The board at a glance

Slack Tide reads the Macro Compass board once a month: what moved, what did not, what is scheduled next. This issue reads the snapshot dated September 4; the monthly pass ran a few days past the September 1 target. Twenty-three gauges in seven trend groups: 3 near-signal, 2 approaching, 1 inverted, 8 quiet, 9 context (each state is defined in the board's about block). Against the August 3 snapshot: two state changes, productivity approaching to near-signal and wage growth vs. unemployment quiet to approaching, and no threshold crossed between the two snapshots. One row could not be refreshed from its usual table this month; the coverage sentence below names it. The lead: the July 31 yen intervention.

The lead: the yen intervention and the dials that would register it

On July 31 the US Treasury bought yen with euros, jointly with Japan's Ministry of Finance: the first US purchase of yen since 1998 and its first yen intervention in either direction since 2011. The operation ran on the Treasury's books, through its Exchange Stabilization Fund, with the New York Fed's trading desk executing as agent. Japan's Ministry of Finance put its own side at ¥15.4 trillion for July 30 through August 26. The US amount has not been published as of this writing; a Senate Banking Committee letter of August 13 asked for it. The Fund's July statement, dated July 31, shows its euro and yen holdings little changed from June; its August statement, due in the coming weeks, is the next place to look.

Two limits before any dial. The board has no row for official currency operations themselves, so it reads this event only through its consequences. And an unchanged reading is not proof of no effect, only that nothing crossed the line a row watches; none of the dials below can attribute a move to this operation on its own, and inconclusive is a valid reading.

• The TIC foreign-official rows. If Japan funded its side by selling Treasuries, the place to look is the country tables, Japan's row, with the two-month lag: the July data, September 16, covers only the operation's first days; the August data, in mid-October, covers most of the window. Japan's row combines official and private holders, so a decline there is evidence to weigh against the Ministry's own disclosures, not attribution, and a flat reading does not rule a sale out. The June data, the last pre-operation month, showed official holders across all countries as modest net sellers of Treasury notes and bonds. The board's share row still reads approaching.
• The swap-lines row, and one line the board does not carry. Swap lines supply dollars when dollars are scarce; the US leg here supplied none, and the row sits at $132 million, de minimis. Japan sells dollars to buy yen; in the same weekly Fed release, the repo facility for foreign official accounts lets a central bank borrow those dollars against its Treasuries instead of selling them. The Treasury is reported to be pressing the Fed to expand it for Japan's yen defense, and that line, which the board does not carry, reads zero today. It sums all foreign official accounts; a draw would be a reason to look, not attribution.
• The term-premium row. The row reads monthly whether fiscal-risk pricing reacts to an operation of this shape: 0.76% on August 31, from 0.84% at the end of July, a move the monthly reading cannot assign to any one of August's events.
• The dollar-system meter, at ~stirring. That tilde-marked estimate is an audited judgment, not a measurement. Its only audited read to date is dated August 17, after the news and before any measured gauge had moved on this news; it has not been re-audited since.

Sources: CNBC (Aug 1 intervention report; Aug 3 explainer on the euro leg; Aug 3 analysis on the Fed's repo facility), Bloomberg (Aug 5, on the same), Council on Foreign Relations (Aug 4, on the 1998 and 2011 precedents), Japan's Ministry of Finance (monthly intervention release, Aug 28), the Senate Banking Committee letter of Aug 13, the Federal Reserve H.4.1 release of September 3, the Treasury's ESF reports page, the Fed's FOMC calendar and the Treasury's TIC release schedule (for the two non-board calendar dates below).

What moved, what did not

Productivity (near-signal, approaching to near-signal). The Q2 print came in at +2.24% year over year, and the same release revised Q1 up to +2.93%; the revised Q2 estimate on September 3 changed nothing. The four-quarter average now reads 2.53%, above the 2.5% bar for the first time. The bar is the clock's start condition, not the row's threshold: the threshold is that average holding above 2.5% for eight consecutive quarters, and the clock stands at one of eight, with 3 basis points to spare. A soft Q3 print on November 5, or a downward revision to any quarter already in the average, could reset it.

Wage growth vs. unemployment (approaching, quiet to approaching). The Atlanta Fed tracker rose to 3.8% in July from 3.6% in June, while unemployment fell to 4.1% in July from 4.2% in June and held at 4.1% in August. One month of that pairing plus a hold is a hint, not a confirmation.

Curve re-steepening (near-signal, unchanged). The 3m10y spread sits at +0.87%, from +0.79% at the August 3 snapshot. Its designated confirmer moved the other way: the Sahm rule, which reads the rise in unemployment that has accompanied past recessions, printed -0.07 for August after -0.03 in July and +0.07 in June. Live, still uncorroborated.

Deficit vs. unemployment (near-signal, unchanged). A quarterly row; the reading from the August 3 snapshot, a deficit of 5.6% to 5.8% of GDP against that snapshot's 4.2% unemployment, stands until the November quarterly walk.

Foreign official Treasury share (approaching, unchanged; basis change). The row now reads long-term Treasuries on both sides of the ratio, because the all-maturities holdings table could not be refreshed this month. Official holdings of notes, bonds, TIPS, and floating-rate notes were 14.0% of the long-term marketable stock in June, from 14.3% in May and 15.3% a year earlier. The earlier 12.5% figure counted bills on both sides; the two levels are not comparable, and the board says which basis it is on. The sides must match: with the Treasury funding heavily in bills, an unmatched ratio would fall on its own. The decline shows on the matched basis too.

Manufacturing construction (inverted, unchanged). $169.8 billion at an annual rate in July, down 21.2% year over year, six straight monthly declines.

Debt and fiscal pressure, otherwise quiet. Net interest as a share of receipts ticked up to 19.75% on the July Treasury statement, from 19.5%. The 10-year real yield reads +2.42%, so the repression gauge still reads not operating. August's 10-year and 30-year auctions both stopped through their when-issued levels, even as the 30-year cleared at 5.216%. The Treasury's August 19 decision to at least double its long-end buybacks, first larger operation September 9, is read through the term-premium and auction rows over the next two refreshes, not as a headline.

Liquidity, context. Fed net liquidity fell to $5,769 billion on September 2, down $71 billion in a month as the Treasury's cash balance rebuilt.

Dollar system, otherwise quiet. Foreign buying of US equities ran $918 billion over the twelve months to June. China plus the Belgium custody line hold $984 billion of long-term Treasuries, down 2.1% year over year, on the same long-term-only basis as the share row. The Fed's own foreign-currency assets show no purchase from the yen operation, which ran on the Treasury's books, not the Fed's. The line read $18.9 billion on July 29, $19.3 billion on August 5, and $19.2 to $19.5 billion on every weekly reading since, through September 2.

Demographics, and trade beyond the construction row: no refresh this month; those rows are quarterly or annual.

Sixteen of the 23 rows refreshed this month. Seven are quarterly, annual, or context rows whose next release falls outside the window; their last readings stand. No scheduled release failed to arrive; the one row whose usual source could not be refreshed is the foreign official share, above. All three estimate meters stand where the August 17 audit set them.

Base rates

The expansion is 77 months old, longer than 8 of 12 postwar expansions and about 1.5x the postwar median. The NY Fed's term-spread model still reads roughly 20% at one significant figure for July 2027; its August input has not been published yet, so the reading carries over. Priors inform context, never triggers.

On the calendar

• Sep 11: CPI, August (BLS)
• Sep 16: FOMC decision; and Treasury International Capital data for July, the first month that could show a Treasury-sales footprint from Japan's side of the operation
• Sep 30: Personal Income and Outlays / PCE, August (BEA)
• Oct 2: Employment Situation, September (BLS)
• Oct 14: CPI, September (BLS)
• Nov 5: Productivity and Costs, Q3 preliminary (BLS), the productivity clock's next test

Board calendar rows, plus two dates the board's calendar does not carry: the FOMC decision and the TIC release. The next issue follows the board's early-October refresh.

What this letter is

One email a month, sent shortly after the board refreshes: what moved, what did not, what data is scheduled next. The full board, with thresholds, sparklines, sources, and the chip legend, lives at craigstoller.com/macro. The subscribe form lives there too, for archive readers and forwards alike.

Everything here is general information and personal research, not investment advice.