The 2024 UK General Election was fought on 650 constituencies, each a battleground representing 75,000 (or thereabouts) voters. Some are tiny, some are as vast as minor medieval kingdoms. Of course, each boast their unique issues.
We know who the constituencies elected. But how did they vote? How did the neighbourhoods vote? Where is this or that party’s strength of support?
Wards – the areas for which local councillors represent a smattering of neighbourhoods – can cover as little as a few thousand voters, and sometimes as many as ten thousand or more. Campaign teams use them as the building blocks for raising support and winning the constituencies outright, and they are the closest we are going to get to a granular look at the 2024 general election results.
But breaking down how Britain voted by ward is a calculation game. What proportion of a party’s vote across the constituency is likely situated within this ward? If the city centre of Chester, for instance, makes up one-fifth of the Chester North electorate, is it the same for the Green vote? Or the Reform vote? Or is the concentration in that ward greater – or less? Fundamentally, this is what I have set out to do.
Almost all councils don’t count constituencies by ward, or divulge when they do, so a degree of triangulation and modelling is required here. Which is what I’ve done here.
The assumptions put into this model as to a party’s vote are derived from a few areas. Demographics is a key driver, absolutely. Housing, age and education are decent indicators as to one’s voting intention. Income, too. But they only tell half the story. Parties overperform and underperform among some demographics. There can be geographical variation. Social housing voters in rural areas are no the same voters as those in urban areas. Historic associations, or even religious upbringing, can have your party over or under perform among certain stratas of society as well. Historically Catholic voters in northern England have demonstrated a stronger loyalty to the Labour Party than their incomes and education backgrounds would have you expect – to name one example.
So while this model employs demographic data, it also relies to a limited degree on historic local election results. We know, for instance, that George Galloway’s Workers Party did very well in central Rochdale in the May local elections – less so on the outskirts. We know, broadly, where his vote will likely be situated. And so that features in this model and presentation.
A very heavy health warning, however. Treat the figures mapped above and in the spreadsheet as estimates, not definitive summations as to who won and by how much. I’ve been privy to more than my fair sharing of samples from the count – from all parties (and thank you very much for them). I know in plenty of wards the actual result (or rather, the sample as taken by party activists) is more than one hundred or so votes off what I’ve been able to model here. In some wards the modelled result puts the wrong party ahead of what the sampling said. In some wards, too, turnout is overstated by a few points. But generally speaking, what I’ve been able to cross-reference shows me an average error for the top two parties of around four percentage points.
So as a literal example: when you look at a ward, I’d advise you to look at them with a margin for error. If it says the result was Labour 40 per cent, and Conservative 38 per cent, treat the probable result as around Con 34-42 per cent, and Lab 36-44 per cent. Yes, Labour probably won that ward. But it wasn’t a guaranteed win. There’s only so much a model can do.