How Black-76 and Bachelier Swaption Algorithms Price Rate Optionality: Forward Swap Rates, Annuities, Lognormal vs Normal Volatility and Negative-Rate Diagnostics

Reader question: A swaption gives the right to enter an interest-rate swap in the future. How can a complicated stream of future fixed and floating cash flows be compressed into a Black-76 or Bachelier option formula — and why did negative interest rates make the distinction between lognormal and normal volatility operationally important?

The central reduction is that, under the standard swaption approximation, the option is written on a forward swap rate and scaled by the present value of the swap’s fixed-leg payment dates, called the swap annuity. Black-76 then assumes the forward swap rate evolves lognormally, while Bachelier assumes it evolves normally. The formulas can give similar prices near ordinary positive rates when volatilities are translated consistently, but they are not interchangeable because their state spaces, volatility units and tail behaviour differ.

What this page owns — and what it does not

This page owns the pricing reduction:

discount curve + forward swap rate + annuity + strike + expiry + volatility convention → payer/receiver swaption value.

It does not replace interest-rate swap valuation, SABR smile calibration, implied-volatility inversion, or Black–Scholes equity-option mechanics.

This is derivatives mathematics, not a recommendation to trade swaptions or interest-rate risk.

Step 1: calculate the forward swap rate

Consider a swap beginning at option expiry T and paying fixed coupons on dates T₁, …, T_m. Let α_j be fixed-leg accrual fractions and P(0,T_j) discount factors.

The fixed-leg annuity is:

A(0) = Σ_j=1^m α_jP(0,T_j).

For a standard par swap, the forward swap rate can be written schematically as:

F = [P(0,T_start) − P(0,T_end)] / A(0),

with multi-curve implementations replacing the simple numerator by the present value of the projected floating leg.

This is why the swap-valuation engine is upstream of the swaption formula.

Step 2: express the option payoff

A payer swaption gives the right to pay fixed at strike K and receive floating. At exercise, its intrinsic value is proportional to:

A(T) × max(S(T) − K, 0),

where S(T) is the then-current forward/par swap rate for the underlying swap.

A receiver swaption has intrinsic value:

A(T) × max(K − S(T), 0).

Under the annuity-measure approximation, the current annuity factors out of the option expectation, leaving a vanilla option on the forward swap rate.

Black-76 payer swaption formula

Let:

• F = current forward swap rate;
• K = strike;
• σ_B = Black/lognormal volatility;
• T = option expiry in years;
• A = current swap annuity.

For positive F and K, the Black-76 payer value is:

V_payer = A[FΦ(d₁) − KΦ(d₂)],

where:

d₁ = [ln(F/K) + 0.5σ_B²T] / (σ_B√T),

d₂ = d₁ − σ_B√T.

The receiver formula is:

V_receiver = A[KΦ(−d₂) − FΦ(−d₁)].

Fischer Black’s 1976 forward/futures option framework is the mathematical ancestor of this widely used interest-rate option formula.

What “lognormal” means here

Black-76 models proportional movements:

dF ≈ σ_BF dW.

The instantaneous volatility scale is proportional to the current rate level. A 20% Black volatility means roughly “twenty percent of the current forward rate per square-root year,” not a twenty-percentage-point absolute rate move.

The unshifted lognormal model also keeps a positive process positive. That property became a practical limitation when many market interest rates moved near or below zero.

Bachelier payer swaption formula

Under the normal/Bachelier model:

dF ≈ σ_NdW,

where σ_N is an absolute rate volatility.

Define:

d = (F − K)/(σ_N√T).

The payer value is:

V_payer = A[(F − K)Φ(d) + σ_N√T φ(d)].

The receiver value is:

V_receiver = A[(K − F)Φ(−d) + σ_N√T φ(d)].

QuantLib’s public Bachelier implementation explicitly warns that normal volatility is an absolute volatility rather than a percentage volatility.

The unit difference is not cosmetic

Suppose a forward swap rate is 2%.

A Black volatility of 25% means a lognormal standard deviation scale proportional to 0.02.

A normal volatility might instead be quoted as, for example, 50 basis points per square-root year:

σ_N = 0.0050.

Putting “25” or “0.25” directly into a Bachelier function that expects absolute rate units can produce a catastrophic price error.

Volatility must therefore carry a convention and a unit, not just a number.

Why negative rates break unshifted Black-76

The Black formula uses:

ln(F/K).

If the forward or strike is zero or negative, that logarithm is undefined in the ordinary real-valued formula.

The Bachelier model has no such positivity constraint because it models additive normal moves. Negative forward rates remain mathematically admissible.

This is one reason normal-volatility quoting became important in interest-rate markets during negative-rate regimes.

Shifted lognormal as a third route

A shifted Black model introduces displacement s:

F’ = F + s, K’ = K + s.

Black-76 is then applied to F’ and K’, with the shift chosen so they stay positive over the relevant domain.

MathWorks’ public Black-model documentation describes this explicitly: the shift moves the lognormal lower bound from zero to −s.

Shifted lognormal retains multiplicative-style dynamics above the shifted floor; Bachelier uses additive normal dynamics with no finite lower bound.

At-the-money comparison

When F ≈ K and volatility is not extreme, the two models can be matched approximately by equating their near-ATM price scales.

A useful intuition is:

σ_N ≈ F × σ_B

near the money, with units handled consistently.

This is only a local approximation. Away from the money, Black and Bachelier assign different tails and therefore different prices even if ATM premiums are matched.

Model choice changes the smile representation

A volatility smile is a map from strike/expiry/tenor to implied volatility.

But “implied volatility” is model-dependent:

• Black implied vol is a percentage/lognormal number;
• Bachelier implied vol is an absolute/normal number;
• shifted-lognormal implied vol additionally depends on the chosen shift.

The same set of option premiums therefore produces different-looking volatility cubes under different quoting models.

This connects to the SABR calibration page, which explains how smile parameters are fitted rather than assuming flat volatility.

Payer–receiver parity

A powerful model-independent diagnostic is payer–receiver parity:

V_payer − V_receiver = A(F − K),

under consistent settlement and discounting conventions.

This follows from:

max(F−K,0) − max(K−F,0) = F−K.

Both Black and Bachelier formulas should satisfy the parity identity numerically. A failure often signals sign, annuity, discounting or option-type errors.

Zero-volatility limit

As volatility tends to zero:

Option value → A × intrinsic forward value.

For a payer:

V → A max(F−K,0).

This is another high-information test because it catches formulas that accidentally preserve time value when uncertainty has been removed.

Large-expiry or extreme-strike behaviour

Black and Bachelier differ materially in their tails.

The normal model permits large negative and positive rate moves symmetrically in absolute units. The lognormal model prevents the shifted process from crossing its lower bound and produces multiplicative tails.

Far-from-the-money options are therefore sensitive to model choice even when near-ATM premiums look similar.

A risk system should not convert a whole volatility cube from normal to Black using one ATM scaling rule.

Cash versus physical settlement

ISDA’s 2021 Interest Rate Derivatives Definitions distinguish physical and cash settlement mechanisms for swaptions.

The simple annuity-times-option formula is a pricing representation, not a replacement for the legal settlement terms. Cash settlement can use elected fallback methodologies if parties do not mutually agree a cash settlement amount, while physically settled swaptions produce the underlying swap according to the contract.

A pricing engine must therefore map the correct contract specification before choosing a formula.

Inputs and outputs

A swaption formula engine can require:

• option type: payer or receiver;
• expiry date/time;
• underlying swap start/end dates;
• fixed-leg accrual schedule;
• discount curve;
• forward curve(s);
• forward swap rate;
• swap annuity;
• strike;
• volatility;
• volatility convention: Black, normal or shifted Black;
• shift where applicable;
• settlement method.

Outputs can include:

• payer/receiver present value;
• intrinsic and time value;
• delta, vega and rate sensitivities;
• implied volatility under the chosen convention;
• parity residual;
• model/quote convention metadata.

Evidence polarity: what supports confidence?

Evidence for a correct implementation includes exact payer–receiver parity, correct zero-volatility limits, agreement with independent QuantLib or analytic formulas, prices increasing with volatility, correct absolute-vs-percentage volatility units, and stable handling of zero/negative forwards under models that support them.

Evidence against confidence includes taking logarithms of nonpositive rates, normal-vol numbers interpreted as percentages, Black and Bachelier prices being identical at all strikes after one simple rescaling, parity breaks, negative time value, or discontinuous prices when a shifted model crosses an arbitrary implementation threshold.

Counterexample: “normal vol is 50%” can be meaningless

A Bachelier volatility is normally expressed in absolute rate units, often basis points per square-root year. Writing “50% normal volatility” without a unit convention is ambiguous and likely wrong.

Counterexample: Bachelier is not automatically better because rates can be negative

The normal model solves the positivity problem, but it also permits arbitrarily negative rates and assumes additive volatility. Those dynamics can be a poor description of some markets and horizons.

Model admissibility is not the same as model realism.

Counterexample: shifted Black does not remove model risk

A displacement can make the Black formula numerically valid for negative rates, but the chosen shift affects smile geometry and sensitivities.

Two systems using different shifts can quote different implied volatilities for the same market premium while both reproduce the price.

Counterexample: an implied volatility can exist while the surface is inconsistent

Each individual swaption premium may invert to a volatility, yet the full expiry–tenor–strike surface can violate smoothness or no-arbitrage expectations.

Single-instrument inversion does not validate the volatility cube.

Weak links in implementation

annuity error. Fixed-leg accrual fractions or discount factors are wrong.

curve mismatch. Forward and discount curves are mixed incorrectly.

volatility-unit error. Normal and Black vols are confused.

negative-rate crash. Unshifted Black receives nonpositive forward or strike.

shift inconsistency. Calibration and pricing use different displacements.

expiry-unit error. Calendar days are inserted where year fractions are expected.

settlement mismatch. A physical-settlement valuation assumption is applied to a contract with different cash-settlement terms.

surface interpolation error. Smile interpolation generates unstable or non-monotone option prices.

Diagnostics: how to test the engine

• payer–receiver parity: verify the difference equals annuity times forward-minus-strike.
• zero-vol test: recover discounted intrinsic value.
• ATM test: compare Black and Bachelier after a consistent local volatility conversion.
• negative-forward test: ensure Bachelier works and unshifted Black rejects rather than silently misprices.
• shift test: verify shifted Black remains continuous as rates approach zero.
• vega test: price increases with volatility for a vanilla swaption.
• implied-vol round trip: price → implied vol → price returns the original premium.
• annuity replay: independently reconstruct the fixed-leg PVBP/annuity.
• surface test: evaluate multiple strikes rather than only ATM.
• independent-library test: compare selected cases with a second implementation such as QuantLib.

What would falsify confidence?

Confidence should be withdrawn if payer–receiver parity fails; if a Black implementation accepts invalid nonpositive inputs without an explicit shift; if volatility-unit changes by 100× produce no obvious error; if implied-volatility round trips fail; or if the pricing result depends on an undocumented quote convention.

Alternatives

Black-76 is transparent and efficient for positive forward rates and lognormal quoting.

Bachelier naturally accommodates negative rates and absolute volatility quoting.

Shifted Black preserves lognormal-style dynamics above a shifted lower bound.

SABR adds a strike-dependent volatility smile.

short-rate and market models can price more path-dependent or exercise-sensitive rate options but require additional calibration and numerical machinery.

No single formula owns every interest-rate option problem.

How this connects to the surrounding knowledge estate

The swap engine supplies the forward swap rate and annuity. The SABR engine supplies smile-consistent volatility across strikes. The implied-volatility solver inverts premium back to a quote. AAD can differentiate the valuation efficiently across many curve and volatility inputs.

Verification and update triggers

Preserve the curve snapshot, annuity definition, volatility convention, quote units, shift, expiry year fraction, settlement method and formula version. Revalidate after market-convention changes, negative-rate regime changes, curve architecture changes, volatility-surface migrations or any disagreement with independent premium/implied-volatility round trips.

Primary and high-quality references

• Fischer Black, The Pricing of Commodity Contracts, Journal of Financial Economics 3, 1976.
• QuantLib, Black and Bachelier formula documentation, including the distinction between percentage and absolute volatility.
• MathWorks, Black and Shifted Black model documentation, for the shifted-lognormal lower-bound mechanics.
• ISDA, Key Changes in the 2021 ISDA Interest Rate Derivatives Definitions, for swaption settlement context.
• Patrick Hagan et al., SABR literature, connected to the dedicated SABR article for smile modelling beyond flat Black/Bachelier volatility.

Educational boundary: This article explains swaption option-pricing formulas and volatility conventions. It does not recommend a swaption, swap or interest-rate position and does not provide personalized financial advice.