SOFR Averages vs the SOFR Index
Both are compounded SOFR, published by the New York Fed each business day. The averages fix the window for you; the Index lets you pick it. Latest: 90-day average 3.70963%, Index 1.26208139 (October 9, 2026).
The SOFR Averages are three ready-made compounded rates, covering the 30, 90 and 180 calendar days that end on the publication date. Use one when a contract needs a single published number, usually fixed at the start of an interest period ("in advance"). The SOFR Index is a running compounding factor that has grown from 1.00000000 on April 2, 2018. Divide the Index on the last date of a period by the Index on its first date, subtract 1 and annualize on actual/360, and you have compounded SOFR in arrears for exactly that period. Both come from the same daily SOFR prints and the same compounding rule, so an average is an Index ratio over a fixed window whenever that window opens on a business day.
What the SOFR Averages measure
The Federal Reserve Bank of New York began publishing the "30-day Average SOFR", "90-day Average SOFR" and "180-day Average SOFR" on March 2, 2020, after consulting on the idea from November 4, 2019. Each is a percentage rounded to the fifth decimal place. They are not arithmetic means of daily SOFR. They are compound interest rates, built with the convention in the New York Fed's reference rate methodology:
Interest compounds on business days only. Over a weekend or holiday, the preceding business day's SOFR earns simple interest for each calendar day, which is what the nᵢ weight does. The day count is actual/360.
The window rule
An average published on a given date starts "exactly 30-, 90-, and 180-calendar days before the publication date, regardless of whether or not that date is a weekend or holiday", and runs through the SOFR published that morning. If the start date falls on a Saturday, Friday's SOFR covers two days; on a Sunday, one day.
Take the 90-day average published on October 9, 2026. Ninety days earlier is Saturday, July 11, 2026, so the window opens with Friday July 10's SOFR of 3.55% applied to July 11 and 12, and closes with the SOFR for October 8, the last rate published by the morning of October 9. That partial-weekend start is the one thing the Index cannot reproduce cleanly, which matters later.
What the SOFR Index measures
The SOFR Index is the value of $1 invested at SOFR on April 2, 2018, the first SOFR value date, compounded on the same convention as the averages. Its starting value is 1.00000000. The first official published value, on March 2, 2020, was 1.04085026, reflecting compounding from April 2, 2018 through that date. It is quoted to eight decimal places, and the latest value is 1.26208139 for October 9, 2026.
For any period from date x to date y, the New York Fed gives this formula:
The dating convention trips people up. The Index (and each average) carries a value date one business day after the last SOFR it includes. The Index dated October 1, 2026 contains SOFR up to and including September 30. Look up the Index on the first and last day of the interest period itself, not on the day before.
Worked example: a 3-month rate from two Index values
Suppose a note accrues from July 1, 2026 to October 1, 2026 and pays compounded SOFR in arrears with no lookback. Both dates are business days, so both Index values are published.
| Index, July 1, 2026 (start) | 1.24923010 |
| Index, October 1, 2026 (end) | 1.26099341 |
| Ratio, end ÷ start | 1.0094164478 |
| Growth over the period (ratio − 1) | 0.0094164478 |
| Calendar days, d_c | 92 |
| Annualized: 0.0094164478 × 360 ÷ 92 | 3.68470% |
| Interest on $10,000,000 (principal × growth) | $94,164.48 |
The period draws on 64 SOFR prints, from July 1 through September 30, 2026. They ranged from 3.53% on July 9 to 3.90% on September 25, 28 and 30.
We checked the result by compounding those 64 daily rates directly, weighting each Friday and pre-holiday rate by its calendar days. That also gives 3.68470% to five decimal places. The calendar-weighted simple average of the same rates is 3.66783%, so compounding adds about 1.7 bp over three months at this rate level. Ignoring it is a common reconciliation break.
The SOFR Index calculator runs this two-number calculation for any pair of published dates. If a contract adds a lookback without observation shift, or a lockout, the Index ratio no longer matches the contract's arithmetic. Use the compounded SOFR calculator, which builds the product day by day.
An average is an Index ratio over a fixed window
When an average's window starts on a business day, the Index reproduces it exactly. On October 8, 2026, the 90-day window started on Friday, July 10 and the 30-day window on Tuesday, September 8:
| Window | Index at start | Index at end | From Index | Published average |
|---|---|---|---|---|
| 30 days, Sep 8 – Oct 8, 2026 | 1.25795382 | 1.26194573 | 3.80800% | 3.80800% |
| 90 days, Jul 10 – Oct 8, 2026 | 1.25036103 | 1.26194573 | 3.70603% | 3.70603% |
The New York Fed warns that, because the Index is rounded to eight places, an average derived from it "may occasionally" differ from the published one at the fifth decimal. When a window opens on a weekend, as the 180-day window did that day (Saturday, April 11, 2026), there is no Index value for the start date at all. The published average is then the only official figure for that exact window.
The New York Fed's guidance on approximating Index values for periods that begin on a weekend or holiday is, per its methodology page, under revision. This site does not offer a method for that case. If a contract's period starts on a non-business day, follow the contract's fallback wording or ask the calculation agent.
When the averages are the right tool
The ARRC's 2021 user's guide to SOFR draws the line plainly. The Index exists "to allow users to calculate a compound SOFR based on ISDA's definitions over any start and ending date", while the averages "are more likely to be used in advance."
In an in-advance structure, the rate for a period is set on its first day from the average published that day. The borrower knows the payment from day one, and systems that expect a single rate per period need no daily compounding. The cost is that the rate describes the previous 90 days, not the period being paid for.
The averages also appear in law. Under the Federal Reserve Board's Regulation ZZ (12 CFR part 253), 1-, 3-, 6- and 12-month LIBOR in contracts of FHFA-regulated entities (other than Federal Home Loan Bank advances) is replaced by 30-day Average SOFR plus the tenor spread. In FFELP asset-backed securities, 3-month LIBOR becomes 90-day Average SOFR plus 0.26161%. The credit spread adjustment guide covers where those spreads came from.
In advance vs in arrears on real 2026 quarters
The table sets the 90-day average published on the first day of each quarter (in advance) against the Index result for the quarter itself (in arrears), on $10 million at actual/360.
| Period | Days | In advance | In arrears | Gap | Interest, advance | Interest, arrears |
|---|---|---|---|---|---|---|
| Apr 1 – Jul 1, 2026 | 91 | 3.67664% | 3.63546% | −4.1 bp | $92,937.29 | $91,896.33 |
| Jul 1 – Oct 1, 2026 | 92 | 3.63493% | 3.68470% | +5.0 bp | $92,892.66 | $94,164.48 |
In the second quarter the in-advance borrower paid $1,040.96 more than an in-arrears borrower; in the third it paid $1,271.82 less. The third-quarter gap has a visible cause in the data. The federal funds target range moved up 25 bp to 3.75%–4.00% from September 17, 2026, and SOFR went from 3.62% on September 16 to 3.85% the next day. The arrears rate captured those last two weeks; the in-advance rate, fixed on July 1, could not.
Note also that the third quarter's in-advance rate (3.63493%) sits within a basis point of the second quarter's in-arrears rate (3.63546%). That is by construction: the 90-day average on July 1 covers nearly the same days as the April–June period. With a 90-day average and quarterly periods, in-advance pricing is roughly in-arrears pricing lagged one period. The two match only when rates are flat. Otherwise the in-advance borrower pays less than an in-arrears borrower while rates are rising and more while they are falling, as the two quarters above show. The arrears conventions guide covers the lookback, observation shift, lockout and payment delay structures that sit between these two extremes.
How the three averages relate to each other and to SOFR
30-, 90- and 180-day Average SOFR, last three years
The longer the window, the more slowly an average reacts. It also keeps reflecting a rate change for the full length of its window after the change. Two dates in the history show how far that lag can run:
- Falling rates. The 90- and 180-day averages peaked on September 19, 2024, at 5.37097% and 5.39812%. That publication still excluded the first SOFR after that month's cut, 4.82% for September 19. By December 31, 2024, the 180-day average was 5.02781%, about 66 bp above the 4.37% SOFR for December 30, the last day it included.
- Rising rates. On December 30, 2022, the 180-day average was 2.89325% while SOFR for December 29 was 4.30%, a gap of about 141 bp.
The order of the three averages shows recent direction. On October 9, 2026 they stood at 3.81569% (30-day), 3.70963% (90-day) and 3.68823% (180-day). The shortest window is highest because only it is dominated by the weeks since the September increase. The 30-day average also absorbs month-end and quarter-end prints more visibly, since one high day is a larger share of a 30-day window. The month-end guide has the pattern; recent published values sit on the SOFR Averages and Index data page.
Choosing between them
| Question | SOFR Averages | SOFR Index |
|---|---|---|
| Period length | Fixed: 30, 90 or 180 calendar days | Any period between two published dates |
| Window ends | On the publication date | Wherever the contract says |
| Typical timing | In advance (rate known at period start) | In arrears (rate known at period end) |
| Precision | 5 decimal places (percent) | 8 decimal places (index level) |
| Weekend start dates | Handled by the published window rule | No Index value; NY Fed guidance under revision |
| Lookback without shift, lockout | Not applicable | Not reproducible; compound daily instead |
Choose the Index when the contract defines compounded SOFR over its own interest period, or over an observation period shifted back a set number of business days. Taking the Index on the shifted start and end dates gives the observation-shift result, because the Index weights each rate by calendar days in the observation period. A forward-looking Term SOFR rate is a third, licensed option that this site explains but does not display.
Choose an average when the documents name one: Regulation ZZ fallbacks, in-advance loan or note structures, or any system that must book one rate at the start of each period. Then check two details before relying on the number: which publication date the contract reads it from, and whether a spread is added on top.