The term premium returns: bear steepener risk in US Treasuries through 2026

What the term premium is and why it went negative #

The term premium is the compensation a marginal investor demands for holding a long dated nominal bond rather than rolling short bills. It is not directly observable. The two workhorse decompositions are the Adrian, Crump, and Moench (ACM) five factor model from the Federal Reserve Bank of New York and the Kim and Wright three factor model from the Federal Reserve Board. Both split the 10 year nominal yield into a path of expected short rates and a residual term premium.

Across 2017 to 2022 the ACM 10 year term premium printed persistently negative, reaching approximately minus 150 basis points in March 2020. Negative readings reflected Fed quantitative easing that had absorbed roughly 5.7 trillion of Treasuries and agency MBS at the SOMA peak in April 2022, pension and life insurer duration hedging, and a reach for yield environment in which long bonds served as a recession hedge given negative stock bond correlation. Kim and Wright tracked the same regime.

Late 2023 broke the regime. The ACM 10 year term premium crossed back into positive territory in the autumn and reached roughly plus 50 basis points in October 2023, contemporaneous with the spike in 10 year yields above 5 percent. The premium has since oscillated between modestly positive and modestly negative readings, but has not returned to the deeply negative pre 2023 regime. The BIS Quarterly Review in March 2024 attributed the shift to larger projected Treasury supply, an eroded safe asset premium, and recalibration of inflation risk.

Quantitative tightening and the duration drain #

The Federal Reserve has been letting its SOMA portfolio roll off since June 2022. The original caps of 30 billion in Treasuries and 17.5 billion in agency MBS doubled to 60 billion and 35 billion in September 2022. On May 1, 2024 the FOMC cut the Treasury runoff cap from 60 billion to 25 billion per month while leaving the agency cap at 35 billion. As of the H.4.1 release in April 2026, total securities held outright sit roughly 1.7 trillion below the April 2022 peak, with Treasuries the bulk of the decline.

QT operates on the term premium through the duration channel. When the Fed lets a 10 year note roll off without reinvesting, the private sector absorbs that duration. Federal Reserve research, including Bonis, Ihrig, and Wei staff papers, has estimated that each 100 billion reduction in SOMA Treasuries lifts the 10 year term premium by roughly 4 to 5 basis points in steady state. Applying that elasticity to the cumulative runoff implies a contribution of around 60 to 90 basis points relative to a flat portfolio counterfactual, directionally consistent with the timing of the 2023 regime shift.

The runoff is on track to continue through 2026 unless the Fed announces an end date. Reserves remain ample by Federal Reserve Bank of New York reserve demand elasticity estimates, but the gap is narrowing. The 2019 repo episode is the cautionary precedent. Markets are watching the federal funds effective rate against IORB and Standing Repo Facility usage as the primary signals.

Treasury supply, the bills share, and the TBAC band #

TBAC guidance has long pointed to a target range of 15 to 20 percent for bills as a share of marketable Treasury debt. Through 2024 and 2025 the bills share consistently exceeded that band. Per the latest Monthly Statement of the Public Debt, bills accounted for approximately 22 percent of marketable debt. The overweight has drawn pointed attention, including a widely circulated Hudson Bay Capital paper in mid 2024 that coined the term Activist Treasury Issuance.

The mechanical question is what happens when bills normalize back toward the band. Each percentage point shift from bills toward coupons over two years would push roughly 270 to 290 billion of duration onto the auction calendar, given marketable debt near 27 trillion. A return to the midpoint of 17.5 percent implies roughly 1.2 trillion of incremental coupon issuance versus the current mix. That supply has to clear at a price, and the price is term premium. The May and August 2026 quarterly refunding statements are the key calendar dates.

Cyclical demand counterweights are real but bounded. Money market fund assets exceed 7 trillion. Stablecoin reserves backed by bills and Treasury repo crossed 220 billion in early 2026 under the GENIUS Act regime. Bank Treasury holdings have stabilized after the 2022 to 2023 unrealized loss episode. None of these channels naturally absorb 10 year and 30 year duration. They take bills.

Foreign holdings, the basis trade, and the marginal buyer problem #

Foreign holders of Treasuries totaled approximately 8.5 trillion in the most recent TIC release covering February 2026. Japan retains the top slot at roughly 1.10 trillion. The UK and China are clustered close behind at roughly 760 billion and 770 billion. The China figure understates true Chinese ownership given migration into Euroclear custody routed through Belgium and into agency MBS, but the direction is unambiguous. Reported China holdings are down from a 1.32 trillion peak in November 2013, and Japan has been broadly flat in nominal terms since 2014.

The marginal foreign buyer is no longer the central bank. It is the leveraged real money investor, including Japanese life insurers running cross currency hedged allocations, European bank Treasury desks, and Cayman domiciled hedge funds. The hedge fund channel is dominated by the Treasury cash futures basis trade, in which a long deliverable cash note is funded in tri party repo and hedged with a short CME futures position. Federal Reserve Bank of New York Liberty Street Economics and staff reports have sized this trade at roughly 1 trillion of notional with leverage near 50 to 1, concentrated in a small number of relative value funds. The OFR and BIS have issued parallel warnings.

The basis trade is a duration sponge in normal times and an accelerant in stress. The March 2020 dislocation, in which the on the run 10 year yield gapped 60 basis points wider in 48 hours, was driven in part by forced unwinds when repo haircuts widened and futures margins were called. A repeat under a bear steepening backdrop, with the long end already under supply pressure, is the canonical tail risk today.

Bear steepener mechanics and the historical playbook #

A bear steepener is a yield curve move in which long rates rise faster than short rates, lifting the 2 year to 10 year and 10 year to 30 year curve slopes. It is the curve signature of supply shock, term premium repricing, or a credible end of cutting cycle. It contrasts with a bull steepener, in which short rates fall faster than long rates as the Fed cuts. The bear steepener channels into the real economy through three rails. Mortgage rates, which key off the 10 year and the 30 year MBS option adjusted spread, climbed back above 7 percent in the Freddie Mac Primary Mortgage Market Survey during the autumn 2023 episode. Investment grade and high yield corporate spreads widen as duration repricing forces total return funds to derisk. Equity duration, the long dated growth stock segment of the Russell 1000 Growth and Nasdaq 100, derates as discount rates climb, replaying the dynamic of the 2022 drawdown.

Historical bear steepeners are a small dataset. The 1981 episode, anchored by Volcker disinflation and a federal deficit running at 2.5 percent of GDP, saw the 10 year yield reach 15.84 percent in September 1981 according to Federal Reserve H.15. The 1994 episode, the canonical bond market vigilante moment, lifted 10 year yields from roughly 5.6 percent in October 1993 to over 8 percent by November 1994, with a peak to trough Bloomberg Aggregate drawdown above 5 percent. The 2003 to 2004 episode, in which the 10 year sold off from 3.1 percent in June 2003 to over 4.6 percent within a year, was driven in part by the carry unwind tied to convexity hedging by Fannie Mae and Freddie Mac.

The 2026 setup rhymes with elements of 1994 and 2003 rather than 1981. Inflation is closer to target than in 1981, but the deficit and supply backdrop are heavier. Convexity hedging mechanics still exist in the agency MBS book and in callable corporate bonds, although Dodd Frank constrained the GSEs. Bond market vigilantism, a market discipline mechanism that imposes a fiscal cost on policy without congressional action, has been visible at the long end during gilt and Japanese government bond stress episodes since 2022 and is the political economy backdrop to any 2026 repricing.

Fiscal trajectory and what tips the regime #

The Congressional Budget Office Long Term Budget Outlook published in 2025 projects that net interest outlays will reach approximately 4.1 percent of GDP by 2034, surpassing total Medicare outlays around the same horizon and rising further over the following decade. Federal debt held by the public is projected to climb from roughly 99 percent of GDP at the end of fiscal 2025 toward 122 percent of GDP by 2035 under current law. The CBO baseline is a current law projection and does not assume tax cut extensions or new spending. The Penn Wharton Budget Model and the Committee for a Responsible Federal Budget have published parallel projections that broadly match the CBO baseline within a band of plus or minus 5 percent of GDP at the 2035 horizon.

Three observable triggers would tip a contained term premium drift into a bear steepener. First, a quarterly refunding announcement that lifts coupon issuance materially above current dealer expectations, particularly at the 10 year and 30 year tenor, would force a duration mark to market. Second, a clear signal that QT continues into the second half of 2026 against deteriorating money market plumbing would test the floor system and could force a return to outright Treasury purchases under stress, an outcome that would itself be inflationary on signaling grounds. Third, a basis trade unwind, triggered by an idiosyncratic hedge fund deleveraging or a tri party repo haircut shock, would create forced selling at the long end. Any single trigger could be absorbed. The combination of two simultaneously is the tail.

For corporate treasurers, the operational implication is to extend liability duration where possible at current rates and to layer interest rate hedges on floating rate exposure ahead of the May 2026 refunding. For asset allocators, the implication is to demand more spread for adding 10 year and 30 year duration and to size for the convex tail rather than the central path. For policy makers, the implication is that the era in which the long end was insensitive to fiscal news has ended. The term premium is back in play, and supply is now priced.

Sources #

• Federal Reserve Bank of New York, Adrian Crump Moench Treasury term premia
• Federal Reserve Board, Kim and Wright three factor term structure model data
• FRED, 10 Year Treasury Constant Maturity Rate (DGS10)
• Federal Reserve, FOMC statement on May 1 2024 reducing the Treasury runoff cap to 25 billion
• Federal Reserve H.4.1, Factors Affecting Reserve Balances
• US Department of the Treasury, Quarterly Refunding Statements and TBAC minutes
• US Department of the Treasury, Monthly Statement of the Public Debt
• US Department of the Treasury, Treasury International Capital data on foreign holdings
• Federal Reserve Bank of New York, Liberty Street Economics on the Treasury cash futures basis trade
• Office of Financial Research, hedge fund Treasury exposure analysis
• Bank for International Settlements Quarterly Review, term premia and Treasury supply
• International Monetary Fund, Global Financial Stability Report
• Congressional Budget Office, Long Term Budget Outlook
• Federal Reserve Bank of New York staff reports on QT and the term premium (Bonis, Ihrig, Wei)
• Freddie Mac Primary Mortgage Market Survey
• Federal Reserve H.15 Selected Interest Rates historical data
• Federal Open Market Committee Summary of Economic Projections
• Hudson Bay Capital, Activist Treasury Issuance paper, July 2024