Dublin Central Bye-Election
A first test of the system
This is the first ever test of our transfer system, we recently tested our core first preference model in action in Scotland & Wales with pretty good results. We’ll work on refining both in time for the next NI Assembly election and Irish General Election.
Our Projection
Our pre-election projection had Ennis, Social Democrat winning with with 39.7% to Boylan’s (Sinn Féin) 31.1%. We expected the Greens to finish third ahead of Hutch.
Ennis started out with a small lead but was able to extend this by collecting more transfers than Boylan from both left-wing and centre-right parties.
Results
Real Life
Ennis (Social Democrats) defeated Sinn Féin’s Boylan by 48.5% to 31.3% after all votes were counted. The Green’s finished in third ahead of Hutch.
Ennis started out with a small lead but was able to extend this by collecting more transfers than Boylan from both left-wing and centre-right parties.
Results
Differences
The biggest differences between real life and our system was the number of votes transferred in general. We overestimated non-transferables consistently, we’ll work on improving this. It’s a fairly simple adjustment to the system.
We also underestimated Fine Gael to Green transfers, the legacy of the last coalition likely plays a role in this. We can adapt to this. A huge part of why we ran the test was to learn for the real deal.
We also counted for one extra round, this was not an error. To speed up the real life count, the proceedure is to simultaneoulsy eliminate all candidates who’s combined vote share can’t catch up with the next lowest candidate, or hit the 5% threshold for holding their deposit. This year the tiny independents were eliminated in one batch. We have no need to speed up the count so we just do every elimination seperately and its an unnecessary extra calculation which would actually slow down our Monte Carlo system.
Reading the charts
Use the play button on each chart to step through the rounds, or use the round selector to jump directly to a specific count. The bar shading shows which candidates are still in contention; eliminated candidates drop to zero before their preferences flow to the remaining candidates.
These are illustrative scenarios, not predictions — first-preference shares and transfer patterns will both depend on who actually stands and how the campaign plays out.