Did Northern Ireland opinion polls turn out to be true?
That's all the NI results in, and it's only proper that we compare the actual results with our last Opinion Panel, published in the Belfast Telegraph last Saturday. First of all, our Opinion Panel was only meant to give an approximate view of the trends as it was based on a small NI sample and couldn't therefore be used to predict e.g. vote sizes, or shares.
Well how did our panel do? Even though they weren't meant to make full predictions, they/we got 16 out of 18 correct.
However more importantly we got the trends correct in 17 out of 18 - which after all was the main purpose of the panel, and what the panel was meant to do on a monthly basis.
In Upper Bann our panel noticed a late swing against the DUP with their probability dropping from 80% to 65% in our April panel poll, but it remained a DUP seat reflecting that final 65% score.
Over our March-April panels the DUP probability score in South Antrim dropped 10% and was trending downwards as we approached the election - ending up at 75%, mostly because the DUP was the sitting tenant holding the seat, and we quite rightly built-in incumbency into our models.
Yes, the UUP actually won South Antrim, but as we say we were tracking the trends so maybe that trend kept going right-up through election day, and ousted the DUP + NB there is delay of about 3 days between polling and the results publication - and then a delay again to the actual election.
Apart from South Antrim, the other seat that our panel got wrong (i.e. if you want to look at it from a seat prediction point of view) was Fermanagh and South Tyrone.
This was always going to be a knife-edge seat, and difficult to predict by any polling methods. Frankly the only way to probably predict this seat accurately would have been with a full constituency poll- similar to our January East Belfast constituency poll which predicted the final East Belfast result to within 1% (in terms of the margin between the DUP and Alliance).
All our other predictions were spot-on, even with seats that were being pushed by pundits as marginal e.g. South Belfast. There we recorded a 10% increase for the SDLP in our April panel poll to 75% and this momentum continued resulting in Alasdair McDonnell holding the seat.
North Belfast we always had as a 90%+ for the DUP even before the Unionist pact, and it went to 95%% after the pact, in a seat that Nigel Dodds was always going to win anyway despite again some pundits saying he was in danger.
The East Belfast result also reflected our Opinion panel with the DUP never going above 75% to win the seat, and the final result reflected this 75% probability score pretty accurately i.e. a 6% vote share victory for the DUP over Alliance. As we say above, our January full constituency poll in this constituency predicted this same margin of victory for the DUP over Alliance (i.e. 6%).
So, all in all, we're fairly happy with our Opinion Panel results, with only one incorrect in terms of both final % probability score, and trending i.e. Fermanagh and South Tyrone. Perhaps this type of close marginal constituency requires a full constituency poll to accurately gauge opinion.
In terms of the national scene I'm not here to defend the big national polling companies - we are a NI based company who only do political polling in Ireland.
However it's important to notice that the national vote share polls from the 'big name' UK polling companies e.g. YouGov, Comres, ICM etc.. were approximately correct in the final 7-10 days building up to the election. But the mistake these polling companies and the media made was trying to project these national vote shares into seat No's - that's got a huge error, maybe +/- 80 seats either way. But seat projections are a lot more interesting to present than percentages.
Incidentally as I said in my election guide on DebateNI yesterday, exit polls are usually very accurate, and this one followed that trend. And one final point for all those bashing the polls - remember the Exit poll is actually carried out by the polling companies as well, + they analyse the results, and then present these results to the network TV companies. So if you're criticising the Polls you're also criticising the Exit poll as well!
Bill White is Managing Director of Belfast polling and market research company LucidTalk, polling partners to the Belfast Telegraph. You can follow LucidTalk on Twitter: @LucidTalk.
Belfast Telegraph Digital