Reader question: A three-month SOFR futures price implies an interest rate. Why can’t a curve-building system simply treat that implied futures rate as the forward rate for the same future period?
Because a futures contract and a forward-style interest-rate agreement do not have the same cash-flow timing. Futures are marked to market through variation margin during the life of the contract. A forward typically settles its economic result at or near the underlying period. When interest rates are stochastic, the timing of those interim cash flows can create a valuation difference.
The numerical bridge between the two is commonly called a convexity adjustment.
What this page owns — and what it does not
This page owns the conversion problem:
SOFR-futures market quote + model assumptions → forward-rate-consistent curve input.
It does not replace SOFR benchmark construction, swap valuation, or general yield-curve bootstrapping. Those pages own the benchmark, valuation and curve machinery around this specific adjustment.
This is mathematical fixed-income education, not a recommendation to trade SOFR futures, swaps or any other derivative.
Start with what a SOFR futures price means
CME Group quotes SOFR futures using an IMM-style index:
Price = 100 − R.
For a Three-Month SOFR futures contract, R is tied at final settlement to the realized compounded SOFR rate over the contract’s reference quarter. CME’s current educational materials state that Three-Month SOFR futures settle to 100 minus the realized compounded daily SOFR over the reference period.
So a price of 95.75 corresponds to an implied contract rate of 4.25%.
That implied rate is an observable market price transformation. It is not automatically the same thing as a model-free forward rate.
Why futures and forwards diverge
Consider two contracts referencing the same future three-month interest period.
- Futures: gains and losses are realized through daily variation margin.
- Forward/FRA/OIS-style exposure: the contract’s economic payoff is concentrated at its contractual settlement/payment time.
If rates move, futures gains may arrive early and be reinvested, while futures losses may require funding early. The value of those early cash flows depends on the path of interest rates.
Therefore:
futures-implied rate = forward-consistent rate + convexity effect
under a chosen sign convention.
The exact adjustment is model-dependent.
Daily margining is the source of path dependence
Suppose rates rise sharply. A short position in an interest-rate futures contract may receive variation margin as the futures price falls. That cash arrives while interest rates are high and can potentially earn more.
If rates fall, the opposite margin pattern occurs.
The correlation between the contract’s mark-to-market changes and the rate at which margin cash is funded or reinvested creates a non-linear expected-value effect.
This is why the adjustment is called convexity-related: it emerges from the interaction of stochastic rates, non-linear discounting and cash-flow timing rather than from a fixed deterministic spread.
CME explicitly treats convexity adjustment as necessary for swap pricing from SOFR futures
In a 2025 CME Group article on pricing and hedging USD SOFR swaps with SOFR futures, CME states that SOFR futures are a major source of short-term interest-rate price discovery and that convexity adjustments are necessary to pricing when translating futures into swap-curve information.
This is a useful boundary condition: the market futures quote is an input, but the translation into a forward projection curve requires additional modelling.
A useful computational pipeline
- Read the futures price.
- Convert price to the contract-implied rate.
- Map the contract to its exact SOFR reference dates and day-count structure.
- Specify the discounting framework and current OIS curve.
- Choose a stochastic interest-rate model or approved adjustment methodology.
- Calibrate model parameters such as volatility and mean reversion to appropriate market data.
- Compute the futures-versus-forward valuation difference.
- Adjust the futures-implied rate into a forward-consistent curve input.
- Rebuild the projection curve and reprice benchmark instruments.
- Check residuals and model sensitivity.
The adjustment therefore belongs inside a calibration-and-validation loop, not in a hard-coded lookup table.
A short-rate-model view
One common framework uses a one-factor Gaussian short-rate model such as Hull–White. Schematically:
dr(t) = [θ(t) − a r(t)]dt + σ dW(t).
The parameter a controls mean reversion and σ controls short-rate volatility. The time-dependent drift function is chosen so the model can fit the initial term structure.
Given the calibrated model, the algorithm can evaluate the expected effect of daily settlement and derive the forward/futures difference for the contract’s dates.
The exact closed-form expression depends on the modelling conventions. A public educational implementation should avoid pretending that one universal formula applies to every futures contract, collateral setup and curve framework.
Why the adjustment usually grows with horizon and volatility
Intuitively, if the reference period is very near, there is little time for stochastic rate paths and margin reinvestment to create a meaningful difference.
If interest-rate volatility is near zero, futures and forwards converge because the path uncertainty disappears.
Longer horizons and greater volatility generally create more room for the margining effect to matter.
Mean reversion changes how shocks persist, so it also changes the adjustment.
But the sign should not be hard-coded as a universal law
Market practitioners often discuss a common sign pattern under standard positive-rate Gaussian assumptions, but a robust algorithm should compute the adjustment from its model rather than encode “always add X” or “always subtract X.”
The sign and magnitude can depend on:
- model structure;
- rate level and volatility;
- contract dates;
- discounting assumptions;
- margin remuneration;
- curve shape;
- calibration data.
The falsifier is simple: if changing the model or parameters can legitimately change the sign, the production code must not treat sign as a constant business rule.
One-Month and Three-Month SOFR futures are not identical products
CME’s current materials distinguish:
- One-Month SOFR futures: based on the arithmetic average of daily SOFR across the delivery month;
- Three-Month SOFR futures: based on compounded daily SOFR across the reference quarter.
The exact reference periods and final-settlement formulas differ. A curve engine must therefore know the contract family before extracting a rate.
Using a Three-Month compounding formula on a One-Month contract is a contract-specification error, not a small numerical approximation.
The futures rate itself also contains benchmark compounding mechanics
CME’s Three-Month SOFR settlement calculation applies daily compounding over the reference quarter using SOFR observations and actual calendar-day spans on a 360-day money-market basis.
Therefore two layers must be separated:
- contract-rate construction: how realized SOFR produces the final futures settlement rate;
- convexity adjustment: how a futures-market expectation relates to a forward-consistent rate before realization.
Confusing those two problems produces double adjustments.
Calibration: where model risk enters
A Hull–White-style adjustment requires model parameters. Those parameters are not observable physical constants.
Possible calibration sources include:
- caps and floors;
- swaptions;
- other short-rate option markets;
- historical rate dynamics, if permitted by the methodology;
- internally governed parameter sets.
Different calibration choices can produce different convexity adjustments even from the same futures price.
Therefore the algorithm should preserve the calibration date, source instruments, objective function, parameter bounds and residuals.
Model risk is not the same as market-data error
If a futures price is stale, the input is wrong.
If the futures price is correct but the model’s volatility or mean-reversion assumptions are poor, the adjustment can still be wrong.
These need separate diagnostics:
- market-data quality;
- contract-specification quality;
- model calibration quality;
- curve-repricing quality.
A practical verification invariant
After adjusted futures inputs are used to build the projection curve, the system should reprice:
- the futures contracts under the chosen conversion framework;
- nearby OIS or swap instruments;
- other curve calibration instruments.
If the resulting curve cannot reproduce the market instruments within the methodology’s tolerances, the convexity treatment, calibration or curve construction needs review.
Inputs and outputs
A convexity-adjustment engine can require:
- futures contract identifier;
- market futures price;
- reference-period dates;
- SOFR contract family;
- discount curve;
- initial projection curve;
- short-rate model choice;
- volatility parameters;
- mean-reversion parameters;
- margining assumptions;
- calibration instruments and timestamps;
- numerical tolerance.
Outputs can include:
- futures-implied contract rate;
- convexity adjustment in basis points;
- forward-consistent rate;
- model parameters;
- calibration residuals;
- sensitivity to parameter perturbations;
- diagnostic warnings.
Evidence polarity: what supports confidence?
Evidence for confidence includes exact contract-date mapping, reproduction of CME final-settlement mechanics, stable adjustment under nearby calibration choices, good option-market calibration, successful curve repricing, small residuals and independent benchmark comparison.
Evidence against confidence includes a constant hard-coded basis-point adjustment, wrong contract family, stale volatility data, large sensitivity to starting values, poor calibration residuals, forward-curve jumps at futures boundaries, or a curve that prices futures but breaks nearby OIS instruments.
Counterexample: a futures price can be accurate and the derived forward can still be wrong
The market futures price may be perfectly observed. If the conversion engine assumes zero convexity adjustment when the effect is material, the forward curve can be biased.
Market-data accuracy does not validate the transformation.
Counterexample: a beautiful model fit can still use the wrong contract dates
If the engine assigns a futures contract to the wrong IMM reference quarter, an optimizer may still find model parameters that minimize errors elsewhere.
The calibration can look numerically successful while the economics are wrong.
Contract metadata should therefore be validated before optimization.
Counterexample: fitting one curve instrument does not identify the model
A single futures quote cannot uniquely determine both a full volatility structure and a mean-reversion parameter.
Convexity adjustment is therefore not an observable derived solely from one price. It is a model-based inference requiring additional calibration information.
Counterexample: replacing every futures rate with an OIS forward may double-count corrections
If a curve framework already embeds a provider’s approved futures-to-forward transformation, applying a second external convexity adjustment can distort the curve.
The system must preserve exactly where the adjustment is applied and ensure it occurs once.
Diagnostics: how to test the engine
- zero-volatility test: send volatility toward zero and verify the convexity effect collapses appropriately.
- short-horizon test: move the contract near its reference period and verify the adjustment behaves sensibly.
- contract-family test: distinguish One-Month arithmetic-average and Three-Month compounded contracts.
- date test: verify exact IMM/reference-period mapping.
- settlement replay: reproduce CME final settlement from realized SOFR observations.
- parameter stress: perturb volatility and mean reversion and observe monotonic/sensible sensitivity.
- model comparison: compare an approved alternative model or numerical method.
- curve repricing: rebuild the curve and check futures plus nearby OIS instruments.
- stale-data test: freeze option-volatility inputs while the futures strip moves and require a diagnostic.
- double-adjustment test: ensure the transformation is applied exactly once.
What would falsify confidence?
Confidence should be withdrawn if the engine cannot reproduce official contract settlement mechanics; if adjustments remain material as model volatility tends to zero without a documented reason; if curve results depend wildly on optimizer starting values; if reference dates are wrong; if calibration residuals exceed governance limits; or if applying the adjustment worsens consistency with observable swaps and OIS.
Alternatives and limits
A deterministic curve can ignore convexity for very short horizons when the effect is demonstrably immaterial under an approved policy. More advanced frameworks can use multi-factor short-rate or HJM-style models. Some institutions can calibrate the entire curve and volatility system jointly.
More complexity is not automatically better. The correct model is the simplest one that meets the intended accuracy, hedging and governance requirements with transparent diagnostics.
How this connects to the surrounding knowledge estate
SOFR benchmark algorithms define the realized reference rate. Day-count algorithms supply the accrual basis. Yield-curve construction consumes adjusted futures information. Swap valuation tests whether the resulting curve prices the surrounding market coherently.
Verification and update triggers
Preserve CME contract specifications, reference dates, calibration market data, model version, parameter bounds, optimizer settings, margining assumptions and curve-repricing diagnostics. Revalidate after CME contract changes, major volatility-regime shifts, OIS-curve framework changes, model-library upgrades, benchmark methodology changes or unexplained futures-versus-swap basis changes.
Primary and high-quality references
- CME Group, Pricing and Hedging USD SOFR Interest Rate Swaps with SOFR Futures, 2025.
- CME Group, Explaining the SOFR Settlement Process.
- CME Group, Understanding SOFR Futures.
- CME Group, Three-Month SOFR Futures Rates and Future SOFR Levels.
Educational boundary: This article explains futures-to-forward modelling and curve construction. It does not recommend derivatives positions or provide personalized financial advice.