Do Communicate Research polls overstate Labour?

Yesterday’s Communicate Research poll in the Independent on Sunday suggesting that Labour is doing even better than at the 2001 General Election has provoked a lot of debate on how the pollster carries out its surveys.

For ever since the firm began polling for the paper we’ve raised questions about the CR methodology which we believe overstates Labour and understates the the Tories and Lib Dems.

A major challenge for telephone pollsters is finding a sample of voters to interview that is representative of the electorate as whole. For every successful interview they carry out most have to make to six calls. Many people are out, many have answerphones or employ some form of phone-screening that stops calls from unknown sources getting through and many, the majority, simply refuse to be interviewed.

So how can they know that those they do talk to are representative?

Could it be that the available interviewees are not a fair reflection of the electorate as a whole? Could it be, say, Labour supporters have different lifestyles from, say, Tories, that makes it more likely that they will be there when the phone call comes and they are more willing to take part?

As well as asking standard demographic questions to ensure the correct age/gender and socio-economic make-up of the group a measure that all the other phone pollsters adopt is to ask how interviewees voted last time.

Because we know that Labour got 42% of the vote in 2001, then clearly something needs to be done when, for example, the March Populus poll found that 51% of those taking part said they voted for the party four years ago. You have to allow for some false memory or faulty recall and in this case Populus weighted their sample so that 46% were Labour voters. In broad terms this means that the Labour totals in the March survey were reduced by about a tenth.

It is this which CR are not doing and why we think their polls are over-stating Labour. We are also concerned that when they put the question CR’s interviewers do not list the party options. In the past it has been shown that this reduces support for the Lib Dems who sometimes get forgotten.

All those interested in election betting and forecasting political outcomes have an interest in knowing how each pollster works. We would love to hear from CR about the the points we raise.

Mike Smithson