The data

The stable identifier of the form CTF-E-nnnnnn. Cite this, not the numeric id, and not the slug: the slug can be rewritten, this cannot.

The URL segment for the record page. Readable, but not a citation target.

The recorded citation address. Three different URL shapes are stored across the table and the majority shape uses a language-prefixed path this site does not serve, so it does not resolve. Repairing the column is a pending data correction; until then, prefer the address shown on the record page.

When the affiliation changed. Always read it together with date_precision: a value of 1930-01-01 with year precision does not mean the first of January.

Empty means no date has been established at all, not that the event is undated in the world.

How exactly event_date is known: day, month, year, circa, range or unknown. This is the single most important field for a reporter. Most of the record is dated to a year only, and a year-precision row cannot support a sentence naming a month or a day.

The literal value unknown means the precision itself was never assessed — weaker than year precision, not equal to it.

The upper bound when date_precision is range.

Empty on every row that is not a range. Not a defect.

The kind of affiliation change: joined_party, left_to_independent, expelled, suspended, party_formation, party_dissolution, merger, caucus_departure, return_to_party, changed_affiliation or other.

The value other means the change did not fit the vocabulary, not that it was unclassifiable in principle.

individual, group or institutional. A group event is one event involving several people; it is the reason an event count and a person count are different numbers and cannot be compared directly.

The editorially confirmed number of people involved, where one has been established.

Empty means no head count has been confirmed. Do not substitute the number of named participants: on group events those differ, which is the whole point of the field.

The record’s own estimate of how many people were involved, for example 17 for a caucus described in a source as "about seventeen members". It is the record declaring an estimate, not a projection computed here.

Empty means no estimate was recorded, which is different from an estimate of zero.

not_applicable, unresolved, partial or complete — how far the work of naming the individuals has got.

unresolved means the event says more people were involved than are named. Rendering such an event with an empty participant list, or counting its participants as its head count, understates it.

unverified, single_source, corroborated or official_and_independent — the strength of the evidence behind the record.

unverified means nobody has checked it against a source yet. Filtering on this field ALONE is unsafe: a withdrawn record can still carry a corroborated level, so exclude withdrawn as well.

legacy_lead, pending_review, verified, disputed or withdrawn — the editorial lifecycle of the record.

legacy_lead means imported and never reviewed. Most of the table sits here. A lead is a research pointer, not a published fact.

When the record was published. The third leg of the publication gate: verified status AND corroborated evidence or better AND this timestamp.

Empty means never published, whatever the other two fields say.

When a human last checked this record against its sources.

Empty means never verified. It does not mean verified long ago.

Foreign key to jurisdictions. Province and territory names come from that table, never from free text — which is why the same province can no longer appear under two spellings the way it did in the legacy model.

Foreign key to legislatures, and the ONLY source of federal / provincial / territorial. There is no level column on the event itself.

Empty means the event has no level at all, so any filter by level necessarily excludes it. Filtering by level and then reporting the result as a total is the easiest way to lose rows silently.

Foreign key to chambers.

Empty means no chamber recorded. Note separately that the chambers table does contain an upper chamber, the Senate of Canada, and that Senate records exist here — none of them published, so they appear only when you ask for unpublished records.

Foreign key to sessions.

Empty on every row in the live database. The sessions table is populated; nothing has been linked to it. Do not build a session-based analysis on this column.

The event title in each language. Both columns are NOT NULL.

Because neither column can be empty, a missing translation shows up as the French being byte-identical to the English rather than as a null. A null-based check will report this table as fully bilingual when much of it is not; the API and the site flag identical pairs for exactly this reason.

A short prose description of the event in each language.

Empty means no summary was written in that language. It is served as absent, never machine-translated on the fly.

Increments on every material change. Paired with the revisions table, it lets you reconstruct what the record said when you cited it.

Link back to the row in the original 25-table model, for anyone reconciling against an older export.

Empty means the record has no legacy counterpart — usually because it was created after the migration.

Foreign key to persons.

Empty means the leg is recorded but the individual is not yet identified.

Foreign keys to political_parties.

Empty is common and meaningful. A member who went Independent has no destination party at all, and a member who came from a body with no party row has no origin party. Neither is missing data; use the affiliation label instead, and never treat an empty to_party_id as "stayed put".

The verbatim label for each side, used where no party row exists — for example Independent.

Empty means neither a party nor a label was recorded for that side.

voluntary, involuntary, mixed or unknown — whether the member jumped or was pushed.

unknown is the default and is the value on the large majority of rows. It means the question has not been answered, not that the move was neutral. Any share computed over this field must be taken over the rows where it is not unknown, and published beside that denominator.

yes, no, unknown or not_applicable — whether an appointment followed the change.

unknown dominates this column. It is the field a "were they rewarded?" story would rest on, and it is mostly unanswered; treat a small known set as a small known set.

to_government, from_government, between_opposition, government_to_independent, opposition_to_independent, unknown or not_applicable.

unknown means the direction was never established, which is the case on most rows.

How the standing of each caucus changed.

Empty on every row in the live database. The arithmetic is obvious and has deliberately not been inferred, because an inferred number is indistinguishable from a recorded one once it is exported.

Foreign key to ridings.

Empty means no riding recorded for this leg.

When this person’s leg took effect, which can differ from the event date when a group moved over several days.

Empty means the leg carries no date of its own; fall back to the event date and its precision.

Order of this leg within its event. It is a display order, not a ranking of importance.

The stable identifier for the person.

The person’s name in each language.

These two are frequently identical, and for a personal name that is usually correct rather than a missing translation. Do not read an identical pair here as an untranslated record without checking.

Years only, never full dates.

An empty death year means no death year is recorded. It does NOT mean the person is living, and must never be used to infer that.

Whether the record has been reviewed under the stricter standard applied to living people.

An unreviewed value means the review has not happened, not that the person is dead.

The address of an image, with its rights status, credit, licence and source page in the neighbouring columns. Data packages carry the address and the credit fields, never the image file.

Empty means no portrait is held, which may be because one was removed for want of a rights basis rather than because none exists.

hansard, journal, official_member_profile, official_party_record, government_record, election_record, news, academic and others.

How much weight the sourcing standard gives this kind of document.

An unassessed tier means nobody has placed the source yet, not that it is untrustworthy.

Where the document can be read.

Empty means an offline source with no address — a bound volume, a microfilm. That is different from a source whose link has never been checked.

A preservation copy of the document and when it was captured.

Empty on every row in the live database: nothing has been archived yet. A source is therefore only as durable as the site that hosts it.

A hash of the source and, crucially, a statement of WHAT was hashed. A checksum without its scope cannot be reproduced by anyone else and so verifies nothing.

Empty means no hash was taken. A hash present without a scope is counted here as unverifiable, not as verified.

Where inside the document the passage sits — a page, a column, a timestamp.

Empty means the citation points at the document as a whole and a reader must find the passage themselves.

What may be done with the source document. It governs the document, not our metadata about it.

An unassessed status means the question has not been settled; treat the document as all-rights-reserved until it is.

en, fr, und (undetermined) or mul (multiple).

und means the language was never determined, not that the document has none.

What the claim is about, by public identifier rather than by numeric key, so a claim survives a re-import.

Which field of the subject the claim asserts.

pending_review, supported, partially_supported, disputed, unsupported, unknown or withdrawn.

pending_review is the value on most claims and means no reviewer has looked at it. A claim in that state is not evidence of anything except that someone recorded an assertion.

high, medium, low or unknown.

unknown means no confidence was assigned, which is not the same as low confidence.

Flags claims held to the stricter standard because they are about someone still living.

False marks a source that CONTRADICTS its claim. Such citations must never be filtered out of a citation count: a record with a contradicting source is better documented than one with none, and hiding them would make the evidence look cleaner than it is.

The passage in the source that carries the claim.

Empty means no excerpt was transcribed. It never means the source was paraphrased into the field.

Which record was corrected, by public identifier.

True when the correction changes what a reader would reasonably have understood — the kind of change that would oblige you to update a story you had already filed.

When the correction was made. Compare it with the retrieval date in your citation to know whether what you cited has since changed.

The revision this correction produced, so the before and after states can both be recovered.

CA for the country, CA-ON, CA-NS and so on for provinces and territories. The safe join key for merging with another dataset.

federal, provincial or territorial. This column is the only place the level lives; events reach it through legislature_id.

lower, upper or unicameral.

A row of type upper exists in the live database: the Senate of Canada. The Senate is present in the record and absent from the published record, because none of the Senate events has passed the publication gate. Treat an empty Senate result as a statement about review, not about the schema and not about whether senators change affiliation.

The party’s own published colour.

Empty on nearly every party. A colour is not invented for a party that does not publish one, so a chart drawn from this column will be mostly uncoloured — by design.

party, parliamentary_group or coalition. A parliamentary group is not a registered party and counting the two together will overstate the number of parties.