A Currency Spike Looks Like One Event. Its Official Data Arrives on Five Clocks.

When a currency pair moves several percent in a handful of sessions, the reflex is to look for the explanation inside the price. The more productive question is which observable input changed, because foreign exchange is unusual in how much of it is documented by official bodies with release numbers and fixed schedules. The spot rate, the interest rate on each leg, whether a government transacted, and how large the underlying market is are all published. None of them on the same clock.

That mismatch is the working problem. A volatility spike resolves in hours. The Federal Reserve's interest rate release updates every business day, its bilateral exchange rate release once a week. Japan's Ministry of Finance discloses a monthly intervention total and withholds the daily breakdown until the quarterly release. The Bank for International Settlements measures the market's size once every three years. A position opened during the move is sized against a record that will not be complete for months.

Knowing the shape of that delay is worth more than any read on direction. It separates what is verifiable now from what will only be graded later, and puts a number on how long the grading takes.

What Is Observable, and When It Becomes Observable

Five primary sources cover most of what matters, and each has a different refresh interval.

• Federal Reserve H.15, Selected Interest Rates. Posted every business day at 4:15 p.m. The release of 21 August 2026 carried the federal funds effective rate at 3.63 percent each day from 14 through 20 August, the 2-year Treasury constant maturity at 4.19 percent and the 10-year at 4.69 percent on 20 August.
• Federal Reserve H.10, Foreign Exchange Rates. Weekly, updated every Monday at 4:15 p.m. The release dated 17 August 2026 covered only 10 through 14 August. On 14 August the yen printed at 159.21 per dollar and the broad dollar index at 118.9028 on a January 2006 = 100 base.
• Federal Reserve G.5, Foreign Exchange Rates. Monthly. The release of 3 August 2026 gave July averages. Its own note states that averages are based on daily noon buying rates for cable transfers in New York City certified for customs purposes by the Federal Reserve Bank of New York — a single daily fixing, not a close, and not a volume-weighted average.
• Japan Ministry of Finance, Foreign Exchange Intervention Operations. A monthly release gives one aggregate figure for a window running from roughly the 27th of one month to the 26th of the next. A separate quarterly release gives the individual dates and daily amounts.
• Treasury and Federal Reserve Foreign Exchange Operations, published quarterly by the New York Fed. The second-quarter 2026 report, announced 13 August 2026, stated that the Federal Reserve and U.S. Treasury did not intervene during the quarter. The first-quarter report, announced 14 May 2026, said the same.

Behind those sits the BIS Triennial Central Bank Survey, conducted once every three years and the only census of how large the market is.

The gap between the top row and the bottom row is roughly a factor of a thousand. Treating the currency market as one well-lit venue means fusing a daily feed, a weekly feed, a monthly feed, a quarterly feed and a triennial feed into a single picture, then reasoning as though all five were current.

The April 2026 Yen Operations Are a Clean Test of the Lag

Japan's framework makes the delay measurable rather than theoretical, because the same event is released twice at two resolutions. The Ministry of Finance monthly release covering 28 April through 27 May 2026 was published on 29 May 2026 and reported a single figure, ¥11,734.9 billion, with no dates and no breakdown. The quarterly release for April through June 2026 was published on 7 August 2026 and disclosed the individual operations.

Date	Amount	Direction
30 April 2026	¥6,278.7 billion	U.S. dollar sold, Japanese yen bought
4 May 2026	¥780.2 billion	U.S. dollar sold, Japanese yen bought
6 May 2026	¥4,675.9 billion	U.S. dollar sold, Japanese yen bought

Three numbers fall out of that pair of releases, and none require interpretation.

1. The first operation was on 30 April. The aggregate became public on 29 May — 29 days later.
2. The daily detail became public on 7 August — 99 days after the first operation.
3. The preceding monthly release covered 30 March through 27 April. The 30 April operation fell outside a window that had already closed, so the release just published did not capture it.

One calendar detail is worth registering without over-reading it. The Federal Open Market Committee met on 28 and 29 April 2026; it holds eight regularly scheduled meetings a year and publishes minutes three weeks after each decision. The first Japanese operation came the day after that meeting concluded. Proximity is not causation and the primary sources establish no link. What the calendar does establish is that a scheduled policy event and an unscheduled official transaction can land inside the same 48 hours, and only one of them was announced in advance.

The U.S. side of the same window is also documented. The New York Fed's quarterly report for the second quarter of 2026, announced 13 August, recorded no U.S. intervention — 44 days after the quarter ended. The New York Fed acts as fiscal agent for the Exchange Stabilization Fund, and the currencies used in U.S. operations have historically been drawn in equal parts from the System Open Market Account portfolio and the ESF, held in euro and yen. So the operations in that window were one-sided, and confirming that took until August.

The implication is narrow. During the move itself, official participation was an inference; it stayed one for four weeks; the full record arrived a quarter later. Any framework that requires knowing whether an official body is transacting while the move happens is asking for information that does not exist yet.

Separating the Bilateral Move From the Dollar Move

One question can be settled immediately, and it is the one most often skipped. When a pair moves, either the quoted currency moved or the dollar moved against everything. The Federal Reserve publishes both legs, so this need not be guessed. Below are G.5 monthly averages for the twelve months through July 2026 alongside the broad dollar index from the same releases.

Month	Yen per U.S. dollar	Broad dollar index (Jan 2006 = 100)
August 2025	147.4786	120.9844
October 2025	151.3545	121.1712
December 2025	155.9150	120.1883
February 2026	155.1016	117.9060
April 2026	159.1173	119.0363
June 2026	160.7700	120.0835
July 2026	162.3295	120.5970

Over those twelve months the yen figure rose from 147.4786 to 162.3295, about 10.1 percent in yen per dollar. The broad dollar index went from 120.9844 to 120.5970, about minus 0.3 percent. Measured against a trade-weighted basket, the dollar finished roughly where it started. The bilateral rate did not.

Chart data: Federal Reserve Statistical Release G.5, Foreign Exchange Rates (monthly), releases of 3 November 2025, 2 February 2026, 1 May 2026 and 3 August 2026. Series: Japan, yen per U.S. dollar; Broad dollar index, JAN06 = 100.

A move concentrated in one pair and a move visible across the whole basket are different exposures, even when the chart on the screen looks identical.

One quotation trap belongs here, because it produces sign errors that survive into spreadsheets. The H.10 release states that rates are in currency units per U.S. dollar except as noted by an asterisk. The asterisked entries — including the euro, the pound, the Australian dollar and the New Zealand dollar — are quoted as U.S. dollars per currency unit. On 14 August 2026 the yen line read 159.2100 and the pound line read 1.3556. Those two numbers point in opposite directions with respect to dollar strength, and nothing in the table layout flags the difference except the asterisk.

The Differential Prices the Forward, Not the Spot

The interest rate leg is where two feeds have to be combined, and where the most common misreading sits. A wide differential says nothing determinate on its own about where spot will go. What it fixes is the relationship between spot and forward.

The BIS described covered interest parity in its September 2016 Quarterly Review as the closest thing to a physical law in international finance: the interest rate differential between two currencies in cash money markets should equal the differential between the forward and spot exchange rates. If it does not, a fully hedged position can be constructed to capture the gap.

Both inputs are published. H.15 put the federal funds effective rate at 3.63 percent on 20 August 2026. The Bank of Japan's statement of 31 July 2026 kept its guideline to encourage the uncollateralized overnight call rate to remain at around 1.0 percent, by an 8-1 vote, after moving to that level on 16 June 2026. The overnight differential is roughly 2.6 percentage points.

Read through the identity, a higher dollar rate means the forward yen-per-dollar rate prints below spot: the dollar at a forward discount, the yen at a forward premium. That is an arithmetic consequence, not a forecast. The carry sits in the forward points rather than in any expectation about the spot path — a distinction that matters for how a hedge is priced and how a rollover behaves.

Two cautions apply to running that arithmetic. Overnight policy rates are not what a dealer uses to price a twelve-month forward; term money-market rates are, and they publish separately. And the residual left after the differential is accounted for is the cross-currency basis, which the BIS documented as persistently negative for dollar borrowing against the euro and the yen since 2007. A non-zero basis means the identity does not close at observed prices, and that residual is a funding-condition signal in its own right.

Turnover Tells You Which Segment a Spot Print Comes From

The last source refreshes slowest and is used least, which is a mistake, because it answers a structural question that daily data cannot. The BIS Triennial Central Bank Survey for April 2025 — the fourteenth edition — reported global foreign exchange turnover of $9.6 trillion per day, up 28 percent from $7.5 trillion in April 2022. The instrument split matters more than the headline:

• FX swaps: about $4 trillion per day, roughly 42 percent of turnover
• Spot: about $3 trillion per day, roughly 31 percent
• Outright forwards: about $1.8 trillion per day, roughly 19 percent
• Options: roughly 7 percent; currency swaps roughly 2 percent

The U.S. dollar was on one side of 89.2 percent of all trades, against 88.4 percent in 2022; the yen appeared in 16.8 percent, essentially unchanged.

Spot is under a third of the market. Most activity is in swaps and forwards, instruments used to move funding across currencies and to hedge rather than to express a directional view. A spike on a spot chart is a price formed in the smaller segment of a market whose larger segment is doing something else.

The survey's own timeline is worth noting as a final example of the clock problem. Data were collected in April 2025. Preliminary findings appeared in September 2025, full turnover data in the December 2025 BIS Quarterly Review, and final figures in June 2026 — roughly fourteen months from the observation month to the settled number.

What Would Invalidate This

This frame concerns the timing of evidence rather than outcomes. Several conditions make it stop being useful.

• Short holding periods. If a position opens and closes inside a day, none of these releases arrive in time. The framework is for sizing and post-trade review, not intraday decisions.
• Currencies without comparable disclosure. Japan publishes intervention figures on a fixed schedule. Many jurisdictions do not, or publish only reserve aggregates on a lag that makes attribution impossible. Where the disclosure does not exist, the absence of a published operation is not evidence that none occurred.
• Managed and pegged regimes. Where the rate is administered, capital controls and official balance sheet operations dominate. The identity still holds arithmetically; the inputs stop being freely determined.
• Reference-rate mismatch. The G.5 and H.10 rates are single daily fixings certified for customs purposes. They are not the rate a position is marked at, and the gap between a fixing and an execution widens precisely on the days volatility is elevated. Using these series to reconstruct a fill is a category error.
• Restated windows. A conclusion drawn from a monthly aggregate alone can be wrong about which days carried the weight. In the April and May 2026 figures, one day carried more than half the total.

Concrete Framework

A repeatable sequence for the next time a pair moves faster than usual. It produces a written record, not a signal.

1. Timestamp the move against the release calendar. Write down when each source will next print: H.15 the same business day at 4:15 p.m., H.10 the following Monday at 4:15 p.m., G.5 early the following month, the Ministry of Finance monthly total once the 26th-to-27th window closes, the New York Fed quarterly report roughly six weeks after quarter end.
2. Pull both legs of the rate differential before forming any view. The H.15 release for the dollar leg; the counterpart central bank's own published guideline or overnight rate for the other. Record the figures and their dates, not a remembered level.
3. Check the bilateral against the basket. Compare the pair's move with the broad dollar index from the same H.10 or G.5 release, and record whether the move is concentrated or general. That determines whether the exposure is to one currency or to the dollar.
4. Verify the quotation convention. On H.10, the asterisk is the only marker separating currency-units-per-dollar rows from dollars-per-currency-unit rows.
5. Write down the intervention hypothesis before it can be confirmed. State it with the date, then leave it open. Close it against the monthly aggregate when that publishes, and again against the quarterly daily detail. Leaving it unresolved for one to three months is the accurate treatment.
6. Size against the slowest input, not the fastest. If a position depends on a fact that will not be published for 99 days, it has to survive 99 days of not knowing. That is the sizing constraint this exercise produces. It also argues for re-checking instrument context whenever the BIS triennial data refresh, since spot at roughly 31 percent of turnover is a different backdrop from spot as the market.
7. Archive the release, not the summary. Keep the release number, date and figures — H.15 of 21 August 2026, H.10 of 17 August 2026, G.5 of 3 August 2026. Secondary write-ups compress and drop the quotation convention, which is where the errors enter.

None of this predicts a currency's direction. It replaces assumptions about what is knowable with a schedule of when each thing becomes knowable — a smaller claim, and a more durable one.

This article describes publicly available statistical releases and is not investment advice.