Having heard the results of the exit polls for Tuesday’s elections, Israelis went to bed thinking that the Likud and Zionist Union were neck and neck. Two TV exit polls (on Channels 1 and 10) showed Prime Minister Benjamin Netanyahu’s Likud and Isaac Herzog’s Zionist Union, the country’s two largest parties, tied at 27 Knesset seats each. The third (on Channel 2) showed the Likud leading the Zionist Union by one seat, 28-27.
Either way it was close, and though Netanyahu was clearly better placed to form a government, the leaders of both parties were said to already be reaching out to smaller parties in bids to form ruling coalitions.
It was a whole other story, however, when citizens awoke on Wednesday to learn that with 99 percent of the ballots counted, the Likud had actually won 30 seats and the Zionist Union only 24. Herzog, who had promised hope and change, conceded defeat and Netanyahu was on his way to forming a right-wing government and another premiership.
Zionist Union campaign strategists were not the only ones being asked what had gone wrong. Questions were being raised as to how the exit polls — announced with great fanfare and little hint of potential fallibility — could have been so far off in predicting the actual six-seat spread between Likud and Zionist Union.
The pollsters offered answers, but it seemed they themselves weren’t completely sure what had happened.
Professor Camil Fuchs, a Tel Aviv University statistics expert who conducted the Channel 10 exit poll, told Army Radio (Hebrew link) that the discrepancy could be attributed to the fact that a larger number of voters refused to participate in his exit poll than in past elections.
Mina Tzemach, who conducted Channel 2’s exit poll together with her partner Mano Geva, insisted her statistical model was good. She also pointed to lower than usual participation in her exit poll as the possible problem.
“Usually we have 7 or 8 percent of voters who don’t agree to participate, but this time is was more like 15 percent,” she told Army Radio.
Tzemach said this lack of participation was greatest at polling places that were known to have many pro-Likud voters, and in particular immigrants from the former Soviet Union countries who are reluctant to take part in exit polls. This under-representation of Likud voters must have skewed the results, she said.
Channel 1 pollster Stella Karayov dismissed a suggestion that exit poll participants could have been untruthful about whom they had actually voted for. According to her, this kind of behavior would have affected results for not only the large parties, but also the smaller ones — whose exit poll numbers turned out to be very close to actual vote counts.
Channels 1 and 2 used similar methods in their exit polls. Both asked voters who had just finished dropping their actual ballots into official ballot boxes to vote again in similar fashion for the exit poll.
Channel 2 set up 60 ballot boxes at polling stations around the country to gather information from 25,000 voters this way. The polling stations were chosen based on socio-economic criteria, as well as on information gleaned from the two previous elections as to the reliability of exit polling at these particular locations. Channel 1 used the same number of ballot boxes, but at polling places that served a total of 30,000 voters.
Instead of using ballot boxes, Channel 10 asked people to stop and fill out a survey after voting. Fuchs told Haaretz (Hebrew link) that he suspected that the lower participation rate that this approach generates (usually around 80 percent) is due to the lack of anonymity. At the same time, however, he said the methodology was preferably in terms of speed and logistical simplicity.
Professor Tamar Hermann, academic director of the Israel Democracy Institute’s Guttman Center for Surveys, does not give much credence to exit polls, no matter how they are conducted, she told the Times of Israel. She prefers to look at the information that long-term surveys provide.
In a report she wrote for the Peace Index survey just a week before Election Day, Hermann predicted that the right bloc would win based on a lack of change in public attitudes in the months since the elections were called late in 2014.
“The results of the present Peace Index survey show that the events of the recent weeks and days had no influence on the Jewish public’s assessment of which bloc has a greater chance to form the next government. Two weeks before the elections, 61% believe the right-wing bloc’s chances are greater (compared to 59% who thought so in January and 60% in December),” she wrote.
Echoing comments by the television pollsters, Hermann said that certain subsections are indeed harder to survey. She confirmed that right-wing voters do tend to participate less in exit polls, as they consider them to be associated with what they perceive as a left-leaning media.
In other types of polls and surveys, she said other groups, such as young people, are harder to reach because methodologies such as calling people at home on telephone landlines are becoming less useful and reliable in the digital age.
The real problem with exit polls may not lie with how voters act as they are surveyed as they exit the polling stations, but instead with how they behave at home after they have voted.
“It’s a problem of impatience and greed,” Hermann said of the inaccuracy of exit polls.
There’s the matter of the pollster’s commercial interests, but also of the voters’ desire to know the elections results the minute the polling places close at 10 p.m.
“We act as though it would really hurt to wait 10 hours to get the results,” Hermann said.
One Twitter user agreed, tweeting, “If the elections are over at 10 p.m. and we have final results at 6 a.m., then what do we need an exit poll for?”
Less patient Israelis (which would probably be the majority, if we were to do a poll) would rather keep the pressure on the polling professionals.
“If I had done a survey as I walked around on [Tel Aviv’s] Gordon Beach, I’d still have come out with a more reliable poll,” tweeted one.