Ninety Days From Today

Count it out on a calendar. Ninety days from this morning is January 30, a Saturday, which means this window closes on the last weekend of the first month of a new year. Between here and there sit two Federal Reserve meetings, on December 8 and 9 and again on January 26 and 27; a midterm election on November 3, two days from now; a Treasury quarterly refunding on November 4; an October employment report on November 6; October inflation on November 10; Nvidia’s quarter on November 17; a Thanksgiving on November 26; November employment on December 4; November inflation on December 10; a government funding deadline on December 11; a Christmas on December 25; a new Congress seated on January 3; and the December employment and inflation reports in the second week of January.

Last month this series told you the ten-year would land somewhere between 5.2% and 5.8% by December 30. It closed October at 4.94% (U.S. Treasury daily par yield curve). That is twenty-six basis points below the bottom of the range, with two months still to run, and it ended a seven-month streak of rising yields. The call was wrong. The way it was wrong is more useful than the call was.

The mistake was not the direction of the war. The war escalated exactly as the bear case assumed. October was the worst month for shipping in the Strait of Hormuz since the conflict began, with at least thirty-five attacks on vessels against roughly thirty in March, and Brent settled October 30 at $117.64, its highest since late April, up 13.6% on the month (UKMTO via Middle East Eye; CNBC). Inflation escalated too: September CPI came in at 3.7% headline and 2.6% core, both hotter than forecast, and September PCE at 3.6% and 3.1% (BLS via CNBC; BEA).

The mistake was assuming the bond market would keep pricing the war. In October it stopped, and it stopped because of something the war did not cause.

On October 2 the September employment report came in at 29,000 jobs against a consensus near 84,000, with unemployment at 4.2% and average hourly earnings up 3.0% on the year (CNBC; The Washington Post). That morning the front end rallied and the long end did not, which was the correct read: a soft labor market argues for a lower funds rate, and it argues for nothing at all about a thirty-year bond issued into a 3.6% inflation rate. The rally that mattered came three weeks later, on October 22, when an AI-led equity rout sent the ten-year down seventeen basis points in a session, its sharpest one-day fall since the regional bank failures of March 2023 (CNBC; MarketWatch).

That is the paradox branch this series has been describing since the spring. It is no longer a description. It fired, and it fired from a direction the previous installment did not name precisely enough. I wrote that the flight to quality could not arrive because the shock was a price event into which the Fed would not cut. The shock that arrived was not a price event. It was an equity valuation event, and the Fed has nothing to say about those (AP).

Where the Market Stands Today

The numbers as of the October 30 close, with the September 30 close beside them for the month’s move.

  • The 10-year Treasury closed at 4.94%, down from 5.29% on September 30, a fall of thirty-five basis points. It touched an intraday 5.3645% on October 7, its highest level since April 2002, and finished the month at its low (U.S. Treasury; Reuters via LiveMint).
  • The 30-year Treasury closed at 5.31%, down from 5.64%, a fall of thirty-three basis points. Its October high was 5.67% on October 7.
  • The 2-year closed at 4.56%, down from 4.88%, or thirty-two basis points.
  • The belly led the rally. The 5-year fell thirty-eight basis points to 4.71% and the 7-year fell thirty-nine to 4.80%, both more than either end of the curve.
  • The 10s30s spread closed at 37 basis points, against 35 on September 30. It has spent two months inside a band of seven basis points.
  • The 2s10s spread closed at 38 basis points, down from 41. The belly-led rally narrowed it slightly rather than steepening it.
  • The 20-year sits above the 30-year, at 5.34% against 5.31%. That inversion has held for months and it is the quietest signal in this article: the market charges more for twenty years than for thirty, which is what a buyer does when the far end of the curve has stopped being about inflation and started being about who will hold the paper.
  • Brent settled at $117.64 for December delivery and $113.43 for January; WTI at $104.12. Brent gained 13.6% in October and WTI 15.2%, the fourth consecutive monthly gain for both (CNBC).
  • Tanker traffic through Hormuz remains about 70% below prewar levels, at 277 transits in the week to October 28, and October was the worst month of the war for attacks on shipping (UKMTO via Middle East Eye).
  • The federal funds target range is 3.75% to 4%. The Fed held on October 28 by a vote of 10 to 2, with Governor Lisa Cook and Cleveland Fed President Beth Hammack dissenting in favor of a quarter-point hike (Federal Reserve statement; CNBC).
  • The market still prices a December hike at roughly 85%, per CME FedWatch as reported after the September PCE release (Reuters via Investing.com).
  • The average 30-year mortgage is 7.24%, down from 7.32% the week before and up from 6.17% a year ago (Freddie Mac).
  • Third-quarter GDP grew 3.4% annualized, above the 3.1% estimate, while personal spending rose 0.5% in September and real spending was flat (CNBC).
  • The University of Michigan’s final October sentiment index came in at 48.0, the second-lowest reading on record, with year-ahead inflation expectations at 4.7% and five-year expectations at 3.5%, both the highest since May (University of Michigan).
  • Gasoline averages $4.20 a gallon, down 43 cents in a month but $1.16 above a year ago. Diesel is $6.32, below its September 22 record of $6.53 and still up about 71% on the year (AAA).

Read the first four bullets and the eighth together. The whole curve fell between thirty-two and thirty-nine basis points in a month when oil rose 13.6%, inflation printed at 3.7%, and the Fed refused to rule out a hike. Nothing about the inflation story improved. What changed is that the market finally found a reason to buy that did not require the inflation story to improve.

The Bid That Finally Came

For nine months this series has argued that the wartime bond market was not behaving like a wartime bond market, because the reflex formed in 1990 assumes a financial shock that a central bank can answer by cutting, and this has been a price shock instead.

That argument was right for nine months and it was wrong for one week, and the week deserves taking apart because it tells you what threatens the next ninety days.

The sequence began on October 22 with Meta. The company beat on revenue and raised its capital spending plan, and the stock fell 7% anyway, on the arithmetic of a business that had just posted its first cash outflow in more than a decade to fund AI capacity (AP). The Nasdaq fell 2.6%, the Philadelphia Semiconductor Index 6.7%, and money that had been in the AI trade went looking for somewhere to wait. It found Treasuries. Yields fell across the curve in the sharpest single session since the 2023 banking panic.

One week later the same thing happened twice more. Alphabet lost about 12% on October 29, roughly $500 billion of market value, after disclosing that AI-generated search summaries were denting its core advertising revenue; Intel fell 28.6% the next day, its worst session since the 1970s, on guidance that missed badly; Amazon held its $220 billion 2026 capital budget and warned of spending more in 2027 while running negative free cash flow for a second quarter (CNBC; NBC News). More than $1 trillion of technology market value went out the door, and the ten-year finished the month at 4.94%.

Now notice what the bond market was doing while all of this happened. Oil was climbing. On October 22, the same day yields fell seventeen basis points, Brent rose 4% to $113.17 on Houthi strikes against Saudi refining and on a New York Times report that the United States and Israel had drawn up plans for a “short but intensive” campaign against Iran after the midterms (AP; NPR). On October 29 and 30 the Houthis claimed attacks on the Ras Tanura and Rabigh refineries and on an Abqaiq production facility, and a tanker in the strait was damaged by limpet mines attached by a small boat, the first successful attack of that kind in nine months of war (CNBC).

So the old correlation broke. For most of this year the ten-year and the front-month crude contract moved together at a rolling correlation near 0.96, because both were pricing the same inflation shock (CNBC). In the last ten days of October, oil went up and yields went down. That divergence is the single most important thing that happened to this market in 2026, and it has a precise meaning: the bond market stopped asking what the war would cost and started asking what the economy would do without it.

That is a healthier question in one sense and a more dangerous one in another. It is healthier because it means the long end can rally on good news about growth rather than only on bad news about peace. It is more dangerous because the thing that produced the rally was a de-rating of the largest sector in the stock market, and de-ratings do not stop on command. The bond market’s gain in October came out of the equity market’s pocket, and the equity market has not finished deciding how much it owes.

What the Curve Did, and Why the Belly Led

The shape is worth its own section, because the previous installment set a specific test and the market answered it.

The test was this: the 10s30s had pinned itself in the mid-thirties and should no longer be read as a signal, so the instrument to watch this quarter was the 2s10s. If it kept steepening while the funds rate stayed put, the market expected the Fed to stop before it won. If it flattened back toward 27 basis points, the market expected the Fed to keep going.

It flattened, from 41 basis points to 38, which by the letter of the test says the market expects the Fed to keep going. But the letter is not the whole answer this time, because the flattening did not come from the front end rising. The two-year fell thirty-two basis points. The ten-year fell thirty-five. The spread narrowed because the long end fell slightly faster than the short end, which is a bull flattening, and a bull flattening in a month when the Fed held and promised price stability is the market saying something more specific: it believes the Fed will hike in December and that the hike will be close to the last one.

The belly tells the same story louder. The five-year and seven-year fell thirty-eight and thirty-nine basis points, more than either end. That is where the market prices the peak of the cycle, and the peak just moved down and closer.

The 10s30s, at 37 basis points, remains almost exactly where it was. Two months, half a percentage point of level, and the spread has moved two basis points. That is a strong confirmation of last month’s conclusion and it is worth restating plainly, because it is the part of the curve most people still misread. The gap between the ten-year and the thirty-year is not currently a measure of anything. It is not a term premium signal, because a term premium that could move would have moved during the worst month of an oil war. It is not a recession signal, for the same reason. It is a number in the thirties that reflects a market pricing one inflation regime across every maturity long enough to be hurt by it, and it will stay an uninformative number until something forces the ten and the thirty to disagree.

Two things could force that disagreement in this window. A Fed hike in December, which hits the ten harder than the thirty, because the ten is the maturity that carries the mortgage market and the corporate borrower. Or a genuine credit event, which hits the thirty harder, because a flight to quality with a recession attached buys the longest duration it can find. Watch which one arrives. The answer will show up in this spread before it shows up anywhere else.

The Buyback Scorecard, Fourth Entry

This section exists to hold one claim accountable: buying your own bonds does not lower the rate at which the market will lend to you for thirty years.

The evidence this month is almost embarrassingly clean. The enlarged long-end buyback program, announced August 19 and running through November 4, did not lower the long end in September. In October the long end moved thirty-three basis points, in the Treasury’s favor, and it did so during the final weeks of a program whose caps are about to expire, while the department said nothing new about whether it would extend them.

Put the two facts together and the conclusion writes itself. The bid that arrived in October was not the Treasury’s. It was a payroll report, an equity de-rating, and a queue of investors looking for somewhere to put money that had been in semiconductors that morning.

That does not make the program worthless, and the dealers who have to intermediate the auctions are worth hearing. Citi’s Jason Williams expects the November 4 refunding to cut 20-year and 30-year auction sizes by $3 billion each and shift the funding into bills, which is the supply-side version of the same admission: less duration in private hands, less term premium, a friendlier long end (CiVL). Goldman Sachs expects the opposite on size, holding coupons steady, and expects any increases to resume in May 2027 concentrated in the two-year to seven-year sector (Goldman Sachs Research). LPL Research put the scale of the whole effort in one sentence: buybacks absorb roughly 14% of ten-to-thirty-year supply, about $128 billion a year against more than $900 billion sold, which makes the program “more of a signal than a solution” (LPL Research).

Fourteen percent. That is the honest number, and it is why this section is now in its fourth installment. A program that retires one bond in seven cannot set the price of the other six, and a market that has just demonstrated it will rally thirty-five basis points on an equity selloff was never waiting for the department to show up.

Watch November 4 for three things, in order of importance. Whether the 10s-20s and 20s-30s caps are extended beyond the current expiry, whether 20-year and 30-year auction sizes come down, and whether the statement’s forward guidance loses the phrase “at least” from its description of how long coupons will stay steady. The third is the one that matters most and it is the easiest to miss. JPMorgan warned in July that removing “at least” “risks pushing the curve steeper as market participants demand higher term premium, similar to what occurred in August 2023, to digest an expected increase in duration supply” (Bloomberg). The midterms are the day before the announcement, so the base case is no change at all. A change in any of the three would be a bigger signal than anything the Fed says in December.

The Governance Question

This series treats the rule of law as a standing input to the term premium rather than a political aside, and October supplied the largest single governance item since the series began. It is not a scandal. It is an audit request, which is worse in one respect: it means the people who wrote the program cannot say for certain what happened inside it.

Roughly $166 billion is being handed back to importers, and the allocation runs by filing date. The Supreme Court held in February that the emergency statute does not authorize tariffs. Roughly $166 billion was collected on about 53 million entry summaries from roughly 330,000 importers between February 2025 and February 2026, and about $100 billion had been refunded as of August (Representative Jill Tokuda). On October 21, Senator Raphael Warnock, Representative Tokuda, and Representative Brad Sherman asked the Government Accountability Office to audit the process, naming the questions that matter: whether controls ensure refunds are “complete, accurate, properly authorized, and protected against improper or duplicate payments,” which importers and industries received the money, and whether any of it reached the consumers who paid the surcharge (Senator Raphael Warnock).

Two weeks earlier, on October 8, Judge Richard Eaton of the Court of International Trade certified an injunctive class covering all importers of record who paid the duties, which opens a refund path for companies that never sued (ArentFox Schiff). The Department of Justice is still arguing that relief belongs only to the named plaintiffs. So the difference between two importers with identical claims is a lawyer and a filing date, and the sum at stake is large enough that the difference will be litigated for years. That is what a term premium is made of: not a bribe, but a rule that produces different answers for identical facts and cannot explain why.

The exemption channel has moved from tariffs to pharmaceuticals. Under Proclamation 11020, patented drugs from most countries face duties of up to 100%, while nine specialty categories, including orphan drugs, gene therapies, antibody-drug conjugates, and plasma-derived therapies, pay 0% when the product comes from one of nineteen named jurisdictions or when Commerce grants a case-by-case “urgent health need” determination (Federal Register; WilmerHale). The categories are defensible on the merits. The discretion is the point, and it is the same discretion the Journal of Financial and Quantitative Analysis study measured in the older tariff exemption program, where contributions to the party in power raised the odds of approval and each grant was worth a median of about $51 million in firm value (JFQA).

And the pressure on the Fed has acquired a courtroom. The White House has convened a committee of its own appointees to examine whether Governor Cook made false statements on mortgage applications, with an in-person hearing set for November 5, two days after the election (CNN). Politico’s framing was blunt and correct: “The real defendant here is Kevin Warsh” (Politico). The chair has so far been the most hawkish voice in the room, holding a 10-2 vote against a president who wants cuts to 1%, and he said at his press conference that politics played no role in the decision (Federal Reserve transcript). The notable irony is that the governor the administration is trying to remove dissented for a hike. A removal proceeding aimed at the board’s least dovish member is not a case about mortgages.

The Ninety-Day Map

Take the premise seriously: no improvement in the war, no breakthrough, no formal reopening of the strait. Here is the path from November 1 to January 30.

November 3 and 4. The midterm election on Tuesday, then the refunding on Wednesday morning. Forecasters have Democrats heavily favored in the House, near 95% on the Silver Bulletin model, and favored in the Senate by a range running from about 64% to 78% depending on the model, needing a net gain of four seats (Silver Bulletin; Election Lab). A contested or slow-counted result during a war adds a premium rather than removing one.

November 4, in Jerusalem. Israel’s election results are certified and submitted to President Herzog the same day. No bloc won a majority on October 27, with Netanyahu’s at 53 seats and the opposition at 48, and the coalition window is forty-two days (AP). A caretaker prime minister is a constraint on a new campaign, which is one of the few forces in this window pulling toward de-escalation.

November 5. The White House hearing on Governor Cook.

November 6 and 10. October employment, then October inflation. This pair is the window’s first real test. If payrolls are soft again and CPI holds near 3.7%, the market gets the configuration it rallied on in October and the December hike becomes the only question left. If payrolls rebound, the bull case from October loses its foundation.

November 17 through November 26. Nvidia reports, and then the country sits down to Thanksgiving with diesel at $6.32 and gasoline at $4.20. This is where the war stops being a market story and becomes a household bill.

December 4 through December 11. The cluster. November employment on the fourth, the Fed on the eighth and ninth with a fresh dot plot, November inflation on the tenth, and the continuing resolution expiring on the eleventh (Congress.gov). A rate decision, an inflation print, and a funding cliff inside seventy-two hours.

January 3 through January 27. A new Congress is seated, the December employment and inflation reports land in the second week, and the Fed meets on the twenty-sixth and twenty-seventh with no projections. That meeting is where the market will decide whether December was the last hike or the second of three.

That is the base case, and it lands the ten-year between 4.6% and 5.2%, the thirty-year between 5.0% and 5.6%, mortgages between 7.0% and 7.5%, and the 10s30s anywhere near where it is now.

Two branches sit on either side of it.

The escalation branch. The reporting from October says the United States and Israel have plans for a short, intensive campaign after the vote, aimed at energy infrastructure, nuclear sites, and missile and drone arsenals, with some refineries deliberately spared so that gasoline does not spike (NPR). If it comes, expect the counterintuitive sequence that October taught: yields fall first, because a shooting war is a risk event and the first instinct is still to buy the safest paper on earth. Then, within weeks, the second move. A three-day campaign that closes the strait again does not lower the price of diesel, and the Fed cannot cut into a 4% inflation rate. The 1979 precedent is the one that applies here, and it is not about the first month. It is about what a central bank does when a supply shock refuses to leave.

The paradox branch. Last month the paradox was a flight to quality. This month it is the opposite and it deserves equal weight: a bond rally that continues because the Fed blinks. The two-year is at 4.56%, the labor market added 29,000 jobs in a month, and the market is still 85% certain of a hike. If November payrolls come in near zero and the December meeting passes without a move, the front end reprices fast and the long end follows, and a ten-year at 4.4% with inflation at 3.6% becomes a real possibility. That is not a bullish outcome for the country. It is a market deciding that the Fed has stopped fighting, and it would be the most expensive kind of rally.

What a Careful Investor Does

None of this is a forecast that yields fall in a straight line. It is a statement about what changed and what did not. What changed is that the long end has finally shown it can rally on something other than peace. What did not change is that the inflation impulse is still running above 3%, the deficit is still being funded into a market that just discovered how fast it can turn, and the war can restart the whole argument in a weekend.

  • Extend duration selectively, and only where you can hold it. The thirty-year at 5.31% pays thirty-seven basis points more than the ten-year for twenty additional years of inflation and issuance risk. That is still a bad trade for money you might need. It is a reasonable trade for a liability twenty years out, which is the only purpose a thirty-year bond serves.
  • Take the belly seriously. The five-year and seven-year led the October rally for a reason: they are where the market prices the peak of the cycle, and the peak just moved closer and lower. A ladder weighted toward three to seven years captures that repricing without betting the whole position on the Fed stopping.
  • Do not buy duration because the equity market is falling. That trade worked for exactly one week in October and it will work again, but it is a trade, not a strategy. The reason to own bonds is the yield, and the reason to own them now is that 4.94% on the ten-year is the most you have been paid since the summer.
  • Own inflation protection while it is cheap relative to the risk. TIPS breakevens are pricing a world in which the war premium fades. The University of Michigan’s respondents expect 4.7% over the next year while the ten-year offers 4.94% nominally. That gap is the case for TIPS in one line.
  • Keep the ladder, and keep some powder. A ladder converts a two-branch map into a schedule. The cash is for the escalation branch, which is the one that arrives as a headline on a Saturday.

The Long Way of Saying It

The bond market is a referendum on whether a government’s promises are worth holding, and this month the vote moved without any of the promises improving.

That is worth sitting with, because it is the opposite of what nine months of this series trained me to expect. I spent the spring and summer arguing that the long end would not rally until the inflation impulse broke, and the inflation impulse did not break. It got worse. CPI printed 3.7%, PCE printed 3.6%, oil gained 13.6%, the worst month of attacks on Hormuz shipping passed without a ceasefire, and the ten-year fell thirty-five basis points anyway.

The lesson is not that the bear case was wrong about inflation. The lesson is that a bond market prices two things and I had been tracking one. It prices inflation, which got worse. It also prices the demand for safety, and in the last ten days of October the demand for safety came from the only place big enough to move it: an equity market that had spent the year borrowing against a future it could not yet invoice. When that borrowing got repriced, the money did not leave the country. It moved across the street.

So the next ninety days turn on a question that has nothing to do with the strait. Does the AI de-rating continue? If it does, the bond market has a buyer it did not have in September, and the long end can hold 5% or better through a hot inflation print and a December hike. If it stops, the buyer goes back to buying semiconductors, and the arithmetic of a 3.7% inflation rate and a Treasury funding a war in its tenth month reasserts itself with nothing in the way.

Storms do not send invitations, but they do leave clues. The clue this month is that the safest asset in the world rallied on the worst inflation news of the year, and it did so because something else broke first. Build the defenses before the lightning strikes, and notice which building the lightning chose.

Sources

PRH | huffmanwrites.org | © Philip Huffman