How this audit works
Campaign documents compress. That is their job. The public-interest test is whether the compression preserves the source's meaning: what was measured, across what period, with what method and with what level of certainty.
For each figure we separate the campaign claim from the strongest primary record we could identify, then state the narrowest conclusion that record supports. A source can be real while the headline built from it remains incomplete.
01 / $15 billion
The claim
Coalition material treats $15 billion as corruption-related spending embedded in Victoria's capital program.
The record
Geoffrey Watson SC derived the figure by applying 15 per cent to an assumed $100 billion Big Build base after seeking views from industry participants. He described the estimate as very rough.
Observer finding
The estimate has an identifiable author and method. It is not an audit, a court finding or a demonstrated $15 billion transfer to criminals.
The source chain matters. The calculation appeared in Watson's draft but was removed from the final Rotting from the Top report after concerns were raised about terms of reference and expertise. Watson later defended the calculation in evidence before Queensland's CFMEU inquiry.
That evidence supports describing the number as Watson's estimate of additional cost attributed to CFMEU-linked crime and corruption. It does not establish that every dollar in the estimate was a corrupt payment, or that the entire sum reached organised crime.
Primary records: final Watson report; inquiry transcript, 11 February 2026; Coalition statements recorded in Hansard.
02 / $22 billion
The claim
A back-office hiring freeze would produce $22 billion in savings.
The record
The Coalition has publicly described this as a Parliamentary Budget Office estimate over ten years, delivered through natural attrition rather than redundancies.
Observer finding
The figure is presented publicly as part of a ten-year economic plan. Any shorter presentation should keep that decade-long horizon beside the number.
This is not a one-budget or one-term saving. It is a cumulative projection whose delivery depends on vacancies, exemptions, workforce composition and the continuation of the policy across a decade.
The fair criticism is therefore not that the timeframe has never been disclosed. It has. The question is whether each prominent use of $22 billion gives readers enough context to understand when—and under what assumptions—the saving is expected to accrue.
Primary records: Coalition economic plan; Victorian budget reply and debate.
03 / 12 per cent
The claim
Regional Victoria receives less than 12 per cent of infrastructure investment despite holding about a quarter of the state's population.
The record
The PBO attributed 12.0 per cent of total asset investment reported in the 2025–26 Budget exclusively to regional Victoria. Another 30.4 per cent could not be categorised exclusively as metropolitan or regional.
Observer finding
The disparity is independently documented, but the result is not a measure of all new funding or a decade of historical spending.
The underlying analysis covers 1,414 new, existing and completed projects reported in one state budget. It uses total estimated investment over each project's life—not simply what will be spent during 2025–26.
The Coalition's proposed 25 per cent floor applies to new Victorian Government infrastructure expenditure. That policy benchmark is not identical to the PBO dataset used to justify it. Both can appear in the same argument, but their scopes should not be merged.
Primary record: Victorian Parliamentary Budget Office, Regional and metropolitan investment: 2025–26 Victorian Budget.
04 / 3,000 police
The claim
A Coalition government would recruit 3,000 more police and use the larger workforce to restore station capacity.
The record
Coalition speakers have described 1,500 positions as filling current vacancies and another 1,500 as workforce growth. Victoria Police currently states that recruitment usually takes three to six months before training.
Observer finding
The promise is measurable only if the baseline, net headcount target, attrition assumptions, Academy capacity and delivery date are published together.
Victoria Police's initial Academy program runs for 25 weeks. Recruits are sworn in during that program, while the complete training, placement and Diploma of Policing pathway runs to week 116.
Those timelines do not prove the target is impossible. They show why “recruited,” “sworn,” “qualified,” “deployed” and “net additional” cannot be treated as interchangeable milestones. Filling a vacancy improves operational capacity, but it is not the same accounting claim as growing the authorised or actual workforce by 3,000.
Primary records: Coalition statement recorded in Hansard; Victoria Police recruitment process; Victoria Police training program.
What still requires an answer
- What exact document version and publication date govern each headline claim?
- Will the Coalition publish the full PBO assumptions behind the $22 billion projection?
- Will the $15 billion be described consistently as an estimate rather than an established loss?
- How will uncategorised and statewide projects be treated under the regional 25 per cent guarantee?
- What is the baseline headcount, annual recruitment schedule and attrition allowance behind the 3,000-police pledge?
Observer conclusion
Four real source trails. Four different kinds of certainty.
The numbers are not interchangeable. One is a contested estimate, one a decade-long projection, one a budget-specific allocation analysis and one a workforce promise whose delivery machinery remains incomplete.
The recurring issue is not fabrication. It is compression: method, scope, timeframe and delivery assumptions are reduced until unlike figures all read as equally settled.
The test for every campaign is simple. Put the caveat close enough to the number that a reasonable voter sees both.