The New York Fed's ACM model put the 10-year term premium at 0.89% on Monday, August 17, 2026, the highest daily print anywhere in its 2026 series. The Federal Reserve Board's own Kim-Wright model, fed by the same Treasury curve, ends its published run on Friday, August 14 at 0.84%, against ACM's 0.85% for that identical settlement date. Neither figure was observed by anybody. Both were manufactured by a statistical model out of yields that were observed, and across 9,141 shared trading days since Tuesday, January 2, 1990 those two models have disagreed on the sign of the number 1,534 times.
Kevin Warsh delivers his first Jackson Hole keynote as Fed chair on Friday, August 28 at 14:00 UTC, nineteen days ahead of the September 16 FOMC decision, and the argument he has spent a decade making is an argument about that unobservable number.
What the Term Premium Actually Pays You For
A 10-year Treasury yield contains two things that no amount of staring at the yield will separate. The first is the average overnight rate the market expects the Fed to set across the coming decade. The second is the extra compensation investors demand for committing money for ten years instead of rolling three-month bills forty times and keeping the option to walk away at every step.
That second piece is the term premium, and in plain English it is the price of not being able to change your mind. With the 10-year settling at 4.69% on Thursday, August 20, the ACM model attributes roughly 3.87 points of that to the expected policy path and the remaining 82 basis points to the fee investors charged for the lock-up.
It can also go negative, because investors repeatedly pay up for the certainty of a locked rate, or for an asset that rallies while everything else falls apart. On the New York Fed's ACM term premia series, 2,146 of 16,261 daily prints since Wednesday, June 14, 1961 sit below zero, and the low is negative 1.65% on Monday, March 9, 2020.
The term premium is therefore the slice of a long yield that has nothing to do with what the Fed is expected to do, which makes it the slice most traders never look at.
The Number Is Modelled, and Two Fed Models Do Not Agree
Nobody can measure the term premium, and the charts never say so out loud. It is the residual left after you subtract an estimate of expected future short rates from an observed yield, and that estimate comes out of a term structure model with assumptions baked into it. Two such models are published by the Federal Reserve System itself, and they do not agree.
|
Settlement date
|
NY Fed ACM
|
Fed Board Kim-Wright
|
Kim-Wright minus ACM
|
|
Tuesday, August 4
|
0.75%
|
0.82%
|
+7 bps
|
|
Monday, August 10
|
0.82%
|
0.85%
|
+3 bps
|
|
Wednesday, August 12
|
0.83%
|
0.84%
|
+2 bps
|
|
Thursday, August 13
|
0.80%
|
0.82%
|
+2 bps
|
|
Friday, August 14
|
0.85%
|
0.84%
|
negative 1 bp
|
Both series were pulled from source at 08:41 UTC on Monday, August 24, 2026, and they do not end on the same day. ACM runs through Thursday, August 20 and carries 0.89% for August 17, 0.77% for August 19 and 0.82% for August 20. The Board's three-factor Kim-Wright model stops at August 14.
August flatters the comparison. Across the full shared history the median absolute gap is 47 basis points and the widest single-day disagreement is 236 basis points on Monday, November 3, 2008, while 2026 has run unusually tight at an average of 9 basis points and a maximum of 25.
The cleanest illustration of what "modelled" means comes from two ordinary 2026 sessions. The 10-year par yield settled at exactly 4.40% on Thursday, April 30 and again on Thursday, June 25, same instrument and same country. ACM split that yield into a 0.72% term premium on the April date and 0.47% on the June date, a 25 basis point difference in a number nobody could see, on two days when the number everybody could see was identical.
Which gives you a rule worth keeping. Never quote a term premium without naming the model that produced it, because a figure with no provider attached is not data.
The Argument Warsh Has Been Making Since Before He Chaired Anything
The Kansas City Fed hosts its Jackson Hole Economic Policy Symposium from Thursday, August 27 to Saturday, August 29, and the published theme is "Financial Innovation: Implications for Payments and Policy," the first time the symposium has put payments and financial technology at the centre of its programme. Warsh takes the podium on the Friday at 10:00 ET, which is 14:00 UTC.
One disambiguation belongs here because search results merge the two constantly. The Wyoming Blockchain Symposium held August 17 to 20 is a separate event with separate organisers, unconnected to the Kansas City Fed.
Warsh's long-standing position is that the Fed's Treasury portfolio suppressed the term premium and that shrinking the portfolio means allowing it back. He has never framed that as a forecast, but as arithmetic about who is bidding for duration, because if the largest price-insensitive buyer steps back the price of duration gets set by buyers who care about price.
His language as chair points at the same channel. At the July 29, 2026 FOMC press conference he said the balance sheet "probably works through some other channels, like signaling and portfolio balance," and told reporters the Committee had taken up the question of "how much accommodation are we getting from the balance sheet."
Portfolio balance is the term premium channel wearing an academic name, the mechanism by which buying bonds pushes investors out of duration and compresses what they can charge for holding it. He has also stood up a Balance Sheet Policy task force co-led by Karen Dynan, Raghuram Rajan and Jeremy Stein.
And that is the whole of what can responsibly be said before the keynote, because no published text exists and no outcome is tradable. The background on what Warsh's arrival at the Fed meant for crypto is worth more than any preview of a speech nobody has read.
For anyone whose search results keep returning 2025 material, Warsh chairs this committee and Jerome Powell holds a plain governor's seat. Powell voted with the majority to hold in July, a 9-3 decision in which Beth Hammack, Neel Kashkari and Lorie Logan each preferred a quarter-point increase.
Why Crypto Sits at the Far End of the Duration Curve
Duration describes how violently a price moves when the discount rate moves. The further out an asset's payoff sits, the more a change in that rate compounds against it, which is why a 30-year bond swings on news that leaves a two-year note untouched.
Bitcoin pays no coupon and has no maturity, so every dollar of its value is a claim on a distant state of the world rather than on cash arriving next quarter. Applied literally the label does not fit, because there are no cash flows to discount, but practically it fits better than it does to anything else in a retail portfolio, including flows into a spot Bitcoin ETFwhose buyers are making a decade-length allocation decision.
Put the two ideas together. The discount rate on a long-duration asset is a policy path plus a term premium, and traders watch the first half obsessively while ignoring the second. A cut priced into the Fed's projection materials lowers the expected-rates half and does nothing at all to the term premium half, which bond investors set for themselves in an era of large deficits and a shrinking central bank portfolio.
That is why a rising term premium is the genuinely unpleasant version of rising yields for crypto. Rates rising because growth is strong arrive bundled with a reason to own risk. Rates rising because investors want more compensation for duration arrive bundled with nothing, and they lift the discount rate on every asset valued off distant outcomes at once.
The counterweight matters just as much. In a piece published on the Phemex blog on Monday, August 24 we walked the Treasury curve through the week of August 17 and found it flattened across the same three sessions Bitcoin ran roughly 21%, with 10s30s narrowing from 59 to 53 basis points and 2s10s from 53 to 50. A handful of basis points cannot explain a 21% move, so treat the term premium as a regime variable rather than a session-level signal.
What 2026 Did to Each Half of the Yield
The two halves went in almost completely different directions across 2026.
|
Measure
|
Friday, January 2
|
Thursday, August 20
|
Change
|
|
2-year par yield
|
3.47%
|
4.19%
|
+72 bps
|
|
10-year par yield
|
4.19%
|
4.69%
|
+50 bps
|
|
30-year par yield
|
4.86%
|
5.23%
|
+37 bps
|
|
ACM 10-year term premium
|
0.80%
|
0.82%
|
+2 bps
|
Yields come from the Treasury Department's daily par yield curve series. Every figure in this article is a settle rather than an intraday print, which matters because a "highest since" headline is almost always quoting an intraday extreme that never closed.
Read that table the way a bond desk would. The long end rose least, the front end rose most, and the modelled term premium barely moved. By ACM's own decomposition the 2026 move in long Treasury yields is a repricing of what the Fed will do rather than a repricing of duration risk, so the thing Warsh has argued should happen has not yet shown up in the data.
One correction, because the wrong version keeps circulating. The 10-year and 30-year did not both hit 20-year highs, and only the 30-year did. Its 5.31% settle on Monday, August 17 is the highest 30-year par-yield close since 5.35% on Tuesday, June 12, 2007, checked against every annual maximum from 2008 through 2025, the strongest of which was 5.11% on Thursday, October 19, 2023. The 10-year is nowhere close, with a Friday, August 21 settle of 4.74% against 4.98% on that same October 2023 date.
Frequently Asked Questions
What is a normal level for the term premium?
Higher than anything printed in the last decade. ACM's daily mean from 1961 through 2007, before quantitative easing existed, is 1.84%, against 0.82% for Thursday, August 20, 2026, so anyone calling the 2026 reading extreme is measuring it against the QE era rather than against history.
Why did the term premium turn positive again after years below zero?
Deficit-driven Treasury supply, a central bank that stopped being an automatic buyer, and inflation that no longer behaves like a settled question all push the same way. ACM's last negative daily print was Wednesday, October 16, 2024, and the 2023 episode widely described as the return of term premium topped out at 0.465% on Friday, October 20 of that year.
Does a higher term premium always hurt Bitcoin?
No, and treating it as a trading trigger is the fastest way to lose money on a good idea. It raises the discount rate on every long-duration asset, which is a headwind in the same sense that gravity is a headwind for a plane, and liquidity, positioning and flows routinely overwhelm it across any horizon a trader actually holds.
How do I check the term premium myself?
The New York Fed publishes ACM as a spreadsheet on its Treasury term premia page and the Federal Reserve Board publishes Kim-Wright as a CSV. Compare the same date on each before trusting any figure quoted without a provider attached.
Bottom Line
Watch the split rather than the yield. If the 30-year climbs while the 2-year sits still, duration is being repriced and that is the version which raises the discount rate under crypto. If both legs rise together the market is repricing the policy path, and the 2-year at 4.24% from Friday, August 21 already sits about 62 basis points above the midpoint of the 3.50% to 3.75% target range.
Three dated events decide which shape shows up. Core PCE and the second estimate of Q2 GDP both land at 12:30 UTC on Wednesday, August 26, the Jackson Hole keynote follows at 14:00 UTC on Friday, August 28, and the FOMC decision lands on Wednesday, September 16.
One structural warning to carry out of this. Both series publish with a lag of days, so the decomposition of any move arrives after the move, and you will never get the reading that explains a selloff until the selloff has been paid for. That is why this belongs in the part of your process that decides how much to own rather than the part that decides when to click.
This article is for educational purposes only and does not constitute financial or investment advice. Cryptocurrency trading involves substantial risk. Always conduct your own research before making trading decisions.
