The BoJ has published Japan's end-June 2026 returns for both BIS international banking statistics
Victoria Shinder2 min readThe Bank of Japan has released the Japanese components of the two BIS international banking statistics for end-June 2026: the Locational Banking Statistics and the Consolidated Banking Statistics. Japan has contributed to both since the data for end-June 1998. The BIS aggregates these returns with those of other reporting countries and economies and publishes the global results on its own site.
The distinction between the two series is the reason both exist. Locational statistics count positions by where the reporting office sits, which makes them the natural measure of cross-border flows through a financial centre. Consolidated statistics count them by the nationality of the parent bank, netting inter-office positions, which makes them the measure of where risk actually lands.
What it means
These are not a story on their own, and that is precisely why they are worth marking. The BIS banking statistics are the best available instrument for a question no US data series can answer: how much dollar credit is created outside the United States, by banks the Federal Reserve does not supervise. Cross-border claims are overwhelmingly dollar-denominated, and the offshore dollar system they describe is the part of the funding market that shows up in policy discussions only when it seizes.
Japan's return matters disproportionately within that. Japanese banks have been among the largest net suppliers of cross-border credit for most of the period these series cover, and the yen-dollar basis is the price at which that supply is hedged. So the end-June figures are an input to a question that will be live in the next quarter: whether the cost of hedging dollar assets has changed what Japanese institutions are willing to hold.
One practical note on the release itself: the Bank publishes the data as archives rather than as a summary, so the figures require pulling the files. Anyone building this into a regular process should note the same lesson the Fed's Data Download Program is currently teaching — a pipeline keyed on an archive that quietly changes shape fails silently, and a fortnightly series that has stopped updating looks exactly like a series that has not moved.