UK + Israel, Pollsters Got It Wrong

Whereas the polls confidently predicted a neck-and-neck finish between the Conservative Prime Minister, David Cameron and Labour's Ed Miliband - and foresaw there would inevitably be a messy coalition squabble including Liberals, the far-right UKIP and Scottish Nationalists.... actually the results on the day were entirely different.

David Cameron swept the floor with Labour (Miliband resigned); the Liberals and UKIP vanished; and the Scottish Nationalists took almost every seat in Scotland.

For Israelis, the 'shock' results reflected our own recent elections, where Likud's Netanyahu was predicted to come in a close second to Labour's Herzog. The actual election results showed a dramatic rise in Likud support and fall for Labour.

I am not a statistician, and I appreciate that the science of polling opinions is just that. A science. It is based on rigorous statistical models, computerised simulations, and decades of experience.

Major polls around elections include huge and varied data sets, where margin for error is very small; furthermore, exit polls (which in Israel also were wrong) then act as a final reality-check.

Human error is a part of everyday life, whereas scientific error raises a different level of concern.

Getting the election results so entirely wrong does not have much consequence - at the end of the day, perhaps the pollster's share value dropped, but both the UK and Israel have fairly elected prime ministers and governments.

However, if these same statistical models, fail-safe predictions and rigorous simulations are used for more critical tasks - such as road safety, flight control systems, medical research, national power grids, internet system protection etc...then the consequences of error could be catastrophic.

It's one thing to make derogatory remarks about the pollsters - another to question the statistical methodology they were using.


Popular posts from this blog

Will Motty Borger’s Suicide Make Any Difference?