Reader question: A Treasury issuance auction sells new government securities to investors. A Treasury buyback does the opposite: Treasury buys outstanding securities back from the market before maturity. How does the algorithm decide which CUSIPs are eligible, which dealer offers are attractive enough to accept, how much of each security to purchase, and what happens after settlement?
The current U.S. Treasury buyback program is a competitive multiple-price purchase process. Treasury defines the operation type and maturity bucket, announces a maximum aggregate purchase amount and an eligible-security list, receives price offers through FedTrade, evaluates those offers against prevailing market prices and relative-value measures, accepts some or none of them subject to security-level purchase constraints, and pays each successful submitter the price at which that offer was submitted. Accepted securities are delivered at settlement and then retired.
The key computational lesson is that this is not the mirror image of Treasury’s single-price issuance auction. The objective function is different, the offer variable is price, Treasury can buy less than the announced maximum, and cross-security relative value matters because several CUSIPs can compete for one aggregate buyback capacity.
What this page owns — and what it does not
This article owns:
buyback purpose + security eligibility + dealer price offers + market/relative-value evaluation + security purchase limits → accepted offers → settlement and retirement.
It does not replace Treasury issuance-auction allocation, repo specialness, Treasury-futures CTD selection, or bond-yield inversion. Those mechanisms can influence eligibility or market value, but Treasury buyback selection is a separate owner.
This is public debt-management mathematics, not a recommendation to submit, buy or sell any Treasury security.
Two buyback objectives, two economic jobs
Treasury’s current FAQs distinguish two operation types.
Liquidity-support buybacks are intended to provide a regular and predictable opportunity for market participants to sell less-liquid, off-the-run Treasury securities back to Treasury. The policy aim is to support secondary-market liquidity and market-making capacity.
Cash-management buybacks are intended to reduce volatility in Treasury’s cash balance and bill issuance, reduce disruptions in bill supply and potentially reduce borrowing costs over time. They are generally concentrated in securities with roughly one month to two years remaining to maturity and can be timed around major tax-payment dates.
The same price-offer engine can therefore serve different portfolio-selection objectives depending on the announced operation.
Step 1: define the operation envelope
Before any offer arrives, Treasury announces parameters such as:
- operation type: liquidity support or cash management;
- operation date and time;
- settlement date;
- security type and maturity range;
- preliminary and then final eligible CUSIPs;
- minimum and maximum aggregate purchase amount.
The operation is therefore a constrained optimisation problem rather than an open-ended instruction to “buy cheap Treasuries.”
Step 2: construct the eligible CUSIP set
Treasury currently says it buys back off-the-run nominal coupon securities and TIPS. It does not intend to buy back Treasury bills, floating-rate notes or STRIPS through the current program.
It also excludes securities whose removal could work against market functioning or other Treasury objectives. Current examples include:
- on-the-run securities;
- securities trading significantly special in repo or otherwise in exceptional demand;
- securities that may be cheapest-to-deliver into an actively traded futures contract;
- recent issues not yet past their first coupon date;
- specified securities around coupon or maturity boundaries;
- certain short-dated issues around major tax dates for cash-management operations.
This means the eligibility layer deliberately uses information from adjacent markets. A security can look attractive on price but still be excluded because removing it would worsen scarcity in repo or disrupt a futures-delivery basket.
Why CTD and repo specialness are exclusion signals
Suppose an off-the-run note is scarce collateral and trades very special in repo. Buying more of it out of the market could reduce free float further and intensify that scarcity.
Similarly, buying a likely cheapest-to-deliver bond could alter Treasury-futures delivery economics.
Treasury therefore treats market-functioning constraints as part of the candidate-generation algorithm, not merely as after-the-fact commentary.
Step 3: calculate security-level purchase capacity
Even when a CUSIP is eligible, Treasury does not intend to buy the entire outstanding issue.
Current Treasury FAQs describe security-level purchase constraints including:
- free float after the operation should remain above specified thresholds;
- SOMA ownership should not exceed 70% of outstanding par after settlement;
- the minimum amount Treasury would purchase of a single security in an operation is at least $10 million par.
Current stated free-float thresholds are $10 billion par for nominal coupon securities and $5 billion par for TIPS. These are policy parameters and should be versioned rather than hard-coded forever.
Free float as a simple constraint
Define:
FreeFloat = OutstandingPar − SOMAHoldings − StrippedAmount.
If Treasury considers purchasing amount x, then a stylised post-buyback constraint is:
FreeFloat − x ≥ MinimumFreeFloat.
This yields one upper bound:
x ≤ FreeFloat − MinimumFreeFloat.
A second bound can come from the 70% SOMA-ownership condition. The actual security purchase capacity is the minimum of the applicable limits and the amount Treasury is willing to buy for relative-value reasons.
Step 4: participants submit multiple price offers
Treasury currently conducts buybacks through the Federal Reserve Bank of New York as fiscal agent using FedTrade or a successor platform.
The process is competitive and multiple-price. The current FAQ states:
- minimum offer amount = $1 million par;
- minimum increment = $1 million par;
- up to nine offers may be submitted per security;
- offers are expressed as price per $100 par.
Because each successful offer is paid at its own submitted price, the buyback is a pay-as-offered purchase mechanism.
Treasury price notation matters
Coupon Treasury prices can be submitted in 32nds and eighths of 32nds. Treasury’s FAQ gives:
102-172 = 102 + 17/32 + 2/(8×32) = 102.5390625.
A parser that reads “172” as 0.172 or 17.2/32 would mis-rank offers before the economic evaluation even begins.
This is a small example of a broader rule: financial notation is part of the data model, not decoration.
Step 5: evaluate offers against prevailing market prices
Treasury states that offers are evaluated based on their proximity to prevailing market prices at the close of the operation as well as relative-value measures.
For one CUSIP, a simple diagnostic quantity is:
OfferPremium = OfferedPrice − ReferenceMarketPrice.
A seller asking far above the prevailing market is less attractive than a seller offering close to or below comparable market value, all else equal.
But Treasury does not publish one universal scalar formula that says “sort every offer by OfferPremium and take the first N.” Relative value across securities matters.
Relative value is the cross-security problem
Suppose Treasury can buy up to $4 billion across a maturity bucket containing ten CUSIPs.
CUSIP A may be offered 0.02 points cheap to its prevailing market. CUSIP B may be offered 0.04 points cheap. CUSIP C may have a smaller price concession but provide more liquidity-support value because it is particularly off-the-run and still satisfies all market-functioning constraints.
The operation therefore involves a portfolio-selection problem:
choose accepted offers subject to aggregate cap, security limits and policy objectives.
The exact Treasury evaluation model is not public and should not be invented. Public articles can explain the observable inputs and constraints without claiming knowledge of proprietary selection code.
Step 6: accept all, some or none of the maximum
An announced maximum is a capacity ceiling, not a commitment to purchase that amount.
Treasury states that it may buy back less than the announced maximum or, in its discretion, buy nothing during an operation.
This is a major difference from an algorithm that is required to clear a fixed quantity regardless of price quality.
Conceptually:
AcceptedPar ≤ AnnouncedMaximum.
Weak offers can therefore leave capacity unused.
Marginal offer allocation within a CUSIP
Published buyback results show the highest accepted price and allocation percentage at that price for each accepted CUSIP. TreasuryDirect’s result notes explain that offers submitted below the highest accepted price are allotted at 100%, while offers at the highest accepted price can receive partial allocations.
This is intuitive for a buyer: lower offered prices are better. If the remaining desired amount of one CUSIP is smaller than the total par offered at the marginal accepted price, the marginal offers need partial allocation under Treasury’s rules and increments.
A stylised one-CUSIP example
Suppose Treasury wants at most $60 million of one security and receives:
| Offer price | Par offered |
|---|---|
| 95.90 | $20m |
| 95.95 | $25m |
| 96.00 | $30m |
If all three price levels remain attractive under Treasury’s broader evaluation but the desired purchase is $60 million, the first $45 million can be accepted in full and only $15 million of the $30 million offered at 96.00 is needed.
The highest accepted price is 96.00 and the marginal price level receives a 50% allocation in this stylised example.
Real operations also enforce $1 million par allocation increments and security-specific/cross-security constraints.
Multiple-price means accepted sellers can receive different prices
If Seller A offers at 95.90 and Seller B offers at 96.00 and both are accepted, A receives 95.90 while B receives 96.00 for the relevant accepted par under the multiple-price process.
This differs sharply from Treasury’s issuance auction, where successful bidders in the current single-price system receive the same auction price derived from the stop-out result.
That distinction is important enough to become a unit test.
Step 7: publish aggregate results
Treasury publishes operation results including quantities such as:
- total par amount offered;
- total par amount accepted;
- number of eligible and accepted issues;
- par offered and accepted by CUSIP;
- highest accepted price;
- weighted-average accepted price;
- allocation at the highest accepted price;
- post-operation privately held par.
Those fields make several parts of the allocation externally testable even though Treasury’s full relative-value decision model is not public.
Weighted-average accepted price
If accepted offers for one CUSIP have prices p_i and accepted par amounts q_i, then:
Weighted Average Accepted Price = Σ(p_i q_i) / Σq_i.
This value should fall between the lowest and highest accepted price for that security.
If it does not, the result data or weighting algorithm is inconsistent.
Step 8: settlement and retirement
Treasury says settlement generally occurs one business day after the operation, although same-day settlement can be specified.
Submitters are responsible for delivering accepted securities free of liens and other restrictions. Failure to deliver can lead to fails charges or additional sanctions under the applicable rules.
Once purchased and settled, Treasury states that the securities are retired. They are not held as a trading inventory and later re-sold.
Cash-management buybacks change the maturity profile of Treasury outflows
Suppose Treasury expects a large cash inflow around a tax date and a large coupon maturity a few weeks later.
A cash-management buyback can use part of the temporary cash surplus to retire some short-dated coupon securities early. That can:
- reduce the later maturity cash outflow;
- reduce swings in Treasury General Account balances;
- reduce the need to make abrupt changes in Treasury bill issuance.
The algorithm therefore links today’s purchase decision with future cash-flow smoothing.
Liquidity-support buybacks target off-the-run market structure
Off-the-run Treasuries can trade less frequently and with wider intermediation costs than the newest benchmark issues.
Regular Treasury purchase opportunities can give dealers another outlet for older inventory and can concentrate trading interest around operation times. Treasury and primary-dealer feedback have described the program as moderately supportive of liquidity in some off-the-run sectors.
That is an empirical objective, not a mathematical guarantee that every buyback will narrow bid–ask spreads.
Current 2026 programme scale is time-sensitive
Treasury’s August 2026 Quarterly Refunding statement said it expected to purchase up to $38 billion of off-the-run securities for liquidity support during the quarter and up to $25 billion in the one-month-to-two-year bucket for cash management.
On 19 August 2026, Treasury also announced that long-end nominal liquidity-support operations in the 10-to-20-year and 20-to-30-year sectors would increase from a current $2 billion maximum to at least $4 billion per operation beginning 9 September 2026 for the remainder of that refunding quarter.
These are operational parameters, not permanent rules. Software should read current schedules and announcements.
The March 2026 rule amendment is another update trigger
Treasury amended 31 CFR Part 375 on 30 March 2026 to reflect expanded direct offer-submission eligibility, update participation certifications and align the regulations with current practices.
This is why operational systems should store an effective rule version rather than assume the 2024 launch configuration is still current.
Evidence polarity: what supports confidence?
Evidence for a correct buyback engine includes a final eligible list matching Treasury’s announcement, offer-price parsing that reproduces decimal values, accepted quantities within aggregate and security-level limits, marginal allocations that reconcile to published results, weighted-average prices between accepted extrema, settlement quantities matching delivered securities and post-settlement retirement.
Evidence against confidence includes on-the-run or strongly special securities entering the candidate set contrary to current policy, an accepted amount above the operation maximum, SOMA ownership exceeding the stated threshold after purchase, free float below the required minimum, a seller paid a uniform price despite a multiple-price operation, or accepted offers that cannot be traced to market and relative-value evaluation records.
Counterexample: the lowest offered price need not consume the entire operation
A cheap offer in one CUSIP can still be limited by the free-float or SOMA constraint for that security.
Once the security-level cap binds, the remaining aggregate capacity may be allocated to other CUSIPs.
The algorithm is therefore constrained optimisation rather than one global sort by price.
Counterexample: the announced maximum is not a target that must be filled
If Treasury receives weak offers relative to prevailing markets, it can buy less than the maximum or nothing.
This falsifies the assumption that every announced buyback automatically injects the full maximum amount of demand into the market.
Counterexample: buybacks are not designed as an emergency market-rescue mechanism
Treasury’s current FAQ says liquidity-support buybacks are not intended to mitigate episodes of acute market stress.
The regular program is a debt-management and market-functioning tool, not a standing emergency purchase facility.
Counterexample: accepted security does not remain Treasury-owned inventory
Purchased securities are retired at settlement.
A system that books them into an investment portfolio for future resale misunderstands the redemption operation.
Weak links in implementation
stale eligible list. The preliminary list is used after the final announcement supersedes it.
price-format bug. Treasury 32nds/eighths are parsed incorrectly.
market-reference timing error. Offer quality is evaluated against a stale market snapshot rather than the intended close-of-operation reference.
free-float omission. Purchase quantity removes too much publicly available supply.
SOMA constraint omission. Post-operation Federal Reserve ownership exceeds the policy threshold.
objective collision. Cash-management and liquidity-support rules are treated as identical in maturity targeting and capacity carry-forward.
uniform-price mistake. Buyback settlement is coded like a Treasury issuance auction.
settlement/retirement mismatch. accepted par is not reconciled to delivered and retired securities.
Diagnostics: how to test the engine
- eligibility replay: rebuild the final eligible CUSIP set from the announcement and market-state exclusions.
- 32nds parser test: reproduce Treasury’s 102-172 = 102.5390625 example.
- minimum-offer test: reject an offer below $1 million par or not on a $1 million increment.
- security-cap test: push free float and SOMA holdings close to their limits.
- multiple-price test: two accepted offers at different prices must settle at their own offer prices.
- marginal-allocation test: only the highest accepted price level should require partial allocation in a simple one-CUSIP monotonic case.
- weighted-average test: independently recompute the published weighted-average accepted price.
- underfill test: weak offers should permit accepted par below the announced maximum.
- settlement test: accepted par equals delivered and retired par after settlement.
- rule-version test: apply March 2026 participation-rule changes only from their effective date.
What would falsify confidence?
Confidence should be withdrawn if the algorithm cannot reproduce public operation results from accepted-offer records; if security limits are violated; if Treasury’s multiple-price rule is implemented as uniform-price; if settlement does not retire the accepted securities; or if market/relative-value evaluation cannot be reconstructed to an auditable input state.
Alternatives and limits
Treasury issuance auctions raise funding by selling new or reopened securities. Open-market operations by the Federal Reserve have monetary-policy or balance-sheet objectives and are institutionally different. Treasury buybacks are debt-management redemption operations under Treasury authority.
The full relative-value selection function used by Treasury is not public. A responsible public model should reproduce the published constraints and results without pretending to know a private optimisation score.
How this connects to the surrounding knowledge estate
Treasury issuance auctions provide the contrast between single-price sale and multiple-price buyback. Repo specialness helps explain eligibility exclusions. CTD algorithms explain another exclusion channel. STRIPS explain why stripped amounts matter when Treasury measures free float.
Verification and update triggers
Preserve the 31 CFR Part 375 version, quarterly schedule, preliminary and final announcements, operation objective, eligible CUSIPs, outstanding par, SOMA holdings, stripped amounts, market-reference snapshots, submitted offers, accepted par, settlement records and retirement records. Revalidate after every rule amendment, quarterly-refunding policy change, buyback-size change or failed-delivery incident.
Primary and high-quality references
- U.S. Treasury, TreasuryDirect, FAQs about Treasury Securities Buybacks.
- U.S. Treasury, TreasuryDirect, Redemption (Buyback) Operations Regulations, including the March 2026 rule update and links to 31 CFR Part 375.
- U.S. Treasury, TreasuryDirect, Buyback Announcements & Results, including security-level prices and allocations.
- U.S. Treasury, Quarterly Refunding documents and tentative buyback schedules.
- U.S. Treasury, Treasury Announces Increased Sizes of Nominal Long-End Liquidity Support Buybacks Beginning September 9, 19 August 2026.
Educational boundary: This article explains public Treasury debt-management algorithms. It does not recommend submitting an offer, trading a Treasury security or taking a market position and does not provide personalized financial advice.