Contract Creep

When an Australian Government agency signs a contract worth more than $10,000, it has to say so. A contract notice goes onto AusTender carrying the supplier, the purpose, the dates and the price. That notice is the public record of the deal, and the price on it is the number that gets quoted.

The price is not final. Agencies amend contracts — options are exercised, terms are extended, scope is added — and each amendment restates the contract’s total value. The data dictionary is explicit that the value at signing “should not include amendments, such as options, extensions, renewals”, and that those “are reflected in the contract value as amendments are made”. The revision is published. What is not published is a second announcement.

This investigation reconstructed every Commonwealth contract first reported between 1 July 2019 and 30 June 2024, then followed each one through every amendment made to it since — 457,606 release records covering 323,211 contracts, pulled from the Department of Finance’s own machine-readable feed. The question is simple: what is the distance between the price at signing and the price at the end?

$109.7 billion#

Those 323,211 contracts were signed at a combined $244.1 billion. After every amendment on the record, they are worth $353.8 billion. The difference is $109.7 billion, an increase of 44.9% on the value that was announced.

The declining upper band in the recent years is a maturity effect, and it is the main reason this analysis stops at June 2024 rather than running to the present. A contract signed in 2019-20 has had six years to accumulate amendments and has grown 59.4%; one signed in 2023-24 has had two, and has grown 30.3%. The 2023-24 figure is not a smaller problem. It is an unfinished one.

Amendment is not the normal case. Only 72,402 contracts — 22.4% — were ever amended at all. But those contracts are not a random sample: they hold 71.1% of all final contract value. Amendment is concentrated in the contracts that matter most.

And when a contract is amended, it does not move slightly. The median amended contract ends at 1.96× its original value, with a middle half running from 1.27× to 2.93×. Roughly half of all amended contracts at least double.

Movement runs both ways. 4,152 amended contracts finished below their original value, releasing $4.6 billion. At the other end, 1,730 contracts finished at more than ten times what they were signed at.

The line at $80,000#

Until November 2025, the Commonwealth Procurement Rules set a procurement threshold of $80,000 for non-corporate Commonwealth entities buying goods and services. Below it, an agency had wide discretion. At or above it, the additional rules in Division 2 applied, an open approach to market was the default, and an agency choosing a limited tender instead had to record a specific exemption or condition. AusTender’s own documentation states the rule plainly: contract notices “with a value at or above $80,000 and a Limited Tender procurement method will require an exemption or condition”. Thresholds are inclusive of GST. The entire window studied here sits under that single $80,000 regime.

A threshold that changes an agency’s obligations is a threshold worth looking at.

The distribution declines smoothly with size, as it should — small contracts are more numerous than large ones. It does not decline smoothly through $80,000. The $79,000–$80,000 bin holds 2,296 contracts where the fitted trend expects 886. The bin immediately above the threshold holds 1,178. Taking the last $4,000 below the line together, 5,797 contracts sit where the trend expects 3,660 — an excess of 2,137 contracts, 1.58× what the surrounding distribution predicts.

A further 629 contracts were reported at exactly $80,000.

That excess is only interesting if the estimator is not manufacturing it. Reported contract values cluster on round numbers everywhere — there are visible spikes at $60,000, $70,000, $75,000 and $99,000 that have nothing to do with any rule. So the identical procedure was pointed at five cuts that are not thresholds.

At $60,000, $70,000, $90,000, $100,000 and $110,000 the estimator returns ratios between 0.80 and 1.07 — no excess, which is what a null result should look like. At $80,000 it returns 1.58. Round-number clustering is real and it is everywhere, but it does not produce this. The bunching is specific to the threshold.

Then there is what happens afterwards. Of the 24,715 contracts signed between $60,000 and $80,000, 16.2% finish above $80,000 — about one in six. They end up at a value which, had they started there, would have required an open approach to market or a written exemption.

What could explain this instead#

The pattern above is a shape in a distribution. Several explanations fit it, and most of them are not misconduct.

Contracts are genuinely estimated near the threshold. Agencies know the threshold and scope work to sit inside it — that is not evasion, it is the rule working. A $75,000 project stays a $75,000 project because the agency deliberately kept it small. The bunching is equally consistent with agencies designing procurements to fit under the line as with agencies disguising procurements that belong over it. This analysis cannot separate the two, and the difference matters.

GST arithmetic puts round quotes on odd numbers. Reported values include GST, so a round ex-GST quote lands on an inclusive figure that looks deliberate. $79,200 is exactly $72,000 plus GST, and 324 contracts sit on that precise value. But GST rounding explains only part of the mass: of the 5,797 contracts in the last $4,000 below the line, 950 are round ex-GST figures. The remaining 4,847 are not, and the placebo test already rules out generic round-number clustering as the driver.

Amendments are how procurement is supposed to work. Multi-year contracts carry option periods. Extending a contract by exercising an option that was competed at the outset is not a failure of competition; it is the contract performing as designed. A contract that doubles because a pre-agreed second term was taken up is a very different object from one that doubles because the work was mis-specified, and AusTender does not record which is which.

Scope changes for reasons nobody controls. Two of the five years in this window were pandemic years. Programs were expanded at short notice, and a large amendment in 2020 or 2021 may record an emergency, not a drafting failure.

Small contracts have more room to multiply. A contract signed at $50,000 that gains $200,000 of work is a 5× growth; the same $200,000 on a $10 million contract is a rounding error. Ratios flatter small contracts. That is why the dollar totals and the ratio distributions are reported separately here, and why the concentration figures below matter more than the medians.

A prediction that failed#

Five hypotheses were written into the project record before any figure was computed. Four survived. One did not, and it was the one with the clearest story attached.

The prediction was that limited tender contracts — awarded without an open approach to market — would creep more than openly tendered ones. The reasoning was that a contract awarded without competition faces less scrutiny at the outset and would be easier to grow later.

It is wrong, and not marginally.

Openly tendered contracts are amended 34.8% of the time; limited tenders, 12.2%. When they are amended, open tenders grow by a median 2.00× against 1.68× for limited tenders. Both measures point the same way, against the prediction.

The likely reason is visible in the same data and undermines any clean reading: the median open tender is signed at $126,750, the median limited tender at $33,518. Limited tenders are mostly small, short purchases that were never going to be extended. Open tenders are the large, long, complex arrangements that attract amendments for legitimate reasons. The comparison is confounded by size and duration, and this analysis does not disentangle it. The prediction failed; the explanation for why it failed is not settled by the evidence here.

Duration is the strongest single predictor of amendment in the data, which points the same way.

Contracts running under three months are amended 3.3% of the time. Contracts running two to three years are amended 61.5% of the time, and when amended they grow by a median 2.65×. A long contract is not a fixed price. It is an opening position.

Where the money is#

The $109.7 billion is not spread across the Commonwealth. It is concentrated to a degree that makes most of the sector-wide averages above slightly misleading.

The Department of Defence alone accounts for $61.9 billion of the $109.7 billion — 56% of all value added across the Commonwealth. Defence signed 126,259 contracts in the window and amended only 13.2% of them, but its contracts are large and long, and they grew 44.9% on original value. The top ten agencies together account for 84.5% of all value added.

Individual contracts concentrate harder still.

Of the 67,694 contracts that grew, the largest 1% carry 66.6% of all the value added. The single largest hundred carry 40.2%. The largest single increase in the window is a Defence contract for conventional war weapons that moved from $426.2 million to $6.68 billion across a single amendment.

Some individual movements are very large relative to where they started. A Defence construction support services contract signed at $247,600 is now recorded at $344.2 million — 1,390× its original value. Every one of these can be read on AusTender, amendment by amendment, and the biggest of them are almost certainly long-running standing arrangements whose original notice was never intended to describe the eventual scale of the work. That is precisely the point: the notice is the public record, and the public record was $247,600.

What this does not show#

No claim is made here that any contract in this dataset was improperly awarded, improperly amended, or represents waste. Amendment is lawful, expected, and frequently the efficient thing to do. Extending a working arrangement is usually cheaper than re-running a tender.

Nor does this measure the whole of Commonwealth spending. AusTender covers reported procurement by Commonwealth entities above the reporting threshold. It excludes grants, most corporate Commonwealth entity activity below its own separate threshold, and reporting practice varies between agencies.

What the data does show is a gap between two numbers that are both official. The first is announced, quoted, and treated as the cost of the decision. The second is recorded quietly, on average 45% higher, and never announced at all. Both come from the same government system. Only one of them gets read.

In November 2025 the procurement threshold rose from $80,000 to $125,000, the first increase in twenty years. If the bunching measured here is a response to where the line sits rather than to the size of the work, it should reappear underneath $125,000. That is a testable prediction, and the data to test it will exist in a few years.