Why it exists at all is the LIBOR story, and the difference is estimates versus transactions. The Federal Reserve convened the Alternative Reference Rates Committee in 2014, in the ARRC's own account, "and tasked the group with identifying an alternative to U.S. dollar LIBOR that was a robust, IOSCO-compliant, transaction-based rate derived from a deep and liquid market. In 2017, the ARRC fulfilled this mandate by selecting the Secured Overnight Financing Rate, or SOFR." The ARRC lists what SOFR has that LIBOR and similar rates did not, and one item is decisive: SOFR "is produced in a transparent, direct manner and is based on observable transactions, rather than being dependent on estimates, like LIBOR, or derived through models." SOFR is built from the Treasury repo market, which the ARRC calls "the largest rates market at a given maturity in the world." The committee itself finished its work and ceased operations in November 2023.
Secured and unsecured are not a technicality; they change what the number measures. LIBOR was a term rate for unsecured bank borrowing, so it embedded a premium for the risk of lending to a bank for one, three or six months. SOFR is an overnight rate fully collateralized by Treasury securities, so it carries almost none of that. Swapping one for the other without compensation would therefore have changed the economics of every existing contract, which is why the transition came with fixed spread adjustments rather than a straight substitution.
For legacy contracts with no workable fallback, Congress and the Federal Reserve substituted the rate by operation of law. The Adjustable Interest Rate (LIBOR) Act, codified at 12 U.S.C. 5801 and following, and the Board's Regulation ZZ at 12 CFR part 253 exist, in the regulation's own words, "to establish a clear and uniform process, on a nationwide basis, for replacing the overnight and one-, three-, six-, and 12-month tenors of U.S. dollar LIBOR in existing contracts" that do not provide for the use of a clearly defined or practicable replacement rate. The regulation sets the LIBOR replacement date as "the first London banking day after June 30, 2023." It deliberately does not disturb contracts that already had a workable fallback, or new contracts, where parties remain free to use any appropriate rate.
For a consumer loan caught by that regulation, the replacement is the corresponding one-, three-, six- or 12-month CME Term SOFR plus a fixed tenor spread adjustment, and the spread phases in linearly over the first year rather than arriving in a single step. The adjustments themselves are set in the regulation and do not change: 0.00644 percent for overnight LIBOR, 0.11448 percent for one-month, 0.26161 percent for three-month, 0.42826 percent for six-month and 0.71513 percent for 12-month.
An overnight rate cannot price a three-month loan, so the market built two ways around that. The first is averaging. The New York Fed publishes SOFR Averages, which are "compounded averages of the SOFR over rolling 30-, 90-, and 180-calendar day periods," together with a SOFR Index that "measures the cumulative impact of compounding the SOFR on a unit of investment over time," set to 1.00000000 on April 2, 2018, the first value date of the SOFR. The second is a forward-looking term rate, which the ARRC set as a goal and which is now published by a private administrator as CME Term SOFR and named as such in Regulation ZZ. The practical consequence for a borrower is that a rate quoted as "30-day Average SOFR plus 2.50 percent" resets from an average of the past thirty days, so it lags a change in short-term rates rather than moving with it on the day.
It can move for reasons that have nothing to do with monetary policy. Because SOFR is a repo rate, it reflects the supply of cash and collateral in that market on a particular morning, including quarter-end and year-end balance-sheet effects. Governor Christopher Waller's account of autumn 2019 is the clearest official description of what that can look like: the Federal Reserve had reduced the level of reserves through its balance sheet normalization program, heavy Treasury issuance arrived in September, and in his assessment "the level of reserves likely went a bit too low," with stresses showing up in money markets. A borrower whose rate resets off a daily print can therefore see a move that no policy decision explains, which is one practical argument for the averaged versions.