What if 2024 had been counted differently?

The problem it tries to solve

In a first-past-the-post election, each constituency elects whoever has the most votes, and the national result is just the sum. Votes for losing candidates have no effect on the outcome, so seat shares can drift a long way from vote shares. In 2024 they drifted a long way:

Political scientists summarise this gap with the Gallagher index: 0 means seats exactly match votes, and higher means a bigger mismatch. For 2024 under first-past-the-post it is about 23.8, or 23.2 if you leave out the small parties and independents that can’t be represented in the model. We computed both from the official results.

How the proposal works

The Green Paper behind this site proposes Regional Direct–Compensatory Representation. It uses a single vote, as now, and three steps:

1. Divide the country into the 12 regions (England’s nine, plus Scotland, Wales and Northern Ireland). Each gets a fixed number of seats, 650 in total. About half are ordinary local constituencies, roughly twice the size of today’s.
2. Vote as normal. You vote for one local candidate. Whoever tops the poll in each constituency wins that seat outright.
3. Top up the region. Add up each party’s votes across the region to work out how many seats it “deserves”. Parties with fewer local wins than that receive extra seats, awarded to their best-placed candidates who didn’t win locally, until the region’s seats are all filled. Parties below a minimum vote share (the proposal uses 3%) don’t qualify for top-ups.

Because the top-up is calculated from the same single vote, there’s no second ballot to game. That is the main design difference from the German-style systems this resembles.

Try it yourself

This runs the allocation in your browser on the corrected 2024 regional vote totals. Move the sliders and watch the 650 seats change. Every square is one seat.

What the independent re-run found

We rebuilt the model from scratch: new data path, new code written only from the paper’s written description, and a tougher robustness test. The votes come from the House of Commons Library’s corrected 2024 results file for all 650 constituencies. We checked it totals exactly 28,809,340 valid votes and that every party’s national total matches the published figure.

Using the Green Paper’s own central local-seat scenario. NI parties and Speaker: SF, DUP, Alliance, UUP, SDLP, TUV plus the Speaker. The Greens gain at 3% because the Scottish and Welsh Greens then qualify, at the expense of one seat each for Labour, the SNP and Plaid.

The main results

• Seats track votes closely. The Gallagher index falls from about 23.8 to about 3.1, or from 23.2 to 1.0 when the unrepresentable residual is left out. Either way, most of the mismatch disappears. The two figures differ because 3.5% of votes went to small parties and independents that the model cannot give seats to.
• No party gets near a majority. Labour’s 234 is 92 short of 326. Labour + Lib Dems reach 312. Adding the SNP gives 329, or adding the Greens gives 353. The right-of-centre parties (Con, Reform, DUP, UUP, TUV) reach 264. Any parliament under this system would be a coalition or an agreement between parties.
• The result barely depends on who wins locally. We tried thousands of alternative local-seat patterns, including ones where Reform and the Greens win local seats they didn’t win in 2024. In seven of twelve regions the final seat totals didn’t change at all, and nationally Labour moved only between 233 and 238. That is the system’s strongest property: it makes the constituency map much less decisive.
• The settings matter more than the maps. Raising the threshold to 7% cuts the Greens from 41 to 23. Moving to a 60:40 local/top-up split lets Labour reach 252 and brings the Gallagher index (excluding the residual) from about 1.1 to 3.2. Around 50:50 to 55:45 the system stays close to proportional; beyond that it degrades.
• Independents lose out. Six independents won in 2024. The model returns none, unless they win a local seat outright, because top-up seats go to parties.

The full method, sensitivity tables and every number are in the report, and the code and data are published so you can check them.

Two models, one conclusion

This site rests on a comparison of two independent builds: the original Green Paper model and an audit, and a separate re-run built without using its code or its transcribed data. Where two different builds agree, that’s better evidence than either alone.

Would it suit the UK?

The numbers above can say what the system does. Whether it is a good idea is partly a matter of values, so we separate what we can show from what we can only argue.

As an electoral system in general

For the UK specifically

• Coalition becomes normal. On 2024 votes no party gets close to a majority. Westminster has coped with hung parliaments in 2010 and 2017, but its habits are built around single-party government.
• Constituencies get bigger. 327 seats of about 145,000 electors, roughly twice today’s size. Rural and island areas feel this most.
• The regional split is a hidden choice. The Green Paper’s regional seat numbers match a split by total population, not by registered voters as now. That moves London from 75 to 85 seats and Scotland from 57 to 52. The effect on party totals is small, but who counts is a constitutional question and should be debated openly.
• Scotland and Wales. The SNP would hold 17 seats on 2024 votes (9 under FPTP), with some protection from wipe-outs. A 3% threshold lets the Scottish and Welsh Greens in, which is a flashpoint with the SNP and Plaid.
• Northern Ireland is the most sensitive region to method and scenario, with seats at stake between unionist and nationalist parties. It needs its own consultation.
• Behaviour would change. Tactical voting is widely reported to have mattered in 2024, and this system would remove much of the reason for it. That could mean more votes for smaller parties than modelled here. We can’t measure that from the 2024 data, so treat the seat totals shown as an illustration, not a prediction.
• Adoption is a separate question. The 2011 AV referendum failed with far less change on offer. This page addresses suitability only.

Data, code and checking our work

Everything is published so you can reproduce it. All vote counts come from the House of Commons Library’s 2024 general election results (corrected edition, changes logged to 5 January 2026).