Why we will not print a vote share

The most common question about this forecast is why it reports seats and never vote percentages. The answer is that almost nobody publishes the percentages.

Israeli pollsters release projected mandates. Of the 13,898 poll observations in our corpus, the only vote percentages we hold are for parties reported as falling below the threshold — the (2.4%) form that says both “no seats” and “this is what they polled”. For every party above the line, the published figure is a seat count that already has the pollster’s own threshold assumption and allocation baked into it.

What inverting it would cost

You can, in principle, work backwards from seats to a vote share. You would be inventing the thing that matters most. Bader–Ofer allocation is a step function: near the 3.25% threshold, a hundredth of a point of vote share is the difference between four seats and none, and the same seat count is consistent with a wide band of vote shares. Inverting the mapping puts a number on the page that looks precise and is mostly an artefact of the assumption used to invert it.

April 2019 makes the point concretely. The New Right finished 1,453 votes below the threshold out of 4.3 million cast. No seat-space model can see a gap that size, and one that claimed to would be lying about its resolution.

What would change our mind

One source publishes the percentages, because the law compels it: polls deposited with the Central Elections Committee under section 16E(e) carry the methodology and the raw shares behind the projection. We read those for 2026. The registry does not exist for earlier cycles on the live site — we checked each one — though the filings for Knessets 21 through 24 survive in the Internet Archive, and there are 218 of them.

If we can measure the yield of that archive, a vote-share track becomes possible with something to calibrate it against. Until then, seats.