[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries

Polls and predictions


The statistics of voting are now topical, so the time is ripe to comment on the remarks journalists have been making. Shortly after the UK election exit poll, a former Lib Dem MP, Ming Campbell, commented to this effect: “One thing is certain, Theresa May will not get the huge landslide result she sought.” 

The presenter replied: “If the exit poll is right,” in not so many words.

The presenter could not conceivably have understood what they were saying on national television. I am not sure where the goal posts were. But we can say May’s objective was to have a massive majority over all other parties. Say her target was 400 seats. In order for May to achieve her goal, despite the exit poll, predicting 314 seats, we’d have had to have the most unlikely saga. The exit poll would have to be wrong to the tune of 86 seats.

We all know from experience that exit polls can be materially wrong. This happened especially in the 1992 General Election when Conservatives won despite an exit poll predicting a Labour victory. It is not clear why exit polls can be significantly wrong. Presumably it was caused by some Conservative voters not admitting they voted Conservative when interviewed by the opinion poll companies. Even so, exit polls are not wrong to the extent of 86 Conservative seats. A pedant might say the former Lib Dem MP should have said: “Unless the Exit Poll is wildly wrong we can say with certainty that May will not achieve the kind of majority she sought by calling the General Election.”

Actuaries should educate the media about statistical inferences. This is just one example of  the sort of misuse of statistics journalists have been making.

Anthony Pepper

12 June 2017