What did the market lose when the benchmark became risk-free?
Compounding conventions are an inconvenience. The credit question is a genuine economic problem, and it is the reason the LIBOR debate never fully ended.
The gap, stated plainly
LIBOR was a rate at which banks claimed they could borrow unsecured. It therefore embedded bank credit risk. When banks got into trouble, LIBOR rose.
SOFR is a rate on lending secured by Treasuries. It contains essentially no bank credit risk. In a flight to quality, demand for Treasury collateral surges and repo rates fall — at exactly the moment banks' own funding costs are rising.
Now put a bank on the wrong side of that. It holds loans paying SOFR plus a fixed spread. Stress arrives. Its asset yields drop with SOFR while its own liabilities get more expensive. Its margin is squeezed by the very event that should have widened it. Under LIBOR, the two moved together.
This is real basis risk, not paperwork. March 2020 was the textbook demonstration: credit-sensitive short-term rates spiked while secured overnight rates collapsed.
The attempted answers
Credit-sensitive alternatives were built to close the gap — AMERIBOR, based on transactions on the American Financial Exchange, and Bloomberg's BSBY, the Short-Term Bank Yield Index. BSBY is the instructive one: Bloomberg announced its discontinuation in November 2023 after IOSCO raised concerns that its underlying market was too thin to support it.
Read that carefully, because it is the whole lesson of Unit 3 recurring: the reason a credit-sensitive term benchmark is hard to build is the same reason LIBOR failed. The market it would need to measure — large-scale unsecured term bank funding — is much smaller than it was. You cannot rebuild a benchmark on a market that no longer trades in size just because the old benchmark was convenient.
Spread adjustments: how legacy contracts converted
Because SOFR sits structurally below where LIBOR fixed — no credit component, and secured — converting an existing contract one-for-one would have transferred value from lender to borrower. So conversions add a fixed spread.
ISDA's fallbacks use the five-year historical median of the LIBOR-minus-RFR spread, fixed once and for all on 5 March 2021:
- 3-month USD LIBOR → SOFR: about 26.161 basis points
- 1-month USD LIBOR → SOFR: about 11.448 basis points
Worked example. A legacy loan paying "3-month LIBOR + 100bp" converts to roughly "compounded SOFR + 126bp." If compounded SOFR over the period is 4.20%, the coupon is about 5.46%. On $50 million for 90 days:
50,000,000 × 0.0546 × 90 / 360 = $682,500
Same intended economics, expressed in a benchmark with transactions underneath it.
The honest summary
The world traded a benchmark with a credit signal and no transactions for a benchmark with transactions and no credit signal. That is a trade, not a free upgrade. Whether it was the right one is still argued about by serious people, and this course does not adjudicate it — the job here is to make sure you can state both sides accurately.
In the data
What was lost leaves one daily trace: unsecured overnight lending minus secured, EFFR minus SOFR. Here it is for five calm weeks, in basis points:
From 18 August to 24 September 2026 it never moved further than 5 basis points from zero, and it changed sign again and again. In calm conditions the credit component of one night's unsecured lending is close to nothing. It is a thin proxy — one night, one market — for what a credit-sensitive benchmark carried across a whole term.
Try it now
- Below are SOFR and EFFR on five days of March 2020, the stress period this lesson describes: the secured and the unsecured overnight rate, side by side. Did the secured rate and the unsecured rate move by the same amount between 2 and 31 March, and which one ended the month nearer zero?
- The table below is every dollar reference rate for 24 September 2026. SOFR90D is SOFR compounded over the previous 90 days, a close stand-in for the three-month compounded figure a converted contract uses. Add 26.161bp to it to see what a converted legacy 3-month contract referenced that day.
- In one sentence each: name one thing SOFR does better than LIBOR, and one thing it does worse. Both sentences should be descriptive, not evaluative.