Forecasting the 2017 Assembly Election

The snap Assembly election called for the 2nd of March 2017 will be the first to return 5 MLAs from each constituency, reducing the total number of MLAs from 108 to 90. This, combined with indications of a fall in support for the DUP and a rise in support for the Green Party and People Before Profit, makes this year’s election less predictable than is usually the case for Northern Ireland elections.

I have built a forecast model in an attempt to predict who will be returned to the next Assembly. To do this, I have taken polling data from Lucid Talk and compared parties’ current polling figures with what they received in 2016 to build a baseline for expected support this year (for example, DUP support appears to be down 14%, whilst support for the UUP is up 12%). I also looked at the transfer patterns from 2016, and used this to build a matrix for how the vote of various parties tends to be transferred under the Single Transferrable Vote system.

This data was used to develop a forecast model for each constituency, with normally distributed values for first preference votes and transfer distributions generated from the adjusted baseline and transfer data observed from 2016.

Multiple simulated counts were run in each constituency, with the surplus votes of the candidate ranked first distributed if they achieved over the quota required for election, or the votes of the candidate ranked last distributed if no candidate reached the quota on each count. Candidates are deemed elected if they reached the quota or were in the last five candidates remaining after other candidates were eliminated. Each constituency model was run 1,000 times, and the data generated from these runs creates a probability for each candidate that he or she is elected.

The forecast for the expected number of seats won by each party in each constituency are shown on the table below.

The model suggests that the DUP will remain the largest party, but that they may just fail to win the 30 seats required to bring a petition of concern without the support of another party. Elsewhere, the Alliance Party, the Greens and People Before Profit look well placed to defend all of their seats in the Assembly, whilst the four largest parties can expect to return fewer MLAs than they did in 2016.

It is extremely unlikely that Sinn Féin will eclipse the DUP to become the largest party in the Assembly. To illustrate why, the following table shows the probabilities for the DUP (in red) and Sinn Féin (in green) for their target seats from most likely to least likely to win. The DUP have a realistic shot at 38 seats and are favoured to win in 29, whilst Sinn Féin have a plausible chance in 30 seats and have a 50% chance or better in 24.

For example, if Sinn Féin were to win 28 seats, then they would have to run the table in all of the contests they were favoured to win, and then win all of the following: a third seat in Newry and Armagh, a seat in East Antrim, a third seat in Mid Ulster and a second in East Londonderry. Even in that case they would need to rely on the DUP losing out on any seats in Foyle or Newry & Armagh, or a third seat in Strangford, all of which they are favoured to win.

The challenges that the move to electing five MLAs per constituency poses to the UUP and the SDLP can be seen in the table below. There are a number of seats which will be difficult to defend for both parties, with few opportunities for pickups.

The Alliance Party face a more favourable environment defending their seats, with seven of their eight MLAs expected to defend their seat with a probability of 97% or higher, with even their most vulnerable seat (South Antrim) successfully defended 79% of the time.

Of course, recent events have shown that attempting to predict elections is generally a mug’s game, and complex elections such as Northern Ireland Assembly elections are especially hard to forecast. There could, for example, be dramatic changes in turnout amongst various sections of the electorate, or there could be a step change in transfer patterns.

However, the model does provide an indication of the daunting task facing the UUP and the SDLP if they are to once again assume their former roles as the largest parties of Unionism and Nationalism respectively.

The incoming Assembly, whilst smaller, is still likely to have the DUP as the largest party and Sinn Féin as the second party. As before, it will ultimately be down to the willingness of these parties alone to do business which will determine whether power sharing has a future.