The United Kingdom votes in a referendum Thursday on whether it will leave the European Union, or "Brexit." The decision could have major global economic and political ramifications.
There have been 65 polls of the British citizenry since HuffPost Pollster started tracking them on April 1, but polls aren’t providing much insight into which way voters will swing. The combination of pre-existing polling issues and a one-time special interest vote is proving particularly vexing to pollsters.
The HuffPost Pollster model estimates that if the election was held today about 46 percent of Britons would vote to leave the EU, and 45 percent would vote to remain. The probability that leave is in the lead is about 78 percent, but it wouldn’t be surprising for the June 23 vote to go either way.
But those numbers have wavered -- just last week, the polls appeared to swing even farther toward “leave.” Polls since Member of Parliament Jo Cox’s murder, which could have been connected to her outspoken support for staying in the EU, have inched back toward “remain.”
But the trajectory of the race is different depending on what polls you look at. There’s a substantial split between live interviewer polls and polls done over the internet, and splitting the polls along that cleavage results in completely different views how opinions have shifted.
In the live interviewer polls, "leave" the EU holds a substantial 6-point advantage -- a complete trend reversal from just a month ago. “Remain” was averaging a 10-point lead in mid-May.
Internet polls show opinion more narrowly leaning toward leaving the EU -- the margin is about 2.5 percent. These polls have been very close since HuffPost Pollster began tracking them in April, with the lines virtually indistinguishable until late May. A shift in sampling methods from YouGov, one of the prolific online pollsters, doesn’t seem to have altered the overall trend.
Another big factor in whether the polls will be accurate is how their estimates of voter turnout will compare to what actually happens on Thursday. Referendum votes occur less frequently than parliamentary elections, and can bring out a different electorate. Pollsters often use past vote history to help ground their estimates for turnout, but past vote information doesn’t exist for most pre-referendum polls because this is only the U.K.’s third referendum in more than 40 years. That infuses even more uncertainty into the estimates.
British polling has been under a microscope since May 2015 when polls didn’t anticipate a sizable Conservative victory. An inquiry report from the British Polling Council showed that the polls weren’t actually much farther off than usual. But by indicating a tied race they essentially “called” the race wrong -- whereas showing a lead for the Conservatives and the election resulting in a much bigger lead wouldn’t have raised so many eyebrows.
Regardless, this is the first major test of the British polls since that perceived failure. And it’s a gigantic test that highlights -- and magnifies -- the major problem in election polling in the modern era: how to get a representative sample of citizens who will actually vote.