The certified value and the booked value are two different records

Every capital project has two running totals of what it has cost so far, and they are kept by different people from different documents.

The first is the project's. It is built from payment certificates: the contractor applies, the engineer certifies, and the cumulative certified value is what the project says has been built and is owed. It is organised by contract, by bill item, by measured quantity, and it is kept by the people on site.

The second is the accountant's. It is a balance in a holding account called work in progress, assets under construction, or construction in progress, depending on the manual. It is built from what the finance system has processed: invoices, accruals, journal transfers, sometimes retention and advance payments netted in ways the project never sees. It is organised by cost centre and general ledger code, and it is kept by people who were not on site and did not read the certificate.

Nothing in either record refers to the other. The standard does not bind them together, because the standard is about what the balance must be, not how two functions arrive at the same number. The thing that binds them is a reconciliation: somebody puts the certified total and the booked total side by side, project by project, explains every difference and adjusts whichever record is wrong. Where that is done monthly, the two records are one number seen from two angles. Where it is not, they drift, and the drift compounds for as long as nobody looks.

What the standard requires, and what it leaves to the entity

IPSAS 45 is precise about the end of the holding period. Paragraph 17 says recognition of costs in the carrying amount ceases when the item is in the location and condition necessary for it to operate as management intends, and this publication has already argued what that does to the transfer date.

The transfer date is a capability test, not a certificate

, read: The date a project becomes an asset is not the date on the certificateThe date a project becomes an asset is not the date on the certificate

What the standard leaves to the entity is the interim: how the balance is supported while the project runs, and how anyone knows that the costs accumulated under a project code in the ledger are the costs of that project and nothing else. That is an internal control question, and the clearest published statement of what the control is comes not from an accounting standard but from an auditor describing what it expected to find.

That is the requirement in its entirety: compare, research, resolve, adjust, and do it before the differences age. It has a cost, which is a person with access to both records and the standing to make either side correct theirs. Whether it earns that cost is the question the rest of this piece is about, and the evidence is unusually good, because the largest asset owner in the world has been audited on exactly this point, by the same two bodies, for a generation.

One auditor, 27 years, the same finding

The US Department of Defense holds real property and construction in progress on a scale nothing else matches, and since the Chief Financial Officers Act of 1990 it has been audited on it annually. The findings below are from GAO and the DoD Office of Inspector General, read in full, and the caveat that must travel with them is stated after.

Read that last sentence twice. Adjusting one record to match the other produces agreement and destroys the information the disagreement carried. It is the reconciliation performed as a ritual rather than as a control, and GAO named it in 1993.

Two of the examples in that report are worth carrying because they show what unreconciled time does. A 76-mile waterway completed in 1968 still had 27m sitting in construction in progress on 30 September 2002, 34 years later. A flood control tunnel completed in 2001 and handed to its local sponsor by official letter in January 2002 still had its 21m in the account nine months after that. Neither was hidden; both were on the file. Nobody's job was to compare the file with the account.

The mechanism is stated plainly in the report. Some 1.022bn of the unsupported value had originated in a predecessor financial system, been carried across at a system change in 1998, and could no longer be traced to source documents. The Corps had six months of audit fieldwork to find them and could not. A migration had turned a reconcilable balance into an unreconcilable one, and the balance had then been carried forward, unchallenged, for five years.

Twenty-seven years separate the first finding from the last. The systems changed, the auditors changed, the regulation was rewritten, and the finding did not move: two records, no reconciliation, a balance that nobody can support.

Which reading this publication takes, and the limit of the evidence

A requirement that is this widely unmet can mean four different things, and the reading has to be stated rather than left for the number to imply.

It does not mean the requirement is unnecessary. The 2003 report's conclusion is the reason: the costs accumulated in construction in progress become the recorded cost of the asset when it is placed in service. A 43% overstatement in the holding account does not vanish at transfer. It is capitalised, and then depreciated, for the life of the asset, which for civil works is measured in decades. The reconciliation is cheap next to what an unreconciled balance becomes.

It does not mean the requirement is unaffordable. The control is a monthly comparison of two lists that already exist.

The reading here is that the requirement is unowned. GAO's phrase for the Army in 2019 was that it had not properly assigned reconciliation responsibilities. The 1993 report found the engineering directorates were not required to tell finance about a project until after construction, so finance booked the costs as expenses and had no construction-in-progress record to reconcile to. In both cases the control failed not because anyone refused to run it but because it sat between two functions and belonged to neither. That is a defect of practice and of organisation, not of the standard.

The limit is that every finding above is from one owner. It is the largest and most audited capital estate there is, its financial systems have a documented history of failed migrations, and it remains the only major US federal agency never to have received a clean audit opinion (GAO-25-108052). Nothing here shows that a ministry or a utility in another country carries a 43% overstatement. What it shows is that the mechanism is structural: wherever the project's record and the accountant's record are produced by different people from different documents, only a reconciliation makes them one number, and the audit record of the one owner examined at this depth shows the balance drifting for decades when that reconciliation is not run. The reader's question is whether it is run where they are, and that is checkable in an afternoon.

What to examine

• Ask for the reconciliation, by project, for last month. If it does not exist, the two records have never been compared and the balance is whatever the ledger says it is.
• Ask who owns it, by name, and which function they sit in. A control that belongs to nobody is the pattern in every finding above.
• Ask for the ageing of the balance. A project with a taking-over certificate on file and a balance still in work in progress is the 1968 waterway, earlier in its life.
• Ask whether any balance is negative. An asset account cannot be, and each one is an unresolved posting error with a date on it.
• Ask how differences were resolved. If the answer is that the ledger was adjusted to match the other record, ask how anyone knew which record was right. That is the Fort Shafter answer, and it was a finding in 1993.
• Ask what happened at the last system migration. Balances carried into a new system without their supporting detail are the 1.022bn the Corps could not find in 2003.

---

Sources. US Government Accountability Office, Financial Management: Army Real Property Accounting and Reporting Weaknesses Impede Management Decision-making, GAO/AIMD-94-9, 2 November 1993, on the Army's fiscal year 1992 financial statements, all from the letter report: page 1 for the 3.6bn unreconciled difference and the estimated 276m of completed cost reported as construction in progress; page 4 for the 300 projects that appeared substantially complete; page 8 for the 24.2bn and 27.8bn balances; page 9 for the Fort Shafter reconciliation and the Corps not comparing accounting with engineer records; page 10 for the 87 civil and 47 military projects, the $560,000 building addition reported twice, and the engineering directorates' reporting gap. Department of Defense Office of Inspector General, Reliability of Construction-in-Progress in the U.S. Army Corps of Engineers, Civil Works, Financial Statements, D-2004-017, 7 November 2003: Table 1 for the 500-item sample and its breakdown, pages 7 to 10 for completed projects, unsupported balances and negative balances, and page 10 for the 4.3bn projection and the conclusion on capitalised cost. US Government Accountability Office, Defense Real Property: DOD-Wide Strategy Needed to Address Control Issues and Improve Reliability of Records, GAO-20-615, September 2020, pages 13 to 15, for the fiscal year 2019 findings of the independent public accountants and the description of reconciliation as a control. IPSASB, IPSAS 45, Property, Plant, and Equipment, paragraph 17, for the cessation of cost recognition. All read in full or in the cited passages on 6 September 2026.