Methodology

How totals are calculated

The headline is gross claimed: the sum of the Amount column for every row in the period, the same convention IPEA uses. Last quarter that was $40.7 million across 248 parliamentarians.

Per-MP totals on the latest-quarter ranking filter to Role = “Parliamentarian”. Former Prime Ministers, post-retirement travellers and surviving spouses are tracked separately. Staff travel charged to an MP’s allocation rolls up into that MP’s total because the allocation is theirs.

What “discretionary” means

Office Facilities (electorate office fit-out and refurbishment) is administered by the Department of Finance, not the MP, and is excluded from any discretionary subtotal. It still appears in the gross headline because IPEA reports it that way.

The “largest discretionary line” chip also excludes Employee Travel, which is staff travel charged to a minister’s allocation and would otherwise dominate every ministerial card without saying anything about the minister’s own behaviour. The full breakdown is on each per-MP page.

Repayments

Repayments are shown as a separate line, not netted off the gross. A repayment’s reporting quarter is when the money was paid back, often a different quarter from the original claim. Netting them would misrepresent both numbers. Repayments are only published from 2022Q03 onwards; earlier quarters show zero.

All-time totals

The all-time leaderboard sums every claim across the 36 quarters in the dataset. Two biases are baked in:

• Tenure. A twenty-year MP mechanically claims more than a four-year MP. The leaderboard defaults to a per-year-in-office sort, which divides each MP’s total by their actual time in Parliament (see Role and time in office below), suppressed for MPs with under a year in office.
• Role. Former PMs, post-retirement travellers and surviving spouses carry continuing travel allowances under entitlement rules. The role filter defaults to Parliamentarians only. Senior and diplomatic roles also run higher international travel, so the all-time view offers an “Exclude overseas travel” switch that drops the International Travel category from the Lifetime column.
• Coverage. “Lifetime” means since 2017Q02, when IPEA’s machine-readable record begins; earlier service does not appear.

Inflation adjustment

Figures across the site default to today’s value (inflation-adjusted), with an “Adjust for inflation” toggle on the leaderboard, each MP page and the compare view to switch back to original dollars. The base period is the most recent quarter with published CPI data, so it advances as the ABS publishes new figures and every real amount is restated against it. The conversion uses the ABS Consumer Price Index, Australia (cat. 6401.0), specifically the All groups index for the weighted average of eight capital cities, quarterly, original series (ABS data key 1.10001.10.50.Q). This is the headline national CPI figure.

The method is straightforward ratio deflation. For each expense quarter q, every dollar amount is multiplied by CPI[base] / CPI[q], where base is the latest quarter for which ABS has published a CPI value. Because the ABS CPI calendar lags the IPEA release calendar by about one quarter, it is possible that the most recent IPEA quarter has no CPI value yet. In that case the base automatically falls back to the prior quarter, and the most recent expense quarter carries a factor of 1.0 (amounts are reported as nominal until the CPI catches up).

What this means in practice. A claim of $1,000 in April 2017 appears as roughly $1,320 in today’s dollars, reflecting that prices in general were about 32 percent lower in 2017 than in early 2026. Long-tenured MPs whose early quarters were in lower-price periods will see a larger uplift on their real totals than recent entrants.

Caveat. CPI tracks the consumer goods and services basket. Parliamentary travel and office expenses do not move in lockstep with the consumer basket. The adjustment is a general purchasing-power guide, not a precise measure of how much more or less it would cost to replicate the same parliamentary activities today. Use nominal figures for direct comparisons against IPEA’s published data.

Role and time in office

To compare MPs fairly, each one is matched to their record in the Australian Parliament House Parliamentary Handbook, which provides the dates they entered and left Parliament and the ministerial or shadow roles they held over time. This lets us do two things the raw expense data cannot.

• Per year in office. Per-year figures divide an MP’s total by the number of quarters they were actually in Parliament within the dataset window, not by the number of quarters they happened to file a claim. A quarter counts as a whole quarter if any part of it was served.
• Like-for-like roles. Each MP is placed, quarter by quarter, into a seniority group judged as at the end of that quarter: Prime Minister, cabinet minister, minister outside cabinet, assistant minister, shadow minister, or backbencher. An MP page compares a figure against the median for that MP’s own group that quarter, so a backbencher is measured against other backbenchers rather than against a pool that includes the Prime Minister.

Caveats. The Handbook interface is an official but beta service, so we keep a local snapshot and refresh it on a schedule. A small number of people in the expense data are not parliamentarians at all (for example surviving spouses of former Prime Ministers, or administrative units); these are left without a role and tenure. Per-year-in-office figures for former Prime Ministers can read high, because their total includes post-office travel entitlements while the years counted are only their time in Parliament; the default leaderboard view shows current parliamentarians only.

The compare page also shows the three most recent Prime Ministers side by side, restricted to the quarters each was Prime Minister. The data window starts in 2017Q02, so this is an indicative snapshot, not a complete ranking: one term is only partly covered, one spans the COVID period, and one is ongoing. Each profile is labelled accordingly.

Gotchas worth knowing

• Description dates can look old. Many rows reference activity dates from prior years (a 2026 Q1 payment with description “Printing and Communications 26 May 20” is a payment processed in 2026 for a service consumed in 2020). This is payments processed this quarter, not services consumed.
• Remote electorates need more travel. A Northern Territory MP’s travel is not directly comparable to a Canberra-based one. We don’t normalise for this.
• Identity stability. From 2022Q03 onwards IPEA emits a stable OfficeCode per MP. Earlier rows match by surname and first name, leaving a small number of cross-period ambiguities.
• Adjustments are not integrated. Corrections IPEA bakes into a re-released CSV flow through; ad-hoc adjustments in the separate file are not.

NDIS scale comparator

Where the homepage hero and all-time hero land a big AUD figure, a one-line comparator sits below it. For example: “About 8 hours of NDIS scheme spend”. The intent is to size IPEA totals against a real federal program, not to invite a budget comparison.

The basis is $46.04 billion paid by the NDIS in 2024–25 (NDIA Quarterly Report Supplement E.111). Divided by 525,960 minutes in a year, that works out to roughly $87,528 per minute. The page picks the unit (minutes, hours, days, weeks, months, or years) that lands closest to a 1-99 reading, then rounds.

Two rules bound this comparator. It only appears on parliament-wide totals (homepage hero, all-time hero), never on individual MP pages. Cross-program framing on a personalised story risks reading as an accusation against the named MP, which the voice rules prohibit. It is a scale comparator, not a like-for-like budget comparison: NDIS scheme spend funds disability supports for participants, parliamentary expenditure funds the operations of the parliament.

NDIS scheme spend

The NDIS section reports program-level spend, not individual MP claims. Two sources back the numbers:

• NDIA Quarterly Report Supplement E.111 (national, annual financial-year totals back to 2013-14) for the headline paid + committed figures and the 12-year trend.
• RoGS table 15A.15 (annual NDIS payments by state and territory, back to 2015-16) for the jurisdiction breakdown. RoGS publishes in 2024-25 dollars (real); the growth figure on the page is nominal because it’s computed from the NDIA actuals to keep reproducibility one click from the source.

Why NDIS is on the site at all. Pollywatch tracks how public money is spent. NDIS is one of the largest and fastest-growing federal programs; tracking its scale is part of the same accountability scope as MP expenditure. Nothing on the NDIS section attributes scheme spend to any individual MP. The framing rule is structural: “the scheme paid”, “the agency committed”. Per-MP pages carry no NDIS comparators.

Refresh cadence. Annual from RoGS each January. Quarterly from Supplement E. The state-specific (F-M), utilisation (Q) and access (R) supplements move to a six-monthly cycle after Q3 2025-26. Flagged here so a future refresh that doesn’t find a new quarterly isn’t mistaken for missing data.

Audit trail. The NDIS audit trail page lists every completed ANAO performance audit of the NDIA, newest first, with the Auditor-General’s headline finding paraphrased and a link to the source report. What we include: ANAO performance audits. What we exclude: ANAO financial audits, internal NDIA reviews, parliamentary committee findings (those may come in a later slice but aren’t ANAO). Verbs are ANAO’s own: “found”, “concluded”, “recommended”. Pollywatch adds no editorial framing on top.

NDIS per-participant cost

The per-participant section on the NDIS page combines two tables from RoGS 2026 Section 15:

• Table 15A.15 provides total NDIS payments by state and territory, by financial year, in real 2024-25 dollars. This is the numerator.
• Table 15A.6 provides the count of active participants (everyone with an approved plan at 30 June) by state and territory, by year. This is the denominator.

The per-participant figure is total payments divided by active participants, so the three figures shown for each state (total paid, participants, per participant) reconcile exactly.

A note on the Productivity Commission’s own figure. The PC also publishes a pre-computed per-participant number (Table 15A.33) that runs slightly higher, by roughly 6 percent in 2024-25. It annualises each participant’s payments, prorating to a full 12-month year based on how many days they held an active plan, which for a growing scheme lifts the figure above a simple division by the end-of-June count. We show the simple division because the two inputs are on the page and a reader can check the arithmetic.

Active participant definition. An active participant is anyone who holds an approved NDIS plan at 30 June, regardless of whether they used any of their funding that year. This differs from “participants who drew funding”, which would be a smaller count and would produce a higher per-person figure. The state-level count of participants who drew funding is not published by the PC; we therefore use the active participant count from 15A.6 throughout.

Dollar basis. All three tables are in real 2024-25 dollars (CPI-adjusted by the PC). The figures are directly comparable across years without further adjustment.

Geography. State and territory figures use the participant’s residing address (where they live), not the provider location. This is the consistent basis the PC uses across all three tables. Years 2019-20 through 2024-25 are shown; 2018-19 and earlier used a different bilateral-phasing basis and are excluded to avoid a series break.

Immigration data

The immigration page publishes Net Overseas Migration (NOM) as the ABS reports it. NOM is the official measure: migrant arrivals minus migrant departures, on the ABS’s “12/16 month rule” residency basis. We draw on two ABS products. The annual financial-year series, the state and country breakdowns, and the arrivals-and-departures decomposition come from ABS Overseas Migration. The headline and the by-quarter view come from ABS National, state and territory population, a separate release.

The headline is the rolling twelve-month NOM: the sum of the four most recent quarters, from ABS National, state and territory population (national NOM, series A2133254C). It is published about twelve weeks after each quarter ends, so it is the freshest official figure. The by-quarter view shows the same series one quarter at a time, with the financial year still in progress shaded lighter. NOM is strongly seasonal, with the September quarter typically the largest, so we never compare a part-year total against a full year; the rolling twelve-month figure stays comparable because it is always a full year.

The annual trend uses ABS Overseas Migration, Table 1.1, by financial year. ABS marks its most recent year as preliminary (an “(e)” flag) because it can revise the figure for up to twelve months; we note that preliminary year on the page and keep the trend, the state and country breakdowns, and every data table to finalised actuals only, so the long series stays stable. The two ABS products measure the same NOM concept but on different vintages and period bases, so recent quarterly and annual figures can differ slightly; we present them separately rather than splicing them. Forward projections from the Centre for Population are not used.

Three further breakdowns come from ABS Overseas Migration. Arrivals and departures come from ABS Overseas Migration DO004, Table 4.1; NOM is arrivals minus departures, and the two net back to the headline subject to ABS rounding. NOM by state and territory comes from DO001 Tables 1.2 to 1.9; the eight jurisdictions are independently rounded and may not sum exactly to the national total. Top countries over time traces the eight largest countries of birth in the latest year across the whole series, using DO001 Table 1.1. Country labels are left as the ABS publishes them.

Four related measures are sometimes confused with NOM. We use NOM and only NOM in the headline:

• Overseas Arrivals and Departures (OAD) counts movements at the border, not people. A single person can be counted multiple times in a year. OAD totals are larger than NOM.
• Census country-of-birth is a five-year stock snapshot of who currently lives here. It does not describe flows.
• Settler arrivals is a historical Home Affairs series of permanent visa holders arriving. It excludes temporary residents who become permanent in-country and excludes departures.
• Migration program planning levels is a Cabinet number (how many permanent visas the government plans to grant). It is a policy input, not an outcome.

ABS sometimes confidentialises small country values to protect identity (Table 1.1 note b). We pass these through unchanged. The country labels (e.g. “UK, CIs & IOM”) are also ABS’s own; we don’t rename them.

Migration Program: planned versus delivered

The “Migration Program, planned and delivered” section on the immigration page covers the permanent visa program, which is separate from NOM. The planning level is the government’s annual target; the delivered outcome is visa grants counting toward the program, including primary and secondary applicants and onshore status changes (people already in Australia). It is not arrivals and not NOM. The Humanitarian Program (about 20,000 places a year) and New Zealand Special Category (subclass 444) visas are excluded from the program.

This data is curated by hand from the Department of Home Affairs annual Migration Program Report PDFs, one per program year, because no machine-readable feed is reliably available. Each figure is read from that year’s report, and the published streams sum exactly to the delivered total. Where a report does not state a figure (for example, Skill-stream planning levels), it renders as a dash rather than a guess. Before 2022-23 the Child category was demand-driven and outside the planning level but is included in the delivered total, as the reports present it; from 2022-23 Child is reported within the Family stream, and the 2022-23 planning level was increased to 195,000 places in September 2022. The section is updated once a year when the next report is published.

The same reports power the skilled visas by occupation breakdown (the report’s Table 2.5, top 10 ANZSCO occupation unit groups in the Skill stream). Those counts are primary applicants only, the person whose occupation is assessed, so they do not add up to the Skill stream total, which also counts partners and children. We show each occupation’s share of the ten listed, not of the whole Skill stream, and pass the ANZSCO codes and names through as the report publishes them. Occupation history starts at 2023-24 because earlier reports do not publish a clean whole-of-Skill-stream occupation table.

Temporary skilled visas by occupation and nationality

The “Temporary skilled visas by occupation and nationality” section covers the subclass 482 visa (Temporary Skill Shortage, and from late 2024 the Skills in Demand stream). This is a temporary employer-sponsored work visa, a separate system from both NOM and the permanent Migration Program, so its occupation figures are not comparable to the skilled-visa figures above. The data comes from the Department of Home Affairs Temporary Work (skilled) visa program dataset on data.gov.au, and counts visa grants in the financial year, primary applicants only, by ANZSCO occupation unit group and citizenship country.

History starts at 2018-19, the first full year of the 482 (subclass 482 replaced subclass 457 in March 2018, so earlier years belong to the closed 457 program and are not shown). The source publishes exact, un-suppressed counts; to protect privacy we combine any citizenship group with fewer than five grants for an occupation in a year into a single “fewer than five each” row, so the listed nationalities plus that row still add up to the occupation total. ANZSCO codes and names are passed through as the department records them.

Public sector workforce data

The public sector page publishes Australian public sector employee jobs as the ABS reports them, with OECD figures alongside for international context. The headline number combines two ABS releases: employee jobs come from Public Sector Employment and Earnings (6248.0.55.002) and the per-1,000 denominator comes from National, state and territory population (3101.0, Table 1). The international comparison comes from OECD’s Government at a Glance 2025.

Three definitions are routinely conflated in reporting on the public sector. They are different counts:

• Australian Public Service (APS) is the people engaged under the Public Service Act in around 102 agencies. It is the federal narrative number reported in the APS Employment Data Release and the State of the Service Report.
• Commonwealth public sector (in ABS PSEE terms) adds the ADF, parliamentary departments, statutory authorities and Commonwealth public corporations on top of the APS. The gap relative to APS is roughly 187,000 jobs.
• General government is the OECD’s unit and excludes public corporations (e.g. Australia Post, government-owned utilities). It is narrower than the ABS public sector total and is what the OECD percentage on this page refers to.

The figures on the page are headcount (employee jobs), not full-time-equivalent. ABS’s federation-wide series is built from Single Touch Payroll administrative data and publishes headcount only. State and territory public service commissions publish their own FTE numbers using methodologies that are not strictly comparable across jurisdictions; those are not aggregated on this page.

The trend chart starts in 2021–22. ABS replaced its survey-based collection with Single Touch Payroll administrative data in 2021-22, which produced a methodology break. Pre-2021-22 numbers exist but on a different basis, and Pollywatch does not splice the two series.

Cross-country comparisons of public sector size are genuinely difficult. Australia delivers public schools and public hospitals through state government employees, which counts as general government. Some OECD members deliver equivalent services through public corporations or contractors that are excluded from the same general-government total. The OECD’s own Working Paper 76 (2024) discusses these scope mismatches. The page uses OECD median and interquartile range rather than naming a single comparator country because every paired comparison can be argued.

Two OECD members are absent from the international panel: New Zealand and Colombia. The 2025 edition of Government at a Glance publishes other indicators for both, but not general government employment as a percentage of total employment. Most likely cause is that their national statistical agencies don’t classify employment by SNA institutional sector in the form OECD needs to harmonise the figure. The chart therefore covers 36 of 38 OECD members.

Pollywatch does not flag outliers on this page. Public sector size is a politically charged subject; the editorial discipline that applies to the IPEA pages binds even more strictly here. The page publishes what ABS and the OECD have published, with neutral framing and source links, and does not editorialise on whether Australia is big or small.

Federal contracts

The contracts page publishes every federal contract notice from AusTender, the Department of Finance system that records every contract awarded by a federal agency at or above $10,000 inc. GST. The OCDS API is the live source of record. A historical archive is documented below for completeness but is not yet merged, so the contracts pages begin in FY2012-13.

• OCDS API at api.tenders.gov.au for 2013 onward, the live source of record. Cursor-paginated, week-window queries, OCDS 1.1 JSON. CC-BY 3.0 AU.
• Historical Australian Government Contract Notice Data on data.gov.au for 1999 to 2012. Annual XLSX and CSV files. Frozen since January 2024; the OCDS API supersedes it for any year from 2013-14 onward. CC-BY 3.0 AU.

The full schema and the era-by-era column aliasing live in data/contracts/SCHEMA.md; the source-research notes in docs/CONTRACTS_DATA_SOURCES.md.

What “awarded” means

We report what agencies awarded under contract notices, not what they paid against invoices. The dollar amount is Finance’s ApplicableValue (life-of-contract total inc. GST); variations and amendments appear as separate rows with their own value. We count each contract in the financial year it was first published on AusTender, which agencies must do within 42 days of awarding it, so the first publication tracks when the contract was awarded. AusTender periodically re-publishes old notices in bulk, re-stamping them with a later date; we use each contract’s earliest publication so a re-publish never shifts it into the wrong year.

Supplier identity

Supplier names are not stable across rows: the same firm appears under multiple spellings even within one FY. We collapse by ABN, which is the stable identifier the data dictionary mandates. ABN-exempt suppliers (typically overseas) are shown but not ranked the same way (they have no stable join key). Big Eight consulting firms span multiple legal entities each, so we hand-curate an ABN allow-list at data/contracts/aliases.csv and treat the set as one firm.

What we don’t do

• No state or local government contracts. Federal AusTender data only. State eTendering systems have their own publishing regimes.
• No per-MP attribution. Contracts are awarded by agencies, not by MPs. The per-MP pages do not carry contracts panels; the contracts pages do not name MPs.
• No invoice-level spend. AusTender publishes awards, not payments. “Awarded” doesn’t mean “spent in cash this FY”.
• No spreading multi-year contracts across years. A multi-year contract’s full life-of-contract value is counted once, in the year it was awarded, not apportioned across the years it runs.
• Below-threshold and intelligence-exempt contracts are not in the data. Procurements under $10,000 inc. GST don’t generate a contract notice. A small number of Defence intelligence contracts under exemption App A:13 are not published at all.

Donations data

The donations page publishes every receipt above the annual disclosure threshold declared to the AEC Transparency Register. Data covers annual returns from 1998-99 onward.

Source

The AEC publishes a single “All Annual Data” ZIP export containing every return type: political party, associated entity, significant third party, political campaigner, and member of parliament. We extract the “Detailed Receipts” CSV, which carries one row per disclosed receipt. The export is re-downloaded on each refresh; the AEC does not provide a Last-Modified header, so we cache by download timestamp.

Disclosure threshold

Donors and recipients must disclose amounts at or above the annual threshold set under the Commonwealth Electoral Act. The threshold is indexed annually (currently around $16,900 for 2024-25). Donations below the threshold are not in the data. This means the total declared is a floor, not a ceiling: the true total is higher.

What we don’t do

• No state donation registers. Federal AEC data only. State electoral commissions publish separately.
• No donor-to-policy inference. A donation does not prove influence. The data says who gave, to whom, and how much.
• No election returns (yet). This page covers annual returns only. Election-specific returns are a separate AEC export with a different schema.

Government projects

The projects page tracks 104 federal and state government projects that an auditor-general has reported with a documented cost increase. It spans 5 jurisdictions: the Commonwealth and the four states with a standing major-projects audit series (New South Wales, Victoria, Queensland and Western Australia). Every project links to the audit report it is drawn from.

The figures are curated by hand from auditor-general reports, because no single machine-readable feed of project cost histories exists across the federation. Federal projects come from the Australian National Audit Office (performance audits and the annual Defence Major Projects Report); state projects come from each state’s audit office major-projects series (VAGO, QAO, the Audit Office of New South Wales, and the WA Office of the Auditor General). Each row records the figure as that report states it.

What “cost increase” means

The overrun is current cost minus original budget, in nominal dollars. The original budget is the first approved or announced figure the audit report cites; the current cost is the latest estimate or final outturn in the most recent report. The percentage is the increase relative to the original budget. Schedule slippage, where the report states it, is shown alongside as a separate delay figure and is never folded into the dollar overrun.

What we don’t do

• Not an exhaustive list. The page carries projects an auditor has flagged with a documented increase, not every government project. A project’s absence is not evidence it stayed on budget.
• Baselines are not uniform. Audit offices set the original budget at different points (first announcement, first business case, or contract award) and use methodologies that are not strictly comparable across jurisdictions. The dollar figures are each internally consistent to their own report, not normalised against each other.
• No inflation adjustment. Budgets and costs are the nominal figures the reports publish. Part of any multi-year increase reflects general price rises, not scope or management.
• No per-MP attribution. Projects are delivered by agencies and governments, not by named MPs. The per-MP pages carry no projects panels; the projects page names no MPs.

News feed

The news feed gathers Australian coverage of the subjects Pollywatch tracks: government spending and integrity, MP expenses, federal contracts, political donations, the NDIS, immigration, the public sector workforce, and major projects. It is a noticeboard, not a publisher. The headlines and summaries are the words of the outlet that published them, shown verbatim and attributed. Pollywatch writes no headline, rates no story, and adds no verdict. Our editorial act is selection and grouping by subject, nothing more.

• Sources are a fixed allowlist. Stories come only from a set list of Australian outlets that have an editorial standards and corrections process: the ABC, The Guardian Australia, news.com.au, The Conversation, Michael West Media, The Mandarin, SBS News, The Australian, the Australian Financial Review, The Sydney Morning Herald, The Age, InnovationAus, Independent Australia, Sky News and The Saturday Paper. The list spans the spectrum without including fringe or hyperpartisan outlets for false balance. A few of these no longer publish a public feed, so their headlines are gathered through Google News and their links route via Google to the source. To keep the mix even, no single outlet takes more than a few slots.
• Selection is by subject, not verdict. Each item is checked for whether it reports an event or finding about one of the subjects above, at the federal level. The check uses keyword rules and, where available, a language model, to decide inclusion only. The model never writes or rewrites any text a reader sees.
• One card per event. When several outlets cover the same story, we show a single card with a count of how many outlets reported it. More outlets means broader corroboration. The headline we surface is the most plainly worded available, then the earliest published.
• Mentions are literal. An article is linked to a parliamentarian only when their full name appears in the headline or summary, and only for sitting members. A shared surname alone is never enough.
• We link out, we do not republish. Every card links to the original article. We show a headline, a one sentence summary, the outlet and the date, then send the reader to the source. A Paywall label marks outlets that require a subscription, so the reader knows before clicking.
• Freshness. The feed is rebuilt through the day, so the relative times read against the most recent build, not the current minute.