×

Brexit is looking increasing unlikely—but should investors trust the polls?

In the final few days leading up to the U.K.'s EU referendum, investors have jumped on polls which suggest the leave campaign is losing steam. But should investors be so quick to accept the data?

After a set of polls early last week suggested the leave campaign was winning over the public, polls by Survation, YouGov and BMG Research released in the last few days have shown a swing back to the remain side. The Financial Times's poll of polls (which has tracked public opinion since 2010 and shows an adjusted average of the most recent polling data) has the two camps tied at 44 percent each.

This polling data are having a big impact of markets. Sterling has rallied against several other currencies, while oil prices have risen.

Christopher Furlong | Getty Images

Chris Weston, chief market strategist at spread betting provider IG, said the polls had increased investors' risk sentiment, but warned that markets were still fragile and could easily turn negative.

"The remain camp has polled well in a weekend Survation poll (for the Sunday Mail). The odds of the U.K. leaving the EU have fallen below 30 percent again, but there are at least five more polls to go," he said in a note published early Monday morning.

"For now though risk appetite has increased with a 1.3 percent increase in GBP/JPY this morning, S&P futures have rallied 0.6 percent, oil is up 0.7 percent and gold 0.9 percent lower."

This recent shift in public opinion from leave to remain is an important indicator of how the country might vote in Thursday's referendum.

"In referenda, what normally happens is in the last few days there is a swing back to the status quo," James Morris, partner and director of the European office at Greenberg Quinlan Rosner, told CNBC Monday.

"Over the weekend, there's been two or three polls that have shown a shift back towards the remain side from a leave win."

However, polling data has come under greater scrutiny this year, after polls failed to predict the results of the U.K. general election in 2015. This has led some to question the value of opinion polls.

"Obviously, there are some limitations to opinion polls. Even though polls based on true random probability samples are the best known method of prediction, the error margins are often larger than we would expect mathematically," said Rainer Schnell, director of the center for comparative social surveys at City University and Kathrin Thomas, a research associate at the center, in a joint statement emailed to CNBC.

Getty Images

"This can have different reasons and is often, but not exclusively, related to omitting relevant sub-samples such as people without phones (non-coverage), selective refusal to participate in surveys (non-response) and errors in the prediction of who actually turns out in elections," they explained.

Morris also explained how failure to contact certain sets of people may have affected the polls' findings.

"Polls contact people in different ways. A lot of them, because they try and get into the newspapers, try and do it over a two-day window. But some of the polls have tried to contact people over seven or eight days," said Morris.

"What they've found is that the people who are harder to contact are more likely to vote remain."

The EU referendum will take place on June 23, with the results expected to be announced early on June 24.

Follow CNBC International on Twitter and Facebook.