Sovereign Stress Index

The number over time

Monthly from 2003-02-01 to 2026-10-07 · 285 reconstructed points · the same engine as the crisis anchors, run on the inputs that exist for each date (coverage 39.9%–69.6% against 100.0% today) · 37 earlier months were rebuilt and dropped as too thin to compare · red dots are the four crisis anchors, the green dot is the live reading. Today-only inputs are excluded, which is why this line sits lower than the headline: it cannot see the debt stock, the effective coupon or auctions from a past date.

Why the line reads below today’s number. The historical line is rebuilt from the inputs that existed on the date being scored — yields, spreads, funding, receipts, the deficit. Inputs that are only knowable today (the debt stock, the effective coupon, bill share, auction demand, the stablecoin float, bond-vol and bank equities) are left out rather than back-filled with present values, because back-filling them is how a chart starts lying about the past. The same four crisis anchors are marked as red dots; the live reading is the green dot at the right edge, and it is the only point that carries the full basket.

Right now

3.9

out of 10.0 · watch

As of 2026-10-07 (weekly)

4-week smoothed: 3.9 (1 weekly readings)

Coverage 100.0% of weights

No action. Review dates, no trades.

First published reading of this model.

Two numbers, on purpose. The headline above is the structural reading - slow-moving, comparable week to week. The acute channels are reported separately, because a plumbing failure is a 48-hour event and averaging it into a slow index is how you miss it:

Acute overlay: 4.2 of 10.0 · firing now: CCC-and-lower high-yield OAS, MOVE bond-vol index

The overlay is 0 at every watch level and 10 when any acute channel reaches its alert level. It raises the flag and gates whether a band may advance; it never changes the headline number, so 4.0 today still means what 4.0 meant last year. Carried-forward data this week: none - every component is this week's data.

What changed since the last reading

No input has moved by 0.3 points or more since the last reading. On a pressure gauge that is itself information: the inputs are quiet.

What would move it next — distance to the alert level

Every scored input, ranked by how close it is to its own alert level. This is the only forward-looking list on the site: it names the inputs that would change the reading, without pretending to know when. "Calm-to-alert" is where the input sits between the two published thresholds.

Input	Now	To alert	Where in the calm-to-alert range
30y Treasury yield	5.64 %	0.11 %	22.0%
Refinancing gap: effective coupon vs 10y	1.74 pp	0.26 pp	26.0%
Net interest / federal receipts	21.13 % of receipts	1.87 % of receipts	31.2%
Japan 30y JGB yield	4.17 %	0.33 %	33.0%
Japan 10y JGB yield	3.1 %	0.4 %	40.0%
Deficit, annualized from monthly	6.15 % of GDP	0.85 % of GDP	42.5%
10y Treasury yield	5.27 %	0.73 %	48.7%
Federal debt / GDP	123.6 %	11.4 %	57.0%

What is driving it, in plain terms

30y Treasury yield — 5.64 %, scored 8.9/10. the price the market demands for long US debt
Refinancing gap: effective coupon vs 10y — 1.74 pp, scored 8.7/10. the interest increase ALREADY locked in as the debt rolls
Net interest / federal receipts — 21.13 % of receipts, scored 8.4/10. Dalio's spiral metric - interest eating the revenue
Japan 30y JGB yield — 4.17 %, scored 8.3/10. long-end JGBs are what force Japanese lifers home; no free history for anchors
Japan 10y JGB yield — 3.1 %, scored 8.0/10. the funding leg of the global carry trade - Japan is the largest foreign holder of Treasuries

The mix is the story. Today the pressure sits in the structural items — what the market charges for long US debt, the interest bill, and how much debt must be refinanced — while the plumbing is quiet: overnight funding sits at the Fed’s floor, credit spreads are calm, and Treasury auctions are being absorbed normally. That is why this gauge reads higher than it did in 2008 and 2020. Those were acute, market-based panics inside a system that was structurally sound; this is the reverse.

What is not stressed — the honest other half

fires first in a bank-confidence problem; was 373bp in Oct 2008 and 210bp in Mar 2020 — currently 3.0 bp.
when repo clears above the Fed's floor, cash is scarce - the Sep 2019 repo spike came from here — currently -1.0 bp.
a dollar token trading below par is a live dollar squeeze - USDC hit $0.88 in March 2023 — currently 1.5 bp below $1.
foreign and fund demand stepping back means the dealers eat the supply - absorption failing — currently 71.5 %.

A gauge that only lists what is wrong is a mood, not a measurement. These are the inputs that would have to break to move the reading into the top band.

The next 12 months — three paths

Window: 2026-10-07 to about one year out. Four paths, cut so that they cannot overlap: what separates them is the end state, not how scary the ride was. These are judgment ranges, not measurements, they are about pressure, and they are not a collapse date.

Method, printed so the method can be attacked rather than the verdict. Two layers. Layer one, the baseline: how often a 12-month window contains a systemic US funding event — roughly one year in five over the last 50 (1987–89, 1998, 2008, 2011, 2019, 2020, 2023), so about 20% for any disorderly event. This is a stated assumption, not a sourced statistic, and it is the weakest number on this page. Layer two, four adjustments from today’s readings: r minus g is −2.8pp on the effective coupon — growth still outruns the rate the budget actually pays, which quietly stabilises debt/GDP and pushes against the disorderly paths; the plumbing is quiet (overnight funding 1bp below the Fed’s floor, auctions clearing at 2.46 times cover with 71% indirect bidders, the acute overlay at 4.2) — also against; the structure is not quiet (net interest 21% of receipts, refinancing gap 1.78pp, CCC spreads 12.1%, Japan’s 30-year at a record) — for; and the tail is fat, because funding failures are step functions rather than drifts, which is what earns Path D its floor. Calibration: each path is graded when its window closes, so the method can be scored rather than believed. The 20% baseline is the target to beat.

48–58% · central 53%

Path A — Orderly. No disorderly event, and no lasting damage

What it looks like: yields stay high and mostly range-bound. Deficits keep growing. Nobody is forced to rescue anything. Credit stays broadly calm. This index wanders between about 4 and 6 all year and never reaches the top band.

Why it is the base case: most 12-month windows contain no systemic event, and every fast-moving stress indicator is currently quiet. Bad structure is survivable for years — that is what the 2010s, and the years since 2023, have shown.

What you would see: bid-to-cover staying above 2.3x, indirect bidders above 65%, CCC spreads under 10%, overnight funding pinned inside the Fed’s corridor, and the acute overlay staying at or near 0.0.

Baseline carried: 20% for any disorderly event, so 53% is the baseline minus the structural adjustment, not a guess made in isolation.

28–38% · central 33%

Path B — Disorderly, then CONTAINED: the backstop works

What it looks like: a fast move, not slow erosion. Weeks, not years. Risk assets and long bonds sell off together, funding spreads spike, policy responds, prices partly recover. Historically such episodes peak in 2–6 weeks and are bought back within a quarter.

Trigger order, most likely first: (1) a yen carry unwind — Japan’s 30-year at a record 4.17%, the BoJ normalising, and speculative yen positioning only mid-range, so there is room for the trade to crowd; (2) a France / ECB event — the France–Germany spread is roughly double where it was weeks ago and the ECB’s crisis tool is blocked by fiscal non-compliance; (3) a failed long auction — dealer takedown is already 10%, and 22% is the alert line.

What you would see first: the acute overlay firing — 10-year velocity above 40bp in a week, VIX/VIX3M above 1.0, CCC above 10%, bid-to-cover below 2.3x, industrial-bank equity breaking down, Japan’s 30-year above 4.5%. The cost this path leaves behind: a permanently higher rate and a bigger interest bill, even after the backstop works. That is what makes it a distinct end state rather than a scare.

Baseline carried: the same 20%, adjusted up because the structure is bad enough that a shock is more likely to start a spiral than to pass through cleanly.

4–10% · central 7%

Path C — Genuine de-escalation

What it looks like: long yields fall because the pressure is genuinely relieved, not because something broke. Inflation cools, growth holds, the interest bill stops eating a larger share of revenue, and Japan and Europe both steady.

What it needs: the 5-year-forward inflation reading staying anchored near target, net interest back below 20% of receipts, and the deficit narrowing through growth rather than accounting. The most honest route is a productivity and investment boom — which is exactly what the AI capital-spending story would have to deliver.

The honest asymmetry: this is the only path that makes a 2027 retirement-and-relocation decision easier, and it has the least support in today’s numbers.

2–6% · central 4%

Path D — Disorderly and UNCONTAINED

What it looks like: a funding or auction failure that the official backstop does not fully contain within the window — the plumbing breaks and stays broken long enough to force restructuring of expectations rather than a repricing.

Why it gets a non-zero floor: plumbing does not fail gradually, and the preconditions are visible now — a deeply negative swap spread (unscored, no free source), drained cash buffers, a 22% bill share and a refinancing gap of 1.78pp. This is the path that would matter most and is least likely; it is on the page so it is never a surprise, and the action ladder already covers it.

Baseline carried: a fraction of the 20%. History says the backstop normally works, which is exactly why this is small and not zero.

The dated decision map — the only place this index is allowed to matter

This instrument never changes a decision. It can only change preparation. These are the four real forks, and the reading that would move each one.

Decision	When	What the index is allowed to change
The tax bill (~$46,500–49,500)	April 2027	Nothing. The date is set by tax law, not by a gauge. The index may only prompt earlier funding of the reserve, never a different sale.
In-system vs out-of-system retirement assets	review ~2031	The timing and the strength of the case being prepared. A band advance only counts here if a structural category (fiscal, funding, absorption) is the one rising — never the fast market or flank limbs alone.
Where to live: Philippines SRRV vs Thai DTV	2027	The order of the paperwork and which deposit is locked. Never the destination itself, and never two deposits.
Crypto as a sovereign hedge	ongoing	The single most decision-relevant sentence on this site: the escape-valve verdict is currently SHUT — BTC fell while gold rose and the dollar firmed, so crypto is behaving as a risk asset, not a hedge. Basis: one 12-month window, and the sign has flipped before.

What these probabilities are not. Not a date, not a collapse prediction, not a trade signal. They are a structured statement of where the pressure most likely goes next. The most useful thing on this page is not the central number — it is the trip-wire list, because that is what tells you which path you are actually in, in real time, rather than after the fact.

Pressure by category

rates

4.3

fiscal

6.2

foreign

2.6

credit

2.7

funding

0.1

absorption

1.1

market

3.2

flank

6.1

Context — reported, deliberately not scored

Reading	Value	Why it is here and not in the score
Gold, 12m change	4.1 % 12m Yahoo GC=F	the classic reserve asset
Silver, 12m change	29.6 % 12m Yahoo SI=F	monetary demand broadening to the higher-beta metal
Bitcoin, 12m change	-30.7 % 12m Yahoo BTC-USD	in 2026 BTC trades as a high-beta RISK asset, not a debasement hedge - the escape valve reads SHUT
Private-credit proxy (BIZD), drawdown from 52w high	-19.4 % Yahoo BIZD	private-credit marks are quarterly; the listed BDC price is the only daily read
Dollar liquidity buffer (reserves + reverse repo)	2.95 $trn FRED WRESBAL + RRPONTSYD	the cash stock that absorbs issuance - but it is ALREADY drained (RRP ~0), so the level is not a clean stress score
Fed USD swap lines outstanding	207.0 $mn FRED SWPT	drawings prove an offshore dollar squeeze is live - coincident, not leading
Crypto escape-valve verdict (BTC vs gold, 12m)	n/a Yahoo GC=F / BTC-USD	ESCAPE VALVE SHUT - BTC -30.7% 12m while gold +4.1%, so crypto is trading as a high-beta risk asset, NOT as a sovereign hedge
US 10y swap spread (dealer capacity)	n/a NO KEYLESS SOURCE - needs a paid feed	the panel's most valuable unavailable input; named here so it is not forgotten

Components — heaviest first

Component	Reading	Score	Band
30y Treasury yield the price the market demands for long US debt · watch 5.25, alert 5.75 · treasury.gov par curve / FRED DGS30	5.64 %	8.9	acute
Refinancing gap: effective coupon vs 10y the interest increase ALREADY locked in as the debt rolls · watch 1.0, alert 2.0 · Treasury FiscalData avg_interest_rates vs FRED DGS10	1.74 pp	8.7	acute
Net interest / federal receipts Dalio's spiral metric - interest eating the revenue · watch 17.0, alert 23.0 · FRED A091RC1Q027SBEA / FGRECPT	21.13 % of receipts	8.4	acute
Japan 30y JGB yield long-end JGBs are what force Japanese lifers home; no free history for anchors · watch 3.5, alert 4.5 · MoF daily JGB CSV (cached)	4.17 %	8.3	acute
Japan 10y JGB yield the funding leg of the global carry trade - Japan is the largest foreign holder of Treasuries · watch 2.5, alert 3.5 · MoF daily JGB CSV (2026/10/6, cached)	3.1 %	8.0	acute
Deficit, annualized from monthly how fast the debt is still being added, on a MONTHLY source rather than the old annual figure · watch 5.0, alert 7.0 · FRED MTSDS133FMS (monthly, annualized) / GDP	6.15 % of GDP	7.9	alert
10y Treasury yield benchmark funding cost for everything else · watch 4.5, alert 6.0 · treasury.gov par curve / FRED DGS10	5.27 %	7.6	alert
Federal debt / GDP the stock that has to be rolled and refinanced · watch 115.0, alert 135.0 · Treasury Debt to the Penny (daily) / FRED GDP	123.6 %	7.2	alert
CCC-and-lower high-yield OAS the weakest borrowers lose market access first - the tail leads the average · watch 10.0, alert 15.0 · FRED BAMLH0A3HYC (history starts Oct 2023)	12.11 %	7.1	alert
Euro area 10y yield (ECB, daily) the euro long end - the other half of the France story, daily · watch 3.0, alert 4.5 · ECB Data Portal YC series (daily)	3.5 %	6.7	alert
MOVE bond-vol index Treasury-market dysfunction, live - no free historical source, so it is live-only · watch 100.0, alert 130.0 · Yahoo ^MOVE	105.2 index	5.9	alert
Private-credit proxy (BIZD) drawdown private-credit marks are quarterly, so the listed vehicle is the only live read on that risk · watch -20.0, alert -35.0 · Yahoo BIZD	-19.4 % from 52w high	4.8	watch
Yen, 3m change vs dollar a FIRMING yen is the carry unwind firing (falling value = stress) · watch -3.0, alert -10.0 · Yahoo JPY=X	-1.99 % 3m	4.3	watch
France - Germany 10y spread the euro core losing its risk-free premium - the belief the whole curve rests on · watch 100.0, alert 200.0 · FRED IRLTLT01FRM156N / IRLTLT01DEM156N MONTHLY (approx 2-month lag)	82.0 bp	3.2	watch
US high-yield OAS credit repricing leads sovereign stress (BB/B dominated) · watch 3.5, alert 5.0 · FRED BAMLH0A0HYM2	3.12 %	3.1	watch
Bill share of marketable debt bills roll every few months, so any funding wobble compounds instead of being absorbed by time · watch 24.0, alert 28.0 · Treasury FiscalData MSPD table 1 (monthly)	22.36 %	3.0	watch
China's Treasury holdings, 12m change the strategic buyer who will not respond to yield - at an 18-year low · watch -110.0, alert -220.0 · FRED FORTREASPOS41408 (monthly, ~6-week lag)	-69.7 $bn 12m	2.6	watch
Auction bid-to-cover, 10y + 30y cover is the first place waning end-user demand shows up · watch 2.3, alert 2.0 · Treasury FiscalData auctions_query	2.46 x	2.3	calm
Effective r minus g the master debt-dynamics condition: above zero, debt/GDP grows even with a balanced budget · watch -1.0, alert 1.5 · Treasury FiscalData avg_interest_rates vs FRED GDP	-2.78 pp	2.0	calm
Regional-bank equity (KRE) drawdown bank-equity stress moves first in a run - the Mar 2023 first mover this index had lost · watch -20.0, alert -35.0 · Yahoo KRE	-10.09 % from 52w high	0.9	calm
5y5y forward inflation long-run inflation anchor - if it de-anchors, inflating the debt away stops working · watch 2.6, alert 3.0 · FRED T5YIFR	2.35 %	0.8	calm
VIX / VIX3M term structure backwardation means dealers are paying up for immediate risk cover · watch 1.0, alert 1.15 · FRED VIXCLS / VXVCLS	0.86 ratio	0.4	calm
Fed primary credit, weekly change vs 13w avg discount-window borrowing is the bank-funding first mover - it jumped within a week in Mar 2020 and Mar 2023, and it is the channel this index was blind to · watch 2000.0, alert 10000.0 · Federal Reserve H.4.1 release (scraped, weekly)	-1867.0 $mn	0.2	calm
Stablecoin float, 90-day change the float IS offshore dollar demand - a contracting float means dollars are scarce and tokens are being redeemed · watch -2.0, alert -8.0 · DefiLlama stablecoincharts (keyless, daily)	1.86 % 90d	0.2	calm
Long-end yield velocity, 5 sessions how FAST long yields are moving - the shape of the LDI and 2020-type events · watch 40.0, alert 80.0 · FRED DGS10 / DGS30	3.0 bp 5d	0.0	calm
Financial CP minus 3m T-bill fires first in a bank-confidence problem; was 373bp in Oct 2008 and 210bp in Mar 2020 · watch 60.0, alert 150.0 · FRED DCPF3M / DTB3	3.0 bp	0.0	calm
SOFR minus IORB (repo plumbing) when repo clears above the Fed's floor, cash is scarce - the Sep 2019 repo spike came from here · watch 10.0, alert 25.0 · FRED SOFR / IORB	-1.0 bp	0.0	calm
Stablecoin peg deviation (worst of USDC/USDT) a dollar token trading below par is a live dollar squeeze - USDC hit $0.88 in March 2023 · watch 100.0, alert 300.0 · CoinGecko simple/price (keyless)	1.5 bp below $1	0.0	calm
Indirect bidders' share, 10y + 30y foreign and fund demand stepping back means the dealers eat the supply - absorption failing · watch 62.0, alert 55.0 · Treasury FiscalData auctions_query	71.5 %	0.0	calm

All 29 components have data.

How a missing feed is handled now. If a series cannot be fetched, the component holds its last published value and is flagged stale, and a degraded banner appears at the top of the page. It is never silently dropped: dropping it would change the basket the composite is computed over, and the first version of this index could read calmer during a feed outage, which is the most dangerous failure a stress gauge can have.

Where today sits — historical anchors

Date	What was happening	SSI then	Today, same basis	Band
2008-10-15	Lehman aftermath model coverage then: 56.9% (12 components unavailable) · no data: Federal debt / GDP, Effective r minus g, Refinancing gap: effective coupon vs 10y, Bill share of marketable debt, US high-yield OAS, CCC-and-lower high-yield OAS, SOFR minus IORB (repo plumbing), Stablecoin float, 90-day change, Stablecoin peg deviation (worst of USDC/USDT), Auction bid-to-cover, 10y + 30y, Indirect bidders' share, 10y + 30y, Japan 30y JGB yield	4.6	4.7	watch
2011-08-05	US downgrade / debt ceiling model coverage then: 56.9% (12 components unavailable) · no data: Federal debt / GDP, Effective r minus g, Refinancing gap: effective coupon vs 10y, Bill share of marketable debt, US high-yield OAS, CCC-and-lower high-yield OAS, SOFR minus IORB (repo plumbing), Stablecoin float, 90-day change, Stablecoin peg deviation (worst of USDC/USDT), Auction bid-to-cover, 10y + 30y, Indirect bidders' share, 10y + 30y, Japan 30y JGB yield	3.4	4.7	watch
2020-03-23	COVID dash for cash model coverage then: 56.9% (12 components unavailable) · no data: Federal debt / GDP, Effective r minus g, Refinancing gap: effective coupon vs 10y, Bill share of marketable debt, US high-yield OAS, CCC-and-lower high-yield OAS, SOFR minus IORB (repo plumbing), Stablecoin float, 90-day change, Stablecoin peg deviation (worst of USDC/USDT), Auction bid-to-cover, 10y + 30y, Indirect bidders' share, 10y + 30y, Japan 30y JGB yield	3.2	4.7	watch
2022-10-12	UK gilt crisis / peak rates model coverage then: 59.6% (11 components unavailable) · no data: Federal debt / GDP, Effective r minus g, Refinancing gap: effective coupon vs 10y, Bill share of marketable debt, US high-yield OAS, CCC-and-lower high-yield OAS, Stablecoin float, 90-day change, Stablecoin peg deviation (worst of USDC/USDT), Auction bid-to-cover, 10y + 30y, Indirect bidders' share, 10y + 30y, Japan 30y JGB yield	2.8	4.3	watch

Read this honestly, and read the coverage column first. The anchors run on a much smaller component set than today - the funding, absorption and credit machinery largely did not exist as data in 2008 - so the “SSI then” figure is not like-for-like with today’s number. That is why every row carries a second figure: “Today, same basis” is today’s reading recomputed on exactly the components that anchor had. That is the only fair comparison on this page. It is also why the 2020 row reads so low: the index is deliberately slow to shout, and a pricing shock inside a structurally sound system is not what it is built to catch.

What to do at each band — the action ladder

0.0–3.9 — calm. Nothing. Look once a week at most.
4.0–5.9 — watch. Review dates and assumptions early. No trades. This is the band where the job is keeping the plan current, not changing it.
6.0–7.4 — alert. The pre-committed review fires on schedule: the in-system vs out-of-system call, with its tax cost priced before anything is sold.
7.5–10.0 — acute. The pre-written plan executes — as written, not as felt.

This index never triggers a trade and never prints a collapse date. It measures pressure, not proximity to an event. Seat 37 rule.

How it is built, and what it cannot do

Each component scores 0 at its calm anchor, 5.0 at WATCH and 10.0 at ALERT, capped at 10 beyond alert. Category weights are listed above; components inside a category share it equally. It is 29 inputs across 8 categories.

14 of those inputs came from a five-lane expert review on 2026-10-07 — fiscal debt-dynamics, offshore dollar funding, Japan and Asian capital flows, bank credit, and market microstructure. Every source was verified live before admission; every addition named a cut that paid for it. The panel’s proposed numbers that failed verification, or whose sign is ambiguous, are printed as context instead of being scored.

Missing data is reported, never guessed. A component whose feed fails holds its last published value and is flagged; coverage and the stale list are printed. An earlier version of this page claimed a failed fetch could never read as a lower score — that claim was false, because dropping a component and renormalising the weights changes the basket the number is computed over. It is fixed rather than reworded.

The 4-week smoothed value is the mean of the last four weekly readings, so one bad week cannot make a trend. Historical anchors use the same code path on FRED history only, so several market-overlay and funding components are absent from the older anchors.

Known lags: the France–Germany spread is MONTHLY (about two months behind), and the China and Japan Treasury-holdings lines are monthly with a ~6-week publication lag, so a fast move in recent weeks is not in them. The Japanese yield curve, the yen, funding spreads, credit spreads and the auction numbers are current. Nothing here is a forecast.