# Israel's Pollsters: Who Calls the Race Before You Vote

My Weird Prompts — an AI-generated podcast. Daniel Rosehill sends in a prompt; AI hosts Corn and Herman discuss it. Verify important claims independently.

- Episode: 6075
- Published: 2026-10-18
- Duration: 21:02
- Audio: https://episodes.myweirdprompts.com/audio/israel-election-pollsters-guide-71772.m4a
- Tags: israel, political-history, taxonomy
- Web page: https://myweirdprompts.com/episode/israel-election-pollsters-guide-71772/
- JSON: https://myweirdprompts.com/api/episodes/israel-election-pollsters-guide-71772.json

## Summary

Israel votes in nine days, and the people who try to call the result before it happens are a small, famous, and fiercely debated cast. This episode maps the polling culture from Louis Guttman's late-1940s research institute through Mina Tzemach, Camil Fuchs, and the pollsters behind every major channel today — including the Channel 14 polls that the rest of the industry watches nervously. It also revisits the nights that made and broke reputations: 1977's Mahapach, the 1996 "went to sleep with Peres, woke up with Bibi" reversal, and 2015's six-seat miss. Along the way: why Israeli polls have consistently underestimated the right, what the refusal rate at exit polls actually tells us, and why a 500-person sample with a four-point margin of error gets argued over seat by seat. With the legal publication blackout about to fall, this is a look at who's on the hook when the numbers land.

## The prompt

> Ahead of the Israeli elections, a deep dive on the main experts who (attempt to) predict election results in Israel. Who are they? Which pollsters are most influential? What predictions have already landed? And when does exit polling come in? Not just who's predicting what, but what the culture of polling actually looks like here and who has influenced it.

## Show notes

Israel polls itself more than almost any democracy, and then spends election night furious that the polls were wrong. That contradiction is the subject of this episode, recorded nine days before Israelis vote.

The culture has deep roots. Louis Guttman set up what became the Israel Institute of Applied Social Research in the late 1940s, generally described as the first public opinion research body in the Middle East. Hanoch Smith was polling on Israel Radio in the 1950s, founded Smith Research in the early 1970s, and ran the first exit poll ever broadcast on Israeli national television. The night that made pollsters famous was 1977 — the Mahapach — when Labor's long rule ended and Menachem Begin's Likud won. Two pollsters, Smith and Mina Tzemach, each claimed that night as their own.

Tzemach became the defining figure: Hebrew University, a Yale doctorate, running the Dahaf institute from 1980, replacing Smith as the national broadcaster's forecaster, and then serving as chief pollster for Channel Two and Yedioth Ahronoth for roughly thirty years. Camil Fuchs, a Tel Aviv University statistics professor, was Channel Thirteen's man on election night for about two decades; his final poll in 2022 had Likud at 31 when it won 32. He died in April 2024.

The misses matter as much as the hits. In 1996, the first direct election for prime minister, exit polls called it for Shimon Peres and coverage effectively declared it around midnight — then the real count turned, and Netanyahu won by under one percent, about thirty thousand votes. The phrase "went to sleep with Peres, woke up with Bibi" comes from that night. In 2015, three channels made the same mistake simultaneously: two showed Likud and the Zionist Union tied, one had Likud up by one. The result was 30 to 24, a six-seat gap nobody showed.

The pattern worth carrying into election night: Israeli polls have tended to underestimate the right — in 2015, April 2019, and even 2022, when final polls put Netanyahu's bloc at 61 or 62 and it got 64. The explanations offered are specific. In 2015, Tzemach said exit-poll refusals jumped to about fifteen percent, double the normal rate, concentrated at Likud-leaning stations including ones with many former Soviet immigrants. Rafi Smith observed that respondents treat pollsters as the establishment. In April 2019, after Channel Twelve's exit poll showed Blue and White ahead 37 to 33 and the actual result was 35 each, Tzemach said on the record that people lied in the sample to screw over the channel. She had been rated the most accurate pollster across three straight elections by an aggregation project called Project 61, and afterward she stepped away from election polling for Channel Twelve.

The current cast is small. Channel Twelve uses Midgam, with an online panel; Kan and Israel Hayom use Kantar; Maariv and Walla use Lazar Research with Panel4All; Channel Thirteen has a consortium; Channel 16 uses Maagar Mochot; i24 uses Direct Polls. Then there's Channel 14, whose polls are run by Shlomo Filber — former director general of the Communications Ministry under Netanyahu and a state's witness in Case 4000. A Channel 14 poll from October 12 had Likud at 29 and the outgoing coalition at exactly 61, while mainstream polls cluster the coalition around 47 to 49. That's a house effect of roughly twelve seats. In 2022 Channel 14 was genuinely close and now markets itself as the only firm that got it right; a Maariv columnist asked this month whether Filber will bury a generation of pollsters the way Tzemach did in 1977.

Underneath it all is a statistical problem the country largely ignores. A typical Channel Twelve poll is about 500 people with a margin of error around four points; a Knesset seat is roughly 0.8 percent of the vote. So the margin of error spans about five seats, and the entire evening is spent arguing over one. Israelis consume polls as seat counts and bloc totals rather than vote shares, because 61 is the only number that matters. Parties, meanwhile, treat internal polls as ammunition: in 2019 Likud was citing internal polls of 35 or 36 seats while published ones had 33.

The law is stricter than most people assume. The Elections Law on propaganda methods bans publishing any new poll from the end of the Friday before election day until polls close — four days of silence. Every published poll must disclose who paid for it, who conducted it, the dates, the sample, how many people answered, and the margin of error, and the pollster must file results and methodology with the Central Elections Committee, which posts them online. The theoretical penalty is up to six months in prison; the famous practical case is September 2019, when the committee chairman asked Facebook to suspend a chatbot on Netanyahu's page for publishing polls on election day. It was back about two hours later after a Likud MK promised it would stop. This August, Haaretz reported that Channel 14 had not submitted data for three of its published polls to the committee.

## Transcript

Corn: Nine days before Israel votes, Daniel wants to know who the people are that try to tell the country the result before it happens.
Corn: Not the numbers this time. The people. Which pollsters actually matter here, who shaped the trade, what they've already got right this cycle, and when the exit polls land on the night. And underneath that, what the culture of polling in this country actually looks like, because he suspects it's its own peculiar thing.
Herman: It is. It is very much its own thing. Israel polls itself more than almost any democracy I can think of, and then spends election night furious that the polls were wrong.
Corn: We put our own forecast down the other day, with three methods and a seat count we're now stuck with. So I'll admit I have a personal interest in who else is on the hook.
Herman: You have a personal interest in who gets blamed when we're wrong.
Corn: Same thing, Herman.
Herman: Fine. History first, because the culture makes no sense without it. Israel had public opinion research almost from the founding. Louis Guttman, the statistician, set up what became the Israel Institute of Applied Social Research in the late forties, and it's usually described as the first public opinion research body in the Middle East. Early on, people met surveys with a mixture of suspicion and resignation, which, honestly, is still roughly how Israelis answer the phone.
Corn: Suspicion and resignation is how Israelis answer everything.
Herman: Then there's Hanoch Smith. Herbie Smith. He was doing polls on Israel Radio in the fifties, founded Smith Research in the early seventies, and he ran the first exit poll ever broadcast on Israeli national television. And the night that made pollsters famous here was nineteen seventy-seven. The Mahapach. Labor had governed since the state began, and Menachem Begin's Likud won.
Corn: Let me guess. Somebody called it.
Herman: Two somebodies, which is the beautiful part. Smith said, for years afterwards, that he was the only pollster who predicted it. And Mina Tzemach's reputation also rests on that same election night, when her numbers showed the upset. So the founding myth of Israeli polling has two people standing on the same patch of ground, each holding the flag.
Corn: Very Israeli. The foundational event and nobody agrees who was there.
Herman: Tzemach is the figure, though. Hebrew University, then a doctorate from Yale in the sixties. She ran the Dahaf institute from nineteen eighty, she replaced Smith as the national broadcaster's election forecaster, and then she became the chief pollster for Channel Two and for Yedioth Ahronoth. For about thirty years, if you watched an Israeli election, the woman telling you the result at ten o'clock was Mina Tzemach.
Corn: The queen of the polls.
Herman: That's what the papers called her. And in two thousand thirteen she became a partner in Midgam, which Mano Geva had founded in nineteen eighty-eight. Midgam is still Channel Twelve's pollster today.
Corn: Every monarchy has a rival court. I'm guessing this one did too.
Herman: Camil Fuchs. Statistics professor at Tel Aviv University, chaired the department, polled for Haaretz and Channel Ten, and then for about twenty years he was Channel Thirteen's man on election night. He died in April of two thousand twenty-four, at seventy-eight. And his last election was a good one. His final poll in twenty twenty-two had Likud at thirty-one. Likud got thirty-two.
Corn: One seat off and he'd still have been annoyed about it.
Herman: Oh, he would have been. These are statisticians who became television characters by accident. That's a big part of the culture, Corn. The pollster here is a personality. Tzemach, Fuchs, Geva. People know their faces.
Larry: Tired of waiting four years to find out what the country thinks? Introducing Poll of the Month Club, the subscription box that delivers a fresh, hand-weighted poll to your door every single month! Bronze members receive one poll that agrees with you. Gold members get a personal exit poll, read aloud at ten o'clock every night, right in the bedroom. Platinum members are guaranteed a sixty-one seat majority, or your margin of error back! To cancel, simply appear in person at an undisclosed polling station, after sundown on a Friday, during the legal publication blackout. Late cancellations are tallied as double envelopes and counted next week. Poll of the Month Club. Because you deserve to win, monthly.
Herman: I have the Gold tier. It woke me up at ten every night for a month and told me I'd won, and I'd only ever voted in the building committee.
Corn: The misses are the part I actually wanted to get to.
Herman: Nineteen ninety-six is the big one. First direct election for prime minister. Peres against Netanyahu. The exit polls called it for Peres, and the television coverage more or less declared it around midnight. Then, around two in the morning, the real count started turning. Netanyahu won by about thirty thousand votes. Under one percent.
Corn: Went to sleep with Peres, woke up with Bibi.
Herman: That's the phrase, and it comes from exactly that night. It's become a sort of national proverb for the polling trade. And then there's twenty fifteen, which in some ways was worse, because three different channels made the same mistake at the same moment. Channel One had Likud and the Zionist Union tied. Channel Ten, with Fuchs, had them tied. Channel Two, with Tzemach and Geva, had Likud up by one.
Corn: Let me guess the real result. Not a tie.
Herman: Likud thirty, Zionist Union twenty-four. A six seat gap that nobody showed.
Corn: Hm. So the error was all in one direction.
Herman: It usually is, and this is the pattern I'd want any listener to carry into next Tuesday. Israeli polls have tended to underestimate the right. Twenty fifteen, April twenty nineteen, and even twenty twenty-two, which most people remember as a good year for the pollsters. Every major channel's final poll had Netanyahu's bloc at sixty-one or sixty-two. It got sixty-four.
Corn: What do they say when it happens? Because I assume nobody stands up and says, we're bad at this.
Herman: The explanations are actually quite specific. In twenty fifteen, Tzemach said refusals at the exit polls jumped to about fifteen percent, roughly double the normal rate, and they were highest at the stations that leaned Likud, including places with a lot of former Soviet immigrants. People just wouldn't talk to the pollster.
Corn: Which makes some sense. The guy with a clipboard outside the school doesn't look like your friend.
Herman: Rafi Smith, Hanoch's son, put it really well. He said respondents treat pollsters as the establishment, even though we're not. And in April twenty nineteen Tzemach went further. Channel Twelve's exit poll had Blue and White up thirty-seven to thirty-three. The result was thirty-five each. And she said, on the record, that people lied to us in the sample, and that the aim was to screw over Channel Twelve.
Corn: That's a remarkable thing for a pollster to say. Your data is fine, it's the public who are wrong.
Herman: She'd earned a bit of bitterness. She'd been rated the most accurate pollster across three straight elections by an aggregation project called Project Sixty-One. And after that night she stepped away from election polling for Channel Twelve.
Corn: You think she was right, though? That right-wing voters were deliberately throwing it?
Herman: Hm. I think... partly. I think the refusal story is better supported than the lying story. Lying is a strong claim. Refusing is easy to measure. What do you think?
Corn: I think if you spend ten years telling a third of the country it's a backward block, some of them will stop doing you favours on the pavement. I don't think you need a conspiracy for that. You need a grudge.
Herman: That's not a terrible theory, and it's roughly what Smith was saying.
Corn: My phone shows me a different poll every morning this month, and I've stopped being able to tell whose is whose.
Herman: You do, and they come from a fairly small cast. Channel Twelve uses Midgam, with an online panel. Kan, the public broadcaster, and Israel Hayom both use Kantar. Maariv and Walla use Lazar Research, run by Menachem Lazar, with Panel Four All. Channel Thirteen has a consortium now. Channel Sixteen uses Maagar Mochot. i Twenty Four uses Direct Polls. And then Channel Fourteen.
Corn: The one that makes everyone else nervous.
Herman: Channel Fourteen's polls are run by Shlomo Filber. And Filber's biography is the whole story. He was director general of the Communications Ministry under Netanyahu, he became a state's witness in Case Four Thousand, and he founded Direct Polls. The Channel Fourteen poll from the twelfth of October had Likud at twenty-nine and the outgoing coalition at exactly sixty-one. The mainstream polls have the coalition clustered around forty-seven to forty-nine.
Corn: That's not a house effect. That's a separate house.
Herman: Well... it's a house effect of about twelve seats, which we weighed up in our own forecast, so I won't relitigate it. The interesting thing is the culture around it. In twenty twenty-two Channel Fourteen was genuinely close, and it markets itself as the only firm that got it right. A Maariv columnist asked this month whether Filber will bury a generation of pollsters the way Tzemach did in seventy-seven.
Corn: So Tuesday decides whether the partisan pollster becomes the new establishment. That's a better story than most of the campaign.
Herman: It might be the best story of the campaign.
Corn: Here's the thing that's always bothered me. Channel Twelve's last poll was about five hundred people with a margin of error around four points. A Knesset seat is less than one percent of the vote.
Herman: Point eight, roughly.
Corn: So the margin of error is about five seats, and we spend the whole evening arguing about one.
Herman: That's the national sport, Corn. The seat count is how Israelis consume polls. Nobody reports vote share. They report seats, and then immediately they add up blocs, because the only number that really matters is sixty-one. So a poll here isn't read as Likud on twenty percent, it's read as Netanyahu is two seats short, and that becomes the news cycle for the whole day.
Corn: Meanwhile every party has a poll in its pocket that says it's doing brilliantly.
Herman: Constantly. In twenty nineteen Likud was citing internal polls of thirty-five or thirty-six seats while the published ones had thirty-three. Internal polls in Israel are less research than ammunition. You leak the one that keeps your voters from drifting.
Corn: What about the law? Because I know there's a blackout.
Herman: There is, and it's stricter than people think. The Elections Law on propaganda methods bans publishing any new poll from the end of the Friday before election day until the polls close. So the Friday papers carry the last polls, and then four days of silence. Every published poll has to say who paid for it, who conducted it, the dates, the sample, how many people answered, and the margin of error. And the pollster has to file the results and the methodology with the Central Elections Committee, which puts them online.
Corn: I'd love to know what actually happens to someone who breaks it.
Herman: Up to six months in prison, in theory. In practice the famous case is September twenty nineteen, on election day, when the committee chairman asked Facebook to suspend a chatbot on Netanyahu's page because it was publishing polls. It came back about two hours later after a Likud MK promised it would stop.
Corn: Only in this country is the scandal a chatbot leaking a poll on election day. And it gets a two hour suspension.
Herman: And it's not purely historic. This August, Haaretz reported that Channel Fourteen hadn't submitted the data for three of its published polls to the committee.
Corn: Fine. What has already landed this cycle? Things people predicted that we can now mark.
Herman: A few. The first is that this election is happening on time, at the latest legal date, the twenty-seventh of October. A lot of people predicted an early election for most of the term. Shas and United Torah Judaism wanted September. It didn't happen.
Corn: Prediction one, wrong.
Herman: The second is the merger rule. There's an old Israeli pattern that two parties merged poll worse than their separate totals. Bennett and Lapid formed Together in April with Bennett at the top, and since then it's slid. It now polls somewhere between seven and thirteen seats.
Corn: The merger rule holds. And the third?
Herman: The new party rule, which runs both ways and confuses everyone. Brand new parties that are fashionable get overestimated, like Kadima, which was polling around forty early on and got twenty-nine. But new parties nobody takes seriously get underestimated. In two thousand six the pensioners' party was polling two seats or less and won seven. In twenty thirteen, Yesh Atid polled eight to eleven and won nineteen.
Corn: So the polls are good at telling you a new party exists and bad at telling you whether anyone means it.
Herman: That's... yes. One more piece on that. Project Sixty-One, the aggregation effort I mentioned, went back over years of Israeli polls and found they overstate the big parties and understate the Haredi and Arab parties. Which party do you think is most at risk of being undercounted this time?
Corn: Shas, if I had to bet. Its voters aren't sitting on online panels answering surveys between meetings.
Herman: That would be my guess too, and Channel Fourteen is the only one with Shas on ten. Everyone else has it on seven or eight. There's a newer aggregator as well, a poll of polls the Middle East Forum launched in September, which takes the latest poll from each house and deliberately doesn't correct for house effects. You see the raw spread.
Corn: An aggregator that refuses to tidy up. I respect that more than I expected to.
Herman: Meanwhile, in this cycle, Blue and White, Gantz's party, has been polling under the threshold in every recent survey, somewhere around one percent. If that holds on Tuesday, it's a party that came within sight of the prime minister's office five years ago going to zero.
Corn: Exit polls. Daniel asked when they come in.
Herman: Polls open at seven in the morning and close at ten at night. At ten, almost on the second, Channels Eleven, Twelve and Thirteen put up their exit polls. That moment is a genuine national ritual. Families gather for it the way other countries gather for a football final.
Corn: Ten o'clock is the easy part. It's what happens after that I've never understood.
Herman: The Central Elections Committee starts posting results by polling station from about one in the morning. By around eight the next morning about ninety percent is counted. Then the double envelopes, which are soldiers, diplomats, prisoners and hospital patients, checked against their home station and counted afterwards. In twenty twenty-two there were around half a million of them. This time they're due on the twenty-ninth, and the official result on the fourth of November.
Corn: Half a million ballots counted two days late sounds like it should matter.
Herman: They can move the thing that matters most, which is who gets over the threshold. In twenty twenty-one the exit polls had Ra'am under the line. Overnight it sat at about three point two four percent, a hair under the threshold, and it ended up with five seats. That one result erased the path to sixty-one that the exit polls had implied for Netanyahu's bloc.
Corn: So the ten o'clock number decides the mood and the morning decides the government.
Herman: And the double envelopes decide whether anyone's allowed to stop arguing.
Hilbert: Herman, what does a pollster do when someone wants to be in the sample? Not refuses. Wants it. Badly.
Herman: Hm. That's a real problem, actually. Online panels have a version of it. The people who sign up are keener than the people who don't. You weight for it.
Hilbert: I'm asking because of Avner Shushan. He runs the juice stand under my flat. Three weeks ago he heard me say the word poll on the phone and he asked to be in one. I said I don't run polls. He said that's fine, he'd pay. Eleven shekels a call. So now I call him every evening at seven and ask him who he's voting for.
Corn: Every evening.
Hilbert: He changes his answer every evening. That's not the problem. The problem is that last Thursday he started telling me he's voting for a list that doesn't exist. The Pomegranate Movement. He's very committed to it. He's printed a ballot slip. Then his cousins heard he was in a sample and they want to be polled as well, and now I call forty-one people a night, at eleven shekels each, on the telephone in my bathroom, because that's where the line comes into the flat. It's a grey Bezeq set with a rotary dial, plugged in under the sink.
Hilbert: Herman, you're too close to the microphone, move back about two fingers... and the handset doesn't sit on a snout, so I hold it against the side of my head like a shell. Avner now insists on a margin of error. I work one out and I write it on the back of his receipt. Plus or minus three cousins. Last night one of the cousins answered and it was his goat, which they keep on the roof in Nachlaot, and Avner said the goat counts because it's registered at the same address.
Herman: That's a weighting question, honestly. If the goat's in the household roster, you've got a design effect.
Hilbert: The goat is for the Pomegranate Movement. Forty-one people and a goat, and I have the Pomegranate Movement on thirty-eight percent in my sample. It's... you know, it's the most interesting thing I've done all year. Watching a number turn into something people believe in. They talk about it at the stand now. People queue for juice and ask what the latest is.
Corn: You've built a party with a polling lead and no candidates.
Hilbert: It has candidates. Avner's brother-in-law is number two. And Herman, for the record, it wasn't about thirty thousand votes in ninety-six. It was twenty-nine thousand, four hundred and fifty-seven.
Herman: Twenty-nine thousand four hundred and fifty-seven. Fine. I'll take it.
Hilbert: Avner pays on the first of the month, in coins, in a jam jar. I did the census in nineteen eighty-three, for about a fortnight, in Beit Shemesh, and nobody ever paid me. Avner answers on the first ring, every night, and he has never once told me the truth, and I'd trust him with my keys.
Corn: There's a version of that in every published poll, though. The people who answer on the first ring.
Herman: That's the whole difficulty with panels. The people who want to be in the sample are not the people who don't.
Herman: The bit that didn't fit anywhere. In two thousand six, in the Kadima leadership primary, an exit poll was published half an hour before voting ended, showing Tzipi Livni with a big lead. She won by four hundred and thirty-one votes.
Corn: An exit poll that arrived early enough to be campaign material.
Herman: Which is exactly why the law now holds everything until ten.
Corn: Where I come out is this. Tuesday night at ten, three channels will tell you a number, and at least one of them is going to be wrong in a way you'll hear about for a decade. What I'll actually be watching is the people who wouldn't talk to the man with the clipboard. Or the man on the bathroom phone.
Herman: And who shows up after midnight in the double envelopes. Thanks, as always, to Hilbert Flumingtop for producing this one.
Corn: This has been My Weird Prompts. If you've got a minute, leave us a review wherever you listen.
Herman: See you soon.