On the eve of the election, the polls were still showing an average lead of around 6 per cent for Conservatives over Labour, which would have led to a small Conservative majority.
As the clock struck ten and the results of the large exit poll were announced, the prediction from that surprised many with the prediction of a hung parliament. As the night progressed, and names from across all parties – from Ben Gummer author of the Tory manifesto, to Nick Clegg and Alex Salmond all found that they would soon have more time to spend with their families , one thing became clear. The final results were in line with the exit poll rather than the opinion polls.
So why the difference?
Well one key factor is that opinion polls are asking people to speculate about what they might do, whereas exit polls are asking a factual question about what they have done. The former is hypothetical and the latter is factual.
The cautionary note for all research studies, beyond political polling, is that responses about stated future behaviour need to be viewed with caution and calibration may be needed to use them as the basis for forecasts.
The statistical modelling involved is something which we have expertise in applying across many research situations, including predicting new product sales, projecting customer recruitment and attrition, and estimating the impact of marketing activities.
For more details of our data modelling services please contact Bob Peters on 01226 767120.