On the eve of the Election on May 7th we published a report of our research with predictions for the overall result and for individual constituencies. We now have the benefit of the Election result to evaluate these predictions.
The overall outcome
We have forecast a Conservative victory (David Cameron as Prime Minister) in each of our weekly surveys since October 2014, so the outcome was no surprise at all.
The experimental constituency forecast
Our experimental constituency level forecast (made only for England) is notable for:
- our prediction of the overall turnout in England (65.5% predicted, actual 65.9%)
- our prediction of the wipe-out of the Liberal Democrats in England. As it turns out, only 6 Lib Dem MPs remain in England, from 43 in the previous Parliament.
- our prediction that UKIP would win no seats at all: in fact they won one, Clacton that, they previously held, with a reduced majority of 3,437.
- our prediction of 8.18m total Labour votes in England; the actual total was 8.09m.
- our substantial underestimate of Conservative vote share and corresponding overestimate of UKIP vote share. This created a major discrepancy in our estimate of the distribution of seats between Conservative and Labour.
We are preparing a detailed report of how we arrived at this and a preliminary investigation of what went wrong, and how we might avert such errors in future.
We are delighted that a methodology in such an experimental stage of development should work so well in a context where the established and far more expensive methods arguably fared worse.
We are now investigating the prognosis for the EU referendum and seeking sponsorship to take the methodology across the Atlantic for the Presidential Election.