Northern Ireland Assembly elections: How did the pre-election polls compare to the results
The election is over and we now have an election break until 2019 - although there is the EU referendum coming up on 23 June.
We were fairly pleased with our pre NI Assembly election polls, all the trends were correct in our last Opinion Panel poll (see previous articles). Yes - some of the figures were a bit out from the actual final figures, but remember our last poll was taken 10 days from the election, so there was still plenty of time for changes of minds.
In any case, we always stress that it's the trends that count - and our final poll showed the correct trends: DUP going up, SDLP & UUP declining, TUV rising (although the TUV had no increase in terms of seats, their NI-Wide vote went up), Green Party and People before Profit going up.
We were also pleased that we detected and predicted the drop in the combined SF/SDLP vote which seemed to raise the ire of some commentators who thought we should have been projecting this score to be over 40%. I think they've been brought back to a world of reality, as the Nationalist/Republican vote has been levelling off in recent elections at around 36-38%.
Interestingly, it was our party leader rating polls which we started in November 2015 that proved to be a very accurate indicator as to the election result we should have been expecting. Peter Robinson scored +22 in our first Leader ratings poll (November 2015), but the following month, with Arlene Foster as the new DUP leader that score jumped to +45 - it was like switching on a light. Mrs Foster continued to score top/near top in our leader ratings poll, right through to the election. Arlene Foster, Steven Agnew, Martin McGuinness were always consistently top in these polls, with David Ford, Colum Eastwood and Mike Nesbitt hitting medium scores on average - the signs were there.
Yes, we have an issue in that we are regularly overestimating the UUP in our party preference polls, but our polls never showed any substantial 'bounce' in terms of the claims that Mike Nesbitt was making. This pattern with our polls has occurred in previous elections e.g. the 2014 European election - although we got the DUP spot-on in that one (within 0.1%), we did again overestimate the UUP and this then had a knock-on impact on other Unionist poll scores e.g. in the Euros we underestimated Jim Allister (TUV). This time, in this NI Assembly election, this UUP overestimation had the knock-on effect of underestimating the DUP by 1-2 points.
However, all the scores were still in the bounds of the stated error (i.e. the disparity wasn't that high) and we say again that it's the trends that are the important thing, and all the up & down trends in all our polls since 2011 have been spot-on accurate. This UUP disparity may indicate that a sizable chunk of the potential UUP vote is soft, with this 'soft group' jumping about from and to the DUP, TUV, and even Alliance (particularly in Belfast). It's also worth noting that the normally rock-solid rural UUP vote is beginning to fray at the edges. All this should worry the UUP.
Paradoxically, although we are not exactly happy with this UUP error, we are relieved that it's a 'constant' error that follows the same trend, and thus we can adjust our sampling to correct this - it would be much more difficult if it was an intermittent error. Our Opinion Panel is divided into 'polling cells' of 7 - 12 members within each of five NI geographic areas, with each cell balanced to be reflective of the demography of the target area. Members of each cell are anonymous to each other, and to all other members of our Opinion Panel. We have 216 of these 'polling cells' currently set-up in NI, and for each Opinion Panel poll project we select a sub-set of these 'poll cells' for polling. We are adding additional cells all the time, and we have added additional cells to rectify this continual UUP poll over-estimation. These updates have been put in place for our current mid-May EU Referendum poll - now running.
On this last point, we also have to admit that our EU Referendum polling up-to-now may be underestimating the Unionist 'Leave' by 2 to 3 points because of this same error - remember the DUP are pro 'Leave' and the UUP pro 'Remain'. Among Unionists, our EU Referendum poll results are currently running at approximately 2 out of 3 for 'Leave'.
This may a good time to say that we do wish people would take polls in the sprit they are intended - they should only be taken as a guide as to opinion at a point in time and not necessarily an election prediction. People do change their minds. Admittedly, polls taken within 1 to 2 days of an election or referendum should approximately reflect the final election result - but none of our polling was within this period.
However polls can show patterns e.g. in terms of the EU referendum, according to our polling, it's the Unionists that make up 70-80% of the current 'Don't Knows/Not Sures' and it's the Unionists that make up the large majority of those who keep changing their minds on this issue. The SDLP/Sinn Fein/Alliance/Green voters are 95% solid in their view, and are showing very little change i.e. they're solidly for 'Remain'. These EU Referendum patterns show the sort of broad information that polls can provide.
So it's onwards and upwards now to the EU Referendum on June 23rd. The results from our currently running EU Referendum poll (17th - 19th May) will be published next week.
Bill White is Managing Director, LucidTalk Polling and Market Research.