It is nearly two weeks after the election – and the big failure of Israeli pollsters to forecast the outcome correctly – but Mano Geva is still angry. The CEO of the Midgam polling organization is not mad at those who refused to participate in the exit polls, nor those who tried to skew the results. Mostly, he is mad at the media.
- Israeli media asking itself how it got the election so wrong
- Why did pollsters get the Israeli election so wrong?
- Election results by city: Likud won in eight out of the 10 largest cities
- How did Netanyahu score such a decisive election victory?
- The six big surprises of the 2015 Israeli election
“The media did a great injustice to the polls and pollsters,” says Geva, who, along with Mina Tzemach, conducted the polls and election night forecasts for Channel 2. “All in all, maybe it was easy for the press to take all the pollsters and blame them for everything that happened. There really was no real reason for it. It was a prediction, after all,” says Geva.
There are two things that are beyond doubt: The first is that the predictions throughout the entire election campaign gave a slight advantage to the Zionist Union over Likud (or, in some exceptional cases, a tie). The last polls published by the press, four days before the election, indicated a four-seat gap in Zionist Union’s favor.
The second is that the exit polls broadcast on election night missed Likud’s big victory, and instead showed a tie – or at most a small advantage to the ruling party. Some called it the “Yom Kippur of the pollsters” [a reference to the 1973 war when Israel was caught unawares by invading neighboring countries, ignoring all the warning signs].
All the pollsters remain firm in their belief that the polls they conducted before the election reflected the reality at that moment. So how can they explain the big gap between their polls and the final result?
The polls not only provided a forecast, they also shaped reality. The pollsters explain that they strengthened Prime Minister Benjamin Netanyahu’s claim that right-wing rule was in jeopardy, and helped contribute to a major swing of votes to his Likud party. This is how the party climbed from 20-22 Knesset seats four days before the election to 30 seats at the moment of truth.
Impossible to predict
Tal Galili agrees. A doctoral student at Tel Aviv University, Galili set up a website, “Bad Poll?” (Seker Ra?), to analyze that very question. “I took the median of the number of seats for every party in the polls that were published two weeks before the election. This number matched almost completely [the results] for every party,” he explains.
Only in two cases were the polls off, notes Galili. “One case was [Eli Yishai’s] Yahad, which received four seats in polls in the two weeks before the election, but did not pass the electoral threshold. That was not a big mistake,” he says.
“The second problem was the big inflation of seats for Likud and the big fall of Habayit Hayehudi,” he adds. “It was possible to start to see the trend in the polls. But if I had plotted a straight line for the phenomenon, I would not have forecast what happened; I would have forecast something more modest. This was a surprise.
“It seems very reasonable that the reason for this surprise was not that there were a lot of Likud voters whom the polls did not find. The reasonable explanation is that Netanyahu had a very effective campaign. This was something that was impossible to forecast,” admits Galili.
The biggest furore concerned the television exit polls. The three forecasts predicted the election results almost perfectly – except for the two largest parties. On Election Day, immediately after the polls closed, Channel 2 aired Tzemach and Geva’s exit poll: 28 seats for Likud; 27 for Zionist Union. True, a small advantage to Likud, but far from the six-seat difference between the parties in the final analysis. The exit polls for Channel 1 and Channel 10 both showed a 27-seat tie between the two parties.
The exit polls had left Zionist Union with some hope. Some left wingers already started sketching out a government headed by chairman Isaac Herzog. Even President Reuven Rivlin rushed to call for a unity government. But the picture changed completely overnight. “They were excellent forecasts, except for this one big thing – the jump in seats for Likud and the drop for Zionist Union,” says Galili. “In statistical terms this was a mistake, but [otherwise] it went quite well for them. In political terms, though, it was a catastrophe.”
“We gave an advantage to Likud. One seat, true, but an advantage,” observes Geva. “The entire [TV] broadcast was built on that. The rest of the forecast is amazingly accurate.”
The exit polls that ran on television included voting up until 8:30 P.M., while the polls closed at 10 P.M. At that point, the results reflected 28 seats for Likud and 26 for Zionist Union, explains Geva. They then decided to be cautious and gave an overall advantage of just one seat to Likud, reporting it as 28-27.
Geva says they were cautious because of the uncertainty and also the low level of participation [people revealing how they voted] in the exit polls, which fell from more than 90 percent to 75 percent, which is significant and influenced the results of their polls.
The exit poll results were actually quite close to the real thing, so there was no problem in the choice of sample polling stations or the statistical model, say pollsters. There were two main reasons that led to the mistakes: Between 8:30 P.M. and 10 P.M., there was a big turnout for Likud; and many Likud voters refused to participate in the exit polls.
“The main recommendation is that we need to provide the forecast in a number of stages,” says Geva.
The first stage should be at 10 P.M., but to be more cautious about late developments before the polls closed. And then, later, to provide a forecast based on exit polls until 10 P.M. Finally, another forecast could be made based on the real results in certain sample polling districts, he believes.
Prof. Camil Fuchs, who conducted the polls and election night forecasts for Channel 10 (and for Haaretz during the election campaign), adds another reason for the errors: Voters who lied and said they voted for Zionist Union – but didn’t.
His conclusions are similar to those of Geva: “I don’t think we can give up on the television exit polls at 10 P.M.” But he too thinks the forecast needs to come in stages: an initial forecast, a later version and then one based on the true results in sample polling places – which is still a forecast, since that, too, can change a bit.