Yonder Seat Projection Model explained

Rick Nye and Andrew Cooper


If the 2024 General Election campaign is remembered at all, it will be as much for the sheer variety of voting intention estimates and the seat projections they have spawned as for any policy announcements or political gaffes. Just because all these estimates point in one direction – a Labour landslide – this should not obscure the fact that even if that landslide comes to pass many of the predictions made over the past 6-7 weeks will have been found wanting.   

Just looking at the polls published in the final few days of the campaign, Labour’s share could be as low as 35% or as high as 42%, the Conservative share anywhere between 18% and 25%, Reform between 13% and 21% (and Reform might be 3% ahead of the Conservatives or as much as 11% behind).  Labour’s lead according to these polls is somewhere between 15% and 22%.  The polls all agree that Labour will win, but they differ significantly on the critical detail of the scale of their victory.

The same is true of the many Multivariate Regression and Post Stratification (MRP) polls that have punctuated the 2024 election and defined a great deal of the media’s narrative on the campaign.  These polls, which extrapolate from very large national poll samples into very precise predictions for every single constituency, all indicate a large Labour majority, with the Conservatives losing many seats in their previous heartland.  But they disagree significantly on the scale of the likely Labour majority and on how the rest of the House of Commons will be divided.  The four MRP polls published in the last three days have a spread of 140 in terms of the size of the expected Labour majority; there is a difference of 75 seats between the lowest and highest projected number of Conservative MPs. Two of them say that the Liberal Democrats will win more MPs, to become the Official Opposition to the new government, while the other two say the Conservatives occupy this position. 

A post-mortem on political polling methods after the dust has settled on the Election seems inevitable. What follows is Yonder’s contribution to that process.

The spread throughout the Election in terms of Labour’s lead in the polls and the findings based on those leads have led us to seek alternative ways both of deriving (rather than asking directly) whether people are going to vote and for whom, and also of taking those responses and building up a constituency-by-constituency analysis on the basis of the answers given.

The decision about which party someone decides to vote for is the consequence of where they come down on attitudinal issues that define and divide parties in an election campaign.   Based on qualitative research and other analysis, Yonder has identified a set of binary questions, capturing the range of attitudinal issues that cause someone to conclude that they will vote for one party than another.  The Yonder 2024 election model goes to the stage before the voting intention question and is based on how voters answer these attitudinal questions The first step was to group constituencies into six crucial ‘battlegrounds’, according to those that shared geodemographic features (identified in our Clockface model) as well as that would play a crucial role in the election (different kinds of swing seats).

The Beaches & the Beeches (Battleground 1)

Coastal England and rural areas in Southern England and Midlands with low degrees of demographic diversity in which Reform will have the greatest impact either by challenging the Conservatives directly or improving Labour’s chances of doing so.

e.g. Clacton, Wyre Forest, Great Yarmouth, Sherwood Forest

The Golden Chain (Battleground 2)

The wealthy circle of Liberal Democrat targets beyond the outer reaches of London mostly spanning Hertfordshire, Oxfordshire, and Surrey.

e.g. Henley & Thame, Godalming & Ash, Wokingham

The Red Rubble (Battleground 3)

The wall of previously Labour-held, less economically secure seats that Boris Johnson knocked down in 2019 across the North, the Midlands, and Wales and which voted to Leave the EU. This will form the backbone of any rebuilt Labour majority.

e.g. Bolsover, Clwyd North, Whitehaven & Workington

The West End, Towns & Suburbia (Battleground 4)

Towns across and around central England (including many new towns) which are demographically fairly typical, as well as central London and its suburbs, which Labour will need to win to secure a significant majority.

e.g. Aylesbury, Hitchin, Stevenage, Wycombe

The Rural Edge (Battleground 5)

Conservative-held communities and towns in and around the fringes of fields, farms, dales, moors, and fens. Somewhat less economically secure than the truest blue areas and vulnerable to Labour.

e.g.  High Peak, Ribble Valley, Monmouthshire, Stafford


Since 2015, the SNP has been dominant in Westminster seats north of the border but its supremacy is based on a large number of small majorities which are within striking distance for Labour as it consolidates the Unionist vote and attracts disillusioned nationalist working class waverers.


The second step identified that different binary questions, relating to different aspects of the election, have particular significance in determining voting depending on the differing characteristics of each of the battlegrounds. Different binaries are influential in different battlegrounds, but we believe battleground seats will behave similarly due to shared geodemographics and attitudes.

The table below shows the importance of the binaries in each battleground. Unlike in 2019, this election is a one-dimensional election around the question of who governs Britain, so the preferred Prime Minister binary is uniformly strong (although not overwhelming). There is though some difference across battlegrounds between how that fundamental question manifests itself. A feeling that both major parties are inadequate carries more weight in the “Beaches and Beeches” battleground where Reform is more pertinent, tactical voting has greater significance in the Liberal Democrat/Conservative “Golden Chain “battleground compared with other battlegrounds and the prospect of a Labour supermajority carrying very little weight in areas where Reform is likely to do well but is being taken into account in more traditionally Tory areas.

There is also the issue of how much of people’s likely voting behaviour is accounted for by their positions on these trade-offs and how much is down to other factors not captured by our model. This varies by battleground as shown below.

So how do these binaries get turned into seat win probabilities from which we derive our estimates of the overall result? This is the process in outline.

We ask the probability of voting for each party on a 0-10 scale. We then build models for every combination of two-way party contest using the binaries as the independent variables and the Clockface to estimate the balance of opinion on the binaries for every census output area. For example, we will have estimates of the percentage of people in each Census output area (small geographic areas with a couple of hundred residents on average) who say prefer Rishi Sunak or Keir Starmer as Prime Minister.

From this we can estimate the probability that someone in each output area will be more likely to support a party in their constituency which we aggregate to arrive at a final constituency-level win chance for each party.

If voting decisions were entirely explained by their response to these binary questions, the model would allocate a constituency to a party on the basis that the model gives it the highest probability of winning.  In practice, as we’ve shown above while the attitudinal drivers that the binary questions expose are significant everywhere, there are many places where there are other factors which come into play (local candidates, residual party loyalty, tactical voting) which mean a seat wouldn’t go to a party simply on the basis of the model giving it the highest probability of winning especially if the second highest probability is close behind.  

In the circumstances of this election, when the defining feature is the biggest ever swing away from an incumbent party, these other factors playing alongside the attitudinal binaries, overwhelmingly point in Labour’s direction.  After the election we will be able to assess how significant underlying attitudes were in defining how people voted – relative to other factors.  

To give an indication of what the outturn might be, we show below how seats may fall to each party across the different battlegrounds we’ve defined – and overall – if we assume that the number of seats won by each party is determined solely by giving each seat to the party having the highest win chance. We have then given an outturn based on the Conservatives having both the highest win chance and double the win chance of the next closest party to offer an estimate if all the non-modelled drivers are working against the Tories.

From these estimates the Red Rubble has already almost entirely gone as has a good chunk of Rural Gateway which together with a good showing in Scotland gives Labour a comfortable working majority to start with. The difference between a two figure Labour majority and a majority of around 200 or more lies largely in the number of further losses Labour are able to inflict on the Tories with the help of Reform in Beaches & Beeches. And the Tories could end up with fewer than 150 seats if they lose more seats in this battleground and yield further to the Liberal Democrats in the Gold Chain.

We are confident that the great majority of seats will fall in accordance with these attitudinal drivers captured in the binary questions.  After the election, when we can see how many fall outside this range and review where this variance occurred, this unique new approach to understanding and mapping elections will be Yonder’s contribution to the debate that will follow about polling methods.

Appendix: Seats by battleground

The Beaches & the Beeches (Battleground 1)

The Golden Chain (Battleground 2)

The Red Rubble (Battleground 3)

The West End, Towns & Suburbia (Battleground 4)

The Rural Edge (Battleground 5)