Don’t trust this Stoke by-election poll until we know more about it

Honesty in public debate matters

You can help us take action – and get our regular free email

Sign up

“UKIP BOOST: Paul Nuttall set for STORMING victory in Stoke by-election, poll reveals”

Daily Express, 3 February 2017

This poll received some attention when it was published. As far as we're aware Leave.EU, the campaign group who carried it out, still haven’t published the information you’d need to say whether or not it’s reliable. We’re waiting to hear back from them.

The Daily Express said this in its report:

“Leave.EU – which is funded by UKIP donor Arron Banks – did not provide details of the poll methodology.

But it insisted it was a "major polling exercise" and said the sample size was greater than 4,000.”

A sample size of more than 4,000 people would be relatively large by the standards of many constituency polls, which commonly interview around 500 or 1000 people.

And Leave.EU has said:

“Our polling method consists of a high volume randomised sample and the use of sophisticated AI social media techniques developed in the United States by renowned public affairs firm Goddard Gunster”.

But without knowing things like exactly how the question was pitched, how the poll was carried out, and whether the responses were weighted to make the results representative of the constituency as a whole, we don’t have the information we need to judge for ourselves whether the results are reliable.

We’ve contacted Leave.EU to ask for the things which the British Polling Council, an association of polling organisations, expects its members to publish alongside their results:

• Dates when the interviews were carried out
• How the interviews were carried out (e.g. in-person, telephone, internet)
• The group the poll is trying to represent (e.g. all adults, eligible voters)
• The percentages of different responses (including 'don't knows' and refusals)
• The number of people who were interviewed and the area the poll covered
• How the response of the sample group was ‘weighted’ to form conclusions about the opinions of the constituency as a whole (i.e. adjusted so that the kinds of people in the sample match up to the population in the constituency)

We also asked for the complete wording of the question, and whether there was a web address where full results could be viewed.

If Leave. EU are able to provide us with this information, we’ll be able to judge the reliability of the poll better. Until then, it’s not possible to say.