Originally from Fermanagh, Eoghan Kelly is a postdoc on University of Edinburgh’s Scottish Election Study and has lectured in Irish politics at Queen’s. Here he introduces a new election projection site for Northern Ireland..
Tools like Electoral Calculus allow anyone with an interest in Westminster elections to convert national polls into seat projections. But no such tool existed for Northern Ireland Assembly elections. The subtleties of STV and a general lack of interest from the GB political class meant we were left guessing: what does a bump in a party’s support in the latest LucidTalk poll actually mean in terms of seats?
There are no projections in our newspapers, none are available online, and as far as we can tell, the parties don’t have one. So, as a newly minted Doctor of Elections with a software engineer wife we decided to build one ourselves. And while we were at it, we created projections for Scotland, Wales, and the London Assembly.
The goal is to lift the lid on our elections and demystify STV. The projection is constituency-based, shows transfer flows, and is fully customisable so you’re not stuck with what we think will happen.
How to use it
We’ve added a few presets to help get you started. These include:
● Previous election results
● The minimum vote share needed for a unionist or nationalist majority
● The threshold for Alliance to finish first
They are examples, we are not saying they’re likely to happen.
Presets
There’s also a Nowcast which is a weighted average of recent polls so should provide a more realistic snapshot of current support. Right now it has Sinn Féin out in front with the TUV making major gains at the expense of the DUP.
The Current Nowcast
You can input your own numbers, experiment with scenarios, and share your results on social media. The system is meant to be quite accurate so you can see how many votes different parties need to start winning certain seats, eg. the TUV were about 1% away from picking up a few seats in 2022 and East Derry is Alliance’s best hope west of the Bann.
If you push a party’s vote share high enough the system will automatically add a bonus candidate, limited to one per constituency. Once nominations close we’ll fix the candidate numbers to match reality.
How it’s done
We started with the 2022 results, adjusted for boundary changes. We made a few editorial calls regarding independents and smaller parties, and smoothed out extreme intra-party vote splits (on the assumption parties learn from their mistakes).
The algorithm then calculates swing for each party based on user input or presets. This is calculated across constituencies, rebalanced so it adds to 100%, alongside a few other tweaks (I’m not going to give the whole game away). This creates projected first preference vote shares for each candidate in each constituency.
First preferences are the main thing in STV because at least 4 of the top 5 candidates after 1st preferences usually win seats, so getting them close is key. After that we do transfers.
Nowcast Vote Transfers in FST
Just like in real life, most transfers go to ideologically similar parties: Sinn Féin to SDLP, TUV to DUP, etc. But there are always voters with… eclectic preferences. We use historical data to distribute transfers, including non-transferable votes to reflect reality.
It includes a map with vote shares of the top five parties and a full seat breakdown.
Full results under current Nowcast
Lessons for a potential border poll
Though this is an Assembly projection, we know most politics here funnels into this question eventually.
Our Nationalist Majority preset gives 46 seats to nationalist parties by giving them about half of the Alliance vote. This would be enough to demand a border poll but doesn’t mean they’d win it. To reach 50% of the vote, you’d need to shift about 75% of Alliance voters and all the Greens to the nationalist cause.
You can also try to see what happens if the much discussed single unionist party happened, although that can fall afoul of the bit of the methodology that only permits one bonus candidate per constituency. The long and short of it is, it’s strategically sensible under Westminster’s First Past the Post system but a bit unnecessary under STV.
Limitations
We’re more confident about the first couple of seats in each constituency but the final seats are hard to predict. Tiny vote differences, transfer quirks, and turnout variations make them tough to model.
Ultimately, we had to decide if it was better to make a projection that would be pretty accurate over a wide range of outcomes or extremely good within a very narrow window. We went with the first option.
So don’t bet based on our projections.
What next?
We’re going to keep adding features, like a tool showing how our coalition is formed, and we’re going to keep refining the methodology.
This is a guest slot to give a platform for new writers either as a one off, or a prelude to becoming part of the regular Slugger team.
Discover more from Slugger O'Toole
Subscribe to get the latest posts sent to your email.
