A number of EU referendum polls have been published in the last 24 hours and they don’t really agree on much. In fact, all the latest round of data shows is that experts are struggling to predict how the June 23 vote will turn out with enough accuracy to be taken totally seriously.
The most recent poll was an OBR survey published by The Telegraph on Tuesday which put the Remain campaign nearly ten points ahead of Leave (52%-43%) — with just 5% of respondents saying they don’t know how they will vote.
Compare this to a ComRes survey released on Monday which gave Leave a seven-point lead (45%-38%) but also said 17% still aren’t sure how they intend to vote. There’s a 12-point discrepancy between the results each research agency published about undecided voters. Twelve per cent is a lot — it could decide a referendum.
The picture becomes even more unclear when you take a look at research conducted by ICM which focused on how voting intention differs between online and over-the-phone polling. According to ICM, its most recent online survey gave Remain a seven-point lead over Leave (48%-41%), but over the phone, more respondents said they planned to vote for Britain to leave the 28-nation bloc (44%-43%).
Online: Remain 43% (50%), Leave 44% (50%), DK 13%
— Martin Boon (@martinboon) April 18, 2016
There is no one reason why online polling differs so much from over types of polling. However it is true that the internet is a facility generally used more by younger people, but 18-34 is the least likely age group to actually go out and vote on referendum day. Only 52% of this age bracket said it would definitely vote on June 23, according to research released by the Observer earlier this month.
There’s other problems with polling methods that need to be considered. Last month, the Financial Times spoke to polling experts James Kanagasooriam and Matt Singh, who pointed out key variants to how polls are conducted which could undermine the accuracy of results.
For example, politicians and the media usually try to get results published as soon as possible, meaning harder-to-reach sections of society often go underrepresented. The phrasing of questions can have a significant impact too, as it can fail to take into the account the “wider social attitudes” of respondents.
It must not be forgotten how wrong most polling services were in the build-up to the 2015 general election. Almost all polls suggested Labour and the Conservatives were level in the final days of campaigning before the Tories won an unexpected majority.