Quant Memo
Core

SOFR and Risk-Free Rate Benchmarks

SOFR replaced LIBOR as the dominant US interest-rate benchmark because it's anchored in actual, large overnight Treasury-repo transactions rather than a bank's estimate of its own borrowing cost.

Prerequisites: Repo and Reverse Repo, How the Treasury Market Works

For decades, trillions of dollars of loans, mortgages, and derivatives were priced off LIBOR — a rate produced by asking a panel of banks what they thought they would pay to borrow unsecured cash from each other, even on days when they'd made almost no such trades. When banks were caught manipulating those submissions, and the underlying unsecured lending market kept shrinking, regulators pushed for a benchmark built the opposite way: from an enormous volume of actual, observable transactions.

SOFR (the Secured Overnight Financing Rate) is the interest rate on overnight Treasury repo, measured directly from hundreds of billions of dollars of real trades every day. It replaced LIBOR because it's based on what actually happened, not a bank's opinion, and because it's secured, it sits below unsecured bank borrowing rates.

What SOFR measures, and what it doesn't

SOFR is calculated from the same repo market covered elsewhere: it's the volume-weighted median rate across a broad set of overnight Treasury-repo transactions. Because that market is huge and transaction-based, SOFR is hard to manipulate and reflects genuine funding conditions in real time.

But that construction also creates a difference from LIBOR that matters for pricing: LIBOR included an implicit bank credit-risk premium — you were lending unsecured to a bank, so the rate embedded some chance of that bank defaulting. SOFR is fully collateralized by Treasuries, so it carries essentially no credit risk. In normal times SOFR sits below where LIBOR would have been; in a banking-stress episode, that gap can widen sharply as bank credit risk rises while Treasury repo stays calm.

LIBOR bank survey, unsecured, estimate not trades SOFR real repo trades, secured, large volume both feed floating-rate loans, swaps, and mortgages
Same job — a floating benchmark rate — built on fundamentally different foundations.

Worked example

A $10 million floating-rate loan resets quarterly at "SOFR + 150bp." Over one quarter, compounded daily SOFR averages 5.05%. The all-in rate for the quarter is 5.05%+1.50%=6.55%5.05\% + 1.50\% = 6.55\% annualized. Interest owed for the quarter (approximately 91 days):

10,000,000×0.0655×91360=165,59710{,}000{,}000 \times 0.0655 \times \frac{91}{360} = 165{,}597

The borrower owes roughly $165,597 for the quarter. If the loan had instead referenced 3-month LIBOR at, say, 5.20% (reflecting a bank-credit premium LIBOR would have embedded), the spread over the risk-free base would typically be set lower to compensate — the two benchmarks aren't meant to produce identical all-in rates, just to be internally consistent within their own market conventions.

What this means in practice

Because SOFR is an overnight rate with no forward-looking term structure of its own, using it in loans that need a rate fixed in advance (like LIBOR's 3-month rate was) requires either compounding it in arrears over the period or using a forward-looking SOFR term rate derived from swaps — a genuine complication that took years for markets to standardize around.

SOFR is a secured rate and LIBOR was unsecured — they are not directly comparable levels, and "SOFR is lower than LIBOR was" is a structural fact about what each rate is collateralized by, not evidence that funding got cheaper.

Related concepts

Practice in interviews

Further reading

  • ARRC, 'A User's Guide to SOFR'
  • Duffie and Stein, 'Reforming LIBOR and Other Financial Market Benchmarks'
ShareTwitterLinkedIn