Back: Articles and information on Clem Sunter

Pollsters should be foxier

By Clem Sunter

Pollsters are being criticised for getting the results of the UK election so wrong. They said the country was heading for a hung parliament instead of which David Cameron won the election outright by a slim majority. They did on the other hand predict a virtual wipe-out for the Liberals and a massive victory in Scotland for the Scottish National Party.

I would prefer to say that they were unlucky rather than wrong on the main issue of the election result because any election has a level of uncertainty that no form of polling technique and interpretation of the sample can overcome. People are often less than truthful about the way they are going to vote and in the UK the term "shy conservatives' is used for those who do not want to reveal their right-wing tendencies either to the pollsters or to their friends.

Moreover, the first-past- the-post principle injects a degree of randomness into the result which is not present in an election based on proportional representation. The other point to make is that the picture presented by the polls themselves could have made people vote differently to the way they originally had thought of doing. For example, the idea of a hung parliament could have propelled potential UKIP voters to vote for the mainstream Tory Party to increase the chances of a definite outcome.

Lastly, the performance of the leaders in the national debate and individual candidates in canvassing the vote in their constituencies has to be taken into account, which can add some variability to the result. The 20 year old Scottish student who has become the youngest MP since 1667 must have done her political homework to perfection!

So what is my recommendation to the pollsters? In the lead-up to the election, do not offer a false sense of accuracy by giving a series of single figure projections on the percentage vote that each party may attract. Rather than a prediction, offer a range of voting percentages from high to low which could be given to each party, with the figure with the highest odds from the sample taken being in the middle. In other words, make allowance for the margins of error as any actuary would do. Then look at the different scenarios for the overall election result that may materialise from different combinations of each party's voting percentages and attach a probability to each scenario.

That will then give the public a much more reasonable estimate in advance of what the election results might be. It will also illustrate the chances of extreme outlier results called in the futures trade "black swan events". As a fox would say, it is much better to be vaguely right than precisely wrong. Pollsters would then not have to apologise or eat their hats for doing something which is impossible in the first place.

Back: Articles and information on Clem Sunter