Israeli pollsters: Our off-mark 2015 surveys were ‘entirely different’ to US election fiasco
As US pollsters lick their wounds after failing to predict Trump victory, Israeli experts insist their models were fine last year; it’s just that Netanyahu electrified his voters at last minute
Marissa Newman is The Times of Israel political correspondent.
When the US presidential election results trickled in on the night of November 8, the increasingly dumbfounded reaction by pollsters and pundits alike and the ensuing mass disillusion with the national polls — which had overwhelmingly predicted a Hillary Clinton victory — was thoroughly familiar for Israelis.
Less than two years earlier, following the 2015 election, many Israelis had turned against their false prophets when the TV prognosticators failed to predict that Prime Minister Benjamin Netanyahu would be reelected and his Likud party voted the largest party by a six-seat margin. Surveys in the run-up to the vote had Likud trailing or tied with the Zionist Union, and the exit polls placed the two parties neck-and-neck.
“When they told me the results, I thought I had died,” Prof. Camil Fuchs, who was in charge of Channel 10’s polling, told The Marker business daily the morning after that election. Channels 1, 2 and 10 all had similar findings from the exit polls, predicting Likud would win 27-28 seats and Zionist Union 27. The actual result was Likud 30 and Zionist Union 24. (Fuchs, incidentally, predicted that Donald Trump would win the presidency in an interview with the Walla news site in early November).
This was certainly not the first time Israeli pollsters were off the mark. Many Israelis still recall the 1996 election shocker when, as the now famous maxim goes, voters “went to sleep with [Shimon] Peres and woke up with Netanyahu.”
But as the dust settles on the US election and its analysts nationwide engage in collective soul-searching, Israeli pollsters are adamant the two elections — and forecasting flops — should not be compared.
“It’s entirely different,” said veteran pollster Dr. Mina Tzemach of the Midgam institute, who conducted numerous surveys for Channel 2, as well as its exit poll, during the 2015 Israeli campaign.
Tzemach insisted that her polling was broadly accurate. What happened, she said, was that the Likud party cannibalized right-wing ally-rival Jewish Home of its voters via a series of last-ditch campaign efforts, including a large right-wing rally held days before the election, and Netanyahu’s election day warning to his supporters that the Arabs were voting in “droves.”
“All the changes in Israel were within a camp, within the right-wing bloc. There were no other changes. The survey offered information that said the Zionist Union would be stronger than Likud, and people were alarmed, and then on the Sunday before the elections there was the Likud rally, and then Bibi spoke about the Arabs heading to the polls in droves, etc. So, simply, some of the Jewish Home voters who were right-wing and wanted the right to remain in power, simply moved to Likud — that’s all that happened,” Tzemach told The Times of Israel.
Tzemach emphasized that she has not dramatically revamped her statistical models since the election because she was largely satisfied with the results of her 2015 polls.
“Listen, we make changes all the time, but it’s not as a result of the election, because we were very pleased with the results of the polls before the election. We were also pleased with the [exit] results from the night of the election — they were very precise — but we close the [exit] polls at 8 p.m., so it’s a problem, because we’re missing two hours of voting,” she said.
“We always check ourselves and change [the models], not because of the last election, but because there are changes in the population and in the behavior of the public,” she added.
‘Every school and academic setting should study his [Netanyahu’s] negative campaign’
Similarly, Dr. Joseph Sarid, the head of the Sarid Research Institute, attributed the shift in tide during the 2015 election to Netanyahu’s “droves” comment, which he described as “very drastic, borderline unethical and immoral,” but added that “every school and academic setting should study his negative campaign.”
Most polls accurately distributed the Knesset’s 120-seats among the political blocs, he maintained, but the”polls couldn’t have known that a few hours before the polls closed, the prime minister would launch a negative campaign par excellence. And that this campaign would be so successful.”
Like Tzemach, Sarid pointed out that the Likud “took the seats from his partners,” on the right, namely the Jewish Home, rather than from the other side of the political spectrum.
“The case of United States is similar to the Brexit in Britain. But it is not similar to what happened in Israel,” Sarid argued, referring to the June 2016 referendum vote in the United Kingdom to leave the European Union, another failure for pollsters.
‘Witnessing the collapse of the paradigm’
Sarid conceded that the field of polling was ripe for a shake-up, noting that with technological and scientific advances, “the industry is a step behind the events.”
“We all rely on the same methodology, the same methods, and on the same paradigm…. We are witnessing the collapse of this paradigm. And it’s happening all over the world,” he said.
At conferences, in the literature, and within the professional community, there is little discussion of far-reaching changes to statistical models, he said, and the “discourse has not ripened.”
“In my view, if there is a change to the paradigm or in the general view on what to do, how to do it, what methodology is appropriate, I assess that if it happens, it will come from the United States,” he said. “Because there, first of all, it’s a huge country, but also there are elections there all the time. And therefore, you can adopt new methods, check them, put them into practice, very often.”
Urging patience and suggesting the need to incorporate other information, including from social media, he added: “The question is how you measure it, what weight you give it, to get answers. This is still unclear. But there is no doubt that we are witnessing the collapse of the paradigm.”
Is big data, monitoring social media the answer?
While pollsters in the US contemplate their future, some companies with access to larger data sets on voters are raising their heads with alternatives.
Nadav Shoval is the co-founder and CEO of Spot.IM, an Israeli firm based in New York that powers the comment sections of over 25 percent of US digital content sites.
Spot.IM’s platform is used by over 4,600 media outlets, including AOL, Time Inc. and Huffington Post, and enables readers to interact with the sites and with each other via comments. There are over 300 million interactions on Spot.IM-sponsored comment forums each month, on articles with over 3 billion page views.
During the election, the company released information on trends in the comments, including their increasingly violent, racist and anti-Semitic nature, and noting that discussion of Trump — at one point — was dropping.
While the comments sections on websites are notoriously rabid and may not seem representative of the broader population, and although the company did not appear to offer a detailed statistical analysis of its findings, according to Shoval, the company foresaw Trump’s victory.
“We predicted before the election that Trump would win, because 60 percent of the commentators that were discussing the election were pro-Trump. So yes, certainly, we have a means to help those future polls, because we have much more reliable information than they do,” he said.
Shoval opined that the polls in 2016 “are becoming less relevant, because people aren’t always truthful, and are sometimes embarrassed to admit the truth.” The only way to draw conclusions on voting trends is to “listen to the masses,” on social media, comment sections, he said.
But he would not comment on whether his company gathers demographic or geographic information to the talkbacks that would allow researchers to parse out information on voting trends. “I can’t reveal things like that,” he said.
Vox Pop Labs runs the popular Vote Compass online quiz, which allows users to assess where they fall on the political spectrum and how their views align with political candidates.
According to Howard Cohen, the Director of Global Outreach, the company in its partnership with Sky News “was one of the few entities to correctly call the Brexit vote — not only overall, but region by region 10 days before the referendum.” Cohen also maintained that he had predicted a Trump win for a graduate-level class taught by CNN’s senior political analyst David Gergen at Harvard University, giving Trump 275 electoral votes to Clinton’s 263 (Trump ultimately won 290, Clinton 232).
“I arrived at this prediction by applying lessons that the Vote Compass team learned from our accurate prediction of Brexit and by taking into account systematic errors that we see in mainstream polling today,” he wrote in an email. “In particular, I took into account the under-representation of the uneducated voter in mainstream polls, the importance of voter enthusiasm (Trump supporters were very enthusiastic about his candidacy, but the same was not true of Clinton supporters) and the systematic under-representation of Trump supporters in mainstream polls owing to the fact that some voters were reluctant to admit their support for him (even to pollsters).”
Cohen attributed the accuracy of the Vote Compass data to both the sample size generated by the participants in the online questionnaire, which he described as “unparalleled,” and to the incentive by users to tell the truth in order to get an accurate assessment of how their political views line up.
“One of the reasons why mainstream polls incorrectly called the US 2016 Presidential election was owing to voters who were too shy to honestly admit their support for Trump to pollsters,” Cohen wrote. “When a pollster calls someone to conduct a survey, there is little reason to tell the truth. By contrast, with Vote Compass, voters respond to the questionnaire to learn about where they fit in the political spectrum. The only way to meaningfully learn about yourself in the political realm is by answering questions honestly.”
“What the US election demonstrates,” he later added, “is that mainstream polling is losing steam and that the future of polling is online.”
