Data and downloads

Every serving table as CSV (UTF-8 with a BOM, so Excel shows Hebrew) and Parquet, a data dictionary, and a zip of the lot.

All tables as a zip

Data version: df9ffc34cb92

Licences and credits

Central Elections Committee, CBS, Population Authority, data.gov.il, bechirot.fyi, English Wikipedia (seats in the 18th–20th Knessets).

Reuse terms for some sources are still being confirmed. Kalpi is a working name.

Conventions

Percentages from 0 to 100.
UTF-8 with byte-order mark, comma-separated, one header row; list columns are joined with ' | '.
External envelopes (soldiers, prisoners, diplomats and others) belong to no locality; locality tables exclude them.

Data dictionary

`elections`

One row per Knesset election: date, threshold, official totals and where they came from.

Grain: election · Rows: 9

Source: bechirot.fyi (K21-K25, citing the Central Elections Committee); English Wikipedia result boxes for K18-K20 totals and seats; K26 from the CEC media26 precinct file (provisional) and Wikipedia seat allocation; checked against the sum of the CEC ballot-box files

Variable	Type	Note
`knesset`	BIGINT	Knesset number (18 to 26).
`source_url`	VARCHAR	Page the list names, votes, threshold and (K21+) seats were read from.
`election_date`	DATE	Date of the election.
`threshold`	DOUBLE	Electoral threshold in percent of valid votes.
`official_valid_votes`	BIGINT	Official count of valid votes in the whole election, including external envelopes (soldiers, prisoners, diplomats and others).
`official_turnout`	DOUBLE	Official national turnout in percent, as published (one or two decimals).
`totals_source`	VARCHAR	Where the official totals were read from.
`lists_passed`	BIGINT	Number of lists that passed the threshold.
`lists_total`	BIGINT	Number of lists that ran.
`seats_source`	VARCHAR	Where the seat counts were read from.
`status`	VARCHAR	final or provisional. Provisional means the CEC precinct file is loaded but the Gazette lock (due 4 November 2026 for K26) has not been applied.
`retrieved_at`	VARCHAR	Date the source pages were fetched.

`lists`

One row per list (party or joint list) in one election. Party letters are not stable across elections, so list_id always contains the election.

Grain: election and ballot letter · Rows: 302

Source: bechirot.fyi and English Wikipedia (K18-K20 seats), with hand-checked fixes in data/curated/list_name_overrides.csv

Variable	Type	Note
`list_id`	VARCHAR	Key: k{knesset}:{letter}, for example k25:פה.
`knesset`	BIGINT	Knesset number.
`letter`	VARCHAR	Ballot letter(s) of the list, in Hebrew, as on the ballot.
`name_he`	VARCHAR	Hebrew name; empty when the source has none.
`votes_published`	BIGINT	Votes as published by the source (compared with our own count by a validation rule).
`share_published`	DOUBLE	Share of valid votes as published, in percent.
`seats`	DOUBLE	Knesset seats won (0 for lists below the threshold).
`passed_threshold`	BOOLEAN	True if the list passed the electoral threshold.
`name_source`	VARCHAR	Where name_he comes from (official results, gazette, a curated correction, and so on).
`name_en`	VARCHAR	English name from the source; empty when the source gives only Hebrew.
`name_en_source`	VARCHAR	Where name_en comes from.
`name_published`	VARCHAR	The name exactly as the source printed it, before any correction.
`retrieved_at`	VARCHAR	Date the source pages were fetched.
`name_known`	BOOLEAN	False for the few tiny lists whose name the source does not give.
`display_name_en`	VARCHAR	English name for display: own name, else the party family's name, else the components joined with ' + '. Empty for lists below the threshold without an English name.
`display_name_en_source`	VARCHAR	How display_name_en was derived: the source, a curated fix, the family, or the components.

`national_results`

National votes per list, split into votes cast in localities and external envelopes.

Grain: election and list · Rows: 302

Source: Computed from the CEC ballot-box files

Variable	Type	Note
`list_id`	VARCHAR	Key into lists.
`knesset`	BIGINT	Knesset number.
`letter`	VARCHAR	Ballot letter(s).
`votes_with_envelopes`	BIGINT	All valid votes for the list, including external envelopes (the official figure).
`votes_locality_only`	BIGINT	Votes cast in localities.
`votes_envelopes`	BIGINT	Votes from external envelopes (soldiers, prisoners, diplomats and others), which belong to no locality.
`share_of_valid`	DOUBLE	Share of all valid votes (with envelopes), in percent.
`share_of_locality_valid`	DOUBLE	Share of valid votes cast in localities, in percent.
`seats`	INTEGER	Knesset seats won.
`passed_threshold`	BOOLEAN	True if the list passed the threshold.

`list_trends`

National result of each party family per election. A family's own list and its appearances inside joint lists are kept apart and never added together.

Grain: party family and election in which it appears · Rows: 139

Source: Computed from national_results and list_family (hand-curated links)

Variable	Type	Note
`family_id`	VARCHAR	Key into families.
`knesset`	BIGINT	Knesset number.
`votes_same`	BIGINT	Votes of lists that were the family's own list (relation 'same').
`votes_joint`	BIGINT	Votes of joint lists the family took part in as one component (relation 'joint_component'). These votes are not the family's alone.
`share_same`	DOUBLE	Share of valid votes of the family's own list, in percent.
`share_joint`	DOUBLE	Share of valid votes of the joint lists, in percent.
`seats_same`	INTEGER	Seats won by the family's own list.
`seats_joint`	INTEGER	Seats won by the joint lists. Not split between the components.
`joint_list_ids`	VARCHAR[]	The joint lists (list_id) counted in votes_joint.
`includes_inferred_link`	BOOLEAN	True if any link behind this row has confidence 'medium', meaning it rests on outside knowledge. Shown as an 'inferred' marker.

`families`

Party families: lists across elections that the project treats as the same political party, by published, evidence-noted links.

Grain: family · Rows: 21

Source: data/curated/families.csv (hand-curated)

Variable	Type	Note
`family_id`	VARCHAR	Stable ASCII key, for example likud.
`name_he`	VARCHAR	Hebrew name of the family.
`name_en`	VARCHAR	English name of the family.

`list_family`

Every link between a list and a party family, with the relation, a confidence level and the evidence note.

Grain: list and family · Rows: 139

Source: data/curated/list_family.csv (hand-curated, sources named in the notes)

Variable	Type	Note
`list_id`	VARCHAR	Key into lists.
`knesset`	BIGINT	Knesset number.
`letter`	VARCHAR	Ballot letter(s).
`family_id`	VARCHAR	Key into families.
`relation`	VARCHAR	'same' if the list is the family's own list; 'joint_component' if the family ran inside a joint list.
`confidence`	VARCHAR	'high': the list name states the components, the same party ran under the same name, or a source was checked and is recorded. 'medium': relies on outside knowledge.
`note`	VARCHAR	Evidence note, with the source where one was checked.

`localities`

One row per locality code that appears in any election's results, with names, CBS classification and a map point.

Grain: locality code · Rows: 1,232

Source: CEC ballot files (codes and Hebrew names), Population Authority and CBS lists (English names, classification), CBS locality points (coordinates)

Variable	Type	Note
`locality_code`	BIGINT	CBS/CEC locality code. Places merge and split, so see locality_lineage.
`name_he`	VARCHAR	Hebrew name, with brackets spaced consistently (a missing space before '(' is added).
`name_en_cbs`	VARCHAR	English name exactly as in the Population Authority list (upper case, cut off at 20 characters); empty when the place is not on it.
`display_name_en`	VARCHAR	English name for display: the Population Authority name in normal capitals with brackets tidied, or a curated name where the source is cut off, wrong or missing. Empty only if neither exists.
`display_name_en_source`	VARCHAR	'cbs' (source name, tidied by rule), 'curated' (data/curated/locality_name_overrides.csv, with a reason per row) or 'none'.
`subdistrict_he`	VARCHAR	CBS sub-district, in Hebrew (reversed brackets repaired).
`subdistrict_en`	VARCHAR	English name of the sub-district, curated in data/curated/geo_names_en.csv (the CBS file gives it in Hebrew only).
`council_he`	VARCHAR	Regional council, in Hebrew, as published.
`council_en`	VARCHAR	English name of the regional council, a curated transliteration (data/curated/geo_names_en.csv); official English spellings to be confirmed. Empty for places without a regional council.
`district_name_en`	VARCHAR	CBS district, as published (the CBS name is kept, including 'Judea and Samaria Area').
`locality_type_name_en`	VARCHAR	CBS locality type, as published.
`municipal_status_name_en`	VARCHAR	CBS municipal status, as published.
`religion_type_name_en`	VARCHAR	CBS religion type of the locality, as published.
`first_knesset`	BIGINT	First election in which the code reports results.
`last_knesset`	BIGINT	Last election in which the code reports results.
`knessets_present`	BIGINT	Number of elections in which the code reports results.
`name_variants`	VARCHAR[]	All spellings of the Hebrew name seen across the sources, separated by ' \| ' in CSV.
`lon`	DOUBLE	Longitude of the CBS locality point (WGS84); empty if there is no point.
`lat`	DOUBLE	Latitude of the CBS locality point (WGS84); empty if there is no point.
`in_cbs_list`	BOOLEAN	True if the code is on the current CBS/Population Authority locality list.
`ended_reporting`	BOOLEAN	True if the code reports no results in the latest election.

`locality_turnout`

Eligible voters, voters and valid votes per locality and election. External envelopes are excluded: they belong to no locality.

Grain: election and locality code · Rows: 10,827

Source: Sum of the CEC ballot-box files

Variable	Type	Note
`knesset`	BIGINT	Knesset number.
`locality_code`	BIGINT	Key into localities.
`series_key`	VARCHAR	Key of the comparable series this row belongs to (see locality_series); equal to the code for most localities.
`locality_name`	VARCHAR	Name as written in that election's CEC file.
`eligible`	BIGINT	Eligible voters registered at the locality's ballot boxes.
`voters`	BIGINT	People who voted at the locality's ballot boxes.
`invalid`	BIGINT	Invalid votes.
`valid`	BIGINT	Valid votes.
`in_locality_turnout_pct`	DOUBLE	Voters as a percentage of eligible voters, in the locality only. Not the official national turnout, which counts envelopes.
`ballot_boxes`	BIGINT	Number of ballot boxes.

`locality_results`

Votes per list in each locality and election (only lists with votes).

Grain: election, locality code and list · Rows: 149,538

Source: Sum of the CEC ballot-box files

Variable	Type	Note
`knesset`	BIGINT	Knesset number.
`locality_code`	BIGINT	Key into localities.
`list_id`	VARCHAR	Key into lists.
`letter`	VARCHAR	Ballot letter(s).
`votes`	BIGINT	Valid votes for the list in the locality.
`share_of_locality_valid`	DOUBLE	Share of the locality's valid votes, in percent.

`locality_family_results`

Votes per party family in each comparable locality series and election, with the family's own list and joint-list appearances kept apart.

Grain: election, locality series and family · Rows: 128,645

Source: Computed from locality_results and list_family

Variable	Type	Note
`knesset`	BIGINT	Knesset number.
`series_key`	VARCHAR	Comparable locality series (see locality_series). Merged predecessors are summed into the successor's series.
`family_id`	VARCHAR	Key into families.
`votes_same`	BIGINT	Votes of the family's own list.
`votes_joint`	BIGINT	Votes of joint lists in which the family took part.
`share_same`	DOUBLE	Share of the series' valid votes for the family's own list, in percent.
`share_joint`	DOUBLE	Share of the series' valid votes for joint lists, in percent.
`series_valid_votes`	BIGINT	Valid votes of the whole series in that election (the denominator of the shares).
`includes_inferred_link`	BOOLEAN	True if a 'medium' confidence link is behind this row.

`locality_series`

Maps each locality-election to a comparable series, so trend lines do not run across merges, splits or changes of area.

Grain: election and locality code · Rows: 10,827

Source: data/curated/locality_lineage.csv through build_lineage.py

Variable	Type	Note
`knesset`	BIGINT	Knesset number.
`locality_code`	BIGINT	Key into localities.
`series_key`	VARCHAR	Series the row belongs to. A code ending in '~pre{K}' is the earlier, different area of a code whose scope changed at election K.
`comparability`	VARCHAR	'direct': the row stands on its own series; 'merged_predecessor': the row is summed into its successor's series; 'pre_break': the area before a scope change, kept as a separate series.
`link_confidence`	VARCHAR	Confidence of the lineage link behind a merged_predecessor row; empty otherwise.

`locality_lineage`

Merges, splits, changes of area and endings of locality codes, with the evidence for each.

Grain: lineage link · Rows: 13

Source: data/curated/locality_lineage.csv, with the evidence ratio computed by the build

Variable	Type	Note
`from_code`	BIGINT	Code the link starts from.
`to_code`	BIGINT	Code the link points to (equal to from_code for a scope change or an ending).
`link_type`	VARCHAR	split_off, merged_into, scope_change (same code, different area) or ended_reporting.
`effective_knesset`	BIGINT	First election affected; empty when no election is affected.
`confidence`	VARCHAR	high, medium or low.
`note`	VARCHAR	Description of the event and its source.
`evidence_ratio`	DOUBLE	Change in eligible voters across the event, as a ratio, where it can be computed.
`evidence_baseline`	DOUBLE	Median change in eligible voters of unaffected localities over the same elections. The event is accepted only if the ratio is within 6% of this.

`series_changes`

Change in eligible voters per locality series and election against the median change, with a flag for abrupt changes that lineage does not explain.

Grain: election and locality series (from the second election of the series) · Rows: 9,580

Source: Computed by build_serving.py from locality_turnout

Variable	Type	Note
`knesset`	BIGINT	Knesset number.
`series_key`	VARCHAR	Locality series.
`eligible`	BIGINT	Eligible voters of the series now.
`eligible_prev`	BIGINT	Eligible voters of the series at the previous election.
`ratio`	DOUBLE	eligible / eligible_prev.
`median_ratio`	DOUBLE	Median ratio of all series between the same two elections.
`ratio_vs_median`	DOUBLE	ratio / median_ratio.
`flagged`	BOOLEAN	True if ratio_vs_median is above 1.33 or below 0.75 and the difference from the expected value is at least 200 voters. It says the series may not be comparable here, not why.

`demographics`

Demographic and socio-economic values per locality (Census 2022 and CBS tables), in long format.

Grain: CBS locality and variable · Rows: 46,415

Source: CBS Census 2022 (data.gov.il), the CBS socio-economic index tables

Variable	Type	Note
`locality_key`	VARCHAR	CBS locality code as text; cast to an integer to join to localities.locality_code.
`variable_id`	VARCHAR	Key into demographic_variables.
`reference_year`	BIGINT	Year the value refers to.
`value`	DOUBLE	Numeric value, in the unit given by demographic_variables.unit.
`value_text`	VARCHAR	Text value for non-numeric variables.
`source_file`	VARCHAR	File the value was read from.

`demographic_variables`

Dictionary of the demographic variables, with a flag for those whose exact definition is not yet confirmed.

Grain: variable · Rows: 43

Source: Hand-written from the CBS dictionary, in ingest_cbs.py

Variable	Type	Note
`variable_id`	VARCHAR	Key used in demographics.
`name_en`	VARCHAR	English name.
`unit`	VARCHAR	Unit of the value.
`variable_group`	VARCHAR	Topic: age, education, household, immigration, income, religion, sector, size, socio_economic or work.
`meaning_confirmed`	BOOLEAN	False if the exact definition (for example the wage year) could not be confirmed from CBS; such variables are marked as approximate.
`note`	VARCHAR	Caveats and definitions.

`correlations`

Weighted Spearman and Pearson of a party family's (or a list's) locality share against a demographic variable, per election. Localities below 200 valid votes are excluded. A locality-level association does not describe how individuals vote.

Grain: subject, election and variable · Rows: 2,350

Source: Computed from locality_family_results (or locality_results) and demographics; weighted by in-locality valid votes

Variable	Type	Note
`subject_type`	VARCHAR	'family' or 'list'.
`subject_id`	VARCHAR	families.family_id, or lists.list_id.
`knesset`	BIGINT	Knesset number.
`variable_id`	VARCHAR	Key into demographic_variables.
`spearman`	DOUBLE	Weighted Spearman rank correlation (average ranks, weights = valid votes).
`pearson`	DOUBLE	Weighted Pearson correlation of the raw values.
`n`	BIGINT	Number of localities used.
`weight_sum`	BIGINT	Sum of the weights (in-locality valid votes).
`min_valid`	BIGINT	Size threshold applied (200).
`reference_year`	BIGINT	Year of the demographic snapshot.

`similar_localities`

Nearest neighbours by voting pattern in the latest election: cosine similarity of party-family share vectors, among localities with at least 200 valid votes.

Grain: locality and neighbour · Rows: 8,752

Source: Computed from locality_family_results in the latest election

Variable	Type	Note
`locality_code`	BIGINT	Key into localities.
`neighbor_code`	BIGINT	The similar locality.
`rank`	BIGINT	1 is the nearest neighbour.
`score`	DOUBLE	Cosine similarity of the family-share vectors, 0 to 1.
`knesset`	BIGINT	Election the pattern is taken from (the latest).
`min_valid`	BIGINT	Size threshold applied (200).