Long-end stress panel

What the panel shows

The Long-end panel surfaces the far end of the government-bond curve — the part the rest of the dashboard did not track. Through late August 2026 the US 30-year yield reached 5.31% (2026-08-17), its highest since June 2007, and six major sovereigns hit multi-year or all-time 30-year highs inside about two and a half weeks. The panel makes that observable from primary sources, in two blocks.

Block A — US long end. The 30Y, 20Y and 5Y nominal yields; the two long-end curve slopes 10s30s and 5s30s (in percentage points, signed so a steepening reads apart from an inversion); the 30Y real yield and 30Y breakeven that split the move into real-rate versus inflation-expectations; the years since the 30Y last printed higher with an all-time-high flag; the 15-year percentile; and the SPY / 30Y correlation regime.

Block B — six-sovereign panel. One row each for the United States (30Y), Germany (30Y), Euro AAA (30Y), the United Kingdom (20Y), Canada (long), Japan (30Y) and Japan (40Y), with the latest level and the years since each last printed higher. A header line carries the two synchrony counts and the panel as-of.

Block C — US supply & auctions. The fiscal side of the long end: the bills share of marketable debt (how much of the deficit is funded at the front end rather than out the curve), the 12-month refinancing wall in percent and dollars, the average interest rate the whole debt stock now carries, gross interest expense fiscal-year-to-date, and total public debt, each stamped against the block’s honest as-of. A compact auction sub-table then carries the demand read from the latest 30Y bond and 10Y note auctions — bid-to -cover (with its trailing-5y percentile), the tail proxy (with its percentile), and the indirect and primary-dealer shares.

Block D — who holds the debt. The demand side, holder by holder: money market fund assets both retail (with its 13-week change) and broad, the Fed’s SOMA Treasuries (with its 13-week change — the QT-runoff pace), the total foreign TIC holdings, Japan (with its 12-month change) and China, and the weekly foreign-official custody (with its 13-week change). Each row carries its own per-cadence as-of, so a weekly carrier is never read as fresh as a monthly one.

The sources, by name

Every number comes from a primary public source:

• US — the 30Y / 20Y / 5Y nominal yields, the 30Y real yield and the 30Y breakeven come through FRED (the St. Louis Fed’s data service; series DGS30 / DGS20 / DGS5, DFII30, T30YIEM), which redistributes the US Treasury par-yield curve.
• Germany — the 30Y Bund yield from the Bundesbank.
• Euro area — the AAA-rated (German-weighted) 30Y from the ECB yield curve, used as a euro-area proxy because there is no clean free daily 30-year French OAT yield.
• United Kingdom — the 20Y gilt yield from the Bank of England (there is no clean free daily 30-year gilt series).
• Canada — the long-term government bond yield from the Bank of Canada.
• Japan — the 30Y and 40Y JGB yields from the Japan Ministry of Finance.

The supply and holder blocks come from the same public register:

• US fiscal supply (Block C — bills share, maturity wall, average rate, interest expense, total debt) comes from the US Treasury’s Fiscal Data service (the Monthly Statement of the Public Debt and Debt to the Penny). The supply block’s as-of is the oldest of its daily carriers, because the monthly MSPD figures lag the daily debt total.
• Auctions (Block C sub-table) come from the US Treasury / TreasuryDirect coupon-auction results — bid-to-cover, indirect and primary-dealer accepted shares, and the stop and median yields behind the tail proxy.
• Holders (Block D) come from primary public sources by holder: money market fund assets from FRED (the retail H.6 series and the broad Z.1 quarterly total), the NY Fed’s SOMA portfolio for the Fed’s outright Treasuries, the US Treasury’s TIC report for foreign holdings (total, Japan, China), and FRED again for the weekly foreign-official custody held at the Fed.

Three caveats travel with these blocks:

• The auction tail is a proxy, not the real tail. No free source publishes the when-issued yield, so the tail is computed as the auction’s stop yield minus its median — a demand-skew proxy against the auction’s own median, not the true when-issued tail. It renders as a dash (never 0) when the median is missing.
• Retail MMF is narrower than the broad total. The weekly retail series ($2-3T) excludes the institutional and government-only share, so it runs well below the broad quarterly total ($8.4T). The two are shown side by side precisely so neither is mistaken for the other.
• TIC lags roughly two months. The foreign-holdings report publishes about two months in arrears, so the total / Japan / China lines share a single month-end as-of. The weekly foreign-official custody line is the timely proxy for the foreign-official bid in between the lagged TIC prints.

The synchrony definitions

The panel’s cross-country reads are counts over the same six-country set — US, Germany, Euro AAA, UK, Canada, Japan at the 30-year point (Japan uses its 30Y):

• Decade-high count — how many of the six sit at a ten-year-or-longer high. See the decade-high count.
• All-time-high count — how many of the six sit at an all-time high in their own series. This header count is the honest cross-country all-time-high read: the per-country rows deliberately show only years since higher, because a foreign row has no per-country all-time-high flag and a long look-back cannot be told apart from a true record without one. Only the US row carries a real all-time-high marker. See the all-time-high count.
• Panel as-of — the oldest of the six countries’ latest observation dates, so the panel never claims to be fresher than its laggard (Japan publishes on Tokyo time and often trails). See the panel as-of.

A missing reading renders as a dash with a muted “pending” hint — never as a zero. A 0 that is the reading (a fresh higher print this year shows years-since-higher = 0; a not-at-a-record 30Y shows the all-time-high flag = 0) renders as the number. That split is deliberate: a stale or absent value dressed up as 0.00% would misread a trading decision.

Why the long end matters here

The long-end move is not a new isolated signal; it is an upstream state that transmits into legs the platform already computes. A long-end selloff feeds the equity channel (the SPY / 30Y correlation flipping positive — stocks and long bonds selling off together, with no bond hedge), the carry channel (a JGB-led shock is the trigger condition for the yen carry-trade unwind the carry_risk category already watches), the funding channel and the credit channel. The panel’s job is to make that upstream state legible next to the legs that react to it — description, not a new prediction.

Where it surfaces

• The dashboard — the Long-end panel beside the Equity & Macro board, carrying Blocks A and B (the US long end and the six-sovereign panel).
• The Engine — the “Long End & Treasury Supply” section in the full breakdown carries all four blocks (A the US long end, B the six sovereigns, C US supply & auctions, D who holds the debt), each level row with a 30-day sparkline where the carrier is tracked. The SPY / 30Y correlation also appears as a row in the Engine’s Correlations table.
• API — the long_end, sovereign_long_end, us_supply and debt_holders typed sub-objects on GET /api/v1/signals/latest. All four are public, mechanical, and vendor-anonymous — panel-only, the same posture as everywhere else this program surfaces.