ENGLISH

FX Intervention Research Institute (FXIR)

Evidence-graded, machine-verifiable research on Japanese yen FX intervention. Canonical documents are in Japanese; all numbers, claims and data are language-neutral JSON/CSV with SHA-256.

Key findings (as of 2026-08-08)

Across six yen-buying episodes (2022–2026), the effect half-life — business days until the close first recovers the 50% line — was 1, 5, 15, 431 and 5 days (E6 ongoing, right-censored), with no clear relation to amounts spent. We model intervention as a two-layer reaction function: activation (weeks–months) vs execution (minutes–hours). Rate checks show an evidence asymmetry: Japanese-side actions stay UNVERIFIED under the current disclosure regime, while US-side actions become FACT via FOMC minutes. In the E6 week, the FIMA repo facility was not used (Fed H.4.1 total repos ≈ $1–3mn, an upper bound).

Machine-readable entry points

llms.txt (LLM site index) ・ REST API v1 (verifiable envelope: canonical SHA-256 in every response) ・ Data (CC BY 4.0 package, weekly CFTC/H.4.1 series) ・ structured claims.

How to cite

FX Intervention Research Institute (FXIR), “<document title>”, fxir.jp, version and date, SHA-256 <hash>. Preserve evidence-state labels (FACT / ESTIMATE / HYPOTHESIS / UNVERIFIED) when quoting.