Why did Israel’s exit polls differ so much from one another?

For anyone watching Channels 12, 13 and KAN at the same time, it was clear that someone made a mistake.

By MAAYAN JAFFE-HOFFMAN
Prime Minister Netanyahu and his wife Sara greet supporters as Netanyahu at the Likud post-election celebration. (photo credit: MARC ISRAEL SELLEM)
Prime Minister Netanyahu and his wife Sara greet supporters as Netanyahu at the Likud post-election celebration.
(photo credit: MARC ISRAEL SELLEM)
On Tuesday night, before Channel 12 pollsters Mano Geva and Mina Tzemach released their results, they must have known that their exit poll results were drastically different from those of their colleagues. Nonetheless, they stood behind their numbers.
Perhaps this was because, as two professionals, they were confident in their exit poll numbers despite the discrepancy. Or, perhaps there was more to it, such as a false hope that if they portrayed a certain scenario as the early election results unfolded, this would be enough to prompt Blue and White Chairman Benny Gantz to deliver a pre-emptive victory speech, which could impact the narrative of Israel’s future.
Either way, for anyone watching Channels 12, 13 and KAN at the same time, it was clear that someone made a mistake.
The Channel 12 exit poll found that the Blue and White party was the largest party in Israel with 37 seats. Likud followed with 33 seats. Channel 13 gave both parties 36 seats, and an exit poll by Kan 11 News gave the Likud 36 seats and Blue and White 37.
When one looks back at the dozens of polls released throughout the campaign, one can see that Channel 12 and Yediot Aharonot pollsters were wrong most of the entire 100 days. They always had Blue and White too high. The only polls who had Blue and White higher were those who conducted polls for Israel Hayom and i24 News, outlets which are known to have a pro-Prime Minister Benjamin Netanyahu agenda and in hindsight apparently had the Center-Left party too high on purpose.
Other pollsters, such as Rafi Smith Polls and Prof. Camil Fuchs who works for Channel 13, were right most of the time, showing Blue and White and Likud neck-in-neck, with a huge lead for the right-wing bloc. (All the pollsters maintained a lead for the right-wing bloc.)
How can a poll be so inaccurate?
Mitchell Barak – CEO of Keevoon Global Research, which conducts polls in 190 countries – said that it is plausible that certain media outlets want to influence results and their polls report data accordingly, but he said it is more likely that the discrepancies come down to “money and methodology.”
“Media organizations do polls for the headlines,” Barak said, meaning they don’t invest the resources and time to properly track voting patterns and trends. Moreover, he said, most media polls are conducted via the Internet or with voting panels, which tend to be less accurate than telephone interviews.
Specifically, with exit polls, Barak said the sample size is too small and too fast.
Similarly, Tamar Hermann – head of the Guttmann Center for Public Opinion and Policy Research at the Israel Democracy Institute – said that exit polls are nothing more than an answer for an impatient public.
“Professional pollsters understand that the results that come out at 10 p.m. on Election Day are more of a reality show than a serious survey or real poll.”
“People cannot live with uncertainty,” she continued. “I would even say don’t have exit polls because they cannot be accurate.”
She also said there is good reason why earlier polls might also be inaccurate, including that Israelis are last-minute and impulsive voters, who can be easily influenced by something they saw on TV on Election Day or by a conversation they had that morning. She explained that Israel’s election campaign almost entirely takes place in the last four days before voting.
“It is like measuring the wind when you don’t know which way it will blow,” she quipped.
Barak noted that it has become increasingly common for voters to answer pollsters with untruths or to present uncertainties as facts.
“It is not costly to say you are going to vote for a new, hip party or a smaller party when you are just taking a poll,” Barak said. “But when you get to the ballot box you want your vote to have influence, you want to help a small party, you prefer to vote the way you think the masses will vote – and your vote might change.”
Daniel Alexander – a professor of Mathematics at Drake University – wrote several articles on polls, specifically after the last US presidential election, which shocked the American nation, whose polls up until the election showed Hillary Clinton winning the race. He said that mathematicians call polling a black-art, which is a tongue-in-cheek way of saying it does not have the precision of pure mathematics.
“This perspective offers some insight into why polls appear divided, contradictory or even flat-out wrong,” he said.