The surplus trick

1. How the budget numbers work

General government (GG) net operating balance, the "operating surplus": revenue minus expenses for core departments only (health, education, police and so on). It is accrual-based and excludes capital spending: building a hospital or rail line is an asset purchase, not an expense, so only depreciation hits the operating statement. This is the headline figure in every media statement ("eighth consecutive surplus") [1][8].
Total public sector (TPS) net operating balance: the same idea, but it adds public non-financial corporations (Synergy, Western Power, Water Corp, ports, PTA, DevelopmentWA) and financial corporations. It is rarely headlined.
Cash surplus/deficit: operating cash flows minus net spending on infrastructure minus dividends. Treasury's own Reader's Guide says a surplus "is generally consistent with declining levels of net debt while a deficit indicates a need for more borrowing" [1]. Net lending/borrowing is the accrual (GFS) equivalent.
Net debt: gross borrowings (including leases) minus cash and liquid financial assets [5].
How a surplus coexists with rising debt: Fact: in 2024-25 the GG operating surplus was $3.65bn, but the TPS cash result was a $1.7bn deficit because of the $13.1bn Asset Investment Program (AIP), almost two thirds of it delivered by public corporations [2]. The 2026-27 Budget forecasts GG operating surpluses every year while TPS cash deficits run $5.1bn (2026-27) and $2.5bn (2027-28), with total public sector net debt climbing from $30.2bn (June 2025) to $44.7bn (June 2029) [3].
Special Purpose Accounts (SPAs) and "funds" (Building Hospitals Fund, Social and Affordable Housing Investment Fund, the Westport land SPA and so on): money is appropriated from the surplus into an account and spent in later years. Fact: In 2021-22 the Government explicitly "applied short-term increases in the general government operating surpluses" to such initiatives, including $1.787bn to fully fund the Women and Babies Hospital and a $400m SPA for Westport land [4]. In 2025-26, $1.4bn of the surplus went into the Building Hospitals Fund, taking it to $3.4bn over four years [8]. The dedicated New Women and Babies Hospital Account is being closed by 30 June 2026: $392m was drawn in 2025-26 and the remaining $1.36bn moved into the Building Hospitals Fund [3]. That makes the hospital's cost harder to track.
Our read: SPA balances held as cash count as liquid assets, so they reduce net debt while they sit unspent. Debt then rises when the money is spent on capital, which never appears in the operating balance.

2. Techniques that flatter the headline

Spending timing. Fact: The 2025-26 surplus beat the Budget estimate, "mainly attributable to timing changes in grants and contract payments" [8]. Fact: 2024-25: the surplus was $1.1bn above the estimate "largely" because a $1bn DisabilityCare Australia Fund grant arrived a year early [2]. Our read: Both are timing effects, not structural savings.
Planned underspends. Fact: The Budget books a "global provision for underspending" because agencies routinely miss their cashflows: $2.4bn assumed for 2023-24 (actual agency underspend $1.6bn), $750m for 2024-25 (actual $1.3bn) [2][5]. The OAG notes that cashflows and completion dates are revised as projects slip [6]. Our read: Underspending makes "debt lower than forecast" a recurring headline that reflects delay, not thrift (see the table below).
Conservative iron ore assumptions, then a windfall. Fact: Treasury assumes the price "will revert to its long-run average" (US$66/t at the time) within months [7]. Actual prices have beaten the Budget assumption every year checked (table below). Our read: This produces a predictable pattern: modest surplus forecast → big "better than expected" surplus → re-allocation to funds and announcements.
GST No Worse Off reliance. Fact: The State says that without the 2018 reforms WA would get about $6bn less a year [9]. 2025-26 MYR: $8bn in GST grants, versus about $9bn under the PC's 2018 recommendation [10]. The no-worse-off top-ups are Commonwealth-funded (about $4.9bn in 2025-26) [11]. The Productivity Commission (Aug 2026) called the deal a costly mistake that should be reversed; the cost was about $23bn versus about $5bn forecast [12]. The guarantee runs only to 2029-30 [11]. Our read: A material share of every "surplus" since 2019 rests on a federal arrangement that is politically exposed after 2030.
Dividends from government businesses. Fact: The GG surplus includes dividends from public corporations. For example, the Perth Mint (Gold Corporation) dividend was withheld while it funded about $34m of AUSTRAC remediation [13]. Our read: Dividends move cash from government businesses' balance sheets (which then borrow for capex) into the GG surplus, flattering the headline without changing whole-of-state debt.
Land transfers to DevelopmentWA. Fact: The OAG found DevelopmentWA's funding model involves land being "written down in advance" on transfer, and said its "clear preference" is transfer at carrying value, so that decision-makers and the community have "full visibility" of development costs and the "loss/return to the State" [14]. Our read: Current practice obscures whether land development actually makes or loses money.
Provisions/contingencies. The Budget includes provisions (AIP underspend, the $1.7bn "Provision for Decarbonisation of the SWIS" in 2023-24 that was then largely unspent: $163m estimated actual, $0 actual line) [5]. Our read: Large unallocated provisions give flexibility to "find" money later.

3. Historical accuracy: original Budget vs actual (ARSF)

Year	GG op. balance, Budget → Actual ($m)	Revenue, Budget → Actual ($m)	TPS net debt, Budget → Actual ($bn)	AIP, Budget → Actual ($m)	Iron ore, Budget assumption → actual (US$/t)
2016-17	-3,914 → -2,474	25,681 → 26,913	33.84 → 31.96	5,862 → 5,137	n/a
2017-18	-2,340 → -618	28,457 → 29,332	37.81 → 34.61	6,001 → 5,052	67.2 → 68.8
2018-19	-906 → 1,317	29,572 → 32,006	39.10 → 35.46	6,248 → 4,965	76.5* → 80.4
2019-20	1,533 → 1,669	31,334 → 32,162	39.53 → 35.45	5,646 → 5,188	73.5 → 92.9
2020-21	1,204 → 5,838	34,139 → 40,151	39.26 → 33.48	7,549 → 5,816	96.6 → 154.5
2021-22	2,791 → 6,019	38,297 → 41,658	32.06 → 29.21	8,021 → 7,181	121.3 → 138.2
2022-23	1,592 → 5,120	38,428 → 43,601	31.11 → 27.36	8,880 → 9,235	77.5 → 109.6
2023-24	3,293 → 4,486	43,218 → 45,947	29.39 → 28.17	10,405 → 11,405	74.1 → 119.4
2024-25	2,603 → 3,652	46,194 → 49,689	32.71 → 30.17	12,079 → 13,132	75.0 → 100.9
2025-26	2,414 → 4,191	50,257 → 52,759	38.97 → 33.41	12,061 → 12,269	77.6 → 104.3

Sources: ARSF 2016-17 to 2025-26 [15][16][17][18][5][20][2][27][28][29]. *2018-19 iron ore is compared with the 2019-20 Budget estimate, as stated in the ARSF.

What the table shows:

The operating balance beat the Budget in 10 of 10 years (2016-17 to 2025-26), by $0.1bn (2019-20, the COVID year) to $4.6bn (2020-21).
Revenue beat the Budget in 10 of 10 years, by $0.8bn to $6.0bn.
Net debt came in below the Budget in 10 of 10 years, by $1.2bn to $5.8bn.
Iron ore actuals exceeded the Budget assumption in 9 of 9 years with data, by 2% (2017-18) to 61%.
AIP: underspent against the Budget in every year from 2016-17 to 2021-22 (6 of 6), by $0.46bn to $1.7bn. Since 2022-23 the actual has exceeded the original Budget figure, but each Budget now nets off a large underspend provision, so the comparison is not like-for-like.

Forecasts vs outcomes:

The 2016-17 Budget forecast 2017 net debt at $33.8bn; the actual was $32.0bn [15].
The Government itself cites a Liberal-era forecast of $43.4bn for 2019-20; the actual for 2020 was $35.4bn [21][17].
The 2025-26 Budget forecast $39.0bn for June 2026; the actual was $33.4bn [19].
The 2026-27 Budget now forecasts $44.7bn by 2029 [3].

Our read: Conservative forecasting is a stated policy and is prudent in itself. But the pattern means "better than expected" is the default outcome. Each year that becomes a political good-news story and new discretionary spending, while the structural trend (debt rising from $27.4bn in 2023 to a forecast $44.7bn in 2029) is driven by capex that sits outside the headline. Forecast debt keeps being "beaten" mostly because of underspending and windfalls, so the beat says little about future debt.

4. "Will it make money?" claims vs outcomes

Claim	What was said	What the record shows	Our read
$3.8bn Synergy renewables / coal exit (2022)	"expected to pay for itself by 2030-31 relative to the status quo of increasing electricity subsidies" [22]	The measure is relative to a counterfactual, not profit. The IPA estimates SWIS replacement capex could exceed $52bn [23] (a partisan think-tank estimate).	Unverifiable until 2030-31; the counterfactual was never published. The business case has never been published.
Westport	Sold as a strategic economic asset	Estimate $4.7bn (2020) → $7.2bn (2024), excluding road and rail	No public revenue/payback figure found. The business case and expected returns have not been published.
Optus Stadium	Economic impact assessment published by VenuesWest [24]	The West reported taxpayers footing an operating cost blowout, with a net subsidy of about $16.8m over three years after $54.5m in higher costs [25]	Runs on subsidy; not self-funding.
Elizabeth Quay	Land sales to return value	The ABC (2016) reported $1.66m of operating costs in the first two months, about $220k a month ongoing [26]	Barnett-era; the land-sale return was not found.
Perth Mint	Dividend-paying government business	Dividend withheld during ~$34m AUSTRAC remediation; the State guarantees its liabilities [13]	Revenue claim reversed for those years; the contingent risk sits with taxpayers.
Transperth fares	—	Fares covered 12.5% of Transperth system costs in 2021-22, 13.4% in 2022-23 and 11.4% in 2023-24 ($181.5m of $1,589.5m), as costs rose 16% and fares fell 1% [30]. The PTA does not publish Metronet fare revenue separately.	Passengers pay about 11 cents of every dollar it costs to run Transperth. Whether Metronet pays for itself can't be tested from published data.