Knesset 2026 poll tracker

How this works

Where do the numbers come from?

Seat projections are taken every hour from the English Wikipedia page Opinion polling for the 2026 Israeli legislative election (text under CC BY-SA 4.0). Polls that do not add up to 120 seats are marked and left out of the average. Every poll carries a link to where it was first published, taken from Wikipedia’s footnotes.

How is the average computed?

Polls from the last 21 days (the window widens automatically when fewer than three are available), one per pollster-and-publisher pair (the most recent), weighted by recency with a seven-day half-life: a poll from a week ago counts half as much as one from today. Lists reported below the 3.25% threshold count as zero seats, but their vote share is kept so the projection can test them against the threshold.

How are seats projected?

The average is converted to vote shares and run through the real allocation rules: the 3.25% threshold, the Bader-Ofer (D’Hondt) method over 120 seats and the surplus-vote agreements signed for this election (Together–Yisrael Beiteinu, Yashar–The Democrats, Joint List–Ra’am). Bloc totals are sums of projected seats.

Who is in which bloc?

Netanyahu bloc: Likud, Religious Zionism–Zehut, Otzma Yehudit, Shas, United Torah Judaism, Noam and Amcha Yisrael (its leader has said he would recommend only Netanyahu for prime minister).

Opposition: Together, Yashar, The Democrats, Yisrael Beiteinu, Blue and White, Reservists–New Economic Party.

In between: Ra’am sat in the 2021 coalition and may join an opposition-led government, so it is shown as a possible partner. The Joint List says it will join no government, so it counts for neither side. Lists whose intentions are unclear (such as Haredi Public) are shown as unaligned.

What do the scenario percentages mean?

Each simulated election draws a vote share for every list around today’s average: a per-list error that grows with the list’s size, correlated errors within each bloc, a national swing between the blocs, and the seven-day trend carried forward at reduced weight. Every draw is run through the seat rules above, and the scenarios count how often each way of forming a government comes out. The four scenarios are exclusive and add up to 100%; the “what if” lines are not. Each percentage is the share of simulations in which the seats would allow that path. It says nothing about whether the parties would actually negotiate it, and it measures how uncertain the polls are, not what will happen.

Where do the error sizes come from?

They are judgement-based assumptions, not a fitted model: a floor of one seat per list plus 8% of its projected size, a within-bloc correlation of 0.3, a national swing of 1.5 points, and the seven-day trend at half weight, capped at three seats. They were chosen to sit in the range of recent Israeli polling misses and are published in full below so you can judge them. An empirical calibration against the final polls and results of the 2019–2022 elections is planned; until then, read the percentages as a sensitivity analysis rather than calibrated odds.

How exactly do seats become vote shares?

Polls report seats, not percentages, so the average has to be turned back into shares before the seat rules can run. A list counts as below the threshold when it is reported under 3.25% in more than half of the weighted polls; it then keeps its reported share. Every other list gets a share proportional to its average seats within the remaining 100%. “Others” are not modelled, and when polls disagree on whether a list passes, the polls that put it below count as zero seats in its average. Because the allocation is proportional, scaling all shares together does not move any seat; what matters is which lists sit near 3.25%, and those are shown with their share rather than a seat count.

Can I exclude a publisher?

Yes. Open “Which polls count” under the poll average and switch any publisher off. The switch works per publisher, the outlet that commissioned the poll, because that is how Wikipedia attributes polls; several pollsters work for the same outlet. All publishers are included by default and none is singled out.

What about polls before 4 September 2026?

Earlier polls used a different party line-up. Columns for lists that later merged are folded into today’s lists (for example Yesh Atid plus Bennett 2026 as “Together” before 26 April); the trend charts start on 26 April 2026 when Together formed. The “All polls” tab can show those earlier periods on request.

What happens on election night?

From the close of polls the page switches to the live count from the Central Elections Committee, refreshed every couple of minutes: seats by the Bader-Ofer rules on the votes counted so far, how many seats could still move, which lists sit near the threshold, and the exit polls for comparison.

Independent project by Yaroslav Mar. No political affiliation. Polls are snapshots, not predictions.