Weekday Japan business intelligence for finance professionals.

Join the list
Tokyo Brief東 京 ブ リ ー フ

Japan's day, wrapped and delivered by morning.

Article

Bank of China auditor flags loan-loss modelling in 2025 accounts

Expected credit losses on customer loans, not the sign-off itself, are the telling detail in the 2025 accounts, with the auditor pointing to judgement-heavy credit-risk staging and model inputs.

Jun 26, 20262 min read
Abstract illustration of bank loans moving through risk bands and reserve models.

Bank of China's 2025 accounts come with a straightforward verdict from the auditor: the group's consolidated financial statements give a "true and fair view". The opinion covers the statement of financial position, profit or loss, comprehensive income, changes in equity, cash flows and the notes for the year ended 31 December 2025. The more revealing line is elsewhere in the report, where expected credit losses on customer loans measured at amortised cost are identified as a key audit matter.

That pairing matters for readers far beyond Japan. The filing says the accounts were prepared under IFRS Accounting Standards issued by the IASB and in compliance with the Hong Kong Companies Ordinance's disclosure requirements. The report also says the audit was conducted under International Standards on Auditing and that the auditor was independent under the Hong Kong Institute of Certified Public Accountants' ethics code. In other words, the Japanese securities report is pointing readers to a cross-border reporting set, and the excerpted audit focus is on credit-loss measurement rather than the accounting rulebook itself.

The audit excerpt is explicit about where judgement enters. It says the measurement of expected credit losses uses significant assumptions, including whether a "significant increase in credit risk" has occurred and how complex models and their many inputs are set. The filing also says those staging criteria may have a significant impact on expected credit losses for loans with longer remaining maturities. The source does not quantify the exposure in this excerpt, but it does make clear that longer-dated loans are especially sensitive to those judgements. For bank readers, that is the useful line to circle: when impairment recognition depends heavily on model design and credit-risk staging, the calm language of an audit opinion only gets you so far.

There is, however, a limit to what this packet can bear. The excerpt provided does not establish whether other key audit matters were listed elsewhere in the report, and it does not include the underlying loan book or impairment figures that would show how those judgements moved during 2025. So the safe conclusion is a narrow one. The auditor signed off on the accounts, and the filing flags expected-credit-loss modelling as one area where assumptions deserve closer attention than the phrase "true and fair view" alone would suggest.