- Growing unease over reliability of public opinion surveys
- Technology, response rates and demography are all at issue
Polls for the 2016 U.S. presidential race have been defying all expectations: Donald Trump as the persistent Republican frontrunner even as he insults large swaths of the country and brushes off policy questions; Hillary Clinton haunted by an email controversy Democrats shrug off while a Vermont socialist keeps gaining on her.
Are the polls correct? While that is hardly a new question, doubts are intensifying after a series of high-profile misfires around the world in the past year, notably in Greece, Israel and the UK. As politics and business lean increasingly on surveys and data, technological and social shifts are combining to challenge polls’ reliability in an entirely new way. Polling professionals have no solution; investors are wary.
"There isn’t a pollster out there who thinks about this seriously who isn’t a little bit uneasy," said Kirby Goidel, editor of the book "Political Polling in the Digital Age." Interviews with more than a dozen pollsters in the U.S. and around the world revealed similar anxiety.
Brad Schruder, a director of foreign exchange at Bank of Montreal, said what many in the investment world have been thinking: "It makes you wonder, how much weight should we attach to these polls?"
The problem stems from a number of causes but begins with a fundamental shift in the public’s relationship with the telephone. For decades, the vast majority of people had landlines that they answered faithfully and, when asked to take part in surveys, mostly did so. Today, home landlines are dying and, when asked over mobile phones to answer questions, a big majority declines.
"Telemarketing, from a pollster’s point of view, poisoned the well," said Charles Franklin, director of Marquette University’s survey operation. "Then came answering machines and caller ID. Most of the time, you never get a human to pick up now."
The paltry response rates come at a time of rising anti-establishment sentiment in U.S. primary elections reminiscent of the unpredictable races in other countries. Such sentiment has never been easy to measure, especially when predicting whether populist anger will turn into real votes.
Last week, Greek voters re-elected Syriza leader Alexis Tsipras by a margin unforeseen by polls. That followed a Greek referendum in July over whether to accept bailout conditions by the European Commission and International Monetary Fund. Polls suggested voters might barely reject it; instead the no’s had it by more than 22 percentage points.
"We may need to change the practices that we have been using for the past 25 years," said Costas Panagopoulos, head of Alco, a Greek polling firm which got its forecast wrong. "It seems that a part of the society thinks that opinion surveys are tools of the establishment and refuses to participate."
Polling in the U.K. in May suggested that Prime Minister David Cameron faced a serious risk of being ousted. His decisive victory led the British Polling Council to launch an inquiry. And in March, polls made it seem that Israeli Prime Minister Benjamin Netanyahu was at risk of losing his job. He won handily.
The U.S. has had its own problems. In 2012, pollsters for Republican presidential nominee Mitt Romney didn’t believe minorities would turn out for the re-election as they had in 2008 when Barack Obama was elected. They were mistaken. Polls in the 2014 midterm elections that cost Democrats control of the Senate under-predicted how well Republicans would do. In most cases, polls did suggest thin Republican victories but in some states, including North Carolina and Alaska, the polling was flat-out wrong.
Plunging participation rates in polling are everywhere. In Britain, ICM Ltd.’s Martin Boon says that when he started the company in 1995, it took 3,000 to 4,000 calls to produce 2,000 interviews. This year, it took 30,000. The cost of polling is therefore soaring.
A single voter
With sample sizes often small, fluctuations in polling numbers can be caused by less than a handful of people. A new NBC News/Wall Street Journal national survey of the Republican race out this week, for instance, represents the preferences of only 230 likely GOP voters. Analysis of certain subgroups, like evangelicals, could be shaped by the response of a single voter.
Shifting demographics are also playing a role. In the U.S., non-whites, who have historically voted at a lower rate than whites, are likely to comprise a majority of the population by mid-century. As their share of the electorate grows, so might their tendency to vote. No one knows by how much, making turnout estimates hard.
Pollsters are finding that the political equivalent of brand loyalty is in decline. “Electorates are more volatile,” said Will Jennings, one of the polling inquiry panel members in the U.K., who teaches politics at Southampton University. “They are de-aligned from parties.” The panel is noting that the British election saw different pollsters trying different methods with different weightings -- and all were wrong in the same direction.
To save money, more polling is done using robocalls, Internet-based surveys, and other non-standard methods. Such alternatives may prove useful but they come with real risks. Robocalls, for example, are forbidden by law from dialing mobile phones. Online polling may oversample young people or Democratic Party voters. While such methods don’t necessarily produce inaccurate results, Franklin and others note, their newness makes it harder to predict reliability.
The young, who increasingly communicate via text, are not only less willing to speak over the phone about political preferences; they are less interested in talking by phone about anything at all.
As response rates have declined, the need to rely on risky mathematical maneuvers has increased. To compensate for under-represented groups, like younger voters, some pollsters adjust their results to better reflect the population -- or their assessment of who will vote. Different firms have different models that factor in things like voter age, education, income, and historical election data to make up for the all the voters they couldn’t query.
"The risk that we all face is that when assumptions trump the raw material, we’re too far out on that limb," said Jon Cohen, vice president of survey research at Survey Monkey and the former polling director for the Washington Post, who is among those using online surveys.
Multi-party contests, more prevalent outside the U.S., are harder to poll than two-party races. This year’s crowded U.S. Republican primary field may create similar polling challenges.
Stan Greenberg, a U.S. Democratic pollster who has also worked for the U.K. Labour Party and the liberal Zionist Union in Israel, said that when the economy is in trouble and anti-elite anger is rising, turnout is hard to predict and polls become less reliable.
"In Israel and the U.K., you have two incumbent governments that were up for re-election,” he said. “And in both cases, they were able to make the elections about the opposition."
While pollsters agree that all these shifts are causing difficulties, there is disagreement over how much of polling reliability has suffered. Despite the drop in response rates, the accuracy of polls using standard methodology has not declined at the same rate.
“Until we have a presidential election where the polls really fail in that context, I don’t think I’m ready to come to a conclusion that something fundamental has changed about polling accuracy,” said Scott Keeter, Pew’s director of survey research.
If anything, the U.S. has a polling advantage over other countries, said James Morris, a U.K.-based partner at Greenberg Quinlan Rosner Research. Unlike Britain, which holds elections every five years, American pollsters get to update their data sets every two years.
"The reason why the U.S. election is unlikely to be like Greece or the U.K. is that in America, you have lots of opportunities to check if your polling is right,” Morris said.
In 2016, however, there is at least one new unexpected and unpredictable variable to contend with.
"And then there’s Trump," Morris said. "No one knows whether that’s real."