Prof Curtice, Senior Research Fellow at NatCen Social Research’s exit poll was the only one which correctly predicted the 2015 general election results. All the others did not.
So he has been analysing recent opinion polls. His conclusion is that you need to ensure that the data collection is right before you even think about analysing it. He has looked closely at the results from the 2015 British Social Attitudes Survey. This records which way people said they were going to vote.
He found that if the polls had been based on the people who answered the door on a first visit then Labour would be six points ahead. But if the polls had been based on those who answered the door only after three approaches then the Tories would have an 11 point lead.
So at his recent briefing in London his great revelation was that, “The people you can easily get hold of are not representative.” This highlighted the importance of ensuring that interviewing quota is properly met.
This is of course music to our ears. We have long believed that traditional face to face interviewing ensures that we actually survey the ‘hard to reach’ participants. We are also sure that questions from properly trained interviewer will improve the accuracy of answers. And in turn, that will mean more accurate analysis.
So we fully endorse Prof. Curtice’s call to spend more time over surveys so that hard-to-reach people can be contacted. This will help to iron out the potential bias. The demand for ‘quick’ results has led to an over reliance on panels and phone methodology. The end result has been unreliable data. Interim results are ‘extrapolated’ (a politically correct term for ‘guessing’) to produce the results.
And, and of course, it’s worse to have inaccurate data than no data at all. Much better to take the time and trouble to get accurate results using tried and tested methods – however uncool.