Download the model here:
Under the first-past-the-post (FPTP) electoral system, the popular vote does not precisely translate into seats in Parliament. As a result, those of us who aim to predict electoral outcomes must use a variety of methods to estimate seat outcomes.
Some seat prediction models use ‘uniform national swing’ (UNS), where they take the +/- for each party nationally and apply it to each seat. Others, such as YouGov and Survation, use multi-level regression and post-stratification (MRP) – a complicated set of equations based on demographic modelling combined with national polling.
This page outlines my seat estimate model.
The Stats for Lefties Model
In one sentence, here is how it works: the model looks at how each party’s support has changed proportionally since the last election, and distributes those changes across all 632 mainland constituencies based on how much each seat contributed to that party’s total vote.
Example
- In the East Midlands Reform got 403,000 votes in 2024. Amber Valley (12,192 Reform votes) provided 3% of the total Reform vote in the region.
- With Reform now projected to gain 373,000 additional votes across East Midlands, 3% of those votes will be allocated to Reform in Amber Valley (11,263) giving them 51% of vote in the seat.
- At the same time, Labour is losing 341,000 votes across the region. Amber Valley provided 2% of the total Labour vote in the region, so 2% of that total (7,124) will be subtracted from the Labour total in Amber Valley, reducing them to 19% of the vote overall.
- We repeat this with all parties for a final result of:
Ref 51% (+22), Lab 19% (-18), Con 17% (-8), Grn 9% (+4), LD 4% (=)
Download the model here:
Stats for Lefties 