Pledged does not grade governments. It records what was promised, what was done, and where the evidence sits. This page sets out the rules.
Pledged records public commitments made by Australian federal politicians and parties. A commitment qualifies if it is (a) made in writing or on the public record by an elected representative, candidate, or party body, (b) specific enough to be falsified, and (c) directed at the federal jurisdiction.
We do not track state, territory, or local promises. We do not track aspirations, slogans, or value statements that cannot be assessed against evidence.
Every promise progresses through up to four states: Made, Funded, Legislated, Delivered. A promise can also be marked Partial or Broken. Each state names the dataset that justifies it — Pledged never moves a promise without an entry in one of these.
Made — the commitment is on the public record with a verifiable source. Source: Parliamentary Budget Office Election Commitments Reports (verbatim title, costing, fiscal-impact breakdown). Where a primary press release or campaign-launch transcript is also recorded, its archive URL is attached.
Funded — money has been allocated in a Budget paper, MYEFO, or appropriation bill, in an amount and timeframe sufficient to deliver the commitment. Source: Treasury Budget Paper No. 2 (Expense Measures), MYEFO, PAES and PSAES — measure title, page reference, and forward-estimates amount. A material gap between the announcement-time costing and the actual allocation is recorded as a partial-funding flag and surfaced for review.
Legislated — the necessary statute, regulation, or executive instrument is in force. Source: APH Library Bills Digests (plain-English measure description), legislation.gov.au royal-assent dates, ParlInfo bill and committee material.
Delivered — the commitment is implemented to the standard described, and evidence of delivery is on the public record. Source: ABS Data API outcome series (SDMX-JSON), ANAO performance audits, departmental annual reports, regulator notices. A commitment is Partial when implementation is underway but falls short on quantum, scope, or timing. A commitment is Broken when the government has formally abandoned it, or when the window has elapsed without delivery.
Every state change requires a primary source: a Hansard reference, a Budget paper, a published Act, an ABS or PBO release, a regulator notice, or an official department or minister statement.
We cite the URL, the publisher, and the date for every event. Secondary reporting (newspapers, broadcasters) is referenced for context but cannot, on its own, move a promise to a new state.
Where a primary source is paywalled, we link the freely accessible record (e.g. Hansard) and note the secondary publication.
Pledged is updated within seven days of a primary source becoming available. Quarterly delivery numbers (e.g. ABS Building Activity) move promises within 48 hours of release.
Each promise carries a "last updated" timestamp and an immutable evidence log. Prior states are visible from the timeline; nothing is silently rewritten.
Errors are inevitable. Where we get a fact wrong, we correct it in place, log the change with a timestamp and a brief reason, and link to the corrected source.
Anyone — citizens, staffers, ministers, journalists — can submit a correction. Submissions are triaged within seven days. We do not negotiate framing or tone.
Pledged is descriptive, not normative. We do not score parties on ideology, intention, or "good faith". We do not assign grades. The four states are evidentiary, not editorial.
A promise marked Delivered is not endorsed by Pledged. A promise marked Broken is not condemned. The figures speak; readers decide.
Quantum-and-quality assessment is hard. Where a commitment hits its number but misses its scope (or vice versa), we mark Partial and explain in the timeline. Reasonable people will disagree at the margins.
Pledged is funded by reader subscriptions and a one-time grant from the Susan McKinnon Foundation. We do not accept funding from political parties, lobby groups, government departments, or vendors that contract with government.
Editorial control rests with the masthead. Every promise is reviewed by at least two editors before its initial publication.
Whenever Pledged totals dollar amounts across promises, four numbers are published, never one. A single "total amount pledged" figure averages over a sign convention, a time window, and a cycle, and the reader cannot see what those choices were. Four labelled totals make every choice visible.
Gross commitment value — the sum of absolute values across the included commitments. Treats expenditure and savings as equally substantial commitments. The lay-reader’s intuition for "how big is this programme".
Total expenditure — the sum of negative impacts, expressed as a positive number. The figure most newsroom and party-comms teams reach for; answers "how much would these promises cost".
Total savings — the sum of positive impacts. PBO ECR routinely contains commitments that increase revenue or decrease expenditure; rendering only expenditure hides them.
Net fiscal impact — the signed sum, matching the figure in PBO summary tables. The four totals are mathematically related (gross = expenditure + savings; net = savings − expenditure) so a reader can sanity-check.
Cycle separation. Each PBO Election Commitments Report covers one parliament and a defined forward-estimates window. Pledged never sums across cycles silently. Cross-cycle views are opt-in, side-by-side, and CPI-deflated to a stated base year using ABS Series 6401.0. Two cycles are never combined into a single number.
Window labelling. Every published aggregate names the financial-year window it covers (e.g. "2025-26 to 2035-36") and the source ECR. Aggregates with mismatched windows are not produced.
Per-politician totals are not published. The PBO Election Commitments Report links commitments to parties, not to specific members of parliament. Until a primary source — Hansard tabling, press-release announcement, or campaign-launch transcript — links a commitment to a politician, Pledged renders an explicit "no data" state for per-politician aggregates and does not infer attribution from URLs or speech metadata.
The full aggregation contract is in docs/12-aggregation.md and is versioned. Disputed totals point at the methodology version and the contributing record IDs, not at Pledged’s editorial judgement.
Pledged reuses public data from Australian government bodies and open civic sources. Each dataset retains its original licence and is attributed to its publisher. Verbatim commitment text is reproduced under the publisher's own licence; Pledged adds structure, linkage, and delivery state.
Pledged's own editorial content (page text, methodology, structure, code) is © Pledged. Compiled datasets are republished under the most restrictive applicable upstream licence; per-record licence is recorded on each promise.
Pledged's codebase is authored using frontier large-language-model coding agents. All code changes are reviewed and committed by a human maintainer.
AI is not used to generate, summarise, or interpret the factual record. Promise text is verbatim from PBO Election Commitments Reports. Status transitions are gated on primary-source evidence per this methodology. Aggregations are deterministic functions of the underlying records, computed by reviewable code — not by any model. The “no editorial voice” principle (§I.4) applies to humans and to AI tooling alike.