Aryeh Leib Thorne Toren

← Israel 2026 · עברית

Strategic voting and the 3.25% threshold

Analysis. Neutral and non-partisan; no list is endorsed.

Four seats or none

Israel elects its 26th Knesset on Tuesday 27 October 2026. Every list needs 3.25% of the valid votes to win any seat. Because 3.25% of 120 seats is 3.9 seats, the smallest list that passes takes roughly four seats. A list one vote under the line takes none. There is no middle outcome, and that cliff shapes how both voters and parties behave in the final two weeks.

What 3.25% means in votes

The formula is simple: T = 0.0325 × V_valid, where V_valid = E × t × (1 − r). E is the number of eligible voters, t is turnout and r is the share of invalid ballots.

The Central Elections Committee register that took effect on 6 September 2026 lists E = 7,331,981 eligible voters (as reported by the Israel Democracy Institute; single source, not yet confirmed). In 2022 turnout was 70.63% and 0.62% of ballots were invalid (CEC final results for the 25th Knesset). Using r = 0.62%, these are estimates, not forecasts:

For comparison, the official final count for 1 November 2022 had 4,764,742 valid votes (CEC final results). 3.25% of that is 154,854.1, so a list needed 154,855 votes to pass. The figure of 154,820 reported on 3 November 2022 came from the preliminary count of that evening (Israel National News). Higher turnout raises the bar in absolute votes, so a list that needs a given number of loyal voters needs more of them when everyone else turns out too.

The lists near the line

Poll figures are paused during Israel's legal pre-election poll blackout and will return after the polls close on Tuesday 27 October.

Where a failed list's votes go

Seats are allocated by the Bader-Ofer method, a form of D'Hondt:

  1. Remove every list below T. Let V_q be the total of the qualifying lists' votes.
  2. Quota: q = V_q / 120.
  3. Each list first gets s_i = floor(v_i / q).
  4. Remaining seats go one at a time to the list (or surplus pair) with the highest v_i / (s_i + 1).
  5. Surplus-vote pairs are treated as one list in step 4, then the pair's seats are split between them by the same rule. A pair counts only if both partners pass the threshold.

The key point is step 1. A failed list's votes are not transferred to its allies; they disappear from V_q. That lowers q for everyone, so every surviving list gains in proportion to its own size. Wasted votes are effectively redistributed to all qualifying lists in proportion to their votes, including the rival bloc's lists. A bloc that loses a partner gives a share of that partner's weight to its opponents.

A worked example (hypothetical)

Invented numbers to show the mechanism. Not a forecast and not based on any party.

Nine lists. Bloc A: A1 1,300,000; A2 750,000; A3 420,000; A4 near the line. Bloc B: B1 1,150,000; B2 700,000; B3 400,000. Others: C1 380,000; C2 200,000. A3–A4 and B2–B3 have surplus agreements.

Case 1: A4 gets 176,000. Valid votes total 5,476,000, so T = 177,970. A4 is 1,970 votes short and drops out, and the A3–A4 agreement lapses. V_q = 5,300,000, q = 44,167. Final seats: A1 30, A2 17, A3 9 (Bloc A = 56); B1 27, B2 16, B3 9 (Bloc B = 52); C1 8, C2 4.

Case 2: A4 gets 180,000. Valid votes total 5,480,000, so T = 178,100. A4 passes. V_q = 5,480,000, q = 45,667. Final seats: A1 29, A2 16, A3 9, A4 4 (Bloc A = 58); B1 26, B2 15, B3 9 (Bloc B = 50); C1 8, C2 4.

Four thousand votes, about 0.07% of the total, move the gap between the blocs from 4 seats to 8. Bloc A gains four seats through A4 but its larger lists each give one back, because q rises. The net effect is +2 for A and −2 for B. The calculation was run in a short script that follows the five steps above.

Why voters hesitate: Duverger's pressure

Duverger observed that systems with high entry costs push voters away from small parties they expect to lose. In Israel the cost is the threshold, and the pressure works through expectations.

Model a voter who prefers small list S but also likes large list L in the same bloc. Let ΔU_S be the value to this voter of S winning its seats, and P_S the probability that their one vote moves S across the line. Let ΔU_L be the value of one more seat for L, and P_L the probability their vote moves that seat. Then:

EV(vote S) = P_S × ΔU_S, and EV(vote L) = P_L × ΔU_L.

For a list far below the line, P_S is near zero, so the vote is "wasted" in expectation. For a list far above, P_S is also near zero, because S is in either way. P_S is largest exactly when S polls near 3.25%, and there ΔU_S is about four seats, not one. That produces the paradox of near-threshold lists: the closer the race at the line, the more a single vote is worth and the more frightening it feels to cast it.

Bloc leaders lean on the first half of this calculation ("don't waste your vote"). Small-list leaders lean on the second ("every vote here is worth four seats"). Both are describing the same equation, and both are partly right. Whether a voter should be persuaded depends on beliefs about P_S, which is exactly what the competing polls disagree about.

A second-order effect is coordination. If many sympathetic voters each conclude that others will defect, S falls short and their fears are confirmed. If they each conclude that others will stay, S passes. Late polls act as the signal, which is why one survey showing a list just under the line can move votes away from it.

Shlomo Filber argued on X on 12 October 2026 that right-wing votes split across six lists put about 3.5 seats at risk of being "burned" (his own count, not a named poll; single source, not yet confirmed). This is the bloc-level version of the same logic.

Recent lists that fell just short

In 2022 the Times of Israel reported that the two camps were separated by 30,293 votes, yet the incoming coalition held 64 seats to 56 (Times of Israel, 3 November 2022). Much of that gap came from the votes cast for Meretz and Balad: 289,410 on the final CEC counts (150,793 + 138,617). The Times of Israel's figure of 288,789 was based on the earlier 3 November count (150,696 + 138,093). That is the worked example played out in real numbers.

Fragmentation: the effective number of parties

The effective number of parties is N = 1 / Σ p_i², where p_i is a list's share. It can be measured on votes (N_v) or on seats (N_s).

A failed list counts in N_v but not in N_s, so the threshold always pushes N_s below N_v. In the hypothetical above, N_v is about 6.5 in both cases. When A4 passes, N_s is about 6.4. When it fails, N_s falls to about 6.0. The Knesset looks less fragmented than the electorate, and the "missing" fragmentation is the wasted vote. A large gap between N_v and N_s after 27 October 2026 would be a direct measure of how many voters went unrepresented. The tracker will publish both.

What to watch

A brief word from Parshat Noach

This Shabbat, 17 October 2026, we read Parshat Noach. The builders of Bavel say, "Come, let us build us a city and a tower, … and let us make a name for ourselves, lest we be scattered upon the face of the whole earth" (Bereishit 11:4). Their fear of scattering led them to fuse into one tower around a name. Their unity came from fear, not from shared purpose, and it did not hold. The threshold creates a similar pull toward building larger and larger lists out of fear of being scattered. Whether a merger rests on shared purpose or only on fear of the line is for each voter to judge. The arithmetic is the same either way; the meaning is not. Ultimately the outcome rests with G-d, and the vote with each of us.

Method: individual published polls only, no averages. Historical figures from the Central Elections Committee via the cited reports. Estimates are labelled as estimates.

Sources

Cited in the text without a link, and so single-source: Israel Democracy Institute (voter register); Shlomo Filber (X, 12 Oct 2026).