How Jamshidian Decomposition Algorithms Turn a Swaption into Bond Options: Critical-Rate Root Finding, One-Factor Monotonicity, Hull–White Calibration and Failure Cases

Reader question: A swaption depends on many future fixed-leg cash flows. Why can a one-factor interest-rate model sometimes price that whole option as a sum of simple zero-coupon bond options?

Jamshidian decomposition works because a one-factor term-structure model can make every relevant zero-coupon bond price move monotonically with the same single state variable. At the exercise date, one scalar “critical state” can therefore make the entire coupon-bond value equal the option strike. Once that state is found, the option on the portfolio separates into options on its individual cash flows.

This article owns the one-factor coupon-bond/swaption decomposition problem: one-factor bond-pricing model + exercise date + deterministic cash flows + strike → critical state, component bond strikes, zero-coupon bond option values and an exact portfolio option value under the decomposition assumptions.

It does not own general multi-factor swaption pricing, smile calibration, Bermudan exercise or the construction of the underlying yield curves. Those are separate jobs in the Bukit Timah Tutor finance-and-banking-algorithms lane.

This is public mathematical and computational education. It is not financial advice or a recommendation to trade interest-rate derivatives.

1. Start with an option on a deterministic-cash-flow bond

Let the option exercise at time T₀. After that date, the underlying instrument pays deterministic positive cash flows:

c₁, c₂, …, c_m

at dates:

T₁ < T₂ < … < T_m.

At exercise, the coupon-bond value is:

B(T₀,x) = Σ c_iP(T₀,T_i;x),

where x is the model’s one-dimensional state at exercise and P(T₀,T_i;x) is the zero-coupon bond price conditional on that state.

A European call on the coupon bond with strike K pays:

max[B(T₀,x) − K, 0].

2. Why the problem looks high-dimensional

The underlying bond contains many maturities. If each zero-coupon bond could move independently, the option payoff would depend on a vector:

(P₁, P₂, …, P_m).

An option on a sum of correlated bond prices would then require their joint distribution.

Jamshidian’s insight is that in a suitable one-factor model the entire vector is driven by one state. Instead of many independent dimensions, every component is a deterministic function of the same x.

3. The monotonicity that makes the trick work

In common one-factor short-rate models such as Vasicek and Hull–White, a zero-coupon bond price at the exercise date is strictly decreasing in the relevant rate/state variable:

∂P(T₀,T_i;x)/∂x < 0.

If all coefficients c_i are nonnegative, then the whole coupon-bond value:

B(T₀,x) = Σ c_iP_i(x)

is also decreasing in x.

That gives a one-dimensional exercise boundary.

4. Find the critical state x*

Solve:

Σ c_iP(T₀,T_i;x*) = K.

Because the bond value is monotone, the root is unique when the strike lies inside the attainable range.

For a coupon-bond call:

• if x < x*, rates are lower, bond prices are higher and the option is in the money;
• if x > x*, bond prices are lower and the option is out of the money.

The entire exercise region has therefore been reduced to one scalar threshold.

5. Turn the critical state into component strikes

For every cash-flow maturity, define:

K_i = P(T₀,T_i;x*).

By construction:

Σ c_iK_i = K.

These K_i are not arbitrary allocations of the total strike. They are the zero-coupon bond prices that occur simultaneously at the common critical state.

6. The payoff identity

When all component bond prices cross their component strikes at the same state x*, they are either all above their strikes or all below them together.

Therefore:

max[Σ c_iP_i − K, 0] = Σ c_i max[P_i − K_i, 0].

This is the decomposition.

The option on a coupon-bond portfolio becomes a weighted sum of options on individual zero-coupon bonds.

It is exact under the assumptions—not an approximation created by linearising the payoff.

7. Why the identity would usually be false

For arbitrary random variables,:

max(X+Y−K,0)

is not equal to a sum of separate calls on X and Y.

The identity works here only because one-factor monotonicity synchronises the exercise decision across every component. A single state makes all component options switch from out-of-the-money to in-the-money together.

This is why the decomposition is structural, not algebraic luck.

8. From coupon-bond options to swaptions

In a traditional single-curve representation, the value at swap start of a floating leg that resets at that date can be related to par. A vanilla fixed-for-floating swap can therefore be rearranged into a difference between a par amount and a coupon-bond-like fixed leg.

A receiver swaption—the right to receive fixed—maps naturally to a call on the fixed-leg bond representation.

A payer swaption—the right to pay fixed—maps to the corresponding put structure.

Jamshidian decomposition then prices the coupon-bond option as a sum of zero-coupon bond options inside a compatible one-factor term-structure model.

9. Hull–White is a natural implementation

In the one-factor Hull–White model, zero-coupon bond prices have the affine form:

P(t,T) = A(t,T) exp[−B(t,T)r_t]

under a common short-rate representation.

For positive B(t,T), the bond price is monotone decreasing in the short rate. This provides exactly the common one-dimensional ordering Jamshidian decomposition needs.

Hull–White also has analytic zero-coupon bond option formulas, so after the critical root is found the remaining components can be valued quickly.

10. The computational pipeline

1. Build the exercise-date deterministic cash-flow schedule.
2. Choose the one-factor short-rate model and calibrated parameters.
3. Implement P(T₀,T_i;x) for every maturity.
4. Define the root function:

f(x) = Σ c_iP(T₀,T_i;x) − K.

5. Bracket and solve f(x*) = 0.
6. Set K_i = P(T₀,T_i;x*).
7. Price each zero-coupon bond option.
8. Weight by c_i and sum.
9. Reconcile against an independent tree, PDE or integration engine.

11. Root finding is small but critical

The root solve is only one-dimensional, but a broken root finder breaks every component strike.

Useful choices include bisection, Brent’s method and safeguarded Newton methods.

Because monotonicity is known, bracketing is especially powerful. Once two states are found with opposite signs:

f(x_L)f(x_U) < 0,

the root is bracketed.

A pure Newton method can be faster but can fail if the initial guess is poor, the derivative is tiny or numerical bond-price errors distort the slope.

12. Root invariants

After solving, verify:

|Σ c_iK_i − K| < tolerance.

Also verify every component:

K_i > 0

for ordinary positive zero-coupon bond prices.

If the strike reconstruction does not match, the decomposition has not actually been implemented.

13. A simple three-cash-flow illustration

Suppose the exercise-date underlying bond is:

B(x) = 4P₁(x) + 4P₂(x) + 104P₃(x)

and the option strike is 100.

Solve:

4P₁(x*) + 4P₂(x*) + 104P₃(x*) = 100.

If the root implies:

K₁ = P₁(x*), K₂ = P₂(x*), K₃ = P₃(x*),

then the coupon-bond call value is:

4C_ZCB,1(K₁) + 4C_ZCB,2(K₂) + 104C_ZCB,3(K₃).

The example hides no multidimensional integration. The high-dimensional-looking payoff has collapsed to one critical state plus simple component options.

14. Why this is valuable for calibration

Calibration repeatedly evaluates model prices while changing parameters. A slow inner pricing engine makes the outer calibration loop expensive.

Jamshidian decomposition can turn a European swaption price under a one-factor model into:

• one scalar root solve;
• a series of closed-form zero-coupon bond option evaluations.

That makes it a useful benchmark and calibration engine for one-factor Gaussian short-rate models.

15. Inputs and outputs

Inputs can include:

• exercise date;
• cash-flow dates and amounts;
• option strike or fixed swap rate;
• discount curve;
• one-factor model parameters;
• zero-coupon bond pricing function;
• zero-coupon bond option formula;
• payer/receiver convention;
• root bracket and tolerance;
• settlement convention.

Outputs can include:

• critical state x*;
• component strikes K_i;
• component bond-option values;
• total swaption/bond-option value;
• root residual;
• strike-reconstruction residual;
• model Greeks;
• benchmark differences versus alternative engines.

16. Evidence polarity

Evidence for confidence includes a unique well-bracketed critical root, tiny strike-reconstruction residual, monotonic component bond prices, agreement with a one-factor tree or PDE benchmark, stable prices as the root tolerance is tightened, and consistent payer/receiver parity relationships under the same conventions.

Evidence against confidence includes multiple critical roots, non-monotone bond values, negative or sign-changing cash-flow coefficients, large dependence on the root bracket, disagreement with an independent model implementation, decomposition residuals caused by multi-curve cash-flow representation, or use on exercise structures the formula does not cover.

17. Counterexample: two factors destroy the common threshold

Suppose bond prices depend on two independent state variables, for example level and slope.

The exercise boundary becomes a curve in two-dimensional state space rather than one scalar x*. One maturity can move above its component strike while another does not.

The synchronized-exercise identity breaks.

Falsifier: vary the second factor while holding the first at the supposed critical value. If the coupon-bond payoff changes sign, there is no single universal critical state.

18. Counterexample: negative cash-flow coefficients

The standard decomposition relies on the portfolio value being monotone in the common state. Positive coupon weights preserve the monotonic direction of the component bonds.

If a structured portfolio contains positive and negative coefficients, the sum can stop being monotone even though each individual bond remains monotone.

Falsifier: evaluate f(x) over a wide state range. If its derivative changes sign or multiple roots appear, a simple Jamshidian decomposition is invalid.

19. Counterexample: modern multi-curve valuation

After the financial crisis, collateral discounting and separate projection curves became standard in many interest-rate markets. The floating leg may no longer collapse to the same simple single-curve identity used in textbook swaption derivations.

A dedicated multi-curve model can sometimes recover analogous formulas or approximations, but the classical decomposition must not be transplanted without checking the cash-flow representation.

Falsifier: independently value the fixed and floating legs under the intended discount/projection framework at exercise. If the supposed coupon-bond representation does not reproduce the swap payoff, stop before applying the decomposition.

20. Counterexample: Bermudan exercise

A Bermudan swaption has several possible exercise dates. Today’s decision depends on future continuation values, not only the payoff at one date.

Jamshidian decomposition is a European-exercise result. It can provide calibration benchmarks or values at individual exercise dates, but it does not solve the backward optimal-stopping problem by itself.

Alternative: use a tree, lattice, PDE or suitable Monte Carlo optimal-stopping method.

21. Counterexample: exact formula, wrong model

Jamshidian decomposition can price a claim exactly inside a one-factor Gaussian short-rate model while the model still fails to capture observed swaption smile, skew or multi-factor yield-curve movements.

Exactness is conditional on the model.

Falsifier: compare model prices across the volatility surface and test hedging errors. If residuals have structure, the decomposition is not the problem—the one-factor model may be.

22. Diagnostics

• Monotonicity sweep: sample B(x) over a wide range.
• Root bracketing: require opposite signs before solving.
• Strike reconstruction: verify Σc_iK_i = K.
• Component parity: check zero-coupon call/put parity.
• Tree benchmark: compare the same one-factor model on a fine lattice.
• PDE benchmark: compare where an independent PDE implementation exists.
• Convention audit: payment dates, day counts, settlement and notional must agree.
• Curve audit: confirm single-curve or multi-curve assumptions explicitly.
• Parameter stress: sweep mean reversion and volatility.
• Cash-flow sign test: flag negative weights before solving.

23. Alternatives

Short-rate trees handle European and Bermudan exercise and are straightforward validation engines.

Finite-difference short-rate PDEs provide another independent numerical benchmark.

Gaussian one-factor integration can price some claims directly by one-dimensional quadrature.

Multi-factor Gaussian models capture more yield-curve movements but lose the one-dimensional Jamshidian threshold.

Market models such as LMM-style frameworks model forward rates directly and can handle richer swaption structures at greater computational cost.

24. Connections to the surrounding Bukit Timah Tutor knowledge estate

The natural model companion is Hull–White calibration algorithms. Jamshidian decomposition provides a fast European swaption pricing route inside compatible one-factor calibrations.

Market quoting connects to Black-76 and Bachelier swaption algorithms, which describe normal/lognormal quote conventions rather than one-factor short-rate dynamics.

The broader no-arbitrage term-structure view is Heath–Jarrow–Morton algorithms.

Modern collateralised curve construction connects to collateralised swap multi-curve algorithms.

The full lane is indexed at Finance & Banking Algorithms | Applied Mathematics in Real Financial Systems.

25. What would falsify confidence?

Confidence should be withdrawn if the underlying cash-flow portfolio is not monotone in one common state; if the critical root is not unique; if component strikes do not reconstruct the total strike; if a multi-curve payoff is forced into an invalid single-curve identity; if the exercise is Bermudan/American; or if an independent implementation of the same one-factor model disagrees beyond numerical tolerance.

26. Verification and update triggers

Preserve the cash-flow schedule, curves, one-factor model parameters, root bracket, root tolerance, critical state, component strikes and component option values for every run.

Revalidate after changes to curve architecture, collateral conventions, swap definitions, model state variable, zero-coupon bond formula, bond-option formula, root solver or settlement convention. Trigger review immediately when negative cash-flow weights, multi-factor dynamics or multiple exercise dates enter the product scope.

27. Primary and high-quality references

• Farshid Jamshidian, An Exact Bond Option Formula, The Journal of Finance, 1989.
• Marc Henrard, Explicit bond option and swaption formula in Heath–Jarrow–Morton one factor model, 2003.
• QuantLib, open-source quantitative-finance library; its pricing-engine family includes a Jamshidian swaption engine that is useful as an implementation reference and cross-check.
• Hélyette Geman, Nicole El Karoui and Jean-Charles Rochet, Changes of numéraire, changes of probability measure and option pricing, Journal of Applied Probability, for the wider measure-change framework behind bond-option valuation.

Educational boundary: Jamshidian decomposition is a precise dimensionality-reduction result under one-factor monotonicity. It should be treated as exact within its assumptions, not as a universal swaption algorithm.