This article was made possible by my amazing patrons! Special thanks in particular go to the following comrades: Alex Wilson, Bert Rothkugel, Chloe Hopkins, Chris Gravell, Daniel Rayner, Konrad, Michael Murray, Mick McAteer, Neil Barton, Patrick Lefevre, Pawel Kaczmarski, Ruth Holliday, Stephen Kaar, Thomas Alexander, Uditha Desilva
To support Stats for Lefties and make articles like this possible, sign up at http://www.patreon.com/leftiestats for as little as £2 a month.
With the publication of the revised constituency boundary proposals for all nations and regions of the UK, it seems like a perfect time to update my seat projection model for UK Parliament elections – both to incorporate these new boundaries, and to outline some other changes I’ve made. Whilst the new boundaries have not been formally approved by MPs, they are unlikely to be rejected.
In this piece I’ll go through the effects of the new boundaries on the parties’ hypothetical 2019 numbers, as well as updates to how I treat Reform UK and the SNP in my model. There are some other small changes too.
Before I go on, I’d like to strongly thank Owen Winter for his expert projections of notional 2019 results under new boundaries, and the person who runs Election Maps UK for sharing map files for the new boundaries. I’d also like to credit Ben Walker of Britain Elects for suggesting that Brexit Party votes should be redistributed to second preferences in any 2024 projection model, as Reform UK is effectively a different party. It was a very good idea, so I’ve followed his lead.
The boundary review
The process of adopting new constituency boundaries has been a turbulent one. The current boundaries were adopted in 2007, and are based on population data gathered in 2005 – meaning that by the time of the next election in 2024, the boundaries would be 19 years out of date. Two boundary reviews have been attempted since 2010, but both have failed: the first because the Lib Dems blocked it, and the second because it was abandoned by the Conservative Party.
Both of these previous reviews had a mandate from the government to reduce the number of seats to 600 (a decline of 50 seats), but following Brexit, Boris Johnson dropped this requirement and legislated for a new review that would keep the number of constituencies at 650. The boundaries have been drawn by four independent Commissions (one each for England, Wales, Scotland and Northern Ireland), and need to be approved by MPs. They almost certainly will be.
The new boundaries
Overall, this new boundary review lacks the dramatic changes of its two failed predecessors. The biggest changes are in Wales (-8 seats) and South East England (+6 seats). Outside of these two areas, no more than 3 seats will be lost or gained. One or two seats would change hands compared to 2019: for instance, North Durham and Kensington are both now notionally Labour.
In terms of the political parties overall, Owen Winter estimates that under the new boundaries, the Conservatives would have won a majority of 92 seats in 2019 (up from 80), gaining 6 seats. Here’s an interactive map showing what the result would have been.
Amongst the major parties, Plaid Cymru would be hardest hit (losing 50% of their seats) with the Lib Dems losing 3 of their 11 seats (27%). Labour would only lose two and the SNP would stay the same.
Now, let’s talk about what this means for my seat projection model.
My existing model
Up until a few weeks ago, my seat projection model was simple:
- I looked at the 2019 results in each constituency;
- Then I looked at the change in national vote share between 2019 and now;
- Then I applied that change in vote share to each individual constituency.
This is known as ‘uniform national swing’ (UNS). It’s an imperfect method but I’ve found it to be useful, aside from remarkable situations like the sudden SNP surge after September 2014 (when one had to look at Scottish-only polls by necessity). But given recent huge changes in polls, and the boundary changes, I decided it was time for an update.
My new model
I will still be using UNS predominantly in my model, but with the following changes:
- Using the new boundaries as a baseline for the 2019 result;
- In Scotland, using a mix of Scottish polls (weighted at 80%) and UNS (20%);
- Redistributing 2019 Brexit Party vote to voters’ second preferences;
- In seats where Change UK / independents did well, adjusting for their absence in 2024.
In short: it incorporates the new boundaries, takes account of Scottish polls, and brings in adjustments that take into account the practical reality on the ground. Here’s an interactive map of what the notional 2019 result looks like, taking into account the new boundaries and my other changes. As indicated on the map, the Tories would notionally gain 7 seats as a result of my adjustments – all seats where the Brexit Party performed well in 2019, and where the Tories benefit from their absence.
Winters estimated that the Tories’ notional 2019 majority is now 92 seats; my adjustments increase this slightly to 108 seats.
So, what does this mean for my numbers going forward? Let’s look at the current polling average for November 2022. The graph below shows a comparison of my projected seat outcome under the old model and the old boundaries (UNS) and the new model (which I will refer to as UNS-Plus).
As you can see, there are essentially no differences – in large part because the slight Conservative advantage from new boundaries and BXP votes is cancelled out by an enormous Labour surge. But if we look at a neutral national environment (i.e. one where the Tories and Labour are tied) the changes become far more obvious, with the Tories gaining 12 seats relative to the old model. This is the result of both notional seat gains and the incorporation of Brexit Party second preferences.
In short, you are unlikely to notice any great difference between my old seat projections and the new ones – because any Conservative gains as a result of methodology changes are being cancelled out by an enormous Labour poll surge. But if the polls start to tighten, the mismatch between popular vote and seats is likely to become ever-more noticeable, with the Conservatives pulling ahead in seats before they pull ahead in vote share.
For now, though, this is just a bit of methodological tidying-up. I hope it was still interesting!