Rebalance of August 31, 2026
An open benchmark for the “risk free” rate
Eleven open US Treasury indices, so that anyone can reproduce and verify.
Free to benchmark against. Paid only to track. No license, no fee and no registration, for any number of portfolios. Naming an index as your benchmark is free too, a prospectus included. A fee applies to one thing: launching a product that tracks one.
What $100 became
Base September 30, 2008 to August 31, 2026
Each line is a total-return index based at 100 at its own base date, so it reads directly as what a dollar invested became. The range above starts at that base, one month before the first published level, which is why it precedes the factsheet range for the same index. The floater index starts in December 2014 because floating-rate notes did not exist before then; nothing is rebased to hide that.
Every line starts at 100 on its base date, the month-end before its first published level. The first month plotted is October 2008, in the weeks after Lehman, and it is the reason the linker line drops before it does anything else: inflation-linked Treasuries returned -8.92% that month as the market priced deflation and forced sellers hit a thin book, against -0.41% for the nominal index.
Four of the eleven, chosen because they behave differently from one another: the nominal notes and bonds that are most of the market, linkers, floaters and the front end. The five maturity bands track the nominal index closely enough to overplot, and the aggregate is close enough to the nominal index to be a second line along the same path. All six have their own page.
| Index | Ticker | Level | Month | Holdings |
|---|---|---|---|---|
| US Treasury IndexNominal notes and bonds, one year and over | SR-UST-TR | 145.41 | +0.33% | 298 |
| 1-3 Year1 to 3 years | SR-UST-TR-0103 | 133.30 | +0.27% | 96 |
| 3-7 Year3 to 7 years | SR-UST-TR-0307 | 152.16 | +0.17% | 90 |
| 7-10 Year7 to 10 years | SR-UST-TR-0710 | 162.48 | +0.22% | 13 |
| 10-20 Year10 to 20 years | SR-UST-TR-1020 | 151.28 | +0.85% | 59 |
| 20+ Year20 years and over | SR-UST-TR-20PL | 149.56 | +0.65% | 40 |
| 1-3 Month Bill1 to 3 month bills | SR-UST-BILL | 128.06 | +0.32% | 17 |
| 1-12 Month1 to 12 months, bills and aged coupons together | SR-UST-SHRT | 130.00 | +0.32% | 88 |
| Inflation-LinkedInflation-linked, one year and over | SR-UST-TIPS | 175.21 | +0.15% | 48 |
| US Treasury AggregateThe whole marketable market: bills, coupons, linkers and floaters | SR-UST-AGG | 143.87 | +0.31% | 447 |
| Floating Rate NotesFloating rate notes | SR-UST-FRN | 130.64 | +0.31% | 8 |
Market value weighted, rebalanced monthly on the last trading day, priced on the bid side of Treasury's own end-of-day file. Levels are chained from 100 at each index's base date. The month column is total return over the month ending at this rebalance, and it chains exactly against the level.
Why an open bond index is unusual
An index priced off licensed marks cannot be given away, because the marks are not the administrator's to give. An index built on a vendor's data and calculated by that vendor inherits the vendor's terms however open its administrator wants to be, and that is the reason open equity indices exist and open bond indices largely do not.
There is no calculation agent here and no licensed input, so there is nobody to pass terms through from. Every input is public and the constituents are published each month, which means the levels can be reproduced rather than trusted. That may be the only way a genuinely open bond index can exist.
Free to benchmark against. Paid only to track. Measuring anything against these indices is free. No license, no fee, no registration, and no permission needed to say that you did. Naming an index as the benchmark of a portfolio, a mandate or a fund, in an investment policy statement, a trustee pack, a client report, a factsheet or a prospectus, is benchmarking. A pension fund whose annual report states that its Treasury sleeve is measured against the Safe Rate US Treasury Index owes us nothing and needs no permission to say it.
A fee applies to one thing: launching a product that tracks an index: an ETF, an index fund, a strategy sold as tracking it, or a structured product whose payoff references it. The test is whether the index is the basis of the product or the yardstick beside it, and in practice it is one word. "Measured against the Safe Rate US Treasury Index" is free in any document. "Tracks the Safe Rate US Treasury Index" is the license.
The point of control is the name, not the numbers. The levels are reproducible by anyone from public data, which is the whole design, so withholding them buys nothing and giving them away costs nothing. What has value is the right to write "tracks" in a prospectus.
| Use | Terms |
|---|---|
| Measure a portfolio, mandate or fund against it | free |
| Name it as your benchmark in an IPS, trustee pack or client report | free |
| Name it as a fund's benchmark in that fund's prospectus | free |
| Receive constituents and returns | free, and derivable from public data anyway |
| Reproduce the index yourself | free, every input is public |
| Launch an ETF, index fund or structured product that tracks it | fee |
| Use the marks in a product's name or branding, or so as to imply we sponsor, endorse or calculate it | license |
Independent evidence
Checked against the funds that follow these bands
Each band is compared against the ETF tracking that part of the market, month by month, using public market prices. The prediction that makes this a test rather than a coincidence is that the gap should be roughly the fee the fund charges. We hold no fee, the fund does, so we should run ahead of it by about that much. Every figure in the table is basis points per month, which is how the comparison is struck: BIL charges 0.1356% a year, so its fee predicts 1.13 basis points a month, and that is the number beside it.
| Band | Fund | Months | Mean gap, bp | Fee predicts, bp | Median, bp | Worst month, bp |
|---|---|---|---|---|---|---|
| US Treasury Index | GOVT | 175 | +0.75 | +0.94 | -0.32 | -88.72021-12 |
| 1-3 Year | SHY | 215 | +1.39 | +1.25 | +1.37 | +40.62022-12 |
| 3-7 Year | IEI | 215 | +1.24 | +1.25 | +0.65 | -92.02022-11 |
| 7-10 Year | IEF | 215 | +1.06 | +1.25 | +0.66 | -128.92022-11 |
| 10-20 Year | TLH | 215 | +0.65 | +1.25 | +3.94 | +204.82021-03 |
| 20+ Year | TLT | 215 | +0.45 | +1.25 | -0.31 | -289.62011-09 |
| 1-3 Month Bill | BIL | 215 | +1.09 | +1.13 | +1.09 | +27.82008-12 |
| 1-12 Month | SHV | 215 | +0.99 | +1.25 | +0.85 | -16.52008-11 |
| Inflation-Linked | TIP | 215 | +1.31 | +1.50 | +0.76 | -146.42022-11 |
| Floating Rate Notes | TFLO | 141 | +1.48 | +1.25 | +1.90 | -57.82015-12 |
Every gap is positive. We run ahead of every fund, which is what holding no fee says should happen. They do not all land on the prediction: 2 sit on it, 2 run above and 6 below. The widest shortfall is TLT at 0.80 basis points under prediction, and the longest bands are the ones furthest under, which is where premium and discount to net asset value are largest and where a fund's price wanders furthest from what it holds.
A fund is not its index. It trails by its fee and moves either side of fair value, so the number worth reading is the center of two hundred months rather than any single one. The worst-month column is there to show how wide a single observation can be, which is the difference between a comparison and a claim.
Expected values are the fee actually in force across each month rather than today's expense ratio, time-weighted over the window. GOVT cut its fee from 0.15% to 0.05% in 2018, so a single ratio would be wrong for most of the history.
Fund prices are public market data. This comparison contains no third-party index values.
IOSCO considerations
The IOSCO Principles for Financial Benchmarks are the standard these indices are built against. No claim of compliance is made: compliance is an assertion an administrator makes about itself, and the governance policy assesses all nineteen principles one at a time and says which are met, which are partial and which are not. The ones worth reading here are where the answer differs from a licensed benchmark's.
Principles 6, 7 and 9: methodology and transparency
Anyone can calculate it, which is the point.
Every input is a public Treasury file, every rule is published, and every holding of every rebalance is on this site with its par, its three float deductions, its prices and its contribution to the month. A reader can recompute a published level from the inputs and check it to floating-point noise. A licensed benchmark cannot offer this at any price, because its inputs are not the reader's to hold.
Principles 10 to 14: data sufficiency and submitters
There is no submitter, so there is nothing to game.
The indices use no submissions, no panel, no expert judgment and no evaluated prices. There is no discretion to exercise and therefore nothing to police: the rules applied to the public file produce one answer. Most of the IOSCO machinery on submitter conduct exists to manage a risk these indices do not carry.
Principle 12: changes to the methodology
Changes are versioned and noticed, not silent.
The rulebook is versioned, every level carries the methodology version that produced it, and changes go through a stated notice period rather than appearing silently. A restatement is republished with its cause rather than overwritten.
Principles 4, 5 and 17: oversight and audit
Anyone outside this firm can look and audit.
Every calculation is reproducible from public data: the inputs are Treasury's own files, the rules are published, and every holding of every rebalance is on this site with the arithmetic that turns it into a level. An auditor needs no access, no license and no cooperation from us to check any figure here, which is not true of a benchmark whose inputs are licensed. What we do not yet have is someone having done it: no commissioned independent review, no external audit, and no oversight function separate from the people who calculate. That remains the largest item on our own list, and it is a gap in exercise rather than in access.
Principle 8: data quality
One calendar year is still unexplained.
2023 carries a residual against fund returns that survives every cause we can vary: the history has been rebuilt with each eligibility rule and each input changed one at a time and none accounts for it. Settling it would need evaluated prices we do not hold. We publish the residual rather than tuning it away, which is the honest treatment but not a resolution.
The rules
Fixed-rate nominal notes and bonds with at least one year to final maturity at the rebalance, and a minimum of $300mn float-adjusted. Bills, inflation-linked securities, floaters and STRIPS are excluded from the coupon families, the sibling indices above are those instruments. Market value weighted on amount outstanding as at the start of the month, not at the rebalance that closes it and not today, so a note reopened during the month is not credited with issuance the index did not hold.
Priced on the bid, dirty. Treasury's file publishes a BUY and a SELL column and labels neither as the bid. SELL is the bid, and that was checked rather than assumed: across 326 quoted nominal notes and bonds, BUY exceeds SELL 273 times, equals it 53 times, and is never below it.
Buybacks are deducted before Federal Reserve holdings, and the two are kept separate. A buyback destroys the debt, at settlement date; the Federal Reserve only parks it. Merging the two adjustments would restore retired par the moment quantitative tightening reversed, which is why they are ordered rather than combined.
Rebalanced on the last trading day of the month, with coupons held as cash until the next reconstitution. A month ending on a bond market holiday uses the last trading day, and a price file with every bid zeroed is rolled back rather than used.
Levels are month-end figures. Curves and prices elsewhere on this site are daily closes, so the August 31, 2026 level sits behind the latest price date by design rather than by staleness, the next level appears at the following rebalance.