Saltire Flags
The flag of Scotland, the Saltire, blows in the wind near Berwick-upon-Tweed on the border between England and Scotland, Sept. 7, 2014. The referendum on Scottish independence will take place on Sept. 18. Reuters/Russell Cheyne

With voter turnout expected to exceed 80 percent in Thursday’s Scottish independence referendum, the outcome of the closely contested vote is becoming increasingly uncertain. The combination of high turnout and a close race is causing concern for many pollsters in the UK, who say they may have gotten their predictions totally wrong.

While polling for the referendum has regularly put those in favor of staying with the United Kingdom ahead, the last few weeks have seen the polls narrow with the pro-independence “Yes” campaign closing the gap in a Tuesday poll to just two percentage points, well within the margin of error.

“We are dependent on a pot of people which is defined, but we don’t know how big it is, and in my view it won’t be big enough,” Martin Boon, director of polling company ICM, told the Daily Mail on Tuesday. “In that lies a real danger for the accuracy of the polls.”

The latest Panelbase poll put the No campaign in front with 50.6 percent compared with the Yes side's 49.4 percent. However, this breakdown does not include people who are still undecided.

The fear, says Boon, is that there will be a surge in support of the Yes campaign among people who don’t normally vote, or conversely that there are more No voters than those who admit it, because people fear being seen as unpatriotic. The combination is misleading pollsters and may be painting a skewed picture of the result.

A similar scenario unfolded in the 1992 UK general election: After 13 years of Conservative rule under Margaret Thatcher, the opposition Labour Party, led by Neil Kinnock, was expected to take power. But the polls got it wrong, and the new Conservative leader John Major recorded a surprise victory. The turnout was 77.7 percent, one of the highest in British general election history.

Moreover, unlike 1992, where polling was conducted face-to-face, the polls this time around have largely been conducted on the Internet and by phone, adding to the uncertainty. There are two reasons for this, says a Pew research paper from 2010. Firstly, not everyone has Internet access, “and those that do,” says the research paper, “are demographically different from the rest of the public.” In Scotland, for example, only 80 percent of adults have access to the Internet, according to 2013 Scottish government data. And even those people may not necessarily be the type that would take part in online polling. The other issue, according to Boon, is that new, young voters -- people as young as 16 can vote this time -- are less likely to sit at a computer answering questions and are also less likely to have their own phone land lines -- so are missed by pollsters.

It’s more likely that those without Internet access are generally from a poorer demographic in the population, says the Pew report -- and people with lower incomes are more likely to sympathize with the independence movement. In short, the polls have not taken into account a lot of people who are likely to favor independence. On the other hand, a lot of people who will vote No often claim they are undecided when asked by pollsters.

So far, the published poll in Scotland with the largest sample has been from a group of around 1,000 people from varying backgrounds, which Boon says is not big enough.