I’ve just finished reading the entrails in Foyle, and my brain hurts. Quite frankly I wouldn’t blame you if you stopped reading right now and just left it to the Almighty in his ineffable wisdom to reveal the final results sometime on May 7th.
To those of you who refuse to heed the warnings, I salute your courage.
We start with the designation shares bearing the, by now familiar, marks of tactical voting in Westminster.
And as usual, these marks become even more obvious when we turn to the party performance chart.
The last Westminster figure is nothing short of spectacular for the SDLP. And the good news for the party is that a large proportion of that increase did not come from tactical voting at all.
It is possible to make a reasonable estimate of the level of tactical support given to the party by voters who otherwise vote unionist or other (including PBP). This is done by comparing the Westminster vote share for those designations with the vote share they received at the closest non-Westminster election. The results are shown in the following graph.
This shows the SDLP (in the solid black line) taking 57% of the total votes cast at the last Westminster election. SF are in dark green, and other nationalists in yellow.
The dotted black line shows the best estimate for the votes that the SDLP would have received had no unionists or other voters lent them their vote. Even relying on nationalist voters alone, the SDLP would have defeated SF heavily by about 44% to 21%.
That 44% would represent 2.6 quotas, which would net them 3 Safe seats, leaving 1 for SF and 1 for the DUP.
On the other hand my Central projection of the Lucid Talk poll suggests a pattern like that below.
This is where we must go into a deeper discussion of how these projections work. Remember, the Central projection is merely the average of two other projections. One is based on the difference between the parties’ shares at the 2019 Local Government election and the poll shares. This produces almost equal vote shares for SF and the SDLP.
The other is based on how 2017 voters have, or have not, changed their minds. The poll shows that SDLP voter retention is poor. Only 64% of their 2017 voters said that they would be voting for the party this time. 13% said they would be switching to SF.
On the other hand, SF voter retention is shown to be particularly high. 83% of their last-time voters will be sticking with them. And only 8% plan to go for the SDLP.
In most constituencies the greater part of SDLP losses is compensated for by gains they make from elsewhere. But in Foyle the combination of a) a high 2017 share (more for the SDLP to lose) and b) a high SF share (less for SDLP to gain) is brutal.
Again, remember that across all constituencies both projections add up to the party shares shown in the polling. This means that if there are factors which mean that SDLP voter retention will be better than average in Foyle, and there may be, then retention must be worse than average in some other constituencies to balance.
Anyway, the Central projection would mean that quotas per candidate might look something like this. Seats won in 2017 are highlighted in gold.
The split of SF vote between candidates is the same as 2017. For the SDLP the split is the same as in 2016 when they last stood three candidates.
Also standing are the IRSP who last fielded a candidate in the 2011 Local Government elections winning 879 votes. If they repeat that next month it would be worth just over 0.1 of a quota.
The candidate with the potential to upset all calculations is the Independent Anne McCloskey. She stood in 2016 on a pro-life, republican platform. Her 1st preference votes then would be worth half a quota today. During the count she accumulated significant transfers and finished runner up. This means her votes were never transferred making it more difficult to judge which candidates have more to worry about from her intervention. A comparison of vote share for 2016, when she stood, and 2017, when she did not, suggests that most of her vote came from SF. The similarity of her programme to Aontú’s suggests that they could be badly affected.
If we feed these assumptions into the Central projection, we see it has the potential to change to look like this:
A great deal would then depend on the fine details of transfers. While it seems unlikely that she would have a path to victory (although it should not be totally ruled out) there are two more likely scenarios.
One is that she is eliminated before PBP or Alliance, in which case her transfers would revert back to those parties remaining in the count. In effect the situation would return to the path of the Central projection, producing 2 Safe SDLP seats and 2 Safe SF.
The final seat would be between PBP, with a Good possibility, and the DUP, with a Lesser possibility. One of the alternative projections gives Alliance a Good possibility, whilst the other suggests it would be a Long Shot.
The second scenario for McCloskey is that PBP is eliminated before her, it is then possible that her transfers could decide whether the DUP or Alliance take the final seat, following the election of 2 SDLP and 2 SF.
If Alliance were eliminated instead of PBP, it is likely that their transfers would favour the SDLP more than PBP, resulting in 2 SDLP, 2 SF and 1 DUP leaving McCloskey as runner-up. Alliance would not have enough votes to bring through both the second SDLP and PBP.
Now go and make a cup of tea or coffee, and thank your lucky stars that you are not a candidate in Foyle (unless, of course you are, in which case the best of luck.)
NOTE ON METHOD:
There is no perfect way to translate a national poll to a local constituency level. Still less so in a PR system. We must recognise a level of uncertainty. I have used the last Lucid Talk poll as my base because, as well as giving party shares, it also tracks how the voters for each party at the last Assembly election intend to vote this time. This allowed me to make two projections for each constituency, one based on vote switching since 2017, the other on changes in party vote shares since the 2019 Local Government elections. (I used the LG elections due to widespread tactical voting at the later Westminster.)
The two different projections mimic two different patterns of changes in party support. In one, a party that is growing strongly will see a bit less of that growth where it is already strong, and a bit more where it was previously at it most weak. Conversely the parties that have lost voters will suffer a bit more in their strongest constituencies.
The other projection has the opposite effect. Either may prove to be a more accurate reflection of reality. And while the differences between the two are not massive (they both must total to the same poll shares across all constituencies) they can still sometimes produce different outcomes.
I should stress that while the Lucid Talk bears the responsibility for the original poll, the projections are entirely my responsibility.
For each 1st preference projection, I have used recent transfer patterns to identify all candidates who have any chance of winning on a scale ranging from Safe to Long Shot.
To avoid burdening you with all this detail I have averaged the results of the two projections and shown them as a Central projection, merely noting where one of the other projections produces a noteworthy difference.
Where a party is running more than one candidate, I have generally split the party vote between them in the same proportions at the last election. I have had to make my own assumptions when a party has a different number of candidates this time.
Michael Hehir is a retired sales and marketing manager. He studied in Northern Ireland but now lives between England and Italy.