Federal Spending Explorer

About · Methodology

How we prepare the data

Every figure on this site comes from a published government document. Where we combine, remove or adjust anything, the figure is labelled and the rule is described here. Source files and every data load are listed on Sources and changes.

Structure

The top figure is total expenses of the Australian Government general government sector, from Budget Paper No. 1. Below it, spending is broken down by portfolio, entity and program using the Portfolio Budget Statements (PBS) program expense line items, published on data.gov.au with each Budget.

Estimates, not actuals

The 2026-27 figures are Budget estimates. The 2025-26 figures are estimated actuals published in the same Budget, so both years use the same program structure. Final outcomes will be added separately and labelled as actuals.

Amounts not counted

The PBS tables list some money in two places, for example once where a department pays it and again where the receiving agency spends it. Counting both would overstate spending. These rules remove the second listing. Each removed line is still shown on its program page, marked “not counted”.

subtotal rows subtotal
A total row inside the table (for example "Totals for Outcome 1"); its parts are already counted.
funding already counted at the paying department corporate-commonwealth-funded
Corporate entity funding paid by its portfolio department. The department already records this as an administered expense, so counting it again at the corporate entity would double count it. In 2026-27: $50.4bn across 79 lines.
Medicare Guarantee Fund credits medicare-guarantee-fund-credit
Treasury credits the Medicare Guarantee Fund Special Account. Health records the medical and pharmaceutical benefits paid out of it. In 2026-27: $59.5bn across 1 line.
tax bad and doubtful debts tax-bad-debts
ATO bad and doubtful debts and remissions. The Budget's government finance statistics treat these as write-downs of assets (an "other economic flow"), not expenses, so they sit outside the Budget Paper 1 total. In 2026-27: $9.6bn across 1 line.

Whole-of-government adjustments

After those rules, program lines total $845.7bn against published total expenses of $833.3bn. The difference, −$12.5bn (−1.5%), is shown as a separate, labelled slice rather than spread across programs. It is likely to include:

We will reduce this figure as specific causes are confirmed, and record each change on Sources and changes.

Budget against actual

Each year has three published figures for total expenses. The year’s own Budget gives an estimate before the year starts. The next Budget, delivered before the year ends, gives a revised estimate. The Final Budget Outcome, published about three months after the year ends, gives the actual. The whole-of-government page shows all three by function for the latest year with an outcome. The difference is the actual less the original Budget estimate.

The figures are as Treasury publishes them in Budget Paper No. 1 and in Table A.1 of the Final Budget Outcome, in millions of dollars. Both use the same functions, so they compare directly. Functions sum to the total within rounding.

Budget estimates include the contingency reserve, under other purposes. It is an allowance for decisions taken but not yet announced and for spending that tends to come in below estimate. The outcome has no contingency reserve, because spending is recorded under the function where it happened. Part of any difference in other purposes is this allowance.

The Final Budget Outcome reports actuals for the whole of government only, by function and by type of expense. It does not report actuals by portfolio or entity, so those pages show Budget estimates and estimated actuals from the Portfolio Budget Statements. Each entity’s actual expenses appear in its annual report.

Sub-functions

Appendix A of Budget Paper No. 1, Statement 6, splits each function into sub-functions. Functions with no sub-functions, such as defence, appear there as a single line. The site checks on load that each function’s sub-functions add up to its total, within rounding.

Budget Paper No. 1 doesn’t split general purpose inter-government transactions, under other purposes. Budget Paper No. 3 does: the states’ GST entitlement, HFE transition payments and other general revenue assistance (Table 3.1), plus the Financial Assistance Grant program to local councils. Those amounts don’t quite add up to the Budget Paper No. 1 figure, and Budget Paper No. 3 gives no reason. The difference is shown as its own row, marked reconciliation, rather than spread across the other rows.

Contracts and annualised estimates

Contract notices come from AusTender, where agencies must report contracts worth $10,000 or more. A notice gives the contract’s total value over its whole life and its start and end dates, not when payments are made. To compare contracts with a year’s spending, we spread each contract’s current value evenly by day from start to end date. Every figure produced this way is labelled estimate: annualised.

This overstates or understates any single year when payments are uneven, and contract values are commitments that may not be spent in full. Contracts are matched to entities by the agency name on the notice; AusTender does not record which program a contract belongs to.

Grants

Grant awards come from GrantConnect, where entities must publish each grant within 21 days of the agreement taking effect. Like a contract notice, an award gives the grant’s total value and its start and end dates, so we spread each award’s value evenly over its term in the same way and label the result estimate: annualised. An amended award is published as a new version; we use the latest. Each box’s “traced to named suppliers and grant recipients” share adds these annualised grant values to the annualised contract values.

A grant “without an open round” is one with no grant opportunity listed on GrantConnect, such as a one-off or ad hoc grant. These are allowed under the Commonwealth Grants Rules and Principles, for example for emergencies or where only one organisation can deliver.

Some awards name no single recipient: agencies may report many small payments as one aggregated award (for example, disaster payments) or withhold a recipient’s name. These are counted in totals but not listed among recipients. GrantConnect has no download service, so we export its award lists by hand; its export repeats blocks of rows in large searches, and we count each award once by its ID.

Contract spending reported by entities

Each non-corporate entity reports in its annual report what it paid in the year on consultancy contracts and on other contracts, and the organisations that received the largest shares. We take these figures from the Department of Finance’s Transparency Portal. They are cash payments including GST, so they differ from AusTender contract values, which are commitments over a contract’s whole life. Corporate entities do not report these figures.

Entities enter the figures themselves, and some have entered dollars where thousands were asked for, which makes a figure 1,000 times too large. We show every figure as published, but check each one, and mark a figure looks wrong with the reason when it:

None of these figures is used in any total, ranking or signal on this site. Where a body reported separately before it was absorbed into another, its figures are not shown on the successor’s page.

Entity names

An entity keeps one page across budget years. Where a department is renamed, or a name is misspelt or shortened in the PBS tables or in AusTender, we match it to the entity by a reviewed list, and the entity’s page shows the other names it is published under. Contracts from an agency that has since been abolished or merged are counted under the entity that took over its work, for example the Administrative Appeals Tribunal under the Administrative Review Tribunal.

Some AusTender agencies are not matched to any entity, so their contracts are not counted in any total:

Limited tender

A limited tender is a procurement where the agency approaches one or more suppliers directly instead of advertising openly. The Commonwealth Procurement Rules allow it in defined cases, such as urgency, a single capable supplier, additional deliveries from an existing supplier, or work exempt under Appendix A (for example, engaging legal counsel). The reason the agency gave is shown on each contract page.

Below the open-tender threshold, agencies may approach suppliers directly without giving a reason, so most notices under the threshold are limited tenders. This is why most contracts by number are limited tenders, while the share by value is lower. The signals count limited tenders only at or above the threshold.

The threshold depends on the agency and what is bought (Commonwealth Procurement Rules, paragraph 9.7):

Amendments

Agencies publish an amendment when a contract’s value or term changes. We record the value first published and every later version, and report the increase from the first published value. Increases can reflect extra work, extended terms or corrections.

Suppliers who are individuals

Some suppliers are people, such as barristers, sole traders and landholders. Their contracts appear on contract and supplier pages under the name AusTender publishes, and count in every total, but they are not listed in rankings such as “largest suppliers” or in the contract lists on signal pages.

A supplier counts as an individual when the Australian Business Register records its ABN as an individual or sole trader. Suppliers with no ABR record, mostly overseas, count as individuals when the name reads as a person’s: two to four words, no organisation words, starting with an initial or a given name common in the register. This test errs toward leaving a supplier out.

Foreign Military Sales

Foreign Military Sales (FMS) is the US Government program for selling military equipment, training and support to allied governments. Under it, Defence doesn’t contract with the US manufacturer. It agrees a case with the US Department of Defense, which buys the equipment on Australia’s behalf. Defence pays through an account at the Reserve Bank of Australia, so AusTender names the supplier as “FMS ACCOUNT RESERVE BANK OF AUSTRALIA”. It is the US Government, not a company, and not the Reserve Bank.

These notices don’t name the manufacturers, and their descriptions are only the category, such as “Missiles” or “Aircraft”, so this spending can’t be traced past the US Government. The agency reference on each notice is the FMS case identifier, such as ATDSAI: AT for Australia, a letter for the US service managing the case (B for Army, P for Navy, D for Air Force), and the case’s own code.

Corporate groups of suppliers

AusTender lists each company under its own name and ABN, so one business can appear many times: a multinational’s Australian and overseas subsidiaries, or a professional firm’s partnership and its service companies. The Suppliers page combines these into corporate groups by default; a switch shows each ABN separately. Each group’s page lists its member companies and why each is included. Groups change only rankings; every total is the same either way.

A company joins a group in one of two ways:

We do not join a company that contracts as trustee or custodian for a trust to its parent, because the trust’s property is not the company’s. Individuals are never grouped. Joint ventures and brands with unrelated names (such as a firm’s acquisitions trading under their old names) are not detected unless the LEI register links them.

AI classification

AusTender’s categories (UNSPSC codes) are broad: “management and business professionals” covers consultants, contractors filling roles, lawyers and auditors alike. To tell these apart, we ask a Claude model to read each contract’s description, agency and supplier and choose one category from a fixed list. It runs in batches with a fixed prompt; each result records the model, prompt version and a fingerprint of the text it read, and is labelled AI-classified wherever it appears.

The model can be wrong, especially when descriptions are short. Results marked low confidence are shown but not counted in totals. The model is never asked to judge value for money or wrongdoing. In a check of 55 results chosen at random across the categories, a reviewer agreed with 91%.

Entity pages set the consultancy an entity reported for a year beside the contracts classified as consultancy for the same year. The two are measured differently and are not expected to match: the reported figure is cash paid in the year on contracts the entity itself counted as consultancies, under the Department of Finance’s definition; ours spreads each AusTender contract’s value over its term, and uses the model’s reading of the description rather than the entity’s choice.

Loans and investments

When the government lends money or buys equity in a company it owns, the budget records an asset bought, not an expense. These amounts are left out of total expenses and the underlying cash balance, but they are added to the headline cash balance, and the government borrows to pay them. Examples are student loans, the Clean Energy Finance Corporation, Housing Australia and Snowy Hydro.

The Loans and investments page shows them as Treasury itemises them in Budget Paper No. 1, in the table that reconciles the underlying and headline cash balances, and the year’s actual from the Final Budget Outcome. Amounts are net: new loans and equity less repayments and returns of equity in the same year. A negative amount means more came back than went out.

Treasury groups some amounts as “Net other”: commercial-in-confidence transactions and decisions not yet announced. We show that line as published. A loss on a loan or investment, such as a write-down of a government company’s value, does not appear in these amounts.

The tables are only published in PDF. We extract them with a script and check that the items add up to the published total in every column.

Project cost estimates

For major projects funded through loans and investments, such as Snowy 2.0 and Inland Rail, we list every published estimate of the project’s cost on the Major projects pages: the first public figure and each official revision. We compile these by hand from ministerial announcements, the delivering company’s reports, audit reports, Budget papers and Senate Estimates. Each estimate links to its source and quotes the words containing the figure. It is dated when it was made if the source says so, for example an estimate made at an investment decision but published years later, and otherwise when it was published. Where a video of the announcement or of news coverage exists on an official or broadcaster’s channel, we link to it. We use a news report only where no government or company source can be found, and label it.

Estimates are shown as published, in dollars of the day, not adjusted for inflation. They often differ in what they cover, for example whether transmission lines or financing costs are included. Where the scope or price basis changed from one estimate to the next, for example from 2010 dollars to dollars of the day, the estimate is marked “not like for like”, and the multiple of the first estimate should be read with that in mind. These estimates are the project’s total cost, which is usually more than the government’s loans and equity in any one year.

Known gaps in the source data