Hull-White Model: Calibration to the Yield Curve
The Hull-White model is one of the most widely used interest rate models in quantitative finance. Introduced in 1990 by John Hull and Alan White, it extends the classic Vasicek model by allowing a time-varying mean reversion level. This single modification enables the model to fit today’s observed term structure exactly — a critical requirement for pricing and hedging interest rate derivatives in practice.
What Is the Hull-White Model?
The Hull-White model is a one-factor short-rate model that describes how the instantaneous interest rate evolves over time. It belongs to the family of affine term structure models, meaning bond prices have a convenient exponential-affine form that allows for analytical or semi-analytical pricing of many derivatives.
The Hull-White model extends Vasicek by making the mean reversion level time-dependent. This allows the model to match the initial term structure exactly — something the basic constant-parameter Vasicek model cannot do.
Also known as the extended Vasicek model, Hull-White is a widely used practitioner model for pricing swaptions, caps and floors, and other interest rate derivatives. Its popularity stems from combining analytical tractability with the ability to calibrate to market observables.
The Hull-White Stochastic Differential Equation
The Hull-White model describes the evolution of the short rate r(t) using the following stochastic differential equation (SDE):
Where:
- r — the instantaneous short rate
- κ — the speed of mean reversion (constant)
- θ(t) — the time-varying mean reversion level
- σ — the volatility of the short rate (constant)
- dW — increment of a standard Wiener process
In Wilmott’s notation, the Hull-White SDE is written as dr = (η(t) – γr)dt + cdX. This is equivalent: η(t) = κθ(t), γ = κ, and c = σ. Both forms describe the same model — the time-varying drift is the key innovation.
Compare this to the basic Vasicek model, which uses a constant θ instead of θ(t). That single change — making the mean level time-dependent — gives Hull-White its defining capability: exact calibration to the initial term structure.
How the Hull-White Model Fits the Initial Term Structure
The fundamental limitation of constant-parameter models like Vasicek is that they cannot reproduce arbitrary yield curve shapes. With fixed parameters, the model generates a specific family of yield curves — if the market’s curve doesn’t match, your hedging instruments will be mispriced, making consistent risk management impossible.
Hull-White solves this by choosing θ(t) to match the observed initial term structure exactly. Critically, θ(t) is not independent of the other parameters — it is determined given the values of κ and σ. The calibration formula derives θ(t) from the initial forward rate curve and includes a volatility adjustment term involving σ.
Hull-White calibration is an interdependent process, not strictly sequential:
- Bootstrap the discount curve from liquid instruments (swaps, OIS rates, futures)
- Choose κ and σ — typically estimated from cap/floor or swaption implied volatilities
- Derive θ(t) — given κ and σ, calculate θ(t) so that model bond prices match the bootstrapped curve exactly
- Price derivatives under the risk-neutral measure using the fully calibrated model
The key insight: θ(t) depends on κ and σ through the calibration formula. Changing the volatility parameters requires recalculating θ(t).
All Hull-White pricing occurs under the risk-neutral measure, where the drift is adjusted so that discounted bond prices are martingales. The θ(t) function is a risk-neutral parameter — it differs from the physical drift that would describe actual rate movements.
Fitting the initial term structure exactly does not mean the model is “correct.” It ensures consistency with today’s market prices for hedging purposes, but says nothing about whether the model will accurately describe future term structure dynamics.
Bond Pricing Under Hull-White
One reason practitioners favor Hull-White is that zero-coupon bond prices retain the exponential-affine form familiar from Vasicek:
Where:
- B(t,T) — has the same functional form as in Vasicek: B(t,T) = (1 – e-κ(T-t)) / κ
- A(t,T) — absorbs the term-structure fitting adjustment from θ(t), including contributions from the initial forward curve and a volatility adjustment
Different texts use different conventions. Here A(t,T) appears in the exponent directly. Some sources write P(t,T) = A(t,T)e-B(t,T)r, where A is a multiplicative factor. Both are equivalent — the key property is that log bond prices are affine (linear) in r.
This structure is what makes Hull-White tractable. The B(t,T) function depends only on the mean reversion speed κ, while A(t,T) encodes the initial curve information through θ(t). Analytical or semi-analytical pricing and efficient tree methods are available for many standard interest rate derivatives.
Hull-White vs Vasicek vs CIR
Understanding when to use each model requires knowing their key tradeoffs:
Vasicek (Constant Parameters)
- Constant mean reversion level
- Highly tractable with closed-form bond prices
- Can produce negative interest rates
- Cannot fit arbitrary initial term structures
- Best for: theoretical analysis, teaching
CIR (Constant Parameters)
- Square-root diffusion prevents negative rates
- Tractable with closed-form bond prices
- Feller condition governs rate behavior
- Cannot fit arbitrary initial term structures
- Best for: scenarios requiring non-negative rates
Hull-White (Time-Varying θ(t))
- Time-varying mean reversion level θ(t)
- Preserves Vasicek-style exponential-affine tractability
- Can produce negative interest rates (Gaussian)
- Can fit the initial term structure exactly
- Best for: practical derivative pricing and hedging
The key insight: Hull-White is preferred by practitioners because it matches market observables. When you hedge a swaption with bonds and swaps, you need your model to price those hedging instruments correctly — Hull-White guarantees this by construction.
Hull-White Model Applications: Swaptions, Caps, and Floors
Hull-White is the workhorse model for pricing interest rate derivatives at many financial institutions. Common applications include:
- European swaptions — options to enter into interest rate swaps
- Caps and floors — portfolios of caplets/floorlets on floating rates
- Callable bonds — bonds with embedded call options
- Bermudan swaptions — swaptions with multiple exercise dates
A pension fund wants to hedge against rising rates by purchasing a 5-year option to enter a 10-year payer swap at a 4% strike. Using Hull-White:
- The desk bootstraps today’s USD swap curve from 3-month SOFR to 30-year swap rates
- They calibrate κ = 0.03 (3% mean reversion) and σ = 0.8% from the ATM swaption volatility surface
- Given these parameters, θ(t) is derived to match the curve exactly
- The model prices the swaption at 285 basis points upfront (2.85% of notional)
If the desk later reprices with κ = 0.05, they must recalculate θ(t) — the entire calibration is interdependent.
For path-dependent or early-exercise derivatives, Hull-White is often implemented using trinomial trees, which efficiently capture the mean-reverting dynamics.
A corporate treasurer needs to value a $500 million 10-year bond with a call provision exercisable after year 3. The bond pays a 5.25% coupon.
- Using Hull-White calibrated to the corporate issuer’s funding curve
- With σ = 1.0% annual volatility and κ = 0.04
- A trinomial tree evaluates the issuer’s optimal call strategy at each node
- The model values the call option embedded in the bond at approximately 180 bps, reducing the bond’s value to investors by that amount compared to a non-callable equivalent
Limitations of the Hull-White Model
Despite its popularity, Hull-White has important limitations that practitioners must understand:
1. Negative rates are possible. Like Vasicek, Hull-White uses a Gaussian (normal) distribution for the short rate, which can go negative. Since 2014, when the ECB, SNB, and BOJ introduced negative policy rates, this has become less problematic — and in some contexts, even desirable for realistic modeling.
2. Calibration instability. The function θ(t) changes whenever you recalibrate to a new market curve. This is expected — θ(t) is designed to refit the current term structure. However, fitted θ(t) can be numerically noisy and has weak structural interpretability, especially when the yield curve has unusual shapes or kinks.
3. One-factor limitation. A single factor cannot capture complex yield curve movements like twists, butterflies, or independent movements at different maturities.
4. Volatility structure. The constant σ assumption means Hull-White cannot capture volatility smiles or skews without extensions.
For exotic or path-dependent derivatives, the one-factor Gaussian dynamics may be too restrictive. Always validate the model against the specific products you’re pricing.
Common Mistakes
When working with the Hull-White model, watch out for these common errors:
1. Confusing Hull-White with Vasicek. Hull-White has a time-varying θ(t); Vasicek has a constant θ. This distinction is what enables term structure fitting.
2. Expecting stable parameters. The calibrated θ(t) function changes with every new market curve. This is normal — it reflects the model adapting to current conditions, not a flaw.
3. Using Hull-White for exotic derivatives without validation. Calibration to vanilla instruments (bonds, caps, swaptions) does not guarantee accuracy for complex path-dependent exotics. One-factor Gaussian dynamics may be too simplistic.
4. Assuming an exact fit to today’s curve means the model is correct. Matching the initial term structure is a calibration choice for hedging consistency — it says nothing about whether the model captures realistic future term structure dynamics.
5. Confusing risk-neutral and physical parameters. The calibrated θ(t) is a risk-neutral parameter used for pricing. It is not the same as the historical drift you would estimate from past rate data. Using historically estimated parameters for pricing leads to incorrect derivative values.
Frequently Asked Questions
Disclaimer
This article is for educational and informational purposes only and does not constitute investment or financial advice. Interest rate models involve simplifying assumptions that may not hold in all market conditions. Always consult qualified professionals and conduct independent analysis before making financial decisions.
