Bill White: Polling predictions - blame it on the weather
After the success of the LucidTalk polling with the NI Assembly election, we carried out three Northern Ireland-wide Westminster ‘Tracker’ polls in April, May, and June.
These results fed into our seat predictor model which turned out to be fairly accurate when compared to the final results, particularly the exit poll.
We also carried out a NI-wide party vote strength polling in the first two of our Westminster pre-election polls, and again all the trends were spotted. For example the DUP predicted to poll at least 4% above their 2015 NI Westminster election performance.
However, although we spotted the Sinn Fein surge in the March NI Assembly election, our polling failed to predict the extent of the DUP surge in the Westminster election.
Yes, we got the DUP upward trend correct, and as we always say trends are never wrong in polling – but we didn’t predict the extent of the DUP surge.
Without making excuses (well too many!), it should be noted that it’s difficult to do broad NI-wide polling as part of a Westminster election due to (a) pacts in seats, (b) parties voluntarily withdrawing in particular seats, (c) certain unionist parties not running in Westminster elections even though they ran in the NI Assembly election like UKIP and the PUP, and (d) the dynamic of the First-Past-the-Post system which drives many voters to go for one of the 'big names' who are perceived to have a realistic chance of winning the seat, rather than the party they would naturally support. However, as with all our systems we’re constantly trying to improve, so we’ll review our poll methodologies and endeavour to correct this for future pre-election projects.
So what about the national UK polls?
Well, if you average all the polls they predicted the Conservative vote share within 0.3% - a brilliant performance, but underestimated Labour by 4-6% on average (not so good).
But how should we judge this, and how should we judge polls overall?
Well we always say polls should be treated like the weather forecast – look for the trends and the broad picture. We hear comments like ‘the polls are wrong’.
But in a sense polls are never wrong i.e. never absolutely wrong. However, they are never absolutely right either. Like the weather forecast they are always somewhere in the middle, never completely right, but never completely wrong.
What the polls (and weather forecasts) do is show the broad picture and the trends. In this context, the UK-wide polls all showed a tightening of the race between the Conservatives and Labour, and with our two ‘Golden rule’ questions i.e. (a) Best party for the economy, and (b) Best leader for PM. Yes, the Conservatives and Theresa May were always ahead of Labour (and Corbyn), but they never scored over 50% with either question. So the signs were there, and in terms of the weather forecast comparison, the polls showed that storm clouds were gathering for the Conservatives and Theresa May.
So in terms of interpreting poll results, as we’ve always said, don’t just read the headline poll results - read the commentaries, analyses, and reviews of the poll results as well. Plus, listen for, and look out for, the signs and trends as to what way the contest is shaping up. If you follow this approach then you’ll find that, like the weather forecast, polls are usually 95% correct.
Bill White, is Managing Director of Belfast based LucidTalk Polling and Market Research. You can follow LucidTalk on Twitter
Belfast Telegraph Digital