Guide

SOFR-OIS spread explained

Harbor Capital runs a $140M operating cash sleeve split between Treasury bills, overnight reverse repo, and a committed commercial-paper backup line. In late March 2024, the desk’s risk dashboard flagged nothing unusual: fed funds effective traded inside the Fed’s target band and SOFR futures implied a smooth path of quarter-point cuts. Yet the SOFR-OIS spread — the gap between compounded overnight secured funding and the fixed leg of an overnight index swap — widened from 4 to 22 basis points in five sessions. CP issuance costs for AA-rated financial paper jumped 14 bp while T-bill yields barely moved. The treasury team had been sizing liquidity buffers off policy-rate instruments alone; they were blind to the unsecured funding premium banks were charging each other beneath the surface.

After the episode, Harbor rebuilt its cash playbook around the spread as a primary stress gauge. When SOFR-OIS exceeds a rolling 75th-percentile threshold, the desk shifts 15% of sleeve notional from unsecured CP programs into T-bills and Fed ON RRP, and extends weighted-average maturity on backup lines. Forecast error on weekly cash drag fell from 11 to 3 bp; CP rollover failures during the next two quarter-ends dropped to zero. This guide explains what the SOFR-OIS spread measures, how OIS is constructed post-LIBOR, what drives widening and compression, how it links to bank reserves and the Fed floor, the Harbor Capital refactor, a technique decision table versus watching fed funds alone, pitfalls, and a production checklist.

What the SOFR-OIS spread measures

The spread isolates the premium for unsecured overnight bank funding relative to collateralized repo. Two rates sit on either side:

SOFR (floating leg) — the Secured Overnight Financing Rate: volume-weighted median of tri-party and cleared bilateral repo transactions secured by Treasuries. It reflects what a borrower pays when posting high-quality collateral.
OIS (fixed leg benchmark) — an overnight index swap where one party pays compounded daily SOFR (or historically fed funds) and receives a fixed rate, or vice versa. The fixed rate at inception is the market’s expectation of average overnight funding over the swap tenor — effectively a clean policy-path bet with minimal credit risk because the floating leg is exchanged daily and collateralized through CCP margin.

The SOFR-OIS spread is typically quoted as:

Spread = OIS fixed rate − expected compounded SOFR

In calm markets with abundant reserves, the spread hovers near zero to a few basis points. When banks hoard liquidity, balance-sheet constraints bite, or unsecured interbank markets seize, the spread widens: lenders demand extra compensation beyond the secured rate to lend without collateral.

Do not confuse this with the term SOFR-OIS basis on multi-year swaps — that captures term premia and supply technicals. The overnight SOFR-OIS spread discussed here is a money-market stress indicator, not a long-horizon yield-curve construct.

How OIS is built in the SOFR era

Before LIBOR cessation, USD OIS referenced the effective federal funds rate. Today most new USD OIS use compounded SOFR in arrears, aligned with interest rate swap conventions on floating-rate notes and loans.

Trade mechanics

Counterparties agree on notional, tenor (often 1 week to 2 years for stress hedges; longer for macro trades), and fixed rate.
Each business day, the floating leg accrues compounded SOFR published by the Federal Reserve Bank of New York.
Net payments exchange periodically (monthly or at maturity) through a clearinghouse with variation margin.
At inception, the fixed rate is set so the swap has zero present value — making the fixed rate a market-implied average SOFR path.

Why OIS is the right benchmark

OIS strips out term credit risk because resets are daily and margining limits exposure. Comparing SOFR to OIS therefore highlights short-horizon funding frictions rather than multi-year bank credit premia (those show up in CDS and bond spreads). For treasury teams, a widening SOFR-OIS gap signals that tomorrow’s unsecured rollover may cost more than the secured curve implies — even if the Fed has not moved policy.

What widens and narrows the spread

Drivers fall into four buckets. Track all four on a dashboard; single-factor stories mislead.

1. Reserve abundance vs scarcity

When the banking system holds large excess reserves, banks lend freely in unsecured markets and the spread stays tight. Quantitative tightening drains reserves; as balances approach institutions’ lowest comfortable levels, spreads often creep wider even without a crisis narrative. Watch aggregate reserves in the Fed H.4.1 alongside the spread — not in isolation.

2. Balance-sheet and regulatory constraints

Leverage ratio, supplementary leverage ratio (SLR), and internal risk limits make some banks reluctant to expand repo books at quarter-end. Foreign banks with limited branch deposits may pay more for dollar funding; that premium appears in unsecured spreads before it moves T-bill yields.

3. Calendar and flow technicals

Month-end, quarter-end, and year-end reporting windows produce predictable widening as banks shrink assets for regulatory snapshots. Corporate tax dates (mid-April, mid-June) drain reserves when Treasury general account balances rise. Treasury bill issuance surges can temporarily lift SOFR relative to OIS if repo collateral is scarce — a secured squeeze, not always a credit event.

4. Stress and flight-to-quality

During bank-failure episodes (March 2023 is the recent U.S. template), unsecured interbank volumes thin and spreads blow out while SOFR may stay anchored by Fed facilities. The spread leads headlines: it moves before equity vol spikes and often before fed funds futures reprice the policy path.

Reading the spread alongside other gauges

No single indicator is sufficient. Pair SOFR-OIS with:

FRA-OIS / LIBOR-OIS (historical) — pre-2023 templates; today SOFR-OIS replaced LIBOR-OIS for USD funding stress.
Commercial paper rates minus T-bills — corporate unsecured funding wedge; moves with SOFR-OIS but adds issuer credit tier detail.
Cross-currency basis (EUR/USD) — foreign banks’ dollar funding pressure; widens when offshore dollar liquidity tightens.
Financial Conditions Index — slower-moving composite; use for regime context, not intraday triggers.
ON RRP take-up — high usage can mean money funds lack attractive alternatives; sudden drops may signal bill supply or rate dislocations.

Harbor Capital’s rule: if SOFR-OIS is wider than 10 bp and three-month AA financial CP is more than 8 bp above T-bills, treat the week as “elevated funding stress” regardless of VIX level.

Harbor Capital liquidity desk refactor

Pre-refactor pain points from the March 2024 quarter-end:

Cash forecast used fed funds effective + 2 bp for all overnight instruments
No live SOFR-OIS feed; risk team learned about widening from a broker email
CP program rolled 40% of maturities into unsecured paper on the widest spread day
Backup line draw fee was fixed; no spread-linked tiering

Post-refactor rules (implemented April 2024):

Subscribe to dealer SOFR-OIS fix; archive 1-year rolling distribution
Green (≤ 6 bp): normal CP and repo mix
Yellow (7–15 bp): cap unsecured CP at 25% of sleeve; add 5% T-bill sleeve
Red (> 15 bp): unsecured CP freeze; 100% government repo and T-bills until spread < 10 bp for three sessions
Quarter-end calendar overlay: auto-yellow two days before month-end if reserves weekly change > $80B

Results over six quarter-ends (Q2 2024–Q1 2025):

Weekly cash drag forecast error: 11 bp → 3 bp (median)
CP rollover failures: 2 → 0
Incremental T-bill carry cost in yellow/red weeks: 4 bp annualized (acceptable insurance)
One red-week activation (March 2025 mini-stress); sleeve avoided 19 bp one-day CP spike

Technique decision table

Approach	Best when	Weak when
Watch fed funds effective only	Abundant-reserve regime; simple policy tracking	QT draining reserves; quarter-end; bank stress episodes
SOFR-OIS spread thresholds	Corporate treasury cash sleeves; unsecured CP programs	Pure secured portfolios (only T-bills and ON RRP)
SOFR futures curve	Hedging expected policy path over quarters	Intraweek unsecured rollover risk (too slow)
Financial CDS / bond spreads	Medium-term bank credit risk in bond portfolios	Overnight operational cash (lags by days)
Cross-currency basis hedges	Foreign banks with USD asset books	Domestic-only corporates with no FX exposure
ON RRP + T-bills only (no spread monitor)	Crisis simplicity; zero unsecured exposure	Yield sacrifice in long green regimes (opportunity cost)

Common pitfalls

Confusing secured and unsecured squeezes — a Treasury collateral shortage can lift SOFR without implying bank credit stress; check repo fails and specialness indicators.
Using stale fixes — SOFR publishes at ~8:00 ET; intraday OIS marks from dealers may lead official prints; document which timestamp your triggers use.
Ignoring tenor — overnight SOFR-OIS differs from 3-month OIS; do not apply overnight thresholds to annual swap spreads.
Single-day overreaction — one wide print on a month-end Tuesday is common; require two-session confirmation or percentile ranks before portfolio shifts.
Assuming the Fed will narrow it — facilities (SRF, discount window) cap extremes but spreads can stay elevated in sub-crisis regimes during QT.
Neglecting issuer tier — AA financial CP widens differently from A2/P2 industrial paper; map your program’s rating bucket to historical beta versus SOFR-OIS.
Mixing EUR and USD conventions — ESTR-OIS in Europe measures a different plumbing stack; do not import U.S. thresholds.
Treating zero spread as permanent — post-2008 and post-2020 QE regimes trained teams that spreads “die”; QT normalization reintroduces variability.

Production checklist

Subscribe to daily SOFR-OIS fix from at least two dealers; reconcile differences > 1 bp.
Archive 252-day rolling distribution; compute 50th and 75th percentiles weekly.
Plot SOFR-OIS against aggregate excess reserves (H.4.1) on the same chart monthly.
Mark quarter-end, tax, and Treasury refunding weeks on the calendar overlay.
Document green/yellow/red thresholds with sign-off from treasury and risk committees.
Backtest CP program costs on historical wide-spread weeks (2019 repo spike, March 2023, recent quarter-ends).
Align triggers with fed funds futures implied path changes — policy surprises can widen spreads independently.
Test failover: if dealer feed drops, switch to Bloomberg or NY Fed public data within one hour.
Review spread behavior after each FOMC statement and reserve management note.
Compare unsecured CP reset costs to model prediction monthly; log attribution when error > 5 bp.
Train ops staff on red-week playbook (who approves CP freeze, settlement cutoffs).
Revisit thresholds after major regulatory changes (SLR, Basel endgame).

Key takeaways

SOFR-OIS measures the unsecured overnight funding premium over secured repo — a real-time money-market stress gauge.
OIS fixed rates embed expected compounded SOFR; widening spreads signal liquidity hoarding, balance-sheet constraints, or calendar stress before policy rates move.
Pair the spread with reserves data, CP-T-bill wedges, and quarter-end calendars — not fed funds alone.
Harbor Capital cut cash forecast error from 11 to 3 bp by tiering unsecured exposure off SOFR-OIS thresholds.
Secured repo squeezes and bank credit stress look similar in headlines but require different responses — check collateral specialness before blaming credit.