A statue of Winston Churchill silhouetted against the Houses of Parliament, London, 24 June 2016.
(AP Photo/Matt Dunham)

By Bernd Debusmann

The last opinion survey on Britain’s EU referendum was conducted on the day citizens cast their votes and was published just hours before polling stations closed on June 23.

The YouGov poll found 52 percent in favor of remaining in the European Union, 48 percent for leaving. A poll earlier in the day by Ipso/MORI showed a comfortable 54-46 lead for “remain.”

The actual result, which shocked both Britain and the world: 52 percent voted for leaving, 48 for staying — the exact opposite of the YouGov prediction.

The two finals polls were not alone in finding that Britons would vote to stay in the EU. Poll after poll – scores of them – predicted that Remain would prevail, narrowly.

What went wrong is a question that will prompt another round of soul-searching in the global polling industry — just as it did after the British parliamentary elections 13 months ago when virtually no poll came even close to predicting a comfortable victory for Conservatives.

In Brexit polls, there was a flawed assumption.

Polls were just as wrong on last year’s Israeli elections, the Greek bailout referendum and Scotland’s independence referendum in 2014.

In the U.S. general election four years ago, Republican Mitt Romney was so confident he would triumph over Barack Obama that he ordered elaborate victory celebrations. His confidence was based on internal Republican party polls and surveys from outside companies, including Gallup, one of the oldest and most respected U.S. polling companies.

Embarrassed by a series of spectacular failures, pollsters point to two trends that have made accurate surveys more difficult over the past few years. One is the explosive growth of cell phones — which make it harder to reach a representative sample of the population — and the other a sharp drop in the number of people willing to answer survey questions.

The response rate in the United States, to use the technical term, dropped from 90 percent in the civic-minded 1930s to eight percent in 2014. And even that small number is shrinking.

In Britain’s Brexit polls, a flawed assumption by many pollsters appears to have played an important role.

Drawing from history, the assumption was that undecided voters would opt for the status quo and vote to remain in the EU. ORB, a London-based polling firm, thought undecided Britons would vote three to one in favor of staying. They did not.

“The news media see every poll like an addict sees a new fix.”

Not long ago, there was consensus that the best way to yield accurate polls is through live telephone interviews. That was considered the “gold standard” of polling.

The British polls have cast doubt over that view. Online polls consistently showed an extremely close result through much of the campaign and gave the Leave camp a small advantage. Telephone polls had the Remainers ahead by a small margin.

Will the spate of polling misfires around the globe curb the demand for predictions? Almost certainly not, despite warnings that polls should be taken with more than a pinch of salt. In the United States, the presidential campaign is likely to set new records for the number of opinion polls and news stories and analyses pegged to surveys.

“Stop the Polling Insanity,” said a headline over an essay in the New York Times newspaper last month by two respected political scientists, Norman Ornstein and Alan Abramowitz.

The pair complained that “in this highly charged election, it’s no surprise that the news media see every poll like an addict sees a new fix.” Many of the polls and their coverage, the two said, have been “cringe-worthy.”

“Horse race” polls have little more than entertainment value.

They also show extraordinary variances.

On June 26, with polling experts still in shock over Brexit, a poll conducted by the Washington Post and ABC gave Hillary Clinton, the presumptive presidential candidate of the Democratic Party, a 12-point lead over her rival Donald Trump (51-39). On the same day, a poll by the Wall Street Journal and NBC saw Clinton leading Trump by five points (46-41).

Those results were based on responses to the question, “If the election were held today, for whom would you vote?”

That question is the reason why many political scientists think “horse race” polls have little more than entertainment value, providing grist for the mill of the 24-hour news cycle.

The answers to the standard poll question might produce an accurate snapshot in time but are of little use as a forecast. Between “today” and November 8, any number of things, from a terrorist attack to stock market gyrations, could change respondents’ mind.

Although media critics and political scientists bemoan the prevalence of shallow horse-race stories, there is no sign that editors and pollsters are reining them in. There is no sign, either, that pollsters are closer to figuring out better ways to predict electoral winners.

Bernd Debusmann is a former columnist for Reuters who has worked as a correspondent, bureau chief and editor in Europe, Latin America, the Middle East, Africa and United States. He has reported from more than 100 countries and lived in nine. He was shot twice in the course of his work – once covering a night battle in the center of Beirut and once in an assassination attempt prompted by his reporting on Syria.