Tracks directional shifts in the US bond market: curve shape, term premium, auction demand, policy events and credit spreads. Every figure is labeled with its date and definition, and conclusions are revised as data change.
The mid-July read was "sticky inflation + hawkish Fed." By mid-August, the data briefly pointed the other way: payrolls turned negative, inflation eased, and September hike odds fell to about 30%. On Aug 28 at Jackson Hole, Warsh pushed it back with a single speech: the September hike probability climbed from about 35% before the speech all the way to About 100%（CME，9/10）；The September 16 FOMC hiked rates 25bp to 3.75%–4.00% as expected, in a 12–0 unanimous vote; the 10-year closed on September 14 at 4.97%, the highest since October 2023; after the US military struck Iranian targets near the Strait of Hormuz on 9/1, Brent spot rose to $109.51 by 9/9. The rise in long-end yields has passed through to households: the 30-year mortgage rate rose to 6.95%, the highest since June 2025. Meanwhile The Sep 4 August payrolls report showed +162K and revised July from −23K to +21K. The "weakening jobs" narrative behind this page for two months was overturned by a single revision.
Each cell showsMost recent change in directiondate and magnitude. The trend line is the indicator's actual series since the start of 2026; the end point is the current value.
Direction labels are based on "the move since the most recent extreme" and are not forecasts. CPI uses the BLS-published non-seasonally-adjusted YoY measure, whose history is not included on this page; buyback size and TGA use are policy events with no corresponding price series, so these three cells have no trend lines.
The items below are not a risk checklist (that's in Section 12), butThe highest-leverage structural nodes, yet widely unpriced. What they share: a very small quantity decides a very large outcome, and most have a clear reveal date. Each one was verified individually for this page using public data.
| Chokepoint | Why it's a chokepoint | Current reading | Release |
|---|---|---|---|
| inflation expectations have not moved at all | The long-end rise is widely attributed to "inflation and fiscal worries", but broken down it is almost entirelyReal rates | 2026 nominal +78bp = real +66bp（85%） + inflation expectations +12bp (9/14); the 10-year breakeven's full-year range is only 32bp | Occurred |
| Real rates at a decade high | What's weighing on the real economy is the actual cost of funding, not inflation; the 7.24% mortgage rate is its direct result | 10Y real yield 2.63%(10/29, FRED DFII10), down 4bp from 2.67% on 10/28, still in the highest range since 2015 (it reached 2.93% on 9/30 and a new high of 2.95% on 10/5) | Persistent |
| Equity-bond volatility divergence at 70th percentile | The "low-volatility environment" only holds for equities. Bond volatility is actually above its median, and the tension between the two has to resolve on one side eventually | VIX 17.31(10/20, FRED VIXCLS close) vs 10-year realized volatility 72bp(No. 71 percentile); ICE MOVE Index, on 9/24, in a single day +9.6% broke above 100 to 104.58, the first time in three months, eased to 96.00(−8.2%), falling back below 100 again, rose again on 9/28 to 101.82(+6.06%), back above 100 again, rose further on 9/29 to 106.60(+4.70%), rose further on 9/30 to 110.45(+3.61%), fell back to on 10/1 108.13(−2.11%), fell further on 10/2 to 107.29(-0.78%), 10/5 rebounded to 113.60(+5.88%), 10/6 fell back to 105.20(−7.39%), fell further on 10/7 to 102.56(-2.51%), 10/8 fell further to 100.70(-1.81%), on 10/9 fell further to 98.47(-2.21%), 10/13 turned up to 101.09(+2.66%), back above 100 again; fell back on 10/14 to 99.06(-2.01%), back below 100 again; on 10/16 rose to 101.1(−1.4% on the day, still up about 21% for the month), rising further on 10/19 to 101.3(+0.2%); climbed further on 10/20 to 101.7(+0.39%), 10/21 fell back to 95.94(−5.67%), dropping back below 100 again, then rebounding sharply on 10/22 to 106.85(+11.36%), breaking above 100 again, pulled back on 10/23 to 103.31(-3.30%), still above 100, 10/26 fell further to 102.19(about -1.08%, CNBC .MOVE, 10/26), 10/27 pulled back to 97.87(about -4.23%), dropping back below 100 again; it fell further on 10/29 to 92.97(-1.68% on the day, -9.70% over the week, -9.46% over the month), remaining below 100, rebounding on 10/30 to 100.21(+7.12% on the day), breaking back above 100 (9/24-10/14, 10/19, 10/21, 10/23, 10/27 from Yahoo Finance; 10/16, 10/20, 10/22, 10/29, 10/30 from TradingView MOVE) | 9/11、9/16 |
| the 5Y is the cheapest segment of the year | Everyone watches 2s10s, but hike repricing actually hits the belly hardest, while the 10-year is propped up by buybacks and duration demand | 2-5-10 butterfly −3.0bp(10/29), down 1bp from -2.0bp on 10/28; the 5-10-30 is -21bp, only the 15th percentile as of 9/3 | Persistent |
| Refinancing gap 101bp (locked in) | No new deficits needed: roll the existing debt onto the current curve along the maturity schedule, and interest costs rise on their own | Outstanding weighted coupon 3.49% vs about 4.5% on blended new issuance. Once fully rolled over, interest will rise from 3.84% of GDP to 5.6% | Realized steadily along the maturity schedule |
| Real-rate spiral | The debt spiral is usually framed as an "inflation spiral," but what drives it isRealrates, and real rates are at a decade high | Federal interest outlays 3.84% of GDP, already exceeding defense (1.04x); total debt $40.07 trillion(9/9), 122.6% of GDP | Persistent |
| Auction concessions quietly widening | Price evidence bid-to-cover can't show: buyers are demanding more compensation | Stop-out above the same-day close: July about +0.3bp → August +1.6bp; 8/19 20-year +3.4bp | On each auction day |
| Buybacks already priced in | The policy works on relative value, not absolute levels; operations only began 9/9, and the spread has already fully retraced | 20Y cheapness on 8/19 announcement day 28.5→25.0bp, has not returned to 28 since; at +24.0 it is now the narrowest of the year | 9/9 operations begin |
| The credit-contagion pathway has moved to private credit | HY OAS only sees the public market; private credit has no daily quote, and by the time it feeds through to public spreads it's usually already mid-cycle | BDC basket median discount 10.7%(vs. 6/30 NAV, 10/30 close), median drawdown 18.0%; the CCC-BB tiering spread 1,131bp, while the HY composite OAS is 326bp (10/29) | Daily (BDC/tiering spread) + NAV revealed each quarter |
| CPI–PCE gap | The Fed uses PCE while market headlines read CPI; for the same month the two gauges point to opposite policy conclusions | August core CPI came in at 2.4% (BLS, 9/11) vs a core PCE nowcast of 3.40% (9/10),Gap of about 1.0pp | 9/11 |
| 30Y auction the day before CPI | $22bn of longest-duration supply has to be priced before the inflation print is known | 9/10 result: bid-to-cover 2.61, indirect 79.3% (Aug: 2.39 / 53.1%) | 9/10 |
| Waller's unobservable threshold | Seen as needing core PCE MoM >0.30% to support a hike, but that data was not yet out on the day of the vote | Nowcast 0.27%(9/10), a 3bp gap, and only an estimate | 9/16 vote |
| Meeting comes two weeks before PCE | August PCE is due 9/30, so committee members can only back it out from CPI/PPI, and this month's mapping between them is distorted by methodology changes | Confirmed by BEA schedule | 9/30 |
| PCE annual revision | Five-year benchmark revision starts the same day; a September hike followed by a downward benchmark revision at month-end would be an awkward combination | Realized: 9/30 core PCE YoY revised down to 3.0%(the new August reading and the revised July value are both 3.0%), closely matching Citi's forecast of a "downward revision of more than 30bp to roughly 3.0%" | Occurred (9/30) |
| Structural shift in auction demand | TIC holdings data lag by two months and include valuation effects;Auction flows are the only high-frequency truth, yet few track it monthly | August total bids did not fall; indirect award 62.8%→55.1%; the first three in September rebounded to 61.7%–79.3% | On each auction day |
| Real policy rate has fallen to zero | The hawks' real argument is not runaway inflation but that policy is no longer restrictive, a number that is rarely reported | 3.75–4.00% minus core PCE = +0.26～+0.66pp | Persistent |
| 20Y–30Y spread | Every report talks about 2s10s; the one actually swinging back and forth around zero is this one | YTD as of 9/3 10 timesCrossed the zero line, inverted ever since, at -3bp on 10/29 (calculated from FRED DGS20/DGS30) | Persistent |
Three kinds of systematic errors this page kept running into while fact-checking:
1. Mixed-up definitions.The classic case is "BofA and Citi have opposite playbooks": their core CPI MoM forecasts are 0.22% and 0.184%, both 0.2% when rounded; with 11 of the 12 months in the YoY already historical data, the disagreement arithmetically shouldn't exceed 0.05pp, let alone 110bp. The truth isCiti's 2.3% refers to core CPI; BofA's 3.4% refers to core PCE, a technical definitional gap dressed up as a bank showdown.
2. Narrative inertia overrides the data.Examples verified on this page: one article said the 10-year hit a "15-year high" (it actually reached 4.98% in 2023 and 4.78% in January 2025); called the curve "extremely steep" (2s10s is at the 32nd percentile since 1990, flatter than normal); said China's US Treasury holdings of $757bn were a "16-year low" (TIC actually shows $652.3bn, the lowest since September 2008); and described Warsh, in office for over three months, as a "potential new Chair."
3. Looking only at levels, not at what they are made of.The August auctions are the best example: bid-to-cover alone says "demand is normal," while indirect award alone says "demand collapsed,"The truth shows only when you read both: total volume did not fall, but the buyers changed. The same goes for term premium: it is the highest since 2015 (supporting the "risk compensation" narrative), but it explains only 53% of the 10Y rise (refuting the idea that "bear steepening = risk compensation").
"Low volatility" only means low volatility in equities.A common combined narrative is 'VIX at 14, MOVE at 72 - markets are too calm.' That narrative has now been thoroughly shattered - on 9/24 the ICE MOVE Index, in a single day, +9.6% broke above 100, closing at 104.58, the first time in three months; on 9/25 it fell back again to 96.00(−8.2%), falling back below 100; rose again to 101.82(+6.06%), back above 100 again; rose further on 9/29 to 106.60(+4.70%); rose further on 9/30 to 110.45(+3.61%); fell back to on 10/1 108.13(−2.11%); fell further on 10/2 to 107.29(-0.78%); 10/5 rebounded to 113.60(+5.88%); 10/6 fell back to 105.20(−7.39%); fell further on 10/7 to 102.56(-2.51%); 10/8 fell further to 100.70(-1.81%); on 10/9 fell further to 98.47(-2.21%); 10/13 turned up to 101.09(+2.66%), back above 100 again; on 10/14 fell to 99.06(-2.01%), back below 100 again (Yahoo Finance); on 10/16 rose to 101.1(-1.4% on the day, still up about 21% for the month, TradingView MOVE); climbed further on 10/19 to 101.3(+0.2%, Yahoo Finance); climbed further on 10/20 to 101.7(+0.39%, TradingView MOVE); 10/21 fell back to 95.94(−5.67%, Yahoo Finance), dropping back below 100 again; rebounding sharply on 10/22 to 106.85(+11.36%, TradingView MOVE), breaking above 100 again; pulled back on 10/23 to 103.31(-3.30%, Yahoo Finance ^MOVE), still above 100; 10/26 fell further to 102.19(about -1.08%, CNBC .MOVE), still above 100; 10/27 pulled back to 97.87(about -4.23%, Yahoo Finance ^MOVE), dropping back below 100 again; it fell further on 10/29 to 92.97(-1.68% on the day, -9.70% over the week, -9.46% over the month, TradingView MOVE), remaining below 100; rebounding on 10/30 to 100.21(+7.12% on the day, -4.72% over the week, -4.24% over the month, TradingView MOVE), breaking back above 100. It's clearer to view VIX and bond-market volatility within their own distributions: VIX 17.31(10/20 close, FRED VIXCLS, still in the low end of the year's 14.2-31.1 range), while the 10-year 21-day realized volatility 72bp at the 71 percentile (range 38-89bp) — the two sit at starkly different positions. The bond market has long since stopped being calm; only the stock market is still pricing in calm, and the direction of MOVE's single-day jump is consistent with realized volatility. If the two events on 9/11 and 9/16 were to transmit bond volatility across, nearly all of the upside room is on the VIX side.
Hike repricing hits the belly, not the long end.2-5-10 butterfly (2×5yr − 2yr − 10yr) rose as high as +1.0bp, matching the level reached on both 9/10 and 9/15 of −1.0bp YTD high(14bp cheaper than 7/17); on 9/25 narrowed to −2.0bp, on 9/28 declined further to −4.0bp, as of 9/29 rebounded to −3.0bp, rose further on 9/30 to +1.0bp, tied with 9/24 for the YTD high, fell back to on 10/1 0.0bp, on 10/2 rose back to +1.0bp, on 10/5 pulled back to −3.0bp, rebounded on 10/6 to 0.0bp, on 10/7 rebounded to +1.0bp, on 10/8 unchanged at +1.0bp, on 10/9 fell to 0.0bp, flat on 10/13 at 0.0bp, on 10/14 fell to −1.0bp, on 10/15 fell further to −5.0bp, flat on 10/16 at −5.0bp, rose back to on 10/19 −4.0bp, rose further to on 10/20 −3.0bp, fell back to on 10/21 −4.0bp, 10/22 fell further to −5.0bp, 10/23 unchanged at −5.0bp, 10/26 rose to 0.0bp, up 5bp from 10/23, falling back on 10/27 to −5.0bp, down 5bp from 10/26, rebounding on 10/28 to −2.0bp, up 3bp from 10/27, falling back on 10/29 to −3.0bp, down 1bp from 10/28 (FRED) —The 5-year's cheapness relative to the wings has been seesawing around the zero line: the belly stayed richer from 10/15-10/23, flattened on 10/26, then turned richer again from 10/27-10/29. Meanwhile, the 5-10-30 butterfly −21bp(only the 15th percentile on 9/3), meaning 10Y relatively rich. Taken together, the picture is very specific: hike expectations hit the front end and belly together, while the 10Y was held up by buyback expectations and duration demand. The 2s10s spread everyone watches happens to average this structure away and shows nothing.
Realized volatility is the 21-day rolling standard deviation of daily DGS10 changes, annualized over 252 trading days, in basis points, used as a computable proxy for the proprietary ICE MOVE index (MOVE is a proprietary index and this page has no continuous feed access; single-point check values for 9/24 (104.58), 9/25 (96.00), 9/28 (101.82), 9/29 (106.60), 9/30 (110.45), 10/1 (108.13), 10/2 (107.29), 10/5 (113.60), 10/6 (105.20), 10/7 (102.56), 10/8 (100.70), 10/9 (98.47), 10/13 (101.09), 10/14 (99.06), 10/19 (101.3), 10/21 (95.94), 10/23 (103.31) are from Yahoo Finance; 10/16 (101.1), 10/20 (101.7), 10/22 (106.85) are from TradingView MOVE; 10/26 (102.19) is from CNBC .MOVE; 10/27 (97.87) is from Yahoo Finance ^MOVE; 10/29 (92.97) is from TradingView MOVE; 10/30 (100.21) is from TradingView MOVE). The butterfly is calculated directly from the Treasury curve, where a positive value means the belly is cheap relative to the wings.
Every figure in this section has a corresponding data source and verification in other sections of this page; "unpriced" means rarely appearing in mainstream coverage and sell-side commentary, and is neither a judgment on market efficiency nor investment advice.
bp = basis point, 1 basis point = 0.01%; OAS is option-adjusted spread. The 10-year, 2-year, and 30-year yields are 10/29 closes (FRED DGS10/DGS2/DGS30); the 2-year-10-year spread is the FRED 10/30 value; the 3-month-10-year spread is the FRED 10/30 value; OAS is the FRED 10/29 value; all changes are relative to 2026-07-17; the mortgage rate is the Freddie Mac weekly value (10/29, FRED MORTGAGE30US, comparison date 7/16).
| Tenor | 10-16 (%) | 07-17 (%) | Change (bp) |
|---|
Earlier, the Kim-Wright model split the 10-year into "term premium + policy expectations," roughly half each. Using a harder measure:Real yields and breakeven inflation quoted directly by the TIPS market, and the answer is no longer half and half:
2026 YTD (through 9/14), 10-year nominal yield up 78bp, of whichReal yield contributed 66bp (85%), inflation expectations only 12bp (15%). And inflation expectations barely moved: the full-year range of the 10Y breakeven was only 32bp(2.18%–2.50%), 5y5y forward expectations range 30bp, currently 2.22%（10/30）。
It's even starker against this year's inflation shock: Brent went from $61 rose to $138(+126%), while headline CPI surged from 2.7% to 4.5%,The 10-year breakeven only moved from 2.27% to 2.38%; and on April 7, when oil prices were highest, 5y5y forward expectations instead posted the year's low of 2.13%——The market read the energy shock as demand destruction, not inflation。
The conclusion is a story commonly told backwards:The long end is pricing a "permanent upward shift in real funding costs", not inflation.10Y real yield 2.63%(10/29, FRED DFII10), down 4bp from 2.67% on 10/28, still in the highest range since 2015 (it reached 2.93% on 9/30 and a new high of 2.95% on 10/5). This is the number holding down the 7.24% mortgage rate and corporate financing costs — and the narrative that "sticky inflation is pushing up the long end" finds no evidence in the TIPS market.
Methodology: DFII10 (10-year Treasury Inflation-Indexed yield) and T10YIE (10-year breakeven inflation), T5YIFR (5-year, 5-year forward inflation expectation) are all FRED daily series; the latest available date for T10YIE and T5YIFR is 2026-10-30, while the latest available DFII10 reading from FRED is 2026-10-29. Breakeven is on a CPI basis, typically about 0.3 percentage points above the PCE basis — meaning the current 2.34% breakeven corresponds to roughly 2% PCE inflation expectations, still at the Fed's target.
This claim can be tested directly. Split the 10-year yield into two parts:Term premium(the extra compensation investors demand for bearing duration risk) andExpected average short-term rate over the next decade(pricing of the Fed path). Add the two together and you get the 10-year itself.
The supporting side: the term premium is indeed at a decade high.It went from the start-of-year 0.58% rose to 0.875%(8/28, still setting new highs), is the highest since 2015, at the 100th percentile; by comparison, the 2015–2025 median was only 0.13%, with a range of −0.66% to 0.81%. Long-term capital is not only demanding more compensation, that compensation is now at its most expensive in a decade.
What needs correcting: it explains only half.2026 YTD 10Y rise 57bp, with term premium contributing 30bp, expected short rate contributes 27bp——Term premium 53%. The result holds in other windows too: 47% since the 6/18 curve low and 50% since the 7/17 previous version. So this leg of the long-end rise ishalf risk compensation, half policy expectations, not a single factor.
There is also a timing issue: this "bear steepening" was interrupted once, then resumed.Early to mid-August saw a long-end-led bear steepening. On 8/28, the day of Warsh's speech, it became a short-end-led bear flattening (spread +47→+39bp) as policy rate expectations took over as the driver. From 8/31 the long end led again, the spread returned to +41bp, and on 9/1 long-end yields hit a new high.Two forces are taking turns in the lead, not just one.Also, +41bp is still only at the 32nd percentile since 1990. The curve's shape itself isn't extreme; what's truly abnormal is the absolute level of the long end, and the decade-high term premium within it.
Term premium uses the Fed's Kim-Wright three-factor model estimate (FRED series THREEFYTP10 / THREEFY10); the latest available date is 2026-08-28, one period behind other data on this page. Term premium is a model estimate; different models (e.g., the New York Fed's ACM) differ in level but usually agree on direction and inflections.
Phase 1 (7/17 – 8/27): bear steepening.The shift is concentrated at the long end;2YUp only 2bp，10Y +12bp、30Y +13bp. The short end was 3.50%–3.75% policy rate anchoring it, the 2-year–10-year spread steepened all the way from its June 18 YTD low of +27bp to August 21's +50bp。
Phase 2 (8/28): hawkish bear flattening.The whole curve moved up on the day of Warsh's speech, but the short end rose more:1Y +11bp、2Y +14bp、10Y +6bp、30Y +3bp, with the spread plunging from +47bp to +39bp. Taking the two periods together (7/17 → 8/28 in the chart above), the curve was actually close toParallel shift up, the biggest gains were actually in the belly:3Y +20bp、5Y +20bp、7Y +19bp。
Phase 3 (8/31 – 9/2): the whole curve rose together; nothing gave way.On Sep 1 the 10Y closed at 4.79%. Across the full FRED series, this is Jan 13, 2025high since then; 30-year 5.27%, level with the 20-year again. But the short end was not idle during this stretch: the 2-year rose from 4.34% to 4.39%, so the 2-year–10-year spread only went from +39 to +40bp.Sep 3 ADP jobs data weaker than expected; short end falls first(2-year back to 4.34%, 1-year −5bp), while the 10-year fell just 2bp to 4.77%, widening the spread to +43bp: a brief bull steepening.then reversed again after the Sep 4 payrolls beat: 2-year +3bp, 10-year +1bp, 30Y −1bp, with the spread narrowing back to +41bp. The reaction that day was concentrated entirely at the front end, and long-end yields fell rather than rose:Pure policy repricing, with no term premium component。
Taking 7/17 to 9/1 as a whole, the biggest rise was inBelly：3Y +25bp、5Y +27bp、7Y +26bp, exceeding the 10-year's +24bp and the 30-year's +21bp;3M–2Yspread widened from +33 to +47bp。This is no longer a pure "long-end story". Hike expectations lifted the entire front half of the curve together.
"Steep" needs a yardstick:+41bp sounds steep, but placed in the distribution since 1990, it sits at the32nd percentile. The 1990–2025 median is +88bp, average +101bp, at its steepest +291bp。The current curve is flatter than normal, not steeper; what is truly abnormal is the absolute level of the long end, not the gap between long and short ends.
Two verifiable highs:30Yhit on August 17 5.31%. Across the complete FRED series, this is Highest since Jun 12, 2007 (5.35%)；10Yhit on July 31 4.75%, at highest since Jan 2025. Another detail is 20Y and 30Yinversion ended on August 13 and returned on 9/10 2bp; this spread has YTD 10 zero-line crossings, is the part of the curve that truly swings back and forth.
Cross-check: US Treasury data on this page come from FRED, with Tushare us_tycr An independent review of 2 / 10 / 30-year yields over the 12 trading days from 8/12–8/27:match exactly day by day and basis point by basis point (0bp difference)。
Looking at the US long end in isolation, it's easy to 5.31% entirely to the US fiscal deficit. But August's long-end selloff wasglobally synchronized: 30-year yields across major developed economies simultaneously hit multi-year highs.
| Market | 30Y yield | Reference |
|---|---|---|
| US 30Y | 5.23% | Reached 5.31% on 8/17, highest since Jun 2007 |
| UK 30Y | About 5.85% | Approaching the May peak, highest since 1998 |
| Japan 30Y | About 4.07% | 8/21 value, near a record |
| Japan 10Y | 2.88% | Reached on 7/9, highest since Sep 1996 |
| Germany 30Y | 3.75% | First time reached since 2011 |
| France 30Y | — | Borrowing costs rise to highest since 2008 |
US figures are from the Treasury curve (8/24 close, 8/17 high, cross-checked against the full FRED series). The rest are values from late-August news reports. Reference dates differ by country, so they are listed side by side instead of sharing one trend chart.
Even as supply rises, the most stable class of buyers is exiting. US Treasury TIC data showChina's holdings of US Treasuries fell to $652.3bn in March 2026, the lowest since September 2008, $659.3bn in May, half its 2013 peak of over $1.3 trillion. Over the same period the PBOC has added gold for many consecutive months. In Japan, with domestic yields back at three-decade highs and the BOJ policy rate raised to 1%, and the yen carry trade that long supported Treasuries is also reversing.Declining marginal demand from the official sector and the new private-sector supply below are two sides of the same problem.
Beyond sovereign debt burdens and inflation anxiety, reports point to a newer factor:AI-related corporate bond issuance is surging, mostly at long durations, are competing with Treasuries for the same pool of long-term money. US companies have issued nearly $1.7 trillionin corporate bonds, up YoY by 27%, a significant portion of it tied to AI infrastructure.
What this means for the dashboard: supply pressure at the long endnot entirely something Treasury can offset. Buybacks can improve Treasuries' own liquidity and maturity structure, but they cannot stop money flowing to the corporate bond market instead. That is one reason the long end fell only a few basis points after the buyback announcement and quickly rebounded.
Placed side by side with China, the US's September 16 rate hike creates a stark contrast:US July CPI YoY 3.4%, China 0.5%；U.S. 3-month Treasury 4.18% (10/29), China 3-month Shibor 1.43%. The two economies' monetary policies are at opposite ends.
China's short-term rates are stuck at 1.4% level, with inflation at only 0.5%, while the US raised the federal funds rate on September 16 to 3.75%–4.00%。The near-285bp short-end rate spread is the underlying driver of the sustained capital flow from RMB assets into USD assets, also providing part of the support for demand at the front end of Treasuries — after the hike takes effect, this gap will widen further.
Methodology: this section isMoney-market short end vs. inflation, not a long-end government bond comparison. The Tushare account used lacks ChinaBond government yield curve (yc_cb) access, so this page does not include China's 10-year / 30-year government bond yields and does not substitute values from other sources. China CPI, Shibor and M2 all come from the official Tushare API.
On Aug 19, Treasury announced that the size of its liquidity-support buybacks of long-end off-the-run Treasuriesdoubled from $2bn per operation to at least $4bn, since Sep 9 onward, running through the end of this refunding quarter (Nov 4). Bessent said the next day that the actual sizecould exceed the $4bn floor, and called the move "Treasury twist operation" (Treasury Twist): buying long bonds and paying for them with short-end issuance.
On Aug 24, two senior Treasury officials told CNBC it may tapthe nearly $1 trillion TGA (Treasury General Account)to fund the buybacks. That day, 10-year and 30-year yields each fell about 4 bp. Treasury's published closing curve confirms the move: the 10Y 4.74% → 4.70%, 30-year 5.27% → 5.23%。
Liquidity-support buybacks targeting long-end off-the-run issues, effective September 9 and covering the refunding quarter through November 4. Long-end yields fell on the announcement day, with the 30-year dropping from 5.28% to 5.19%.
The market then questioned the buyback size and its policy impact; bonds failed to extend their initial gains and yields rose instead. On August 21 the 30-year returned to 5.27%, 10-year 4.74%。
The two officials said TGA funds are "currently available for use" but did not disclose the amount or timing. If this happens, the market's view of Treasury's "ammunition" would change, and it could ease concerns that Treasury will ask the Fed to help buy bonds.
3M–10Y spread from 8/21's +86bp narrowed to +78bp, 2-year–10-year spread +47bp. Treasury has not yet formally explained where the funds would come from.
The reference line is the upper end of the prior administration's $550–600bn target range reported by CNBC. The TGA balance fluctuates with tax receipts and issuance timing rather than accumulating in one direction; the April 22 YTD high of $1,006bn reflects tax-season inflows.
Tallying the fiscal consequences of higher long-end yields (all data are verifiable official figures):
| Item | Amount | of GDP | Source |
|---|---|---|---|
| Total federal debt | $40.07 trillion | 122.6%（Q1） | Treasury Debt to the Penny, 9/9 |
| Federal interest outlays | $1.247 trillion | 3.84% | BEA A091RC1Q027SBEA |
| Defense spending | $1.198 trillion | 3.69% | BEA FDEFX |
| Medicare | $1.334 trillion | 4.11% | BEA W824RC1Q027SBEA |
Interest expense now exceeds defense(1.04x), this point holds and carries a lot of weight. But "interest exceeds defense and MedicareCombined" claim does not hold. Combined, the two $2.53 trillion, interest is only 49%. The often-cited "interest at 3.5% of GDP" is also too low. On a Q2 2026 annualized basis it is actually 3.84%。
Link it to this page's TIPS decomposition, and the nature of the spiral changes.The commonly cited 'debt spiral' implies inflation: more debt → print money → inflation → higher rates. But breakeven inflation has moved only 12bp YTD, with the 5y5y anchored at 2.35% --The market is not pricing in inflating away the debt. The wheel that is actually turning is the other one:RealRates rise → the real interest burden grows → the deficit widens → issuance increases → real rates rise again. The 10-year real yield is 2.90% (9/28, a fresh high since 2015, above the prior high of 2.85% on 9/24 and 2.68% on 9/18), while the weighted-average coupon on outstanding debt sits far below this,Every maturity refinanced pushes the burden higher。
BofA models "each 1pp rise in debt/GDP lifts rates by 1, 2 or 3bp". What this channel is worth can be calculated directly: at $40.07 trillionoutstanding, each 1pp equalsAdditional interest of $4–12bn per year(0.012%–0.037% of GDP); even with a cumulative 10pp rise over a decade, it is only $40.1–120.2bn。
But there is a second channel, of an entirely different magnitude. The Treasury's publishedWeighted average coupon on outstanding debt is only 3.490%(2026-08-31), while current issuance costs are:
| New issue tenor | Current yield | Gap |
|---|---|---|
| 3M | 4.00% | +51 bp |
| 2Y | 4.56% | +107 bp |
| 5Y | 4.75% | +126 bp |
| 10Y | 4.95% | +146 bp |
| 30Y | 5.37% | +188 bp |
By category it's more direct: outstanding Notes coupon 3.345%, new 5-year 4.75%,Gap 141bp; Bonds outstanding 3.453%, new 30-year 5.37%,Gap 192bp. Using a conservative 4.5% blended refinancing cost, the overall gap is about 101bp。
The two channels differ by nearly an order of magnitude.If the average maturity of the stock is 5–7 years, each year roughly $5.7–8.0 trillionrepriced at maturity; once the gap is realizedAdditional interest of $57.9–81bn per year(0.18%–0.25% of GDP) —a single year's volume equals ten years of cumulative volume through the debt-elasticity channel。
Push it to the extreme and it gets clearer: even if total debt didn't grow by a cent and market rates didn't rise at all, simply rolling the existing stock over at the current curve would take interest expense from $1.247 trillionrose to about $1.80 trillion——rising from 3.84% of GDP to 5.6%, an extra 1.7 percentage points, the equivalent of conjuring up another defense budget. This is not a forecast. It is arithmetic already written into the maturity schedule.
The weighted average coupon on outstanding debt is from the US Treasury's Average Interest Rates on U.S. Treasury Securities (2026-08-31): total interest-bearing debt 3.490%, Notes 3.345%, Bonds 3.453%, Bills 3.788%. Total debt is from Debt to the Penny (9/9, $40.074 trillion), interest expense from BEA A091RC1Q027SBEA (Q2 annualized $1.247 trillion), and GDP for the same period is $32.486 trillion. The 5–7 year average maturity is an assumed range, not an official figure, used to indicate magnitude rather than as a precise forecast.
This also explains why BofA thinks debtTotalis not the main driver of the long-end rise. What really moves the bond market isDeficit expectations and issuance pace. Holdings are the slow variable and flows are the fast one. The auction data in Section 10 of this page are a high-frequency reading of that flow.
If funded by issuing bills(what most market participants had assumed): this is purelyTerm structure swap. Swapping short-end supply for long-end buybacks pushes down the long-end term premium but adds short-end supply, and it does not change net Treasury issuance. On 8/25 the short end rose and the 10Y–3M spread narrowed 8bp, consistent with this path.
If the TGA is used: Treasury cash held at the Fed flows to the private sector, equivalent toNet liquidity injection, exerting stronger pressure on the long end without adding bill supply.
Note: the TGA use comes from anonymous sources, and the TreasuryNot yet officially confirmedas the funding source. Officials have not ruled out issuing bills either, and neither the amount drawn nor the announcement timing has been disclosed. The doubling of buyback size was formally announced by the Treasury and is confirmed.
At the Jul 28–29 meeting, the Fed held the federal funds target range at 3.50%–3.75% unchanged, butThe vote shifted from a unanimous 12–0 in June to 9–3: Cleveland Fed's Beth Hammack, Minneapolis Fed's Neel Kashkari and Dallas Fed's Lorie Logan argued forImmediate 25bp hike. Minutes released Aug 19 showed that "many participants judged that tightening policy may be necessary if inflation does not decline."
After the votes turned hawkish, two data releases briefly bent the direction back, until August 28, when Warsh himself bent it a third time.
The statement reiterated that "inflation remains somewhat elevated relative to the 2% objective, partly reflecting supply shocks including energy," and cited the Middle East conflict as a source of uncertainty. Chair Warsh also floated cutting the number of meetings per year from 8 to 6.
Consensus +83K, actual −23K, government −53K; prior two months revised down by a combined 103K, with the three-month average falling to +20K. Unemployment fell to 4.1%, but only because the labor force shrank by another 264K and participation dropped to 61.4% (lowest since Feb 2021). Average hourly earnings growth slowed to 3.2% YoY.
MoM only +0.1%; YoY continued to ease from June's 3.5%. The energy-driven price shock since the start of the year is fading, but the Middle East situation keeps prices volatile.
CME FedWatch shows the probability of a hold at the Sep 16 meeting rising to 68.4%, implying hike odds of about 32%. On jobs report day, the probability of a hold had already jumped from 45% to 60%, whereas a week earlier the market gave a hold only a one-in-three chance.
Headline PCE YoY 3.70%, MoM +0.2%, 0.1 percentage point above expectations; core PCE YoY 3.34%，exactly flat with June, also above the 3.2% expectation. CPI is falling, but the gauge the Fed actually watches is not. These were the very numbers Warsh cited in his speech two days later.
far exceeding about 53Kexpectations, with unemployment holding at 4.1%; June and July combined revised up by 55K，July revised from −23K to +21K, with the three-month average moving from +20K back to +71K. The hike probability then rebounded from about 52% to 59–63%. The bond market reaction was entirely at the front end: 2Y +3bp, 10Y +1bp, 30Y −1bp, and the 2Y–10Y spread narrowed to +41bp.
The Sep 3 ADP employment report was weaker than expected, and the 2Y yield fell 5bp to 4.34%, the 10-year fell only 2bp, and the 2-year–10-year spread widened to +43bp. The same day, the Freddie Mac 30-year fixed mortgage rate was 6.71%, at highest since Jul 31, 2025. The long-end rise has fully passed through to housing finance costs.August payrolls due Sep 4, the last major jobs report before the September 16 meeting.
CME FedWatch puts the probability of a 25bp hike on Sep 16 at about 68%, nearly double the ~35% a week before the speech. The same day, US forces struck Iranian targets around the Strait of Hormuz, and Brent closed at $96.02, +9.4% over four trading days. The 10-year closed at 4.79%, the highest since January 2025.
The September hike probability went from about 35% jumped to about 56%(quotes range 55.7%–60%), with CNBC calling the September meeting "a coin flip". The bond market reaction concentrated at the short end: the 2-year–10-year spread narrowed in a single day from +47bp to +39bp, 3-month–10-year spread flat at +83bp. Stocks fell, S&P −0.3%, Nasdaq −0.5%.
The federal funds rate target range was raised to 3.75%–4.00%, the first hike since 2023; the three hawkish dissenters (Hammack / Kashkari / Logan) who pushed for an immediate hike in July ultimately won over the full committee, with the vote shifting from 9–3 on 7/29 to12–0 unanimous vote. The statement said "inflation remains elevated"; the dot plot showed 16 of 18 participants expecting one more hike this year, with a year-end rate projection range of 4.1%–4.4%.
Path: mid-July ~25% priced for the July meeting → 8/20 ~32% for the September meeting → 8/27 ~35% → ~56% after the 8/28 speech → 9/1 ~68% → 9/4 ~58% → 9/10: ~100%. A week before the speech, that probability was only about 35%. The September 16 FOMC delivered as expected, hiking 25bp to 3.75%–4.00% (12–0 unanimous).
On Aug 28, Warsh gave his first Jackson Hole keynote as Chair. He did not commit to any specific rate action, promising only "discipline, not a particular decision," but he made the bar clear:"We must be confident that underlying inflation is moving clearly, and at a sufficient pace, toward our target. Otherwise, we have work to do."
The figure he cited was PCE YoY 3.7%, six-month annualized 4.1%, and over the past six months 49% of PCE components rising faster than 3% annualized; the milder summer inflation readings "did not tell him the underlying trend has materially improved". These figures were confirmed just two days earlier by the July PCE report —Core PCE 3.34%, not a single basis point changed from June(re-checked against the FRED series). He also said responsibility for 65 months of persistently elevated inflation lies "entirely with the central bank".
On employment, his reading is the opposite of this dashboard's earlier narrative.unemployment at 4.1% and initial jobless claims near multi-decade lows, he attributed weak job growth toVery little labor supply growthrather than weakening demand, concluding that current conditions are "consistent with full employment".Data a week later backed his side: August nonfarm payrolls, released September 4, came in at +162K, with June and July revised up by a combined 55K and July flipping from negative to positive. The case against hiking on "weak jobs" grounds has essentially collapsed.
He also drew a contrast with two predecessors, describing theKohn era's hikes as a "steely marathon death march," and the Bernanke era's as "a much more leisurely pace"; i.e., Kohn = relentless back-to-back hikes, Bernanke = a relaxed pace. (Chinese-language paraphrases often swap the two paths and describe Kohn as "waiting for inflation to blow up, then hiking hard," which does not match the original remarks.) BofA's September read: Warsh's emphasis on trends over single data points puts the onus back on himself; unless August jobs and inflation data are very weak, not hiking in September would erode the credibility he earned with the speech. He also signaled a shift in communication: he opposed forward guidance, saying it creates a "hall of mirrors problem," called for "a quieter Fed, communicating more purposefully, better able to achieve its goals," and said the Fed should not indulge a setup in which "market participants mainly watch the Fed for their next trade."This means fewer policy signals before upcoming meetings, so price discovery relies more on the data itself.
Quotes are paraphrased from the speech text on the Federal Reserve website; PCE figures and component shares are cited from the original speech.
Policy weight of this dataUnusually high, because of a scheduling fact:August PCE is not released until Sep 30. That is two weeks after the Sep 16 meeting, and the same day BEA starts its 2026 annual update (national, industry and regional data begin on the same day for the first time, including historical NIPA revisions).FOMC members won't have the official August PCE in hand on voting day, and can only be backed out from CPI and PPI. Yet according to Citi, the usual CPI→PCE mapping is unreliable this month precisely because of a methodology change.
This has two consequences: first, Waller's reported threshold of "core PCE MoM >0.30% to support a hike" is an invisible number on voting day. The Cleveland Fed's 9/4 nowcast gives 0.28%, just 2bp from the threshold, and this is only an estimate; second, the murkier the data,the more weight the Chair's ability to rally votes carries。
| Reading | Implication | Market path |
|---|---|---|
| ≤ 0.20% | Pause probability rising | 2Y falls and the curve turns to bull flattening. Growth stocks get a breather, but long bonds still need a separate term-premium read |
| 0.21 – 0.29% | Coin flip; most likely outcome | Watch PPI and components; either rally votes for a hike or deliver a "hawkish hold" signaling Oct/Nov; two-way volatility in stocks and bonds |
| ≥ 0.30% | Hike probability jumps sharply | 2Y spikes first and the curve shifts from bear steepening to bear flattening. Richly valued assets come under pressure |
| ≥ 0.30% and 10Y/30Y break out at the same time | The real red light | Short end trading the Fed again, long end still trading fiscal risk and term premium: hit from both ends |
relayed secondhand as "BofA forecasts core CPI MoM at 0.22%, annualized rate will rise to 3.4%" cannot be true alongside this, since the two numbers conflict. The YoY change roughly equals "this month's MoM minus the same month's MoM last year": for core CPI YoY to go from July's 2.5% to jump to 3.4% within a month (+0.9pp), last August's MoM would have to be −0.68%. A single-month core CPI drop of 0.68% has almost never happened.
More direct evidence is the two forecasts themselves:0.22% and 0.184% both round to the same number: 0.2%, and since 11 of the 12 months in the YoY figure are known history, with only 1 month set by this print, the two firms' annual-rate disagreement is arithmeticallyshould not exceed 0.05 percentage point, and cannot possibly be 110bp. Comparing with the Cleveland Fed's 9/4 nowcast puts each in its place:Citi's 2.3% is the core CPI annual rate(forecast 2.32%),BofA's 3.4% is the core PCE annual rate(forecast 3.49%). The so-called "opposite bank playbooks" dresses up a technical definitional difference as a showdown —The real opposition isn't between the two banks; it's between the CPI and PCE gauges themselves。
Section 1 of this page already put this contradiction side by side: CPI YoY at 3.4% labeled "peaking and falling," core PCE at 3.34% labeled "stuck in place." The mechanism is weighting:Shelter is about one-third of the CPI basket, but far less in PCE, cooling housing was the main driver of lower core CPI over the past year; conversely,PCE puts a higher weight on healthcare and covers Medicare, Medicaid and employer-paid portions, while CPI counts only consumers' out-of-pocket spending, so rising medical costs push up core PCE more. Warsh has confirmed PCE is the Fed's preferred gauge. So September will be decided by the line that did not come down.
Hike → short end catches up → bear steepening ends and turns into bear flattening;No hike → the market decides the Fed is behind the curve on services inflation → term premium rises further → bear steepening intensifies. Both paths mean losses for the long end, just in different places. The breakdown in Section 3 of this page measures exactly this: term premium 0.875% is already the highest since 2015; how much more upside it has determines the cost of the "no hike" path.
Policy rate 3.50%–3.75%(raised to 3.75%–4.00% on 9/16), core PCE 3.34%(July actual) to 3.49%(September forecast) —The real policy rate has narrowed to just +0.01 to +0.29 percentage points (rising to roughly +0.26 to +0.66 percentage points after the hike took effect). The hawkish argument is not "inflation is out of control" but"Restrictiveness may have already disappeared"; the so-called path to 5% reflects rebuilding restrictiveness, not panic pricing. This also explains why strong employment actually supported the hike — with no recession risk to offset the real rate falling to zero, it was ultimately delivered on September 16 in a 12–0 unanimous vote.
① Core CPI MoMUnroundedvalue is ≤0.20%; ② Shelter component(determines whether the CPI–PCE gap widens further); ③ Medical services component(determines the direction of the PCE mapping); ④ Core services ex-shelterMoM (the supercore measure); ⑤ energy's contribution to the headline reading. Both expect an energy rebound to push headline CPI MoM to 0.34%–0.37%, if the headline number runs hot while core is only 0.2%,Warsh's framing at the press conference will matter as much as the decision itself. Current VIX 16.07(9/28 close, still in the lower end of the YTD range of 14.2-31.1); in a low-volatility environment, the market impact of a communication misstep is more elastic.
Verification notes: core CPI MoM and YoY from FRED series CPILFESL (July MoM +0.22%), average hourly earnings from CES0500000003 (August YoY 3.09%, consistent with "3.1%"), VIX from VIXCLS; Cleveland Fed nowcasts from its website reading of September 4 (core CPI YoY 2.32%, core PCE YoY 3.49%, core PCE MoM 0.28%); the BEA release schedule and annual update arrangements have been verified. The claim that "supercore PCE has risen to 3.9%" could not be independently verified on this page; note that the Cleveland Fed'sOverall PCE YoY forecast is exactly 3.91%. The two figures are identical, so the numbers may have been mixed up in reporting. Separately, Citi says the 9/30 annual update may revise the core PCE annual rate down by more than 30bp to about 3.0%. A September hike followed by a large downward revision to the baseline at month-end would be an awkward combination, and that is the strongest argument for "waiting one more month."
The credit market not only failed to react to July's jobs shock but kept tightening: OAS went from July 17's 273bp narrowed steadily to the August 27 level of 263bp，has matched the YTD low set on June 17; the YTD high was on March 30 at 346bp。
The buffer is at its thinnest.With employment turning negative, corporate bond issuance up 27% YoY, and the FOMC having raised rates 25bp to 3.75%-4.00% on September 16 — with all three holding true at once, high-yield spreads have widened from their tightest levels of the year (326bp on 10/29, 66bp above the 260bp low on 8/28, 12bp wider than 314bp on 10/28), and the cushion for further credit deterioration remains thin. This is the only risk item that has kept worsening this period.
Rather than betting on a "countdown," watchFalsifiable signals. The table below shows current readings and alert thresholds for seven indicators. Those with a data source are connected and update daily. Those without one are labeled as such and not filled in with estimates.
| Indicator | Current | Alert line | Status |
|---|---|---|---|
| Auction bid-to-cover (full curve) | Aug 2.39–2.71 | consistently below 2.30 | Normal · all tenors in line with YTD means; total demand hasn't fallen |
| Auction indirect award (full curve) | August average 55.1% | consistently below 50% | Watch · July was 62.8%; 6 of 7 tenors below their YTD average |
| MOVE Index | 100.21 | Sustained above 100 | Broke above · on 9/24 jumped +9.6% in a day to break above 100 at 104.58, the first time in three months, rising about 29.7% at one point that week; on 9/25 it fell back to 96.00 (-8.2%), dropping back below 100; on 9/28 it rose again to 101.82 (+6.06%), breaking back above 100; on 9/29 it extended its rise to 106.60 (+4.70%); on 9/30 it rose further to 110.45 (+3.61%); on 10/1 it fell back to 108.13 (-2.11%); on 10/2 it continued down to 107.29 (-0.78%); on 10/5 it rebounded to 113.60 (+5.88%); on 10/6 it fell back to 105.20 (-7.39%); on 10/7 it continued down to 102.56 (-2.51%); on 10/8 it continued down to 100.70 (-1.81%); on 10/9 it continued down to 98.47 (-2.21%), dropping back below 100; on 10/13 it turned up to 101.09 (+2.66%), breaking back above 100; on 10/14 it fell back to 99.06 (-2.01%), dropping below 100 again (Yahoo Finance); on 10/16 it rebounded to 101.1 (-1.4% on the day, still up about 21% for the month, TradingView MOVE), breaking above 100 again; on 10/19 it continued up to 101.3 (+0.2% on the day, Yahoo Finance ^MOVE); on 10/20 it continued up to 101.7 (+0.39% on the day, +3.28% over the past week, +33.44% over the past month, TradingView MOVE); on 10/21 it fell back to 95.94 (-5.67% on the day, Yahoo Finance ^MOVE), dropping back below 100; on 10/22 it rebounded sharply to 106.85 (+11.36% on the day, TradingView MOVE), breaking above 100 again; on 10/23 it fell back to 103.31 (-3.30% on the day, Yahoo Finance ^MOVE), still above 100; on 10/26 it continued down to 102.19 (about -1.08%, CNBC .MOVE), still above 100; on 10/27 it fell back to 97.87 (about -4.23%, Yahoo Finance ^MOVE), dropping back below 100; on 10/29 it continued down to 92.97 (-1.68% on the day, -9.70% over the week, -9.46% over the month, TradingView MOVE), remaining below 100; on 10/30 it rebounded to 100.21 (+7.12% on the day, -4.72% over the week, -4.24% over the month, TradingView MOVE), breaking back above 100 |
| High-yield OAS | 326 bp | 500 bp | Far from trigger · 174bp from the warning line, 66bp from the YTD tightest (260bp on 8/28), 12bp wider than 314bp on 10/28 (FRED, 10/29 value) |
| BDC discount rate (private credit's shadow price) | Median -10.7% | Median discount ≥ 20% | Watch · All 7 are measured against 2026-06-30 quarterly NAV (verified against original SEC 8-K filings); PSEC -71.8%, FSK -39.7%, and OBDC -27.1% have crossed the line, but the median has not yet; the basket's median drawdown from its 52-week high is 18.0% (Yahoo, 10/30 close, updated daily) |
| CCC − BB tiering spread | 1,131 bp | ≥ 1,200 bp | Watch · CCC 1,325bp / BB 194bp (FRED, 10/29, updated daily); same-day HY composite at 326bp — the tiering gap is 3.5x the composite spread |
| Private credit redemption/gate news flow | 30 days 2 gating incidents | ≥2 semi-liquid fund redemption restrictions in the same month | ⚠ has been crossed · September alone already saw 6 pro-rata gatings, the most this year: Blackstone Private Credit Fund (BCRED)'s 9/30 Class I/S/D share NAV fell to $23.51/share, continuing its slide from $23.60 at 8/31 and $23.65 at 6/30; Blue Owl Credit Income Corp (OCIC) received $3.1bn in Q3 redemption requests (as a share of 16.8%, down further from 18.8% in Q2), maintaining the 5% cap, repurchasing about USD 900M, with about 30% of requests fulfilled (10/2); Blue Owl Technology Income Corp (OTIC)'s Q3 redemption requests were USD 1.1bn (share of units 39%, roughly flat with Q2's 38.1%), kept the 5% cap, repurchased about $135M, and fulfilled about 13% of requests (10/2); KKR FS Income Trust Select (K-FITS)'s Q3 (through 9/29) redemption requests came to 5.06% of shares, slightly above the 5% quarterly cap (versus 3.43% last quarter); management invoked its existing authority and the 2% flexibility under Rule 13e-4(f)(1) to fulfill 100% of all requests rather than prorating, with net assets of about $996M (10/5). Separately, on 10/5 third-party Cox Capital launched a secondary market tender offer for Blue Owl OCIC Class I shares at $7.31/share, a discount to the 8/31 NAV of $9.14 of 20%, with an initial size of $10M (expandable to about $206M), through 2026-11-25. Over 90 days, total 16 incidents; 30-day total of 5 incidents (manager announcements and regulatory filings, verified weekly, each with a link) |
| St. Louis Fed Financial Stress Index | −0.58 | Persistently above 0 | Normal · 10/23 weekly value (FRED STLFSI4, updated weekly); a negative value means stress is below the historical norm |
| SOFR − IORB | +4 bp | Sustained above +10 bp | Normal · 10/29 value (SOFR 3.94% - IORB 3.90%, NY Fed / sofrrate.com), YTD range -15...+10bp |
| September FOMC | Hiked 25bp to 3.75%–4.00% | The event itself | September 16 · 12–0 unanimous, delivered |
| Brent crude | $113.43 | — | 10/30 close, up 2.4% from $110.80 on 10/29 (Investing.com) · still above the 9/9 high of $109.51; still below the YTD high of $138.21 on 4/7; intraday on 9/21 it extended its slide to about $97.72, before turning higher |
The Treasury's announced schedule squeezes this page's two monitoring lines into the same week:
| Date | Event | Corresponding indicator on this page |
|---|---|---|
| Sep 8 | 3Y auction, $58bn(constant tier all year) | Bid-to-cover, indirect award, concession |
| Sep 9 | 10Y auction, $39bn(small tier; $42bn in August) | Same as above; August was 2.53 / 61.0% |
| Sep 10 | 30Y auction, $22bn(small tier; $25bn in August) | Same as above; August was 2.39 / 53.1% |
| Sep 11 | August CPI | Four core MoM red lines |
| Sep 15–16 | FOMC | Hiked 25bp to 3.75%–4.00% (12–0) |
The sequence is worth noting:All three auctions come before CPI. If August's across-the-board drop in indirect award was one-month noise, this week's three auctions should come back near their YTD averages (3Y 54.5%, 10Y 62.2%, 30Y 57.6%). If they come in below average for a second straight month, only then is the "real demand deterioration" criterion triggered, and it would happen before the inflation print is outbeforeto see it.
Another implication is absorption risk:The Sep 10 30Y auction has to price the day before CPI is released, bidders must absorb $22bn of the longest-duration supply without knowing the inflation print. Such "pre-data long-end auctions" have historically been where tails most easily blow out.
One more fact makes this week's test cleaner: all three auctions are in the year's smallest size tier.Checking all 2026 auction announcements, the 10-year was in $39bn / $42bnalternating between two sizes, with the 30Y at $22bn / $25bnback and forth; the pattern is Larger sizes in the Feb, May and Aug refunding months, smaller sizes in other months; the 3-year stays constant at $58bn all year. This week's three auctionsAll small tranches(10Y $39bn, 30Y $22bn, both about $1.1bn below the YTD average).
This forms a natural control variable:If demand indicators still weaken in the lightest supply month, it can't be blamed on a "supply surge"; it can only be demand itself, so the concurrent drop in indirect allotment in August looks more like trend than noise. Conversely, if this week's three auctions improve, that should be discounted too: small-tier months are easier to absorb and are not enough to overturn August's signal outright.
Auctions: total demand hasn't fallen, but its composition is shifting.Attributing all of August's 7 auctionsLooking at coupon auctions together (including the three month-end auctions: 2Y on 8/25, 5Y on 8/26, 7Y on 8/27), the picture is clear:Indirect bidder allotment averaged 55.1% in August vs. 62.8% in July, of the 7 tenors, 6 below their respective YTD averages, with the largest drops in the 20-year (−5.5pp), 30-year (−4.5pp) and 7-year (−4.1pp); only the 2-year bucked the trend, 5.5pp higher.
butBid-to-cover barely moved: 2Y 2.60 (avg 2.63), 5Y 2.37 (2.33), 7Y 2.50 (2.49), 10Y 2.53 (2.49), 30Y 2.39 (2.42). Total bids have not shrunk. What changed isWho is buying: demand shifted from indirect bidders (the main channel for foreign and official institutions) to domestic direct bidders and primary dealers. This is consistent with the official sector reducing Treasury holdings, but by the "only sustained deterioration counts" standard,For now it's a one-month structural shift, not a trend. The YTD low for indirect award was 49.1% in February, and the lowest August auction came in at 50.7%.
Incidentally, auction indirect allotment is better suited than TIC holdings to answer whether it is "genuine selling or valuation effects": the former isFlow, available on each auction day; the latter isOutstanding, including valuation changes from FX and prices, and lagging by about two months.
Funding markets: no sign of drying up.SOFR minus IORB is +4bp(10/29), within the year's range at −15 to +10bp range. Genuine liquidity stress would show up as SOFR persistently above the reserve rate; there is none of that now.
Oil: the last version called it a "disinflationary factor"; that flipped within a week.Brent on August 26 touched $87.77, closing September 1 at $96.02——9.4% rebound in four trading days. The trigger was US strikes that day on Iranian targets around the Strait of Hormuz (after two tankers were attacked), followed by harsh rhetoric from both sides.
But the time scale needs to be clear: the YTD high was April 7's $138.21, this year's inflation shock came mainly from January $61.08 doubled by the April round; as of the 9/24 close, $106.39 still below the peak by 23%, rebounding 8.9% from the intraday low of about $97.72 on 9/21. So the current state is —Oil price direction has been choppy, driven entirely by geopolitical events, and its YoY contribution could widen again at any time. This is now the most volatile indicator on this page and the one least suited to extrapolation.
Credit and event risk:High-yield OAS at 326bp (10/29), 174bp from the 500bp crisis-pricing line — 12bp wider than 314bp on 10/28, still near the YTD tightest range; the cushion is thin, not thick.Sep 16's FOMC raised rates 25bp to 3.75%-4.00% as expected (12-0 unanimous);October 28's FOMC held that range steady, with a 10-2 split vote; the hawkish split persists.
The MOVE index is a proprietary ICE index not covered by FRED, and this page has no continuous feed access; the figures in the table are single-point check values from Yahoo Finance (^MOVE) (9/24 closed at 104.58, +9.6% on the day, the first break above 100 in three months; 9/25 closed at 96.00, -8.2%, falling back below 100; 9/28 closed at 101.82, +6.06%, breaking back above 100; 9/29 closed at 106.60, +4.70%; 9/30 closed at 110.45, +3.61%; 10/1 closed at 108.13, -2.11%; 10/2 closed at 107.29, -0.78%; 10/5 closed at 113.60, +5.88%; 10/6 closed at 105.20, -7.39%; 10/7 closed at 102.56, -2.51%; 10/8 closed at 100.70, -1.81%; 10/9 closed at 98.47, -2.21%; 10/13 closed at 101.09, +2.66%; 10/14 closed at 99.06, -2.01%), 10/16 closed at 101.1 (-1.4% on the day, TradingView MOVE); 10/19 closed at 101.3 (+0.2% on the day, Yahoo Finance ^MOVE); 10/20 closed at 101.7 (+0.39% on the day, TradingView MOVE); 10/23 closed at 103.31 (-3.30% on the day, Yahoo Finance ^MOVE); 10/26 closed at 102.19 (about -1.08%, CNBC .MOVE); 10/27 closed at 97.87 (about -4.23%, Yahoo Finance ^MOVE); 10/29 closed at 92.97 (-1.68% on the day, -9.70% over the week, -9.46% over the month, TradingView MOVE); 10/30 closed at 100.21 (+7.12% on the day, TradingView MOVE), not updated daily. Auction data is sourced from the U.S. Treasury's TreasuryDirect auction-results API; the indirect-award ratio = indirect bidder awarded amount / total awarded amount.
In Moody's Credit Outlook published August 31, the macro section's baseline for the Fed ishold steady, with modest cuts in 2027, while at the same time the market put September hike odds at about 68%. This is not a subtle disagreement. The two point in opposite directions, and they are worth comparing point by point.
| Topic | Moody's (originally 8/27) | This page's data / market (as of 9/2) |
|---|---|---|
| September FOMC | Hold steady, modest cuts in 2027 | Hike probability about 68% (CME FedWatch, 9/1) |
| Reversal conditions | "Persistently elevated gasoline and diesel pricescould swing enough FOMC votes toward a hike" | Brent closed at $96.02 on 9/1, +9.4% in four trading days |
| Long-end yields | "The jump in government bond yields is a key development, and the resulting higher borrowing costs pose a new risk to growth" | 10Y at 4.79%, highest since Jan 2025 |
| Other central banks | ECB on hold; Bank of JapanAcceleratingPolicy normalization | Japan 10-year 2.88%, highest since September 1996 |
| G-20 growth | 2.5% in 2026 and 2.6% in 2027, largely unchanged from the May outlook | — |
| US economy | Supported by consumption and AI investment, despite slowing job momentum and sticky inflation | August payrolls +162K (9/4), core PCE flat at 3.34% |
The macro article's labeled original publication date is Aug 27, compared with Warsh's Jackson Hole speech (8/28, hike odds jumping from ~35% to ~56%) by one day, and ahead of the military escalation around the Strait of Hormuz (9/1, Brent +9.4% in four days) by three trading days.
In other words,The "reversal condition" Moody's itself wrote down, persistently high fuel prices, began materializing just two days after publication. So this is not a case of "Moody's got it wrong" but of a baseline finalized before the catalyst. Its real value is that it spells out the conditions:Whether the energy shock is transitory or persistent decides a hold or a hike in September. This test is easier to check than any directional call, and the Brent row in this page's monitoring panel gives its live reading.
One more point: the "slowing job momentum" pillar of Moody's baselinewas also undermined by the September 4 nonfarm payroll revisions: August +162K, and July flipped from negative to positive. Both legs, energy and jobs, are now tilting toward a hike.
The credit consequences of AI capex.Moody's expects over the next 12–24 months, free cash flow at US and Chinese hyperscalers will weaken, debt and lease obligations will rise, and financial leverage will climb, with only existing cash balances as a buffer. Section 4 of this page, fromSupply sidemade the same point: AI-related long-duration corporate bonds compete with Treasuries for long-term money. Rating agencies, meanwhile, fromIssuer creditperspective reached a directionally consistent conclusion.
Risk is relocating.Another article notes that private credit has accumulated a record $56bnof dry powder, while banks absorb these risks through warehouse financing and insurers through rated-note structures, "adding a layer ofOpacity". Compare it with this page's high-yield OAS 326bp(10/29, 66bp from the YTD tightest) taken together:What warrants caution is not just how thin spreads are, but where the risk has moved: public-market spreads may no longer fully reflect how credit risk is distributed.
The above summarizes and paraphrases key points from the macro and Credit in Depth sections of Moody's Credit Outlook of August 31, 2026; text in quotes is translated directly from the original, and the original publication dates in parentheses are taken from each article's label. Moody's views do not represent this page's position; they are listed to allow a testable comparison with market pricing.
August payrolls +162K, July revised up to +21K, three-month average back at +71K, unemployment 4.1%. The revisions overturned the earlier "weakening demand" reading, and Warsh's view that jobs are "consistent with maximum employment" won out.
OAS at 326bp (10/29), 66bp above the YTD low (260bp on 8/28), 12bp wider than 314bp on 10/28. With employment turning negative and corporate bond supply surging while spreads are only just beginning to widen from their YTD lows, credit markets' earlier underpricing of weakening fundamentals is starting to correct.
UK, Japanese, German and French 30-year yields hit multi-year highs in tandem, while AI corporate bonds divert long-term money. Buybacks can hedge Treasuries' own term structure, but not the diversion of funds.
On 9/1 US forces struck Iranian targets, and Brent rebounded 9.4% in four days to $96, rising further on 9/9 to $109.51. Energy is the main driver of this year's inflation shock, and its resurgence directly narrows the Fed's room to hold off on hiking.
After Warsh's speech on 8/28, the probability of a September hike rose from about 35% to about 56%, reaching about 100% by 9/10;The September 16 FOMC formally hiked 25bp to 3.75%–4.00%, with the vote shifting from 9–3 in July to a 12–0 unanimous vote—the three hawkish dissenters who pushed for an immediate hike in July ultimately won over the full committee. He had earlier explicitly judged employment to be "consistent with full employment," removing the case for holding steady on the grounds of labor-market weakness.
This entire page is aboutYields, spreads and curve shape, but holding returns are not the same as yield changes. Below are the four most common pitfalls when converting the former into the latter, each tested with this page's data.
1. Curve-shape labels do not tell you the direction of P&L."Bear flattening is good for long bonds" is wrong: short end +20bp and long end +5bp is also bear flattening, and long bonds still fall. Measured over 2026, the 38 bear-flattening sessions: on 29 of them (76%), 30-year prices fell, daily average −0.33%, cumulative −12.67%; on the worst day, March 11, 2-year +7bp, 10-year +6bp, 30Y +8bp, long-end yields actually rose the most, with a single-day price change of −1.24%. Likewise, "the 30-year yield not rising after a hike" can be consistent with "improved credibility", but could equally stem from growth worries or safe-haven buying. Performance alone cannot establish causality.
2. Convexity units must be standardized first.Based on a 30-year bond with a 5% coupon and 5% yield to maturity, paying semiannually:Modified duration 15.45 years, convexity C = 352.1. At +50bp in yield, the duration term is −7.73% and the convexity term +0.44%, for a total of −7.29% (exact repricing −7.31%). If convexity is wrongly taken as 3, the second-order term shrinks to just 0.004%. If you get 0.04% somewhere, the convexity implied by that result is about 32, 11 times smaller than the true value, a classic case of inconsistent quote scaling. Conclusion:At 50bp, convexity is already worth 0.44%. There is no need to wait for 150–200bp to account for it。
| Δy | Duration term | Convexity term | Total | Exact | Error from duration alone |
|---|---|---|---|---|---|
| +10 bp | −1.545% | +0.018% | −1.528% | −1.528% | +0.02% |
| +50 bp | −7.727% | +0.440% | −7.287% | −7.306% | +0.42% |
| +100 bp | −15.454% | +1.760% | −13.694% | −13.838% | +1.62% |
| +200 bp | −30.909% | +7.042% | −23.867% | −24.945% | +5.96% |
3. A "carry cushion," not a "coupon shield."Dividing the one-year holding return by duration givesBreakeven yield change: duration 15 years, carry 5.2% → about 35bp; duration 2 years, carry 4.5% → about 225bp. The division itself is right, butThe numerator must be a reasonable estimate of the one-year holding return, not the coupon rate or a fund's distribution rate. Discounts and premiums, fees, curve roll-down and duration changes over the holding period all alter the result. And 225bp is already a large move, so the "tipping point" from linear extrapolation with a fixed duration is only a rough guide, not a hard loss threshold. Term premium is also risk compensation built into the yield, not the same thing as coupon.
4. For CNY returns, watch USD/CNY, not the dollar index.The conversion for unhedged holdings is 1+RCNY ＝ (1+RUSD bonds) × end USDCNY / start USDCNY. This is not a formality: 2026 YTD,USDCNH fell from 6.9681 to 6.7074 (−3.74%, CNY appreciation), whilethe broad dollar index fell only 0.72%. Over the same period, the correlation of daily changes between the twoNear zero(133 overlapping trading days). Using the dollar index to infer CNY returns means proxying with a variable nearly uncorrelated with the target. If the fund hedges FX, hedging costs must be counted separately.
Duration and convexity are analytical calculations for a hypothetical bond (discounted cash flows, semiannual coupons), using ΔP/P ≈ −D·Δy + ½·C·(Δy)², with yields entered as decimals; the "Exact" column re-discounts all cash flows at the new yield. A bear-flattening day is defined as 2-year and 10-year yields both rising, with the 2-year rising more; long-bond price changes are estimated linearly using a modified duration of 15.45. FX data: USDCNH from Tushare fx_daily; broad dollar index from FRED DTWEXBGS (latest 8/28). The CNY is a managed float, so the correlation conclusions above cannot be extrapolated to other currencies.
The previous version said "hike risk is skewed lower."After Aug 28, this statement has to be withdrawn—with a single speech, Warsh pushed the odds of a September hike from roughly 30% back up to near even, and then to about 100% ahead of the meeting — a reversal driven by the central bank itself, not by new data.The September 16 FOMC delivered on this pivot: hiked 25bp to 3.75%–4.00%, with the vote shifting from 9–3 in July to a 12–0 unanimous vote.
One point to add: the "inflation has peaked" narrative the market reads is based on CPI (3.5%→3.4%), whereas the Fed reads core PCE, which in June and JulyUnmoved, stopping at 3.34%. The two gauges are not telling the same story, and that is the root of the disagreement.
The current mix is three directions each going their own way: the short end has already priced in a rate hike (25bp on September 16 to 3.75%-4.00%); the long end is supported by Treasury buybacks yet still weighed down by global long-end and AI corporate-bond supply; and credit spreads have widened from their tightest point of the year (326bp, versus a low of 260bp, widening 12bp from 314bp on 10/28).This is not a market you can express with a single duration view.On 10/22 a bout of safe-haven buying appeared: the 10-year yield broke below 4.95%, down about 17bp from 5.13% on 10/21, the largest single-day drop since March 2023 (Benzinga, 10/22); the same day the MOVE index jumped 11.36% to 106.85, with the driver shifting from inflation expectations to stock-market worries over whether AI capex can be sustained — the 10/28 FOMC decision landed: the rate target range held steady at 3.75%-4.00%, with a 10-2 split vote (officials Cook and Hammack favored a 25bp hike); after the decision the 2-year-10-year spread widened 9bp in a single day to 38bp, interrupting the curve-flattening trend. Whether this driver can persist got a partial answer on 10/29: the Q3 GDP advance estimate came in at 3.4% (above the expected 3.2%, prior 2.2%), core PCE MoM annualized fell to 2.5% (prior 3.30%), with the growth-inflation mix overall leaning toward a soft landing (Tushare eco_cal, 10/29); the 10/29 7-year Treasury auction delivered its result: a winning yield of 4.894%, down sharply 19.1bp from September's 5.085%, lowering issuance cost (Tushare eco_cal); still to watch: the 11/2 ISM Manufacturing PMI and the 11/6 nonfarm payrolls.
Allocation implications adjust accordingly:Duration's value for money has worsened, not improved, since 8/28; the middle of the curve (3–7 years) was also dragged down in the hawkish shock, just less than the short end. The long end's odds still hinge on two unresolved things —Whether buyback funding comes from the TGA, andWhether global long-end and corporate bond supply keep diverting long-term money. On credit, the original view stands and is strengthened:Spreads too tight; no moving down in credit quality。