# AUS elections pack handoff

Written 2026-09-03 to `/workspace/elections-ingest/aus/`.

## Files / counts

| File | Rows (excl. header) | Path |
|---|---|---|
| elections.csv | 3 | /workspace/elections-ingest/aus/elections.csv |
| people.csv | 226 | /workspace/elections-ingest/aus/people.csv |
| money-edges.csv | 14999 | /workspace/elections-ingest/aus/money-edges.csv |
| polls.csv | 82 | /workspace/elections-ingest/aus/polls.csv |
| results.csv | 1576 | /workspace/elections-ingest/aus/results.csv |
| SOURCES.md | — | /workspace/elections-ingest/aus/SOURCES.md |
| STATS.json | — | /workspace/elections-ingest/aus/STATS.json |
| HANDOFF.md | — | /workspace/elections-ingest/aus/HANDOFF.md |

## Cycle status (verified against the AEC)

- `aus-hor-2025` **sitting** — 2025-05-03, `confirmed_official`. AEC's federal-elections index states "2025 federal election – Saturday 3 May 2025". 48th Parliament sitting.
- `aus-senate-2025` **sitting** — 2025-05-03, `confirmed_official`. Half-Senate election held the same day; AEC FAQ: 40 of the 76 Senate places are contested at a House + half-Senate election.
- `aus-hor-next` **sitting**, `date_confidence=constitutional_due`, **`election_date` deliberately blank**. AEC media release 2026-07-23 says the next federal election "is due to be held between August 2027 and May 2028" and that the House stays at 150 seats. No polling day has been announced (the Governor-General sets it on the government's request), so no date was invented. The window is in the row's `notes`.
- No officially announced date contradicts the brief.

## Join keys

- `people.csv.person_id` = **Parliament of Australia PHID** (the `MPID` in `aph.gov.au/Senators_and_Members/Parliamentarian?MPID=…`). Alphanumeric, not numeric-only: e.g. `316915`, `R36` (Albanese), `00AOU` (Wong). Store it as a **string**.
- `people.csv.election_id` → `elections.csv.election_id`: the 150 House members attach to `aus-hor-2025`, the 76 senators to `aus-senate-2025`.
- `money-edges.csv.election_id` → `aus-hor-2025` for every row (join anchor only — see money caveats).
- `results.csv.constituency_id` = **AEC DivisionID**; first-preference and two-party-preferred rows for the same division share it.
- `polls.csv.election_id` → `aus-hor-next`.

## Licences

- AEC material (elections.csv sources, money-edges.csv, results.csv): **CC BY 4.0 International** per https://www.aec.gov.au/footer/copyright.htm (excluding the Coat of Arms, AEC logos, AEC maps and third-party content).
- people.csv: Parliament of Australia (aph.gov.au), © Commonwealth of Australia. The exact CC variant on the aph.gov.au copyright page could **not** be machine-verified from this box (Azure WAF blocks plain fetches of aph.gov.au), so the `license` column says only "© Commonwealth of Australia" — confirm before commercial redistribution.
- polls.csv: each row carries the pollster's own copyright string; Newspoll / Resolve / YouGov toplines are transcribed from The Conversation (CC BY-ND) because the primary tables are paywalled.

## Bars

- Sitting members: **150 House + 76 Senate = 226**, i.e. the full chamber floors, no vacancies implied (150 divisions, 12 senators × 6 states + 2 × 2 territories).
- Polls: **82 option-rows** from **8 surveys** across Roy Morgan, Newspoll, Resolve Political Monitor, YouGov (bar was ≥40).
- Money: **14,999** AEC Transparency Register rows, AUD as filed, non-empty.
- Results: official AEC 2025 returns (no CLEA fallback needed).

## People

Incumbents only (`role=incumbent`, `incumbent=true`); no next-election candidate list exists because
the election is not called. `bio_text` is **empty for all 226 rows** — aph.gov.au parliamentarian
pages publish roles, committees and contact details, not a biography paragraph, and nothing was
authored. `photo_url` and `official_url` are the official APH URLs recorded verbatim.

`party_code` is the AEC `PartyAb` matched to the APH party string using official AEC 2025 party
strings (exact match, then a case-folded match after dropping parenthetical/branch suffixes).
It is blank for the one **United Australia Party** member, where no unique AEC abbreviation matched —
not guessed.

Party mix (APH strings): ALP 124, Liberal Party of Australia 38, Liberal National Party of Queensland 18,
Independent 12, The Nationals 11, Australian Greens 11, One Nation 6, and one each of Katter's Australian
Party, Centre Alliance, Jacqui Lambie Network, Country Liberal Party, Australia's Voice, United Australia Party.

## Money

AEC Transparency Register bulk downloads, **AUD exactly as filed** (no conversion, no rounding).
Two cycles are present in the `cycle` column:

- `2024-25` — 13,007 rows, the latest published **annual** returns (donor returns' "Donations Made",
  party / associated-entity / significant-third-party "Detailed Receipts", donor and third-party
  "Donations Received", "Capital Contributions"). Sum of `amount`: **A$1,368,599,887**.
- `2025 Federal Election` — 1,992 rows, the **election** returns for the 2025 federal election
  (donor "Donations Made", candidate and Senate-group donations, donor "Donations Received").
  Sum of `amount`: **A$47,103,053**.

`edge_type` uses what the regulator actually published: `donation` 7,935 · `other_receipt` 6,670 ·
`donation_received` 331 · `subscription` 60 · `capital_contribution` 3. "Other receipt",
"subscription" and "public funding" are the AEC's own **Receipt Type** values in `Detailed Receipts`;
they are *not* donations and must not be displayed as such. Nothing is labelled PAC — Australia has no
US-style PACs.

**Every money row has a blank `person_id`.** AEC returns identify people and organisations by name (and
an AEC donor code), never by APH PHID, so there is no id-level join to `people.csv`.
**Do not name-match money to the roster** — "SMITH, John" in a candidate return is not evidence about
any particular member, and candidate returns include the ~1,000 candidates who lost.

Two more money warnings:

1. **The register publishes both sides of some gifts.** A single donation can appear once in a donor's
   "Donations Made" row and again in the recipient's "Detailed Receipts" row. Both are kept because both
   are what the AEC published; `as_filed_note` names the return each row came from. Do **not** sum across
   the whole file and call it total donations.
2. Aggregate returns (party/associated-entity/donor totals, MP returns, third-party totals) and all
   expenditure files (candidate/Senate-group expenses, media returns) were **excluded** — they are not
   donor→recipient edges. In particular `MemberOfParliamentReturns.csv` (12 rows for 2024-25) reports an
   MP-level donation total by name with no donor detail and no PHID; it was left out rather than
   name-matched into the roster.

## Polls

`polls.csv` is long format, one row per response option, attached to `aus-hor-next`.

- **Roy Morgan** (4 weekly surveys, 3–30 Aug 2026, from Roy Morgan's own releases): primaries
  (ALP, One Nation, L-NP Coalition total, Liberal, National, Greens, Independents/Other), plus
  **first preferences and two-party-preferred kept strictly separate** in `question_text`:
  respondent-allocated ALP-v-Coalition 2PP, 2025-election-preference-flow ALP-v-Coalition 2PP, and
  ALP-v-One Nation 2PP. The published "can't say" share is its own row. Roy Morgan prints no MoE →
  `moe` blank.
- **Newspoll** (3–7 Aug, n=1,242; 24–28 Aug, n=1,243): primaries only. The source states explicitly
  that **no two-party estimate was released**, so none is recorded.
- **Resolve Political Monitor** (9–15 Aug, n=2,256): primaries only. The only 2PP hint published was
  "by 2025 election preference flows, Labor would lead the Coalition by under 52–48" — not a printed
  number, so no 2PP row.
- **YouGov / News24** (4–10 Aug): primaries plus both respondent-preference 2PPs. `sample_n` is **blank**
  because the source gives the sample only as "about 1,500".
- No MoE was printed for any of the eight surveys, so `moe` is blank throughout. Nothing was estimated.
- **Essential Report, Freshwater Strategy, and the Resolve and YouGov own-site tables produced no usable
  published numbers** (JS-only chart embeds, numbers only inside images, dead host, redirects). See
  SOURCES.md for the exact URLs and failure modes. No Wikipedia anywhere.

## Last-cycle returns

`results.csv` — **1,576** rows, all from the AEC's official 2025 federal election downloads
(event 31496, FinalResults phase), `election_id=aus-hor-2025`:

- `office=lower_chamber_first_preference` — **1,126** candidate rows (150 divisions). `votes` = AEC
  `TotalVotes` (ordinary + absent + provisional + pre-poll + postal). `seats` = 1 where AEC `Elected=Y`,
  else 0 (150 winners). `vote_share` is **derived** — candidate votes ÷ that division's formal
  first-preference total — because this AEC download publishes no per-candidate percentage; the `source`
  column says so on every row.
- `office=lower_chamber_two_party_preferred` — **300** rows (2 per division: Australian Labor Party and
  Liberal/National Coalition). `votes` and `vote_share` are the AEC's **published** TPP votes and
  percentage (percentage ÷ 100). These are a different measure from first preferences: never mix them
  into one "vote share".
- `office=lower_chamber_informal` — **150** rows, one per division, the AEC's published informal ballot
  count. `party`, `vote_share` and `seats` are blank; informal ballots are not part of the formal vote
  base.
- `person_id` is blank on every results row: the AEC's `CandidateID` is a different registry from the APH
  PHID in `people.csv`, so joining them would be a guess. The AEC CandidateID is preserved inside
  `result_id` (`aus-aec-2025-fp-{DivisionID}-{CandidateID}`).

## Known holes

- **Senate results are not in this pack** (the brief asked for House first preferences and TPP). Senate
  first preferences by group/state remain available from the same AEC event-31496 download directory.
- `aus-senate-2025` carries all 76 senators for joinability, but only ~40 of those places were actually
  contested in 2025; the continuing state senators were elected at the 2022 half-Senate election. Do not
  read "senator attached to `aus-senate-2025`" as "elected on 2025-05-03".
- Money rows for financial year 2024-25 are attached to `aus-hor-2025` purely as a join anchor. The
  annual-return money is *not* election-period money; use the `cycle` column, not `election_id`, when
  describing it.
- New federal donation-disclosure rules commence 1 January 2027 (AEC media release 2026-07-23), so the
  shape of this file will change for the next cycle.
- `elections.csv` has no `aus-senate-next` row: half-Senate timing differs from the House and no date or
  window specific to it was published by the AEC.