The Treasury Signal That Lasted One Session

The announced change is at least $14 billion of incremental capacity, not $4 billion. Seven scheduled operations fall on or after 9 September in the affected sectors — four in the 10-to-20-year bucket and three in the 20-to-30-year bucket. Treasury increased the per-operation maximum from $2 billion to at least $4 billion, so at a minimum $2 billion of additional capacity each, the increment is at least $14 billion. Every figure derived from it is a floor, not a point estimate.

The policy change is a reallocation, and the figure describing it has not been published. Within the August refunding quarter, long-end liquidity support capacity rises from $16 billion to at least $30 billion while total liquidity support capacity rises from $38 billion to at least $52 billion. The long end’s share of the programme rises from 42.1% to at least 57.7%, assuming the currently published schedule remains otherwise unchanged.

This is not quantitative easing, on statutory, mechanical and institutional grounds. The two operations derive from different statutes, are decided by different bodies, are funded differently, and end differently: the Federal Reserve holds what it buys as an asset; Treasury retires it. Neither eliminates the government’s borrowing requirement — QE leaves it untouched, while Treasury buybacks must be accommodated within it.

But the scale argument does not support that conclusion — and is frequently stated backwards. At the peak of QE3, the Desk scheduled long-end operations with an expected size of $1.25–$1.75 billion. On a minimum-capacity basis, Treasury’s $4 billion figure is 2.3 to 3.2 times that range, and its minimum long-end capacity is approximately 83% of the midpoint of QE3’s peak monthly long-end pace.

The Federal Reserve is expanding its balance sheet, not shrinking it. Runoff concluded and reserve management purchases began in December 2025. Roughly $250 billion of Treasury bills were purchased between January and mid-2026. This materially loosens the front-end absorption constraint that most commentary treats as binding.

The first observable test was unfavourable. The 20-year auction cleared later that afternoon, nine basis points into the resulting rally, and still came soft: 2.53x bid-to-cover against a 2.64x average, 65.58% allotted at the high, and dealers left with 12.49% against a 10.82% average. The announcement moved the secondary market without visibly improving primary demand.

The credible risk is that the announcement raises term premium rather than lowering it. The change was made off-cycle, two weeks after the schedule had been published. Part of the premium embedded in Treasury yields compensates for the expectation of not being surprised.

At approximately 8:40 a.m. Eastern on Wednesday 19 August 2026 — before the equity open, and one day after the 30-year Treasury yield reached its highest level since 2007 — the U.S. Department of the Treasury issued a four-paragraph press release.1

Treasury is increasing, by at least double, the maximum size of liquidity support buyback operations in two nominal coupon sectors: 10-to-20-year and 20-to-30-year. Treasury increased the per-operation maximum from $2 billion to at least $4 billion; the announcement states a floor on the new maximum, not the maximum itself. The change takes effect 9 September 2026 and runs through 4 November 2026, the date of the next Quarterly Refunding. Treasury attributed the change to consistent strong sponsorship from market participants, evidenced by the volume of high-quality offers received in longer-dated operations. An updated tentative schedule was promised separately.1

That is the entirety of the official record. Everything else written about this announcement — including this report — is inference, and is labelled as such.

Reporting on 19 August circulated three different totals — $63 billion, $69 billion and $83 billion — without reconciling them. All three are correct. They measure different windows. The reconciliation is where the actual policy change becomes visible.

The 5 August tentative schedule lists twenty operations between 6 August and 5 November 2026. Two of them — 6 August (1-month-to-2-year, $4 billion) and 11 August (10-to-20-year, $2 billion) — are footnoted on the schedule as falling within the May 2026 refunding quarter. That $6 billion is the entire difference between the $63 billion and $69 billion figures.3

Table 1. Buyback capacity reconciliation. CALCULATION, derived by summing the twenty operations on the published 5 August schedule and independently verified at four checkpoints. Post-announcement figures are floors: Treasury set the new per-operation ceiling at “at least” $4 billion, so each assumes the minimum and assumes the schedule is otherwise unchanged.

The $38 billion liquidity support figure ties exactly to the Quarterly Refunding Statement, which anticipated up to $38 billion for liquidity support and up to $25 billion in the 1-month-to-2-year bucket for cash management.5 The $69 billion figure used in wire reporting is the same schedule measured across the full calendar window.4

Within hours the announcement was being described as "mini QE." The claim is testable against the documented record of both institutions. It fails — but not for the reason usually given, and the usual reason is stated backwards.

Table 2. FACT. Federal Reserve asset purchase programmes. Source: Federal Reserve Bank of New York, Large-Scale Asset Purchases archive, and Board of Governors balance sheet policy timeline. Announced and executed figures differ where the FOMC revised a programme in flight — agency debt purchases under LSAP1 were reduced from $200bn to $175bn on 4 November 2009 owing to limited availability of eligible securities.

Three details in that table are frequently misreported and are worth isolating. First, the Maturity Extension Program was announced at $400 billion but executed at $667 billion of purchases following its June 2012 extension.6 Second, LSAP3 began in September 2012 as an MBS-only programme; Treasury purchases were added only from January 2013, after the MEP concluded.6 Third, the pandemic programme was the only one ever announced without a cap: on 23 March 2020 the FOMC directed the Desk to purchase in the amounts needed to support smooth market functioning, removing all volume limits.8

For completeness of the cycle: tapering of the pandemic programme was announced on 3 November 2021 (a reduction of $10bn Treasuries and $5bn MBS per month) and doubled on 15 December 2021.11 Balance sheet reduction began on 1 June 2022 under monthly caps of $30bn Treasuries and $17.5bn agency debt and MBS, rising to $60bn and $35bn from September 2022.12

Table 3. FACT, except the final row (each institution’s own stated objective). Sources as cited in Sections 3.2 and 3.3.

Two rows carry most of the weight, and both come from Treasury’s own operational documentation rather than from commentary. First, repurchased securities are retired on settlement.17 The Federal Reserve warehouses what it buys; Treasury extinguishes it. There is nothing to unwind. Second, Treasury’s Office of Debt Management states that amounts spent on buybacks are treated like any other source of borrowing needs.17 A Treasury buyback must therefore be accommodated within Treasury’s own financing requirement.

The distinction is not that quantitative easing eliminates that requirement. It does not, and the Federal Reserve is explicit on the point: purchases of Treasury securities from the public are not a means of financing the federal deficit, and the Federal Reserve does not participate in competitive bidding at Treasury auctions.15 What Federal Reserve purchases do is transform privately held Treasury securities into reserve liabilities on the Federal Reserve’s balance sheet, transferring duration risk from the private sector to the central bank while leaving the government’s borrowing requirement untouched. Treasury buybacks retire the securities purchased and must themselves be funded. Neither operation makes the deficit disappear. They differ in what the public ends up holding, and in which balance sheet absorbs the duration.

Neither operation constitutes direct monetary financing of Treasury. The Federal Reserve is barred by statute from purchasing directly from Treasury,13 and Treasury’s repurchases are funded from its own borrowing rather than from money creation.14

A definitional note on the debt-held-by-the-public row. Federal Reserve holdings are counted within debt held by the public, which Treasury defines as federal debt held by individuals, corporations, state and local governments, Federal Reserve Banks, foreign governments and other entities outside the U.S. Government.16 Quantitative easing therefore leaves that measure unchanged and reduces only the privately held share. Commentary describing QE as retiring or absorbing public debt is using the term loosely.

Table 4. CALCULATION. Fed figures are expected purchase sizes from published Desk schedules; Treasury figures are par-amount ceilings from the 5 August schedule. See the qualifications below.

On a minimum-capacity basis, Treasury’s $4 billion ceiling is 2.3 to 3.2 times the Federal Reserve’s $1.25–$1.75 billion expected long-end operation range at the peak of QE3, and 3.6 to 4.7 times the $0.85–$1.10 billion range the Desk was scheduling by May 2014. The two quantities are not like for like and the ratio is stated as a range rather than a midpoint for that reason: Treasury has published a floor on a ceiling, while the Federal Reserve published an expected purchase size it then executed. On the published schedules, Treasury’s minimum long-end capacity of $30 billion across the refunding quarter is approximately $10 billion a month, or about 83% of the midpoint of LSAP3’s peak monthly long-end purchase pace. Were Treasury to set the ceiling at $5 billion rather than $4 billion, the long-end pace would exceed LSAP3’s peak outright. The comparison demonstrates why operation size alone cannot determine whether an asset-purchase programme constitutes quantitative easing.

Two qualifications, both running against this report’s own argument. The Fed’s 2036–2043 bucket represented 23 to 30 years remaining in 2013, not a precise match for Treasury’s 20-to-30-year bucket. And Federal Reserve figures are expected purchase sizes the Desk executed, whereas Treasury’s are par-amount ceilings with a zero floor. Both qualifications make the Federal Reserve larger than Table 4 implies.

For the outer bound: on 13 March 2020 the Desk conducted two operations in the 20-to-30-year sector on the same day, of approximately $4 billion each — $8 billion before the close.9 Ten days later it was buying roughly $75 billion of Treasuries across all sectors every business day.8 Treasury will conduct three $4 billion operations in that sector over an entire quarter. A single Wednesday in March 2020 was about three and a half weeks of Treasury’s new maximum pace.

This is not an improvised instrument. Between March 2000 and April 2002, Treasury conducted 45 buyback operations retiring $67.5 billion of outstanding debt — nearly five times the increment announced last week. Then-Secretary Summers gave three rationales at adoption: enhancing the liquidity of benchmark securities; preventing what would otherwise be a costly and unjustified increase in the average maturity of the debt; and making more effective use of excess cash.21

Note the second rationale. Maturity management has been an explicitly stated objective of Treasury buybacks since 2000. Anyone arguing that duration management is a novel or covert motive in 2026 is arguing against Treasury’s published position of twenty-six years’ standing. No buyback operations were conducted between 2003 and 2013; the programme resumed on a test basis in 2014 and as a regular programme in 2024.22

The comparison also cuts against Treasury. The 2000–02 operations were financed by budget surpluses and genuinely reduced debt held by the public. The 2026 operations are financed by new issuance and reduce nothing. The instrument is the same; the fiscal position it operates within has inverted.

Treasury must fund these repurchases. If it funds them by issuing bills, the operation shortens the weighted average maturity of marketable debt and transfers duration risk from long-end books to the front end. Ryan Swift of BCA Research identified the ceiling on the day: Treasury’s ability to suppress long-dated yields by shifting issuance to the front end is limited by T-bill/OIS spreads, which he described as already stretched, with a shift back toward coupons likely by early next year.23 That observation converts an open-ended intervention into a finite one.

One element of that analysis does not survive contact with the Federal Reserve’s own data, and the error runs in Treasury’s favour. Swift suggested the Fed’s trend is toward shrinking its balance sheet. It is not, and has not been since last year. Runoff concluded after securities holdings had fallen by more than $2.2 trillion since June 2022.24 In December 2025 the FOMC judged reserves had reached ample levels and directed the Desk to begin reserve management purchases.25 The July 2026 implementation note directs the Desk, when appropriate, to increase SOMA holdings through purchases of Treasury bills and, if needed, Treasury securities with three years or less remaining.26

Per the Federal Reserve’s July 2026 Monetary Policy Report, the SOMA portfolio purchased nearly $250 billion in Treasury bills between early January and mid-2026 — roughly $160 billion of reserve management purchases and $90 billion of agency reinvestment — expanding Federal Reserve assets by about $150 billion to $6.7 trillion.27

A concrete illustration from 19 August. The Federal Reserve took a $2.06bn non-competitive add-on at that afternoon’s 20-year auction, rolling over maturing principal in line with the standing directive to roll all Treasury principal payments at auction.2634 Reserve management purchases are confined to bills and short coupons, but rollovers are not: they are allocated in proportion to what is being issued on the maturity date, so the Federal Reserve continues to take down long-dated paper passively. This does not make the Fed an active buyer of duration, and it should not be described as one — but it does mean SOMA is not absent from the long end.

For the official rationale. Michael Lorizio of Manulife Investment Management noted that the Treasury Borrowing Advisory Committee had previously advised that liquidity operations at the very back end of the curve had room to be increased, and that this advice predated the current level of yields.23 TBAC’s Q3 2025 charge flagged higher offer-to-max ratios in long-end buybacks and recommended weighing changes to maximum purchase amounts against the effect on weighted average maturity.29

Against it. Swift’s first observation is that the data do not indicate that rising long-maturity yields were driven by deteriorating liquidity.23 Were liquidity the constraint, distress would appear as widening bid-ask and off-the-run spreads. It has instead appeared as a rising term premium — a price phenomenon, not a plumbing one. Tuesday’s $2 billion operation is the cleanest available test: liquidity was supplied on schedule and the selloff continued.4

Both are true. Treasury had a standing, TBAC-endorsed case for larger long-end operations, and chose to execute it the morning after a nineteen-year yield high. The pre-existing justification is real; the timing is not a coincidence; and the second fact does not falsify the first.

The strongest institutional argument against this announcement is not that it will fail to lower yields, but that it may raise them. Thomas Simons of Jefferies argued that Treasury’s decades-long commitment to predictable communication is itself a source of value, and that an off-schedule change undermines it — short-sighted, on his reading, if the objective was to compress term premium, because part of that premium reflects the market’s expectation of not being surprised.23

This is a testable proposition with a clear mechanism. Term premium compensates investors for uncertainty about future short rates and future supply. If Treasury has demonstrated that its published schedules are revisable mid-quarter in response to price action, every future schedule carries an uncertainty that did not exist on 18 August. The nine-basis-point rally is the visible effect; a wider distribution of expectations around future issuance is the invisible one, and it works in the opposite direction.

Treasury’s own documentation complicates the "reaction function" reading. Both the buyback FAQ and the Office of Debt Management’s programme details state that Treasury does not intend to use buybacks to respond to episodes of acute market stress.3017 Any argument that Wednesday revealed a willingness to intervene against disorderly conditions must be made against that stated position, not in ignorance of it. Joseph Purtell of Neuberger Berman framed the market’s inference correctly — a soft line in the sand, with the open question of whether an extra $2 billion per operation was worth nine basis points against the supply-demand mismatch that created the problem.23

On 3 August, Treasury estimated privately-held net marketable borrowing of $739 billion for July–September — $68 billion above its May projection, and $87 billion above it excluding the higher-than-assumed opening cash balance — assuming an end-September cash balance of $950 billion. It projected a further $628 billion for October–December.31

On 18 August, gross federal debt outstanding reached $40,047,425,768,420.22, with debt held by the public at approximately $32.2 trillion and intragovernmental holdings near $7.8 trillion.32

Treasury will raise more than $1.36 trillion in privately-held net marketable debt over the second half of the calendar year. Against that, it has expanded maximum repurchase capacity by at least $14 billion — roughly 1% of the financing requirement — using an instrument that, by Treasury’s own accounting, adds to that requirement rather than reducing it. Anshul Sharma of Savvy Wealth reached the proportionate conclusion: the move does not resolve deficits, inflation or supply, but demonstrates that Treasury has tools it is willing to use.4

Table 5. FACT. Sources: Treasury SB0607; Debt to the Penny; TreasuryDirect auction records; FOMC minutes of 28–29 July 2026; market levels per Reuters and CNBC.

Table 6. FACT. Bidder shares are percentages of total competitive accepted ($15,809,192,000) and sum to exactly 100%. Source: Treasury auction results release for CUSIP 912810UX4, except the twelve-auction comparators (Helious, enriched from TreasuryDirect records) and the when-issued level (secondary-market quote).

The announcement’s most measurable effect on the auction was to keep it off the record. When-issued levels had indicated around 5.27% in the days before the sale, which would have set an all-time high for the tenor. The morning announcement rallied the sector by roughly eight to nine basis points, and the auction cleared at 5.204% — 4.1bp inside the October 2023 record of 5.245%.41 Pre-auction commentary that described the sale as likely to set a record was reporting a conditional projection, not an outcome; that projection was overtaken by Treasury’s own intervention earlier the same day. The distinction matters, because the two readings support opposite conclusions about how the auction went.

On the two size figures in circulation. Both are correct under different definitions, and the official release reconciles them exactly: $16,000,036,500 was accepted from the public, the Federal Reserve added $2,056,891,600 as a rollover, and total accepted was $18,056,928,100.35 Reporting that cites $16 billion refers to the public offering; reporting that cites $18 billion refers to the total including the SOMA add-on. Any analysis comparing this auction to prior ones must hold the definition constant, since the SOMA component varies with what is maturing in the Federal Reserve’s portfolio and has nothing to do with private demand.

A terminological caution for readers of secondary coverage. The 62.93% figure is the indirect bidder share, which includes foreign central banks and domestic institutions bidding through intermediaries. It is routinely and incorrectly reported as the "foreign" share. The two are not the same, and the distinction matters when the number is used to argue about overseas appetite for U.S. duration.

Read as a unit: the fiscal authority eased financial conditions at 8:40 in the morning, and the monetary authority disclosed a little over five hours later that several of its members wanted to tighten. The dollar fell, gold rose roughly 3%, and bitcoin traded to an intraday high of $69,749 — a hawkish central bank record landing on a market that had spent the session repricing toward easier conditions, and being largely ignored.334443 Brian Jacobsen of Annex Wealth Management drew the implication directly: even if the Fed hikes, Treasury is pushing more money-like short-term debt into the system.23

The initial relief proved short-lived. The 30-year yield, which closed at 5.196% on 19 August, rose more than seven basis points on 20 August to as much as 5.27% — the level at which it had been trading immediately before the announcement — before paring to around 5.25%. The 10-year moved back above 4.70%. On 21 August the 30-year was broadly flat at 5.2371%.38 The entire announcement effect in the 30-year was retraced within one session.

The reversal does not invalidate the liquidity rationale, and one session is not a verdict. But it is consistent with the reading advanced in Sections 6 and 7: that the announcement’s yield effect was a repricing of positioning rather than a durable change in long-end supply and demand. ING characterised the intervention as suggesting discomfort about borrowing costs and raised the prospect of repetition; JPMorgan’s Maia Crook argued that such interventions do not address the underlying structural challenges.40

Treasury’s own response to the reversal is the more consequential development. On 20 August, Secretary Bessent said the programme “could be more than $4 billion per issue,” and that Treasury has “a big toolkit.”39 That is the clearest confirmation available that $4 billion is a floor rather than a ceiling, and it vindicates the “at least” construction used throughout this report. It also moves the reaction-function question from inference to something closer to the record: a Secretary responding to a one-session reversal by publicly raising the prospect of larger operations is describing a reaction function, whatever the buyback documentation says about acute market stress.

One caution. A statement in a television interview is not a change to the published programme. Until a revised schedule or a further release specifies a larger per-operation size, the operative figures remain those in Table 1, and the analysis in this report is unchanged.

Thesis (INTERPRETATION). The 19 August announcement is a maturity-structure policy, not an explicit yield-targeting policy, and not quantitative easing. It may well suppress long-end yields through reduced duration supply and additional official demand — that is the mechanism Swift describes — but suppression as an effect is not the same as targeting as a policy. Its effect on long-end yields is temporary and its capacity is bounded by front-end absorption — though less tightly than commonly assumed, given the Federal Reserve’s ongoing bill purchases. The durable consequence is a modest increase in term premium arising from the loss of schedule predictability, partially or wholly offsetting the mechanical benefit of reduced long-end supply within one to two quarters.

Horizon. Through the 4 November 2026 Quarterly Refunding.

Confirming evidence. The 30-year returns to and sustains above 5.30% despite the enlarged operations from 9 September; the 3-month T-bill/OIS spread continues to widen; Treasury signals a shift back toward coupon issuance at or before the November refunding.

Falsifying evidence. The 30-year holds below 5.00% through October with stable or narrowing T-bill/OIS spreads and no deterioration in auction internals; or Treasury conducts the enlarged operations without incident and November formalises the change on schedule, restoring the predictability premium.

What would break the framework entirely. An FOMC decision to extend reserve management purchases beyond bills and short coupons into the long end. That is the point at which the QE question stops being definitional and becomes real. Nothing in the July 2026 implementation note contemplates it — the Desk is confined to bills and Treasuries with three years or less remaining.26

The 30-year around 5.30%. Not a ceiling — there is no evidence of a yield target and Treasury’s own documentation argues against one — but 5.337% is where the announcement was triggered, and the market now knows it.

The 3-month T-bill / 3-month OIS spread, read alongside the Desk’s monthly reserve management purchase amounts. The constraint and its offset move together.

The 10 September and 24 September operations. First tests at the new size. Watch offer-to-max ratios: if offers do not scale with the ceiling, the strong-sponsorship rationale weakens materially.

20- and 30-year auction internals, specifically dealer take-down, allotted-at-high and the high-to-median spread. The 19 August 20-year left dealers with 12.49% against a twelve-auction average of 10.82%, with indirect participation at 62.93% against a 64.87% average, 65.58% allotted at the high and a 5.7bp high-to-median spread. The next 20-year reopenings fall on 15 September and 21 October, both inside the enlarged-buyback window, with a new issue on 18 November after it closes. Further deterioration from the 19 August baseline while buybacks run larger would be the clearest evidence the programme is not working.

Weighted average maturity of marketable debt outstanding. TBAC’s own recommended metric, and the one that distinguishes liquidity support from duration management.

The 4 November Quarterly Refunding. Whether the increase is made permanent, extended, or quietly reversed will settle the interpretive question this report cannot.

Treasury did not fix the bond market on 19 August, did not claim to, and did not conduct quantitative easing. It expanded a debt-management tool that has existed under the same statute since long before QE was invented, that retires rather than accumulates what it buys, and that must be funded from Treasury’s own borrowing rather than from money creation.

But the reflexive dismissal on grounds of size is wrong, and wrong in a way that matters. In the sector it targets, Treasury’s minimum per-operation ceiling is more than twice the expected operation size the Federal Reserve used at the peak of its third asset purchase programme. The distinction between this and quantitative easing has to rest on mechanics — on reserves, on retirement, on which balance sheet absorbs the duration — and it does rest there securely. It does not rest on the dollar amount.

The $14 billion is immaterial against a $1.36 trillion financing requirement. What is not immaterial is that the long end’s minimum share of Treasury’s liquidity support programme would move from 42% to 58% under the currently published schedule, changed by a single unscheduled press release, and that the market must now price a probability it did not price on 18 August: that the published schedule is revisable when the tape gets bad. The rally it bought lasted one session. The precedent will outlast it, and whether that precedent was worth nine basis points of relief is the question the next three months will answer.

The twenty operations on the tentative schedule published 5 August 2026. LS denotes liquidity support; CM denotes cash management. Operations marked * fall within the May 2026 refunding quarter. Every operation carries a minimum purchase amount of $0.

Table A1. FACT (schedule) and CALCULATION (arrows indicating the effect of SB0607). Source: Treasury tentative buyback schedule, 5 August 2026.

Verification checkpoints — all four tie exactly:

Liquidity support operations excluding the two May-quarter operations — $38.0bn, matching the Quarterly Refunding Statement

Cash management operations (3 and 9 September) — $25.0bn, matching the Quarterly Refunding Statement

All twenty operations, 6 August to 5 November — $69.0bn, matching published wire calculations

Seven affected operations at a minimum $2bn incremental capacity each — at least $14.0bn, matching the published increment and its described composition

Table B1. CALCULATION. The 18 August operation was conducted at $2.0bn before the change took effect on 9 September and is therefore carried at the pre-announcement size in both columns. Post-announcement figures assume the minimum $4.0bn ceiling and are floors.

Every figure in this report traces to a primary source or a named market participant. The following items are qualified rather than asserted.

1. Auction internals for the 19 August 20-year bond — resolved to primary source. This item was previously carried as an outstanding gap. It is now closed. Every figure in Section 9.1 is taken from Treasury’s official results release for CUSIP 912810UX4 and corroborated against the Treasury Fiscal Data auctions record. The release is internally consistent at seven independent checkpoints, and the three bidder shares sum to exactly 100% of total competitive accepted.34

1a. A correction to an earlier draft of this report. A prior draft treated the $18 billion size figure circulating in some coverage as erroneous. That was wrong. The official release shows $16,000,036,500 accepted from the public and $18,056,928,100 accepted in total including a $2,056,891,600 SOMA add-on. Both figures are correct under different definitions; neither is an error. The earlier characterisation is withdrawn.35

1aa. A correction to an earlier draft: “highest since 2020.” An earlier draft described the 19 August high yield of 5.204% as the highest for the tenor since its 2020 reintroduction. That was wrong, and the error is instructive. The claim was inherited from pre-auction reporting, which stated that when-issued levels of approximately 5.27% would set a record if realised. They were not realised, because the buyback announcement rallied the sector before the auction cleared. The result is the second-highest on record, 4.1bp below October 2023’s 5.245%. A conditional projection published before an event must never be carried forward as a description of the outcome.41

1b. A comparison that must not be mistaken for a tail. Figures circulating on 19 August comparing "5.204% versus 5.163%" reflect an economic-calendar convention comparing the result to the prior auction (22 July, 5.163%), not to when-issued. The actual tail was 0.5bp against a 5.199% when-issued level. The 4-basis-point difference between those two comparisons is not a measure of auction weakness and must not be presented as one.42

1c. The when-issued level remains the one non-primary figure. Treasury does not publish when-issued yields, so the 0.5bp tail is necessarily derived from a secondary-market quote. Treasury’s own dispersion measures — 65.58% allotted at the high and a 5.7bp high-to-median spread — are published, are primary, and point the same way.3634

2. The updated tentative buyback schedule promised in SB0607 had not been published at the time of writing. The at-least-$14 billion increment is derived by applying the announced change to the schedule published 5 August, at the minimum permitted per-operation size, and requires confirmation against the revised schedule when issued.1

3. Reserve management purchase totals are as reported in the Federal Reserve’s July 2026 Monetary Policy Report and cover early January to mid-2026. More recent H.4.1 data may revise the running total.27

4. Announced versus executed programme sizes. Table 2 distinguishes the two throughout. Where a single figure is cited in the body text, it is the executed figure. Agency debt purchases under LSAP1 were reduced from $200 billion to $175 billion on 4 November 2009; the Maturity Extension Program was announced at $400 billion and executed at $667 billion following its June 2012 extension.6

5. Sector definitions are not identical across institutions. The Federal Reserve’s long-end purchase bucket in May 2013 (15 February 2036 to 15 May 2043) represented 23 to 30 years remaining; Treasury’s bucket is 20 to 30 years. Comparisons in Section 3.4 are close but not like-for-like, and the difference favours the Federal Reserve being larger than stated.

Superscript numerals in the text refer to the entries below. Primary sources — statute, government press releases, published operation schedules and official statistical releases — are given in full so that every claim can be independently re-verified.

1.U.S. Department of the Treasury, "Treasury Announces Increased Sizes of Nominal Long-End Liquidity Support Buybacks," press release SB0607, 19 August 2026. home.treasury.gov/news/press-releases/sb0607

2.CNBC, "30-year Treasury yield tops 5.33%, highest since 2007," 18 August 2026.

3.U.S. Department of the Treasury, "Tentative Schedule of Treasury Buyback Operations," August 2026 Quarterly Refunding, published 5 August 2026. home.treasury.gov/system/files/221/Tentative-Buyback-Schedule.pdf. All operation dates, sectors, and minimum/maximum purchase amounts used in this report are taken directly from this document.

4.Reuters, "Treasury doubles US long-bond buybacks in the face of surging yields," 19 August 2026.

5.U.S. Department of the Treasury, Quarterly Refunding Statement, press release SB0590, 5 August 2026. home.treasury.gov/news/press-releases/sb0590

6.Federal Reserve Bank of New York, "Large-Scale Asset Purchases," Programs Archive. newyorkfed.org/markets/programs-archive/large-scale-asset-purchases. Primary source for all LSAP and Maturity Extension Program figures in Section 3.1 and Appendix A.

7.Board of Governors of the Federal Reserve System, "Timeline: Balance Sheet Policies." federalreserve.gov/monetarypolicy/timeline-balance-sheet-policies.htm

8.Board of Governors, FOMC statement, 23 March 2020. federalreserve.gov/newsevents/pressreleases/monetary20200323a.htm; Federal Reserve Bank of New York, Operating Policy statement, 23 March 2020.

9.Federal Reserve Bank of New York, Operating Policy statement, 13 March 2020. newyorkfed.org/markets/opolicy/operating_policy_200313

10.Federal Reserve Bank of New York, Operating Policy statement, 10 June 2020. newyorkfed.org/markets/opolicy/operating_policy_200610

11.Federal Reserve Bank of New York, Operating Policy statements, 3 November 2021 and 15 December 2021.

12.Federal Reserve Bank of New York, "Statement Regarding Plans for Reducing SOMA Holdings," 4 May 2022. newyorkfed.org/markets/opolicy/operating_policy_220504

13.12 U.S.C. §355(1), as amended by Pub. L. 96-18, 8 June 1979. The 1979 amendment struck the proviso permitting Reserve Banks to buy or sell direct obligations of the United States directly from or to the Treasury. uscode.house.gov

14.31 U.S.C. §3111, as cited in U.S. Treasury Fiscal Data, "Treasury Securities Buybacks." fiscaldata.treasury.gov/datasets/treasury-securities-buybacks/

15.Board of Governors of the Federal Reserve System, Current FAQs, “How does the Federal Reserve’s buying and selling of securities relate to the borrowing decisions of the federal government?” federalreserve.gov/faqs. The Federal Reserve states that it does not purchase new Treasury securities directly from the U.S. Treasury, that purchases of Treasury securities from the public are not a means of financing the federal deficit, and that it does not participate in competitive bidding at Treasury auctions.

16.TreasuryDirect, “FAQs about the Public Debt.” Debt held by the public is defined as all federal debt held by individuals, corporations, state or local governments, Federal Reserve Banks, foreign governments and other entities outside the United States Government, less Federal Financing Bank securities.

17.U.S. Department of the Treasury, Office of Debt Management, "Regular Treasury Buyback Program Details," April 2024. home.treasury.gov/system/files/221/TreasurySupplementalQ22024.pdf

18.Federal Reserve Bank of New York, "Tentative Outright Treasury Operation Schedule," May 2013. newyorkfed.org/markets/tot_operation_schedule_130531.html

19.Federal Reserve Bank of New York, "Tentative Outright Treasury Operation Schedule," March 2014. newyorkfed.org/markets/tot_operation_schedule_140331.html

20.Federal Reserve Bank of New York, "Tentative Outright Treasury Operation Schedule," May 2014. newyorkfed.org/markets/tot_operation_schedule_140530.html

21.Remarks by Assistant Secretary for Financial Markets Josh Frost, U.S. Department of the Treasury, press release JY1757. Source for the 45 operations, $67.5bn total, March 2000–April 2002, and Secretary Summers’ three stated rationales.

22.U.S. Treasury Fiscal Data, "Treasury Securities Buybacks" dataset (coverage from 9 March 2000; no buyback operations conducted 2003–2013).

23.Reuters, "Instant View: Yields fall after US Treasury says it will double some bond buybacks," 19 August 2026. Comments by Ryan Swift (BCA Research), Thomas Simons (Jefferies), Michael Lorizio (Manulife Investment Management), Joseph Purtell (Neuberger Berman), Brian Jacobsen (Annex Wealth Management), Peter Cardillo (Spartan Capital Securities).

24.Board of Governors, "Policy Normalization." federalreserve.gov/monetarypolicy/policy-normalization.htm

25.Federal Reserve Bank of New York, "The Implementation of Reserve Management Purchases to Maintain Ample Reserves," 31 March 2026; Operating Policy statement, 10 December 2025.

26.Board of Governors, Implementation Note issued 29 July 2026. federalreserve.gov/newsevents/pressreleases/monetary20260729a1.htm

27.Board of Governors, Monetary Policy Report, July 2026, Part 2. federalreserve.gov/monetarypolicy/2026-07-mpr-part2.htm

28.Treasury Borrowing Advisory Committee, "Bill Purchases and the Consolidated Balance Sheet," Charge 1, Q1 2026. home.treasury.gov/system/files/221/TBACCharge1Q12026.pdf

29.Treasury Borrowing Advisory Committee, "Treasury Buyback Program Enhancements," Charge 1, Q3 2025. home.treasury.gov/system/files/221/TBACCharge1Q32025.pdf

30.TreasuryDirect, "Buyback FAQs." treasurydirect.gov/help-center/faqs/buyback-faqs

31.U.S. Department of the Treasury, "Treasury Announces Marketable Borrowing Estimates," press release SB0584, 3 August 2026. home.treasury.gov/news/press-releases/sb0584

32.U.S. Treasury, Bureau of the Fiscal Service, "Debt to the Penny," data as of 18 August 2026, published 19 August 2026. fiscaldata.treasury.gov/datasets/debt-to-the-penny/

33.CNBC, "Treasury doubles debt buybacks as Bessent moves to steady bond market," 19 August 2026; Reuters Instant View, 19 August 2026 (dollar index and equity index levels).

34.U.S. Department of the Treasury, Bureau of the Fiscal Service, "Treasury Auction Results," 20-Year Bond, CUSIP 912810UX4, 19 August 2026 (R_20260819_2.pdf), together with the Treasury Fiscal Data auctions record for the same CUSIP. The release is internally consistent at seven checkpoints: competitive, noncompetitive and FIMA tendered and accepted each sum to their subtotals; subtotal plus SOMA equals the totals; the three bidder classes sum exactly to total competitive tendered and accepted; and the published bid-to-cover equals subtotal tendered divided by subtotal accepted.

35.Both circulating size figures are correct under different definitions and are reconciled from the official release: subtotal accepted from the public was $16,000,036,500, the SOMA add-on was $2,056,891,600, and total accepted was $18,056,928,100. Pre-auction reporting citing $16 billion (Bloomberg, 16 August 2026) refers to the public offering; reporting citing $18 billion refers to total accepted including the Federal Reserve rollover.

36.When-issued level of 5.199% immediately before the auction: investingLive, 19 August 2026, 17:17 GMT. The when-issued yield is a secondary-market quote, not a Treasury-published figure, and no official source for it exists; the derived 0.5bp tail therefore carries this provenance. Treasury’s own published dispersion measures — allotted at high and the high-to-median spread — are used alongside it and are primary.

37.U.S. Department of the Treasury, "Tentative Auction Schedule of U.S. Treasury Securities," covering August 2026 to February 2027.

38.30-year yield path: closed 5.196% on 19 August (CNBC); rose more than seven basis points on 20 August to as much as 5.27%, the level immediately preceding the announcement, paring to about 5.25% (Bloomberg, 20 August 2026); 5.248% intraday per CNBC, 20 August 2026; 5.2371% and broadly flat on 21 August (CNBC, 21 August 2026). Ten-year: 4.647% on 19 August, above 4.70% on 20 August, 4.6882% on 21 August.

39.Treasury Secretary Scott Bessent, interview on CNBC, 20 August 2026: the buyback programme “could be more than $4 billion per issue,” and “we have a big toolkit.” Reported by CNBC and Yahoo Finance, 20–21 August 2026.

40.ING research note, 20 August 2026, describing the intervention as suggesting discomfort about longer-term borrowing costs and raising the prospect of repetition; Maia Crook, senior research analyst, JPMorgan Chase, client note, 20 August 2026. Both reported by CNBC, 20 August 2026.

41.The record high auction yield for the 20-year tenor since its 2020 reintroduction is 5.245%, set in October 2023; the 19 August 2026 result of 5.204% is the second-highest. Pre-auction when-issued indications of approximately 5.27% (Bloomberg, 16 August 2026) would have exceeded the record had they been realised. Record status can be re-verified against Treasury’s Record-Setting Treasury Securities Auction Data dataset (fiscaldata.treasury.gov).

42.Prior-auction comparators (22 July 2026: 5.163% high yield against 5.158% when-issued, a 0.5bp tail; 2.64 bid-to-cover; 69.1% indirect, 16.2% direct, 14.7% dealer) and twelve-auction averages (2.64 bid-to-cover; 64.87% indirect, 24.32% direct, 10.82% dealer): Helious auction desk, enriched from TreasuryDirect records.

43.Board of Governors, "Minutes of the Federal Open Market Committee, July 28–29, 2026," released 19 August 2026. federalreserve.gov/monetarypolicy/fomcminutes20260729.htm

44.Forbes, "Bitcoin approaches $70,000 after Treasury announces buyback expansion," 19 August 2026; TheStreet, "Gold, Bitcoin rally as U.S. Treasury makes unexpected bond market move," 19 August 2026.