Britain's pollsters are fighting about who is more inaccurate

Peter Curran / YouTube

Former YouGov president Peter Kellner claimed the polling form was overstating public support for a Brexit

A dispute is brewing between Britain's top pollsters over the methodology of surveys used to predict the outcome of the EU referendum.

The spat began on Friday and revolves around the methodology YouGov uses for online polling.

Peter Kellner, YouGov's former president, posted a blog in which he criticised the research firm for over-representing UKIP supporters in its EU referendum voting intention surveys.

YouGov admitted earlier this week that it had been using samples which contained too many UKIP supporters and this error led to Brexit support being overstated. However, Kellner's blog said that this flaw hadn't been fully rectified.

He said:"In its (YouGov's) latest poll it has tweaked its methods with the aim of eliminating its overstatement of Ukip's support. However, even after the change, its 18% for Ukip still looks too high."

This claim caught the attention of YouGov's current CEO Stephan Shakespeare, who tweeted shortly after Kellner's blog was posted describing his former president as the person who got the general election wrong.

Then YouGov published an article of its own, in which it defended the accuracy of its sampling methods and accused phone pollsters of being skewed towards the university-educated. 

It was sub-headed: "Phone polls have too high a percentage of graduates - YouGov's online polls don't."

In the article, YouGov claims that phone pollsters were surveying too many graduates in their EU referendum polls - a social group which is generally more likely to vote against a Brexit.

The article says that survey group Populus was using samples where up to 46% of the respondents had degrees, despite the fact that only around 27-33% of the British population has degrees.

If correct, this would mean that the ORB, Ipsos MORI, and ICM phone polls published this week, which gave big leads of up to 18 points to Remain, could be inaccurate because they overstated public support for Britain to stay in the EU. 

This graph shows how polls published in the last week pointed towards an increased lead for Remain.

What Uk Thinks

In fact, YouGov - which usually conducts its surveys online - did a phone poll of its own, using a sample containing a number of graduates that is much closer to the national share. The results of this poll were: Remain 36 / Leave 39.

Given that all the other phone polls published this week showed such big leads for Remain, the results of this YouGov phone poll are surprising.

Polling analyst Matt Singh told Business Insider on Monday that the drastic differences between recent opinion polls mean that some pollsters are on course to making very inaccurate final EU referendum predictions.