-election swings at a 0.08 weight; that signal was retired in the v2.4 reconciliation and no direct adjustment is currently applied. The relationship now runs the other way: each contest gets a forecast frozen before polling day and scored against the declared result. Our calibration work on 2019-parliament by-elections supports this caution — challenger surges do not transfer to general elections; the only component with out-of-sample support is governing-party underperformance.

MakerfieldAberdeenArbroathGortonRuncornGE average
Turnout58.8%37.4%31.4%47.5%46.2%~60%
Con vote share2.2%49.5%19.0%1.9%7.2%~17%
Lab-Reform gap+20.3pp-3.2pp-2.9pp-3.3pp0.0pp~-3pp

An earlier version of this note called the Conservative vote share “the starkest illustration of by-election squeeze”. Aberdeen South ended that generalisation: where the Conservatives began as the credible local challenger they polled 49.5%. The squeeze is a function of local competitiveness, not a universal by-election law.