Christ. Where to begin.
A couple of weeks ago I had a look at the Dublin Bay South by-election here, and concluded the following:
“This, until someone polls the constituency directly and demonstrates a huge shift, remains FG’s to lose. Their generic “cushion” of nearly 30 points will likely not manifest at anywhere near that level on polling day, but overturning that is a huge task”
Well, now someone (IPSOS) has directly polled the constituency and yep, they’ve demonstrated a huge shift – to Labour. If this poll is anywhere near accurate (and there’s possibilities that it may not be, which I will discuss), Ivana Bacik is poised to run Fine Gael’s James Geoghegan close, if not beat him.
Firstly, let’s chart out what the poll shows:
That’s… something alright. On these numbers it’ll come down to transfers, and given the escalating series of blunders that have led to Geoghegan imploding in one of the easiest constituencies in Ireland for Fine Gael, you wouldn’t bet against Bacik closing that gap. What’s even more interesting is that the poll also asked for generic party support in DBS and, yes , this can absolutely be pinned on Fine Gael and their candidate providing a beautiful example of how not to run an election campaign. See below the comparison between candidate support and generic party support:
As you can see here, the results match fairly closely except, obviously, for Geoghegan and Bacik. Geoghegan is down by 10% on his party, Bacik is up by 12% on hers; its almost a mathematical certainty that what we are seeing here is by-and-large Fine Gaelers unhappy with their own candidate going to what they see as the next best match for them – Labour.
Now, this might seem a bit odd – surely Fianna Fáil, or even the Greens, would be better matches? There’s a couple of factors to consider in this. Firstly, this constituency isn’t really one where FF have been strong historically, and their candidate has massive issues of her own, as I discussed in my initial article on the by-election. The FG vote in DBS tends to be of the “socially liberal, fiscally conservative” variety – as evidenced by the huge support for marriage equality and repeal in the constituency. The last one is of particular significance, given that Geoghegan’s role in founding Renua have raised questions of his views on reproductive rights, and given Bacik’s long history of campaigning on the issue.
Finally, it’s also worth noting that there seems to be a cohort of FG voters with particularly guilty consciences who are willing to lend their votes to centre-left parties. The split between Bacik and Geoghegan mirrors exactly the split between the Green Party’s Eamon Ryan and the two FG candidates in 2020. If you consider Lucinda Creighton’s voters in 2016 as another split off from FG’s core vote – this time of a more conservative bloc – the numbers add up perfectly in that cycle as well; FG with a broad “base” in the region of 35-40%, with around 10% of those willing to lend their votes left.
Given that the generic party polling indicates that these voters have come back to FG – but not to Geoghegan – I would speculate that the exact same cohort that lent votes to Ryan in 2020 is doing the same in this by-election, but for Bacik. I would thus caution that Labour should not put too much stock of this as a sign of some kind of sea change, as some of their more enthusiastic members are already doing on social media.
What’s also, in theory, really nice about this poll is that it includes 2nd and 3rd preferences (albeit only for some candidates, because this wouldn’t be an Irish Times poll if they didn’t insist on restricting some of their data to the only five parties they are interested in) as well as the FPV figures. This pans out like this:
Now, understandably, Green Party supporters on social media are getting very, very excited about this – those 2nd and 3rd preferences look excellent for Claire Byrne, and they are. However, unfortunately, all of this data is completely useless because it doesn’t tell us who the transfers are coming from.
If Byrne is getting those transfers from candidates that will be eliminated before her, it would put her in the mix for sure. However, if they are primarily the next preferences of Geoghegan and Bacik supporters – and given general transfer patterns, they almost definitely are – they aren’t going to help her at all. Similarly these numbers certainly look like Bacik would overhaul Geoghegan, but equally if a pile of her next preferences are Geoghegan FPVs, they won’t help her.
Without knowing for certain, this data has little-to-no practical application, though the low number for Sinn Féin adds salt on what must already have been a disappointing poll for them. As I discussed in the prior article, they needed a number of things to break heavily their way to close the gap here, and other than Geoghegan significantly under-running his party’s support, they don’t appear to have manifested yet.
Possible problems with the poll
So is that it? Game over and a two-horse race? Maybe. As I mentioned at the start, there’s possible problems with this poll. That doesn’t mean to say that they definitely exist, but there are some issues that go beyond the general challenges polls face.
Firstly, there are a number of issues that stem from the sample size of 500, the most obvious of these is a 4.4% MoE. You can smooth out these issues by using a large number of polls to effectively increase the data set – this is what the RPA I use for modelling does – but it’s a lot more challenging when it’s a single poll. A 4.4% MoE here means that theoretically you could be looking an eight-point swing either way between Geoghegan and Bacik. That is a large amount – the difference between Geoghegan being 14 points ahead of Bacik (where he would win comfortably) or 4 points behind (where he would more than likely lose).
That might sound like it’s rationalisation, but it’s not – in the run up to GE 2020, TG4 did 500-sample individual polls of Kerry, Galway West and Donegal (IPSOS was also the pollster for these). Some candidates they absolutely got right, but others they got very wrong – some of these were off by as much as 8%. That’s not to say that it is going to be the case here, and even if it is it will not be for all candidates, but it’s worth bearing in mind that this sample size can be very challenging.
Part of this is also geography – which was likely a factor in the aforementioned TG4 polls. I am open to correction on this, but I don’t believe pollsters weight for this factor. DBS is, as discussed in the prior post, an extremely divided constituency geographically, and there may be an under-sampling of votes in some parts of the constituency. Unfortunately for left-wing parties, I don’t believe that’s likely in South East Inner City – Mannix Flynn on 5% would be very hard to reconcile with an inner-city under-sample, although its certainly not impossible. An under-sample outside of SEIC would theoretically favour FG and FF the most, and Labour and the Greens to a lesser extent.
Finally, this poll was conducted face-to-face, which, given how many apartment blocks are gated, may have resulted in less of DBS’ substantial population of apartment dwellers and renters being surveyed than they account for of the voting population. This is likely overall a minor factor, but it still is one that could come into play.
Overall however, while these problems and challenges are all real – though the extent to which they will end up impacting the accuracy of the poll remains to be seen – it would be a genuinely staggering error if we don’t see Geoghegan and Bacik finish in the top two spots on polling day.
A quick note on the model
Forgive me for a brief digression into a little bit of self-indulgence here, but I want to make a quick note on what this implies for the modelling I’ve been using. I’ve been open that it would have limited utility for a single-seat by-election, so I’ve compared the calculations I had made based on RPA for DBS to the party support numbers for the poll, and, well:
I’m actually fairly happy with the overall accuracy of this. The biggest gaps are Labour and FG, which I think is still to some degree attributable to the candidates, and everyone else is within 2%. So there’s certainly more work to be done refining things, but I’m happy with how close they are overall from the first few months of work.
Maybe next time, Fine Gael shouldn’t bully Kate O’Connell out of running.