About Skarim

About Skarim

Skarim collects publicly released political poll results, preserves them in a structured historical database, and connects views of individual polls, parties, publishing channels, sources and aggregate statistics.

Skarim is not a polling institute and does not conduct the polls displayed on the site. Rights in and responsibility for the original publication remain with its publisher and pollster. If a Skarim-originated poll is ever stored in the database, its origin will be identified explicitly in that poll record.

Data sources

Data comes from identifiable public releases by television and news channels, newspapers, polling organizations, and other public sources represented in the database. Skarim does not invent source details that are unavailable.

An individual poll page identifies the publishing channel and the pollster when they can be resolved from the source data and the site’s maintained definitions. Because that directory is maintained over time, it is shown on a dedicated page instead of being copied here as a fixed list.

View the poll-source and pollster directory

What Skarim adds

Skarim’s added value comes from turning separate publications into a consistent database that can be searched, compared and examined over time. Every derived view is generated from the structured records in that database.

Poll summaries are programmatic, data-driven analyses built from fixed templates. Wording is selected deterministically from poll data and assembled inside the application from structured inputs.

• Converting published results into structured records and preserving their history.
• Normalizing party and source identities where the maintained mappings permit it.
• Comparing polls and publication sources over time.
• Calculating polling averages, trends, ranges and party statistics.
• Linking individual polls, parties, channels, sources and aggregate views.

How the polling average is calculated

By default, the polling-average page uses the first five records after sorting the poll records available to the view from newest to oldest by poll date. The initial server render supplies the five latest records in the database’s publication order; when the view holds additional records, the same date ordering determines the sample. Same-date ties also use the record’s stable ordering value — its identifier in the public data. Every poll record in the sample has equal weight. Records are not grouped by day or source, so two separate records count as two polls. Selecting a channel restricts the sample to that channel, and the page controls offer other available window sizes.

Party identities are joined using the current mappings: a known identifier or slug, canonical name, stable catalog identifier, or logo-file identity. A leader suffix introduced by wording such as the Hebrew equivalent of “led by” is removed for matching. If more than one row in one poll matches the same identity, the highest value in that poll is used.

Each party’s mandate mean is calculated across every poll in the sample. If a party is absent from one sampled poll, it contributes zero for that poll. Numeric inputs are ingested as whole mandates; a missing or invalid value becomes zero, and a poll record without a valid date is excluded from the public dataset used by the calculation.

To display a whole-seat distribution, the existing method first floors every raw mean. Remaining seats toward 120 are then assigned, at most one additional seat per party, in descending order of fractional remainder. A party’s displayed number is therefore not an independent conventional rounding of its own mean.

Comparison with the previous poll

On an individual poll page, the comparison point is the previous available poll from the same publishing channel, using the normalized channel identity and chronological order. The page looks for the next older record within up to ten recent records available for that channel. If a predecessor is unavailable in that window, no comparison is shown.

The mandate difference is calculated for every party present in the current poll. If that party is not found in the previous poll, its comparison baseline is zero. The expanded text analysis uses broader identity matching and may also note a party that disappeared from the current poll when it had at least four seats in the previous one.

Party statistics

The historical figures on a party page are separate from the current polling average. Historical statistics use every valid numeric record available in that party’s full history. Special or event records and values that cannot be parsed as numbers are excluded; numeric values are constrained to 0–40. A numeric official Knesset result stored as a regular record can also be included in these historical statistics.

• Historical average: arithmetic mean of all valid records, rounded to one decimal place.
• Historical high and low: the maximum and minimum in the same valid-record set.
• Latest result: the newest valid numeric record in the history ordering.
• Recent trend: valid records with the same date are grouped and averaged to one decimal place; the chart shows the latest 30 date groups in chronological order.
• Current polling average: the separate polling-average calculation, based by default on the five latest poll records.

Political bloc calculations

Coalition, opposition, Arab-party and “unknown” totals are produced by summing party mandates according to the classification stored for each party in a poll record. Poll views also recognize equivalent wording for right/coalition, center-left/opposition and Arab parties; anything unmatched is assigned to “unknown.”

The blocs page also shows Haredi and national-religious subtotals recognized by the current party-name and identifier mapping. These are subtotals and can overlap coalition or opposition totals. Its summary is the arithmetic mean of the latest ten polls, with each bloc total rounded separately to a whole number.

A bloc label reflects the site mapping used when the data is processed; it is not an immutable political fact. A party’s classification can change over time, and such a change requires a data-classification update rather than an alteration of the originally published poll result.

Parties missing from a poll

Treatment depends on the view. In the polling average, a party present in at least one sampled poll but absent from another receives zero for the latter. In a previous-poll comparison, a party present now but absent previously is compared with zero. A party present previously but absent now has no result row in the current poll card, although the programmatic analysis may mention its disappearance under its threshold rules.

A zero used in a calculation is not a separate claim that the party had exactly zero support. It represents the lack of a result for that party in the available poll record, which may reflect non-display, a joint list, the electoral threshold, or the way the source published its data.

Renames, mergers and splits

Israeli parties and lists may rename, merge, split, run jointly, or later appear as a different entity. Skarim maintains identifiers and mappings for names, slugs, catalog records and relationships between entities to preserve history and reduce duplicates where possible.

Those mappings do not claim that every pair of historical entities is politically or organizationally equivalent. Where no reliable match exists, records may remain separate rather than creating an artificial continuous history.

Accuracy and limitations

Skarim republishes and processes polling data published by third parties. A poll is an estimate, not an election result. Every poll has its own methodology, sample, margin of error, timing and publication context, and a difference between two polls does not necessarily show real voter movement.

Calculations and displays depend on the completeness and accuracy of the data available in the database. Normalization, averages and comparisons make the records easier to read consistently, but they do not remove uncertainty from the original polls or make different sources methodologically identical.

Corrections and updates

If you find an incorrect poll value, missing source, wrong party mapping or another factual issue, contact Skarim with the page address and the original publication. A proposed correction is checked against the available source before the database is updated.

Contact support: myvote19@gmail.com