With the UK in the grip of election fever, there is much discussion this week about differing methods of measuring voter intentions and why they give different results (here). So should we trust the opinion polls?
It’s certainly difficult to know whether we should. Even if an opinion poll is based on a random sample the numbers we get are subject to statistical uncertainty and this is reflected in a ‘margin of error’ (formally known as the confidence limit). Of course responsible pollsters report the margin of error when generalising the results to the population, but the fact of the matter is that most opinion polls are not truly random – every member of the population does not have an equal chance of being selected. Even those polls which select phone numbers using random dialing software are not truly random; many people don’t pick up the phone or refuse to take part in telephone surveys. And there are other concerns, for example, is the sample actually drawn from the correct population? There is a difference between the voting population and those who actually turn out to vote on voting day. Even if you screen respondents on whether they intend to vote, it could be that more people say they will vote than actually do. And finally, just to add to the pollsters woes, what about that thorny issue of response bias? Apparently, some people say they will vote for a particular candidate but change their mind when they get to the voting booth (who knew?).