Senedd Forecast
The first election under Wales’s new proportional system. A three-way race between Reform (30 seats), Plaid Cymru (35), and Labour (15). No single-party majority is expected — 49 seats are needed of 96.
New electoral systemHow this election works
tap to toggle
How this election works
- Wales has 16 constituencies, each electing 6 MSs.
- You vote for a party, not a candidate (closed list).
- Seats in each constituency are allocated proportionally using the D'Hondt method.
- A party needs roughly 14% of the vote in a constituency to win one seat.
- Total Senedd size: 96 MSs (up from 60 under the old system).
- Majority = 49 seats. Most likely outcome is a coalition or minority government.
Central estimate — 96 seats
Paths to a 49-seat majority
Probability that each combination reaches or exceeds the 49-seat majority threshold across 5,000 simulated outcomes.
D'Hondt seat allocation per constituency
Each constituency elects 6 MSs. Mean projected allocation shown; bar integers rounded from fractional means using largest-remainder.
Afan Ogwr Rhondda
Bangor Conwy Môn
Blaenau Gwent Caerffili Rhymni
Brycheiniog Tawe Nedd
Caerdydd Ffynnon Taf
Caerdydd Penarth
Casnewydd Islwyn
Ceredigion Penfro
Clwyd
Fflint Wrecsam
Gwynedd Maldwyn
Gŵyr Abertawe
Pen-y-bont Bro Morgannwg
Pontypridd Cynon Merthyr
Sir Fynwy Torfaen
Sir Gaerfyrddin
Senedd voting intention
Single-vote polls under the new closed-list PR system. Pre-reform two-vote polls are excluded (not comparable).
MethodologyHow this forecast is built
senedd-v1-westminster-mc5000
How this forecast is built
Electoral system
The 2026 Senedd election uses a completely new electoral system: closed-list proportional representation with D'Hondt allocation. Wales is divided into 16 constituencies (each pairing two adjacent Westminster constituencies), each electing 6 Members of the Senedd, for 96 seats total. This replaces the previous Additional Member System (40 FPTP + 20 regional list = 60 members). This is the first election under the new system — every forecast is operating without precedent.
Poll aggregation
Polls are weighted by three factors combined multiplicatively: recency(exponential decay with a 10-day half-life, matching our GE model’s snapshot mode); sample size(square-root-of-n, clipped to a 0.5–2.0 range); and pollster accuracy (equal weights for now — no devolved-specific pollster accuracy database exists yet).
Projection method
The model starts from Senedd-specific polling, applies geographic variation across the 16 constituencies using Westminster 2024 GE results as a geographic anchor, and runs D'Hondt allocation in each constituency. 5,000 Monte Carlo simulations produce probability distributions.
Plaid — asymmetric heartland amplification. Plaid’s positive geographic deviations (in its North and West Wales heartlands) are amplified by 1.6×; negative deviations are left unchanged. This reflects Plaid’s historic over-performance at devolved elections.
Conservative and Lib Dem — dampened geographic variation.These parties’ Senedd support is less geographically concentrated than their Westminster support. Geographic deviations are dampened (Con at 0.15×, Lib Dem at 0.15×) to prevent the Westminster anchor from pushing them above D'Hondt thresholds in too many constituencies. Calibrated against the YouGov MRP which projects Con at 1 seat and Lib Dem at 0.
Greens — damping plus urban boost. Green geographic deviation dampened to 0.5× (sparse Westminster baseline) with a +4.5pp urban boost in Cardiff, Swansea, and Newport constituencies.
Comparison with YouGov MRP
The YouGov MRP (March 2026, n=2,978) projects Plaid 43, Reform 30, Labour 12, Green 10, Conservative 1, Lib Dem 0. Our model is close on Reform (30 vs 30), Labour (14 vs 12), and Greens (9 vs 10). The main divergence is on Plaid (38 vs 43) and Conservatives (4 vs 1).
Under D'Hondt, seats allocated to smaller parties mechanically reduce the allocation to larger parties. The remaining Con/Lib seats in our model come at Plaid’s expense. The 6th seat in each constituency is highly sensitive to small vote share changes.
Monte Carlo and candidacy
5,000 simulations with 2.5pp national polling-error shock, per-party vote shocks, and 2.5pp per-constituency local noise (wider than Scotland to reflect the novel system). D'Hondt allocates 6 seats per constituency with list-length caps from confirmed nominations.
All six major parties submitted lists in all 16 constituencies. Reform has 5 candidates in one constituency (capped at 5 seats). Minor parties (Gwlad, Heritage Party, Propel) included as “other”.
Key uncertainties
Plaid vs Reformfor largest party is genuinely close (polls 28–33% vs 23–31%). A single poll shift of 2–3 points could flip the outcome. Conservative near-extinction — both our model and the MRP project Conservatives reduced to 1–4 seats from 16 in 2021. Green breakthrough— 8–10 projected seats would be a first for the Greens in the Senedd; under PR with 6-member constituencies, 11% of the vote reliably translates into seats.
Data sources
- Senedd-specific polls from YouGov, Beaufort Research, More in Common, FindOutNow (25 polls, scraped from Wikipedia).
- Notional 2021 results calculated by Dr Jac Larner (Cardiff University) on 2026 constituency boundaries.
- Westminster geographic anchor — 2024 GE results for the 32 Welsh Westminster constituencies, mapped to 16 Senedd pairings.
- Candidate data — Democracy Club (nominations closed 9 April 2026).
- External benchmark — YouGov MRP (March 2026, n=2,978).
Uncertainty note
First election under a new system — no historical results to validate against. The notional 2021 baselines involve significant assumptions about how AMS votes would translate to a single-vote PR system. Wide confidence intervals are the honest answer.