Australia treated the veteran compensation backlog as a batch of claims to be cleared with money and staff, and nothing was built that would notice it forming again — no maximum decision time, no published target for claims on hand, no staffing rule tied to intake

What is happening?

The Australian Department of Veterans Affairs decides compensation and rehabilitation claims under three separate statutes — the Veterans Entitlements Act 1986, the Safety, Rehabilitation and Compensation (Defence-related Claims) Act 1988, and the Military Rehabilitation and Compensation Act 2004. In 2019 the Productivity Commission reported that this three-scheme arrangement was out of date, overly complex and poorly administered, and recommended consolidating it into two schemes by July 2025. That recommendation was not implemented on that timetable.

In August 2022 the interim report of the Royal Commission into Defence and Veteran Suicide placed the claims backlog among 13 urgent recommendations. The Commonwealth attached 298 million Australian dollars over four years across the 2022-23 and 2023-24 budgets and added 500 claims staff. On 2024-02-13 the responsible minister announced that the backlog the Royal Commission had identified — 41,799 claims — had been eliminated ahead of the deadline the Commission had set for 2024-03-31.

The stock then grew back. Claims on hand ran 82,645 at 2025-06, 95,650 at 2026-02 and 103,074 at 2026-06. The sharper movement is inside that total. Claims with no delegate assigned to them at all, recorded as unallocated, ran 7,784 at 2025-02, 29,989 at 2026-02 and 32,925 at 2026-06 — more than four times the February 2025 figure in sixteen months. Over the same period the staff assigned to process claims fell from 973 full-time equivalent at 2025-06 to 867 at 2026-02, a reduction of about 11 percent.

Demand rose on both counts that determine workload. Lodgements reached 101,157 in the 2024-25 financial year, 13 percent above the year before, and 109,221 in 2025-26 to June 2026, 8 percent above the same point a year earlier. The average number of conditions claimed in a single claim moved from 2.6 at 2022-05 to 4.55 at 2025-06, easing to 4.1 at 2026-01 — so each claim carries roughly one and a half times the assessment work it carried four years earlier, even after the recent fall.

Average processing time for an initial liability claim under the 2004 Act improved for three years and then reversed: about 435 days in 2022-23, about 368 days in 2023-24, about 315 days in 2024-25, and 364 days in 2025-26 measured to June 2026. The last of those four is a partial-year figure sitting beside three full-year averages, and it is not measured the same way as the three before it. From 2025-04-01 the department stopped the processing clock for the period during which it is waiting on medical records requested from the veteran. The opposition spokesperson on veterans affairs said the change distorts comparison with earlier years, and a veteran advocate said it shows only the flattering side of the record.

The structural remedy commences on 2026-07-01. The VETS Act, passed on 2025-02-13, folds the three statutes into a single amended 2004 Act. It is not retrospective. Claims lodged before that date continue under the law they were lodged under, so none of the 103,074 claims on hand at June 2026 is governed by it.

Whose problem is this?

The body that decides the claims is the same body whose staffing level determines how fast they are decided, and neither the volume of claims arriving nor the number of staff available to meet it is set by anyone who waits in the queue.

Where does this problem end?

The boundary here separates a stock of claims from the condition of the people waiting inside it. This document measures how many claims are pending and how long they take, because those are the quantities the department publishes. It does not measure what the waiting does, and no source opened here quantifies that either.

What is the state now, and what should it be?

Now

Needs a new measurementthe target state: no source opened here names a level of claims on hand that the department is working toward, a maximum time within which a claim must be decided, or a staffing level that intake is meant to be matched against. The one target that ever existed in this record was a fixed historical batch of 41,799 claims with a date attached to it, set by a Royal Commission rather than by statute, and it expired when that batch was cleared. What should replace it is exactly the thing every opened source is silent on, and that silence is not a gap in the research. It is the finding.

How big is it?

103,074 claims were on hand at June 2026, and 32,925 of them had no delegate assigned. That second number is the one that describes the state most directly, because an unallocated claim is not being worked on by anyone. It stood at 7,784 in February 2025 and at 29,989 in February 2026.

A claim is not a person and no source opened here converts one into the other. The department publishes counts of claims, and the affected population is a count of people, and nothing available bridges the two. The average of 4.55 conditions per claim at June 2026 measures how many separate conditions are named inside one claim, not how many claims one veteran lodges, so it cannot serve as the conversion. The person-level figures that do exist point at a different question — around 247,000 veterans receiving departmental support of some kind, more than 165,000 veterans and family members supported through compensation in 2024-25, and 355,848 having accessed a departmental service by June 2025 — because all three count people already being served rather than people waiting to be decided. The impact population is therefore recorded as not derivable, and the missing term is stated with it.

The wider frame is easier to size. The 2021 census recorded 581,139 people with Australian Defence Force service experience, comprising 84,865 serving or in the reserves and 496,276 who had left. The only person-level figure in this document that came from a source read directly is older and narrower: departmental support reached 166,000 veterans and 117,000 dependants in 2017-18, at about 13.2 billion Australian dollars a year.

Under what conditions does it arise?

1. The backlog was defined as a batch, not as a flow. What the Royal Commission named and the government funded was a specific set of 41,799 claims. Clearing a fixed batch is a project with an end date, and it ended. Nothing in that definition addressed the rate at which new claims arrive against the rate at which they can be decided, so the arithmetic that produced the batch in the first place was left running.

2. There is no maximum time in which a claim must be decided. An agency with no statutory deadline cannot be late, it can only be slow. No source opened here identifies a legislated processing time, a target level of claims on hand, or any consequence attached to exceeding either. Delay in that arrangement is not a breach of anything, so it generates no event that anybody has to answer for.

3. Capacity and demand moved in opposite directions. Lodgements rose 13 percent in 2024-25 and a further 8 percent in 2025-26 to June, and the average claim carried more conditions to assess than it did four years earlier. The claims workforce fell from 973 full-time equivalent in June 2025 to 867 in February 2026. Nothing links the size of the workforce to the volume arriving, so the two are free to diverge without any decision that looks like a decision to let the backlog grow.

4. Staffing arrived as a project and left the same way. The 500 additional staff were attached to a four-year funding allocation aimed at the batch. Funding that is scoped to a backlog ends when the backlog does, and the capacity it bought leaves with it, which is what the reduction from 973 to 867 records.

5. The measure changed while the thing measured did not. From April 2025 the processing clock stops while the department waits on medical records from the veteran. Time the veteran spends waiting is real whether or not it is counted, and after that change a figure describing performance and a figure describing experience diverge without either being wrong on its own terms. Comparison with earlier years now requires a correction that no opened source supplies.

6. The structural fix does not reach the stock. The VETS Act consolidates the three statutes from 2026-07-01, and it applies to claims lodged from that date. Every one of the 103,074 claims on hand at June 2026 continues under the law it was lodged under. The remedy and the problem therefore do not overlap at any point in time.

What has been tried?

Two directions were pursued and they did different kinds of work. One cleared a defined batch of claims with money and people, and it succeeded on its own terms and then expired. The other rewrote the law the claims are decided under, and it takes effect only for claims that have not yet been lodged. Neither of them changed the rate at which claims arrive or the number of people available to decide them, which is the quantity the record since 2024 has been tracking.

What was found?

Why is it still unsolved?

Institutional gap — the backlog was treated as a batch to be cleared, and nothing was built that would notice it forming again.

A one-off clearance and a standing capacity are different objects, and only the first was funded. The Royal Commission named 41,799 claims, the government attached money and 500 people to that number, and the number went to zero. Everything about that sequence worked. What it never contained was a rule about what happens next, because the target was a stock and the thing that produces a stock is a flow. Lodgements kept rising, the assessment burden per claim rose with them, and the workforce that had been added for the batch came back down. Sixteen months after the declaration the unallocated pile alone was four times what it had been, and no threshold had been crossed, because there was no threshold.

The second reason is that delay here is not a breach of anything. No opened source identifies a statutory time within which a claim must be decided, a published target for claims on hand, or a consequence for exceeding either. Where no deadline exists, a slow decision and a fast one differ in degree and not in kind, and a backlog does not generate the kind of event that forces an answer. The Royal Commission supplied a deadline once, from outside the statute, and it applied to one batch and expired with it. What the record shows since is a series of measurements with nothing attached to them.

The third reason is that the instrument built to fix the underlying complexity was pointed away from the people who are waiting. The Productivity Commission diagnosed the three-statute structure in 2019 and the VETS Act consolidates it from 2026-07-01, which is a real answer to the diagnosis. But the commencement rule applies the new law to claims lodged from that date, so every claim in the current stock stays under the law it arrived under. The reform and the backlog occupy the same department and never touch. For a veteran whose claim was lodged in 2025, the consolidation is an event that happens to other people.

The fourth reason is what keeps this stable rather than merely bad. The measurement changed in April 2025 so that time spent waiting on medical records no longer counts, which means the published average and the veteran experience now move apart without either being false. A department can report improvement while the stock grows, and both statements can be defended. When the indicator and the condition can diverge that way, the signal that would otherwise trigger a response stops being reliable, and the response that never comes looks like a response that was never needed.

What observation would mean it is solved?

Candidates — (a) claims with no delegate assigned fall and stay below their February 2025 level of 7,784 across several consecutive reporting periods (b) a maximum decision time is written into law or published as a departmental commitment, with something attached to exceeding it (c) average processing time for initial liability claims falls on a measure that is stated to be comparable with the years before April 2025.

(a) is the closest thing to a direct test and it can be met by the calendar. An unallocated claim becomes an allocated one the moment a name is attached to it, and that is an administrative act rather than a decision. A fall in the unallocated count is consistent with claims being distributed across the same overloaded workforce, in which case the stock has moved between two columns of the same table. Read alone the number would show improvement in the place where improvement is cheapest to produce.

(b) is the observation that would show the pattern had actually changed, because it is the thing this record has never contained. Every measure tried so far has been a quantity of money, of staff or of cleared claims, and each of them ended when its funding did. A standing rule about time is the only candidate here that would still exist after the attention moves on. It is also the least likely to appear, since it would create a form of failure that the current arrangement does not have.

(c) is the one that cannot currently be read at all. From April 2025 the clock stops while the department waits on records from the veteran, so a figure from 2026 and a figure from 2023 are not the same measurement. No opened source quantifies how many days that change removes from the average, which means the improvement from 435 days to 315 and the subsequent rise to 364 cannot be decomposed into what the department did and what the counting rule did. Until somebody publishes that decomposition, this indicator answers a question about arithmetic rather than about waiting. The three have to be read together, and (b) is the one that would say whether anything structural had moved.

What is it connected to?

Fills with researchclaims and disability determination backlogs in comparable administrations, the appeal and review stage that sits downstream of the decisions counted here, the health service access that a compensation decision unlocks, defence workforce separation rates as the upstream driver of lodgement volume, and the broader implementation record of the 122 Royal Commission recommendations. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• When the original backlog was actually cleared. The ministerial announcement of 2024-02-13 describes elimination in February 2024, ahead of a deadline of 2024-03-31. A departmental claims data page from June 2026 describes the same 41,799 claims as all completed as at 2025-12-31. Whether the two statements refer to the same batch, or to a batch that was redefined after 2024, could not be settled because neither page opened directly.
• How much of the processing time improvement is the counting change. No opened source quantifies how many days the April 2025 clock-stop removes from the reported average, and none states whether figures published after that date can be compared with figures published before it. The improvement from about 435 days to about 315 and the rise to 364 therefore cannot be separated into performance and measurement.
• What the claims workforce baseline actually was. One opened source records 412.8 full-time equivalent staff at May 2022 and another records 544 at June 2022. Neither defines what it counts, and no opened source reconciles them, so the size of the 500-staff increase relative to the starting point is unresolved.
• How many distinct veterans the 103,074 claims represent. Only claim counts and an average of conditions per claim are published. No opened source states a ratio of claims to claimants, and none gives a person count for the stock.
• Which of the 122 recommendations have been implemented. Secondary reporting refers to a number completed by the end of 2025, and the primary implementation record could not be opened, so no item-by-item comparison was possible and no assessment of which items affect claims capacity could be made.
• What delay in deciding a claim does to the people waiting. The Royal Commission and ex-service organisations describe an effect on wellbeing. No source opened here connects days of delay to any measured outcome, so this document carries the delay and does not carry a consequence.
• Most of the primary record, because it did not open. Repeated automated requests to the departmental, audit office, ministerial and service commission pages timed out or were refused. Ten of the rows below are blank for that reason. Where a secondary source and a government statement could differ in wording or in scope, this document has no way to detect it, and the figures marked low in the findings table above are the ones that depend on it.

See the evidence

No Australian government source was read directly except the Productivity Commission inquiry report. That page supplies the 2019 diagnosis of the three-statute arrangement, the recommendation to consolidate by July 2025, and the only person-level support figures in this document that come from a primary source, which are for 2017-18 and are therefore nine years old against a status check date of 2026-08-15. Everything else about the current state reaches this document through secondary sources citing departmental data, because every departmental, audit office, ministerial and service commission page attempted in research timed out or was refused. Ten rows above are blank with the reason recorded rather than filled with a substitute link, and the two most load-bearing claims in this document sit on two of those blanks. The declaration that the 41,799-claim backlog was eliminated in February 2024 comes from a ministerial release that did not open, and the June 2026 position of 103,074 claims on hand with 32,925 unallocated comes from a departmental page that did not open. Both are graded low in the findings table for that reason, and the second is where the source record contradicts itself: the same unopened June 2026 page is reported as describing the 41,799 claims as completed at 2025-12-31 rather than in February 2024, and this document names both accounts and asserts neither. The spine that did open is narrower and it is what the argument rests on. A single secondary source, read across two of its pages, carries claims on hand at June 2025 and February 2026, unallocated claims at February 2025 and February 2026, the workforce reduction from 973 to 867, the lodgement growth for 2024-25, the conditions-per-claim series and the three full-year processing averages. The counting change of April 2025 and the criticism of it were read directly in the press report and are the best-evidenced items here. Where sources disagree the disagreement is left visible: the claims workforce baseline for mid-2022 appears as 412.8 full-time equivalent in one place and 544 in another, with no opened source defining either scope, so neither is used as the denominator for the 500-staff increase. This is a Path A output, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range Not derivable

The reason and what is missing are listed under “What is missing” below