LONDON (Reuters) - British pollster YouGov said on Friday that online polls were doing a better job than phone polls of finding representative samples of voters before a referendum on whether Britain should leave the European Union.
A persistent divergence between telephone polls, which suggest the "In" campaign is ahead by a wide margin, and online polls, which suggest a tight race that "Out" could win, has flummoxed pollsters in the run-up to the June 23 referendum.
The polls are being closely watched by financial markets and have repeatedly caused sharp moves in the value of sterling against other currencies, but the confused picture they have created has added to jitters about the outcome of the vote.
YouGov, which produces online polls, said it had conducted new research that showed telephone polls were based on samples of the population that over-represented graduates, who are more likely to be in favor of remaining in the EU. (http://bit.ly/22hY2dd)
Rob Ford, a political scientist from Manchester University with expertise on polling methods, said an online pollster's view on phone polls should be taken with a pinch of salt but YouGov had nevertheless put forward a plausible hypothesis.
The entire British polling industry failed to predict the Conservative victory in last year's national election which Ford said had created a "nightmare scenario" for pollsters because they were now under pressure to do better on the EU referendum that was much harder due to a lack of precedent.
"I'm sympathetic to their problem. The pollsters are sincerely trying to get the right answer but ultimately we just don't know," he said.
People have also been closely watching betting action as a reliable guide.
William Hill has cut its odds of Britons voting to remain to
1/5, the lowest level to date and indicating an implied
probability of 83 percent. Implied probability of an "In" vote rose 9 percentage points to 79 percent from around 70 percent last week, according to Betfair.
In an article by Andy Morris, its chief innovation officer, YouGov said it had analyzed the data from two telephone polls by rival firm Populus, conducted in February and March, which had 44 percent and 46 percent of graduates respectively.
Morris wrote that if the Populus data was adjusted to match the proportion of graduates in the population at large, estimated at 27 to 33 percent, the 11-point lead for "In" in the February poll shrank to just one to four points.
The same calculation on the March poll shrank a 14-point lead for "In" to seven to nine points, he wrote.
YouGov said it had tested its theory by conducting simultaneous online and telephone polls of its own, in which it weighted the samples to reflect the proportion of graduates in the population.
It said the results were a two-point lead for "Out" in the online poll and a three-point lead for "Out" in the telephone poll.
"This result is extraordinary but not, on reflection, surprising," wrote Morris.
"Sample composition on education level is absolutely fundamental to referendum polling. With phone polls under-representing the less well educated so dramatically it is not surprising that they are skewing toward ‘remain’," he wrote.
(Reporting by Estelle Shirbon; Editing by Angus MacSwan)