The funding chain ends at a milestone payment. Every step in it moves money for building something.

The scale is substantial: $84 billion of Australian Government funding across 81 projects, with $13.5 billion flowing to states and territories in 2026-27, under the Federation Funding Agreement Schedule on Land Transport Infrastructure Projects 2024-2029.

Then a state runs the service, alone, for the life of the asset, out of an annually appropriated operating subsidy that competes against schools and hospitals — because farebox recovery in Australian cities runs at roughly $0.20 to $0.25 in the dollar of operating expenditure. Three-quarters of the cost of running the network is a state liability that no Commonwealth program touches.

This produces a predictable distortion, and it is not anybody's fault. If the only external money available is capital money, then the proposals that attract external money are the capital-intensive ones. A new rail line has a large capital component and a Commonwealth partner. Running more buses on roads that already exist has no capital component and no partner at all — it is pure operating cost, funded entirely by the state.

This entry deliberately does not claim that more buses would be better value than more rail. That is the claim a reader expects here and the evidence assembled for this page does not support it: the standard Australian reference on the question, BITRE's Benefits of Urban Public Transport Subsidies in Australia, works from 1982-83 data and in fact concluded that subsidies might be more efficient with lower frequency and lower fares. The honest claim is narrower and does not depend on which option wins: one option cannot enter the competition at all, because there is no instrument that could fund it.

So the instrument selects the answer. Not by anyone deciding that rail beats buses, but by there being a funding pathway for one and not the other. This database has recorded the same shape twice before: deferred maintenance, where operating expenditure loses annually to capital and creates a bigger capital bill a decade later; and Rent Assistance and the housing queue, where a Commonwealth payment stops at exactly the point a state asset starts. Three domains, one boundary.

Filed as an idea, deliberately. The reason for the boundary is real: recurrent Commonwealth contributions to state services create permanent liabilities, cost-shifting arguments and a ratchet that is very hard to reverse. That is why the boundary exists and why it has held.

But the shape of a narrower version is visible. Not open-ended operating subsidy — something bounded and outcome-tied:

1. Let a non-capital option win an evaluation. Where a proposal to Infrastructure Australia is compared against alternatives, a service-based alternative should be capable of being the answer — and currently it cannot be funded even if it is the better option, which means in practice it is not seriously assessed as one.
2. Time-limited service funding attached to a capital commitment, so the Commonwealth is buying patronage rather than a structure — the same logic as an activity-based hospital payment rather than a hospital building grant.
3. Publish what is actually being bought. The Commonwealth publishes its capital contribution project by project; nobody publishes the recurrent cost the state has taken on as a consequence.

Who acts: the Commonwealth, and only the Commonwealth. This is a decision about the shape of its own program.