Whole country
Hover a list to see where its voters went; click to pin it. Width = number of people.
Cross-block moves
Which lists make up each move between blocks
Transfer table
How this is estimated
Votes are secret, so nobody knows how any individual moved. What is known is each ballot box's result in both elections. Boxes are grouped into clusters that cover the same voters in 2022 and 2026: polling places (the CEC's “ריכוז”) matched by their id, then by shared box numbers; places that can't be matched are pooled per locality, or per regional committee. Double-envelope ballots (soldiers, diplomats, hospitals, prisons) form one cluster of their own.
Only the vote files are used. Each cluster is treated as one population: in 2022 everyone voted for a list, cast an invalid ballot, or didn't vote; in 2026 the same. The files give each cluster's registered voters in both years, so only the net change is known (turned 18 or arrived, minus died or left). Where a cluster grew, the growth appears as new voters (net) on the 2022 side; where it shrank, the difference is removed evenly from its 2022 groups. People who died are therefore treated as replaced by new voters. Lists under 0.2% of the national vote are grouped as “Other”.
A model then finds flows that reproduce every cluster's results in both elections at once. The models differ in what they assume about voters inside a cluster:
Does it recover known flows?
We planted known flows in synthetic data with the same clusters and sizes, aggregated them to box level, and asked each model to recover them. Lower is better: the share of a 2022 list's voters sent to the wrong 2026 destination, averaged over lists.
Caveats
- This is ecological inference: aggregate results can't pin down individual moves, so every number here rests on the model's assumptions. Treat flows under a few thousand people as noise.
- Deaths aren't modelled: about 3% of 2022 voters died before 2026, and their places are taken by new voters (shown net). This can shift a list's row by a few points.
- People who moved home between elections are lost from one cluster and counted as new voters in another.
- Double-envelope voters are a different population in each election (most 2022 soldiers have been discharged). They also stay on their home box's roll, so their numbers are removed from home-box abstainers, in proportion to registered voters.