# What the Market Was Actually Pricing on Hike Day — Notes on MacroVoices #550

> Notes on MacroVoices #550 with Harley Bassman: ten-year breakevens, mortgage convexity, and three traps hiding inside ETFs — what the bond market is really pricing. Educational, not investment advice.

Published: 2026-09-18
Locale: en
Tags: Federal Reserve, inflation expectations, bond market, mortgage convexity, ETFs


![An early-morning corridor inside the Federal Reserve, light falling from high windows onto an empty wooden desk and a brass balance, the hallway receding into depth](/covers/macrovoices-2026-09-17-macrovoices-550-harley-bassman-in-fed-we-trust-cover.png)

> There are three things by which a state is governed: law, trust, and authority.
>
> —— *The Book of Lord Shang*, "On Cultivating Authority" (Warring States period; translation mine)

## What this episode is about

MacroVoices episode 550, recorded 17 September 2026. Eric Townsend talks to Harley Bassman — the man who invented the MOVE index, known on the street as the convexity maven. Hours before the recording the Fed had hiked a quarter point and signalled through the dot plot that another hike was likely before Christmas. Markets took the hawkish half badly.

Bassman opened by moving the question. Twenty-five, fifty, a hundred basis points — none of it changes inflation much, he said. The point of hiking lies elsewhere: trust in government and in institutions has a hole in it, and the Fed's job is to patch that hole. He came back to the same line all hour — I don't care about the numbers, I care about trust. His own preference was a cleaner version: fifty, one and done, the market had already priced it, and dragging everyone through twelve more weeks helps nobody.

What stayed with me: a man who spent thirty years in fixed income and built the index that measures bond volatility ends up attributing the rise in yields to something nobody can measure.

## The main points

**One, the hike is withdrawal, not tightening.** Bassman compared getting off forward guidance to quitting cigarettes — he started to say heroin and remembered this was a family broadcast. Markets spent more than a decade being fed guidance until they lost the habit of pricing risk themselves, and now every change of tone throws them over their skis. He wants a return to markets assessing and managing their own risk, and he expects the withdrawal to be volatile.

**Two, the ten-year breakeven sits at 234 basis points.** That is the spread between TIPS and nominal Treasuries; the four-year average is 235. Yields rose 150 basis points over that stretch and the number did not move. Crazy town, he called it — if the market feared inflation, that spread would be climbing. It isn't, so something else is pushing yields up. He named trust, then added that he might be wrong, and that if another guest comes on in two weeks to explain why the chart is meaningless, he'll accept it.

![A subtraction of two bars: the tall bar on the left is built from a real-rate lower segment and an inflation-expectation upper segment, and subtracting the middle short bar, which is the same height as that lower segment, leaves the bar on the right representing the breakeven inflation rate.](/figures/breakeven-two-quotes-subtracted-en.svg)

**Three, he raised the objection against himself.** Eric pressed: TIPS key off CPI, and CPI has been criticised for decades. Bassman agreed outright — after the Clinton-era rejigger with hedonics and substitution, holding down reported CPI was a policy goal, because Social Security payments index to it. His position: cooked or not, CPI is still what the Fed cites in its meetings and what the market prices off, so the chart still speaks.

**Four, hyperscaler borrowing ignores the price of money.** The deck projects 750 billion dollars of borrowing, closing in on a third of Treasury issuance, all competing for the same pool. The Fed cannot stop it, Bassman said, because these firms will borrow at 6, 8, 10, 12 percent — they are in an existential race where nobody knows who wins and stopping guarantees you don't. His analogy was the dotcom boom: ten companies crowded into one niche, one of them wins. Pets.com didn't, but some pet company did.

**Five, the bonds are money good; the equity is another story.** On AI-related debt now exceeding the municipal bond market, his read: Meta, Google, Amazon, Microsoft, Oracle all sit on underlying cash machines that can cover coupons and repay principal. Share prices are a different question — spending compresses profits, and he declined to comment on the equity. When Eric raised the scenario of Chinese models being open-sourced for free, Bassman called it Global Crossing redux: they laid the cable expecting X dollars a megabyte and got a tenth of that. The people who build get hurt; the people who use get it cheap.

![Two horizontal bars of different lengths: the upper one represents Treasury issuance, and the lower one, only about a third as long, represents the scale of hyperscaler borrowing.](/figures/hyperscaler-borrowing-vs-treasury-en.svg)

**Six, mortgage bonds are getting harder to hold.** This is his home turf. A mortgage bond is a covered call — buy a ten-year at par, sell a three-year call struck at 105. Upside capped, downside open. Three years ago 71% of the market carried coupons of 3.5% or lower at an average dollar price of 78.94, miles from the strike, so the embedded option was worth little and spreads traded tight. Since then the stack has been recouponed: low-coupon paper paid down, refinanced, or prepaid because someone divorced, died, or moved. New production carries high coupons — Fannie 5.5s alone now exceed a trillion. An option is most convex at the strike, so the index as a whole turned more negatively convex and the spread widened from 95 to 110. His technical marker: when the curve flattens or inverts, mortgage spreads widen. The curve has been flattening for three months.

**Seven, three traps inside ETFs.** Leverage splits in two — linear leverage (futures inside the wrapper, two to one, buy and hold) is fine; daily-reset percentage leverage does not survive past a week or so. Buy at 100, up 20% to 120, down 20% leaves 96; run that up and down and volatility drag burns the thing toward zero. Return of capital is the second trap: when junk yields 6 to 8% and a fund distributes 10% or more, part of that distribution is your own money coming back, and the symptom is a price that grinds lower month after month with nothing else changing. The third is liquidity mismatch — an ETF is a hybrid between open- and closed-end, market makers transact at closing NAV, and when assets get sold below NAV the fund absorbs the difference at the expense of everyone who stayed.

## Going further

### "The data says inflation is fine — so why is my mortgage rate still climbing?"

This was the first thing I wanted straightened out. The inflation narratives in the headlines all sound plausible, and prices refuse to follow them.

Unpacked, Bassman's method runs like this. Two instruments quote at the same time. A nominal Treasury quotes what you demand to lend to the US government for ten years. A TIPS quotes what you demand when inflation is already compensated for. Subtract one from the other and you have the market's inflation expectation. It sits at 234 and won't move, while total yields rose 150 basis points — none of those 150 belong to inflation.

![A comparison of two stacked bars, four years ago and now: the lower segment representing inflation expectations is exactly the same height in both, while the upper segment representing other compensation is taller now, and the extra piece is labelled one hundred fifty basis points.](/figures/yield-rise-has-no-inflation-share-en.svg)

Which turns the question into: compensation for what? His answer is the cost of holding dollar assets at all. No war, no recession, unemployment at 4.1%, and a deficit still running 6% of GDP. The market isn't pricing default — the US can print. It's pricing the rent on keeping dollars as a core asset.

The general-purpose version: when a claim makes you uneasy, ask which price would move if it were true, then go look at that price. If it hasn't moved, either the market disagrees with the claim or you're looking in the wrong place. No guarantee you end up right — Bassman said outright he might be wrong — but it converts a vague worry into a number you can check.

### "My high-yield ETF pays every month and the price keeps falling"

Offering a high yield is eye candy, Bassman said, and people can't help themselves. That line made me laugh.

His test is plain enough: are the underlying assets generating enough income to cover the distribution? Whatever isn't covered is your own principal handed back. The visible symptom is a price stepping down month after month while nothing happens underneath. His rough threshold: in a world where junk yields 6 to 8%, a distribution above 10% probably involves return of capital.

![A height gap between two bars: on the left is the income the underlying assets earned, on the right is the larger amount paid out, with the excess piece shaded separately, and a NAV line on the right side stepping down month after month.](/figures/return-of-capital-nav-drain-en.svg)

What got me about this section is that nobody comes to tell you. The distributions land, the cash flow on the statement looks healthy, and the damage hides in the price column where it reads like a market decline. Separating the two takes one action: pull up comparable assets over the same window. They fell and you fell — that's the market. They held and you fell — that's your money coming home.

### "AI has read everything everyone ever wrote. Is my judgement worth anything?"

Eric took Bassman outside his wheelhouse here, and it's the most interesting stretch of the episode. His framing was sharp: every word of this podcast gets transcribed and absorbed into every training run within two days, along with everything Bassman ever said publicly, every page he wrote across a career, and every colleague's opinion on anything. People with less raw intellect are already competing with him because AI carries them. What happens next?

Bassman paused, then drew a line: these models still can't predict the future. What they can do is build the best portfolio given a set of parameters and react as news arrives. Plenty of systems already trade off keywords in Fed statements — they react to the number, they don't forecast it. Until something can know in advance, he thinks humans do fine.

![A vertical line splits the frame in two: the left half is packed with dense horizontal lines representing everything already written down, and the right half is an empty dashed box representing what has not happened yet.](/figures/known-filled-unknown-empty-en.svg)

Eric pushed toward the endgame: if all portfolio management goes to AI, everyone gets the efficient result and there's no money left to make. Bassman's answer circled back to his theme — it's always about character, and it's never different this time. Cars arrived a hundred years ago and the saddle and buggy-whip makers found their way through. It took a while.

What I take is the line itself: reacting versus predicting. A system that has read everything known will beat me inside the known. Outside it, it knows no more than I do. Applied to my own work that means handing off the organising of what's already written and spending my hours on the part nobody has written down.

## Where to look next

- The episode page and this week's slide deck: macrovoices.com (the download button above Harley's picture on the homepage — no registration wall any more)
- convexitymaven.com, where twenty years of Bassman's commentaries sit free, including "Looking Under the Hood of ETFs" and the mortgage convexity piece referenced here
- The ten-year breakeven is on FRED under T10YIE — the 234 in this piece
- For convexity itself, his definition in the episode is the clearest one I've heard: win a dollar lose a dollar is zero convexity, win two lose one is positive, lose three win two is negative

## One thing to take with you

**One idea: every price answers a question, and your first job is to work out which question.**

What Bassman does all hour is split "yields are rising" into answers to two separate questions — how high will inflation run, and what rent do I charge for holding dollars. He checks the breakeven, sees the first answer unchanged, and concludes the whole move belongs to the second. Same number, readable only once you've matched it to its question.

**Something I've tried that has nothing to do with markets**: pick one thing you keep worrying about — whether your job is safe, whether a relationship is cooling, how your parents' health is holding — and write one sentence: "If this were actually happening, which checkable fact would change first?"

The second step is the hard one. Worried about the job, my first answer is usually "my manager's tone" — that's my reading, not a fact. Push one layer down: is the team still hiring this quarter, is my project in next quarter's plan, has anyone been moved off it. Those are checkable. Then go and check one.

![Three stacked horizontal bands narrowing from top to bottom: the widest top band is a vague worry, the middle band in a warning colour is only an interpretation, and the narrowest bottom band is a fact you can go and check.](/figures/worry-to-checkable-fact-en.svg)

Sometimes the fact has already moved and you know months earlier than you would have. Sometimes nothing has changed, and what you've been carrying is a story you built yourself, which you can set down. Both beat another lap around the inside of your head.