Q: How are the stats generated?
A: A full explanation of the methods behind the statistical model can be found here. This page will update with any changes in methodology. In short, the model runs 1,000 simulations of Irish elections based on current polling, past election results and candidate data. The results are averaged to produce probability figures.
Q: How accurate are the figures produced?
A: This is first general election where a statistical model of such complexity has been used. We will only know its accuracy following the results on February 27th.
Q: Is the model purely statistical, or are other factors at play?
A: In explaining the model, I note how “candidate weights” require an element of subjective political judgement. The balance of candidates on party tickets will change between elections. Some candidates posses strong local/national popularity which has not yet been captured in previous elections results. Adjustments of this nature are made in the model, but do not impact the overall strength of parties in a given constituency.
Q: Are candidates listed in order of votes?
No. I discuss interpretation of results on the constituency page where I note the difference between probabilities and votes. The probability figure encompasses votes, transfers and then sums the number of times a candidate was elected across 1,000 simulations.
If a candidate is listed in the top three in a three seat constituency, it does not mean they are likely to be elected. If two candidates of the same party are in fourth and fifth place, and the sum of their probabilities is higher, then that party is more likely to be ahead. We are just unsure which of the two candidates will take the seat.
Q: Why do some seat figures appear to not add up?
A: Rounding can act as a distortion. In the parties for example, we round the seat number to obtain a rough projection for the party in each constituency, however you will find that the sum of these rounded figures does not equal the seat number forecast on the main page. This is because the main page is a sum of probabilities. It adds each candidates score together, regardless if they were elected or not. If five candidates register 40% probability, the sum of these five gives 200%, or two seats. So while normally we may deem each candidate unlikely to win a seat (0), the sum of probabilities gives us 2. Bear this in mind for all data throughout the site. This concept emerges regularly.
Q: Why do the likely seats fluctuate without large changes in support?
The Irish PR-STV electoral system is one of the most dynamic in the world. Likely seats for a party can rise or fall just by changes in support for the other parties.
The list of candidates for the election has changed over time. The addition of new candidates to the model over the past number of weeks has the potential to alter the model. For example, Independent TD Tom Fleming made a late decision to withdraw from the campaign (thus making the election of Michael Healy Rae more certain), and the departure of Sean Conlon TD from Fine Gael has altered the state of play in Cavan Monaghan.
There is also a tendency for parties to have certain thresholds of support, beyond which they earn a number of extra seats, before not earning any more seats until another higher point of support is reached. When Labour polled c. 6% in late 2015, the model consistently forecast low single digit seats. In early 2016 the party has risen marginally to c. 8% but now commands over 10 seats in most model simulations. A hypothetical scenario of 11% support returned 18 seats for the party.
Questions are not frequently asked until they are asked frequently. Please leave a comment below if you need to know more about the model.