The Price of Thirty Years

Start with the question a sensible person asks and a textbook never answers. If a three-month Treasury bill pays 3.92 percent and a thirty-year bond pays 5.27 percent, who in their right mind takes the thirty-year? The extra is a point and a third. The cost is thirty years of not being able to change your mind. Put that way it sounds like a bad trade, and this week a great many people with real money decided it was.

But the framing hides the lesson. A bill and a bond are not two prices for the same thing. They are two different products that happen to be sold by the same seller, and they protect you against opposite risks.

The bill protects you from price. Hold it ninety days and you get your money back at par; nothing that happens to interest rates in between can touch you. What it does not protect you from is reinvestment. In ninety days you have cash again, and you earn whatever the world is paying that morning. If the Federal Reserve cuts to 2 percent in 2028, a bill investor earns 2 percent in 2028. The 3.92 was never a promise about anything but the next quarter.

The bond is the mirror image. It locks 5.27 percent for thirty years no matter what the Fed does, which is exactly what someone who owes fixed payments in 2050 wants. What it does not protect you from is price. A thirty-year bond at these yields has a duration of roughly fifteen, which means its price moves about 15 percent for every one-point move in yield, before convexity softens the arithmetic. Yields go from 5.27 to 6.27 and you are down roughly 14 percent. Yields fall to 4 and you are up about a fifth. Same instrument, same coupon, and a swing of a third of your capital depending on which way the world turns.

A bill investor rents the rate for three months. A bond investor buys it for thirty years and carries the price risk of owning it.

So the 135 basis points are not a bonus for patience. They are the price the market charges to take the second product instead of the first, and that price has a name: the term premium. It is what you are paid for accepting thirty years of price risk while everyone else keeps the option to leave. Whether the premium is fair is the only question that matters at the long end, and it is the question the last three weeks have been arguing about out loud.

Here is the Treasury par curve at the close on September 2, 2026, straight from the Treasury’s own daily table.

Three things in that picture are worth reading slowly. First, the bill yields above the top of the Fed’s own target range, which is 3.50 to 3.75 percent. A three-month bill paying 3.92 is the market pricing a rate hike before that bill matures. Second, the curve is steep from two years out: 88 basis points between the two-year and the thirty-year. Third, and this is the part the headlines are about, the thirty-year had already been higher. It touched about 5.33 percent on August 18, a level not seen since 2007, before a surprise from the Treasury pulled it back.

Two things move the long end. One is the path of short rates, which is the Fed’s business. The other is the term premium, which is nobody’s business and everyone’s opinion. When the thirty-year rises faster than the two-year, the second thing is doing the moving. That is what happened in August. The market did not change its mind about next year. It changed its mind about how much it wants to be paid to think about 2056.

A common guess is that big institutions hold long bonds because some rule tells them to: a 60/40 mix, a mandate, a box on a form. It is not a ratio. It is a match. A pension owes a stream of payments to retirees that stretches decades out. A life insurer owes annuities. The way you make sure a rate move does not blow a hole in your funding is to own assets whose payments arrive on the same schedule as your obligations, so that when rates fall and your liabilities balloon, your bonds balloon with them. That is called liability matching, and it is the only reason anyone with a thirty-year horizon voluntarily owns a thirty-year bond.

Here is the part that explains this month. Once the duration of what you own roughly equals the duration of what you owe, you are done. Buying more long bonds does not reduce your risk; it adds a new one. You are full, and no yield makes you less full. Higher rates over the past three years actually improved pension funding, because liabilities are discounted at higher rates and therefore shrink, and a great many plans used that improvement to lock in, shifting into bonds and de-risking. The marginal buyer who might have absorbed this year’s supply already bought last year.

Run down that roster and the picture is not one of a market on strike. It is a market where everyone who holds long bonds for structural reasons is already holding what they need, and the only marginal buyer left is the one who buys for return. That buyer is a speculator in the honest sense: someone betting that yields will fall from here. If yields drop from 5.27 to 4, a thirty-year bond gains about a fifth of its price. Long bonds are the instrument for betting on rate cuts.

And nobody wants that bet while the chairman of the Federal Reserve is standing in Wyoming talking about hikes.

On the other side of the table sits a seller who cannot walk away. The Congressional Budget Office now puts the fiscal 2026 deficit at about $2.1 trillion, up $200 billion from its February estimate, largely because the Supreme Court’s February ruling on tariff authority removed roughly $250 billion of expected revenue. Ten months into the fiscal year the deficit already stood at $1.8 trillion. Every dollar of that must be borrowed, and some of it must be borrowed for a long time, whatever the market’s appetite that week.

A seller who has to sell into a market with no structural buyer has one lever: price. That is the whole mechanism. The term premium rises, the long bond cheapens, until the marginal buyer — the one who was waiting — decides the reward is worth the risk. Nothing is wrong. The market is doing exactly what a market does when supply is inelastic and demand is not.

Then, on August 19, the Treasury did something unusual. It announced it would at least double its liquidity-support buybacks in the ten-to-thirty-year sectors, from a maximum of $2 billion per operation to at least $4 billion, effective September 9. The ten-year fell 6 basis points that day and the thirty-year fell 9, to 5.196 percent. The secretary said the next day it could be more than $4 billion.

When the seller starts buying back its own product, it is telling you what it thinks of the demand.

Read that plainly. The government is issuing long bonds with one hand and buying them back with the other because it would rather pay a bill rate on the difference than let the thirty-year set its own price in public. It is a rational thing to do. It is also an admission.

Nine days later, on August 28, Chairman Warsh gave his Jackson Hole speech, his hundredth day in the job. He called the 2 percent target “a firm, fixed target,” put twelve-month PCE inflation at 3.7 percent and the six-month rate at 4.1, noted that more than half of the 199 components of the index had risen faster than 3 percent over the year, and counted “65 months of sustained, elevated inflation.” He did not promise a hike. He did say that if the summer’s better readings do not change the underlying trend, “we have work to do.” The front end sold off hard on the day. A speculator thinking about catching the long bond at 5.3 heard that and put the knife down.

It helps to see the sequence in order, because the popular version compresses it into “bonds are selling off” and loses the mechanism.

This is the question that stumps people, and it should. Gold pays nothing. A thirty-year bond pays 5.27 percent. In a hawkish market, with a chairman talking about hikes, the opportunity cost of holding a lump of metal instead of a coupon is at its highest. In 2022, when the Fed last tightened hard, gold went nowhere for exactly that reason. So why is it sitting at $4,334 an ounce on September 2, up from $4,111 a month earlier, after touching about $4,550 on August 21?

First, be precise about what gold is not doing. It is not at a record. The record was about $5,590 in late January 2026, and the metal is roughly 22 percent below it. The official bid has cooled, too: central banks bought 1,092 tonnes in 2024, 863 in 2025, and only 345 in the first half of this year, the weakest first half since 2022. Whatever is holding gold up in the $4,300s against a 5 percent Treasury yield, it is not a record-chasing frenzy and it is not central banks.

It is the question of why rates are high. Rates rising because the economy is strong and the central bank is credibly in control are bad for gold: the 5 percent is real, it beats inflation, and it gets paid back in dollars worth what they are worth today. Rates rising because the deficit is $2.1 trillion, inflation has been above target for 65 months, and buyers of long paper are being paid more and still stepping back is a different thing entirely. That second kind of high rate is not a reason to leave gold. It is the reason to hold it.

The case against gold assumes the 5 percent is real. Gold buyers are betting that one of the words in “real” fails.

The tell is a combination a healthy market never prints: long yields up, the dollar down, gold up. Higher yields are supposed to pull money into dollars. Over August the ten-year climbed from about 4.68 to 4.79 percent and the dollar index slipped anyway, closing the month lower and touching its weakest level since May on August 22. When yields rise and the dollar falls together, it means the higher yield is not attracting the marginal foreign dollar; it is failing to keep it. Some of that money goes into other currencies. Some goes into the one reserve asset with no issuer, no coupon to doubt, and no deficit behind it.

A caveat, because the picture is not clean. The sharpest leg of the dollar’s August drop came on the buyback news, when yields were falling, not rising; the dollar read the buyback as the government leaning on its own rates. And gold’s bounce from $4,111 to the mid-$4,300s is a recovery from a summer lull, not a breakout. The honest claim is narrower than the headline version: in a month when the long end cheapened and the Fed turned hawkish, gold held and the dollar did not. That should not happen if the 5 percent were being taken at face value.

That is the position the long bond is in on the eve of the jobs report: it needs the labour market to be bad in precisely the way that does not frighten the Fed, and the width of that target is narrow. This is the discomfort behind three weeks of headlines. It is not that no one will buy a thirty-year bond at 5.27 percent. It is that the buyer who would wants more than 135 basis points to be wrong about 2056, and the seller would rather buy back the bond than pay it.

Already true

The thirty-year has traded at its highest since 2007. The Treasury has doubled its long-end buybacks. Foreign official holdings are $233 billion off their February peak. The chairman has put twelve-month inflation at 3.7 percent and declined to promise anything but discipline. None of that is forecast.

What has to happen for the long end to settle

One of three things. The deficit path narrows enough that supply stops being the story. Inflation falls toward 2 in a way the Fed believes, which unlocks the speculators’ bet on cuts. Or the term premium rises far enough that the return buyer is paid to step in without either of those. The third is the one the market is choosing by default, and it is the one that hurts everything priced off the long rate: mortgages, cap rates, equity multiples.

Where I am probably wrong

The buyback could work. Four billion an operation is small against a $2 trillion deficit, but the announcement alone moved the thirty-year 9 basis points, and a Treasury willing to lean on the long end has more tools than a blog post can list. If the thirty-year is back under 5 by October with the deficit unchanged, the term premium story was overdone and the August high was a positioning squeeze, not a verdict. Watch the dollar: if it recovers while yields fall, I was wrong about what the money leaving was saying.

Strip away the week and five things remain true at the long end in any year.

1. A bill and a bond are different products. One carries reinvestment risk, the other carries price risk. The yield difference is the price of switching from the first risk to the second. Never compare the two as if the higher number were simply better.
2. The natural holders of long bonds are full, and no yield changes that. Liability matching is a match, not a ratio. Once matched, the pension stops. The marginal buyer is always the one who wants a return, and that buyer wants a term premium.
3. A forced seller sets price with the only lever it has. A $2.1 trillion deficit into a market without a structural bid means the long bond cheapens until someone is paid to hold it. That is the mechanism, not a malfunction.
4. Buybacks are an admission. When the issuer starts purchasing its own long paper, it is telling you the demand it sees. Read the action, not the press release.
5. Gold answers a different question than yield. Not “what does this pay?” but “is the payment real?” When yields rise, the dollar falls, and gold holds, the market is answering the second question, and the answer is not yes.

The thirty-year bond is the most honest instrument in the market because it forces the buyer to hold an opinion about 2056. Bills let you postpone the opinion for ninety days at a time. This month, given the choice, the money postponed. That is the story behind everything your feed is showing you about bonds, and it is not complicated. It is just long.