England local elections: seats by party

Starting probabilities model (May 2026 England local elections)

This market uses a swing-first multilevel simulation model for party seat totals across 5,014 seats (134 councils voting on 7 May 2026; ordinary elections only; Surrey shadow elections excluded).

Key charts

Model structure

1. Party swing layer (hierarchical): each party swing depends on polling swing vs baseline, incumbency conversion tendency, defended-seat exposure, governing-party cycle effect, family random effects, party random effects, and election-level shocks.
2. Compositional normalisation: shifted baseline seat shares are normalised so all party shares sum to 1 each draw.
3. Overdispersion + seat allocation: Dirichlet draw-level noise and Hamilton largest-remainder apportionment produce integer seats that always sum to 5,014.
4. Market ladder estimation: ladders are Monte Carlo estimates of P(seats >= threshold).

Data and reproducibility assets

• Download the Python script used for this upload
• Download model summary JSON (expected seats and 90% intervals)