The Hull–White model is a one-factor short-rate model used to value interest-rate dependent securities and derivatives. It models the instantaneous short rate as a mean-reverting Gaussian (normally distributed) process with a time-dependent drift chosen so the model exactly fits the current term structure of interest rates. Its flexibility and (relative) analytic tractability make it a standard choice for pricing bond options, swaptions and many other interest-rate derivatives.
Key ideas (short)
– Short rate dynamics: the short rate r(t) follows an Ornstein–Uhlenbeck (mean-reverting Gaussian) process.
– Time-dependent drift: a deterministic function θ(t) is used so the model matches the initial yield curve.
– Closed-form bond prices: zero-coupon bond prices are affine in r(t): P(t,T) = A(t,T) exp(−B(t,T) r(t)).
– Single-factor: all movements in the term structure are driven by one source of randomness.
– Normal distribution: rates can become negative in the model (small probability depending on parameters), unlike lognormal models.
Model formulation (standard one‑factor Hull–White)
The short rate follows:
dr(t) = [θ(t) − a r(t)] dt + σ dW(t)
where
– a > 0 is the mean-reversion speed,
– σ ≥ 0 is the volatility parameter,
– θ(t) is a deterministic function chosen to fit the initial forward curve f(0,t),
– W(t) is a standard Brownian motion under the risk-neutral measure.
Useful closed-form components
– B(t,T) = (1 − e^{−a (T−t)}) / a
– Zero-coupon bond price:
P(t,T) = A(t,T) exp(−B(t,T) r(t))
where A(t,T) is chosen so the model reproduces the market P(0,T) (it depends on f(0,·), a, σ). In practice A(t,T) is computed from the initial forward curve and the model parameters.
– θ(t) that exactly fits the initial forward curve f(0,t) can be written (one convenient form):
θ(t) = ∂f(0,t)/∂t + a f(0,t) + (σ^2/(2 a)) (1 − e^{−2 a t})
(References at the end show derivations; forms vary with sign conventions.)
Why use Hull–White?
– Calibrates exactly to the observed initial yield curve (via θ(t)).
– Analytic (or semi‑analytic) pricing for many vanilla instruments (bond options, certain swaptions) and efficient Monte Carlo for more complex payoffs.
– Simpler than multi-factor models (fewer parameters to estimate).
– Gaussian short-rate makes some computations tractable and fast.
Limitations and special considerations
– Single-factor: cannot capture all shapes of volatility term structures or correlated movements that require more than one factor.
– Normality: the Gaussian nature implies possible negative short rates. This is less realistic in some environments but can be useful where negative rates are plausible (recent years have shown that possibility).
– Volatility structure: a single σ is often insufficient to reproduce the full cap/floor or swaption volatility surface. Practitioners often extend to multi‑factor versions of Hull–White or use time-dependent σ or piecewise parameters.
– Model risk: differences between model output and market prices can be material for exotic payoffs — be explicit about calibration choices and parameter stability.
Who are Hull and White?
John C. Hull and Alan D. White developed the model in 1990. John Hull is a finance professor at the University of Toronto known for textbooks on derivatives and risk management. Alan White is an academic in financial engineering; together they produced the influential Hull–White short-rate model (see Hull & White, 1990).
Practical implementation: step-by-step
Below is a practical workflow for using the Hull–White model to price or risk-manage interest-rate derivatives.
1) Gather market inputs
– Current zero-coupon or discount curve P(0,T) or spot/Par/zero rates across maturities.
– If calibrating to volatility data: market prices or implied volatilities for caps/floors, swaptions, or bond options you want to match.
– Historical short‑rate or term-structure data if you plan to estimate parameters statistically.
2) Choose model specification
– Single-factor Hull–White (standard) is dr = [θ(t) − a r] dt + σ dW.
– Consider multi-factor versions if the volatility surface or correlation structure demands it.
3) Fit the initial term structure (compute θ(t))
– Compute the instantaneous forward rate f(0,t) from the discount curve:
f(0,t) = −∂ ln P(0,t) / ∂t.
– Use the analytic expression for θ(t) (given above) to ensure the model reproduces P(0,T).
4) Calibrate volatility / mean-reversion (a, σ)
Two common approaches:
– Calibrate a and σ to market option prices (caps, swaptions) by minimizing error between model and market prices (or implied vols).
– Estimate from historical short-rate data (less common for pricing; market calibration preferred for derivatives).
Practical tip: a is often relatively stable and can be constrained to plausible values (e.g., 0.01–1.0 per year). σ typically determines short-term implied volatility.
5) Validate bond pricing formulas
– Use closed-form bond expressions: P(t,T) = A(t,T) exp(−B(t,T) r(t)). Implement and check that P(0,T) matches market P(0,T).
– Confirm A(t,T) computed with your θ(t), f(0,t), a, σ is consistent.
6) Pricing standard derivatives
– Bond options and some swaptions have semi-analytic or analytic solutions under Hull–White because of the Gaussian structure and affine bond prices.
– For European options on bonds or swaps, derive the relevant normal distribution for the underlying and apply the appropriate option formula (often a normal Black variant or Jamshidian decomposition for certain payoffs).
– For path-dependent or highly exotic instruments (e.g., MBS prepayment models, complicated barriers), use Monte Carlo simulation of r(t).
7) Monte Carlo simulation (if needed)
– Use the exact discretization of the Ornstein–Uhlenbeck process for numerical stability:
r(t+Δt) = r(t) e^{−a Δt} + μ(t,Δt) + ξ
where μ(t,Δt) = ∫_{t}^{t+Δt} e^{−a (t+Δt − s)} θ(s) ds (compute numerically or analytically if θ is simple),
and ξ ∼ N(0, v(Δt)) with v(Δt) = (σ^2 / (2 a)) (1 − e^{−2 a Δt}).
– This produces exact conditional Gaussian increments and avoids Euler discretization bias.
8) Compute Greeks and risk measures
– Greeks: for analytics, derive sensitivities from closed-form formulas when possible; for Monte Carlo, use pathwise or likelihood ratio methods for efficient Greeks.
– Stress tests and scenario analyses: check behavior when short rates move substantially or when parameters vary.
9) Reporting and governance
– Document calibration choices (instruments used, objective function), parameter stability, and known model limitations.
– Backtest: compare model prices/hedges against realized P&L or market prices.
Example practical calibration sketch
– Instruments: choose a set of at-the-money swaption prices across maturities.
– Objective: minimize sum of squared differences between model swaption prices (or implied vols) and market quotes.
– Parameters: solve for a and σ (or a time-dependent σ schedule) while θ(t) is set by the initial curve.
– Output: calibrated a, σ; inspect residuals and, if too large, consider multi-factor extension.
Extensions and variations
– Multi-factor Hull–White: add additional independent OU processes to capture richer principal components of rate movements.
– Shifted or displaced models: sometimes used to ensure positivity or better fit to skewed vol surfaces.
– Stochastic volatility / local volatility variants: incorporate time-varying or stochastic σ to capture smiles/skews.
When to prefer Hull–White versus alternatives
– Prefer Hull–White when: you need an analytically tractable short-rate model that fits the current curve and gives closed-form bond pricing; negative rates are acceptable or plausible.
– Consider Vasicek if you want a simpler constant-drift version; consider CIR if positivity of rates (square-root diffusion) is critical.
– Use HJM or BGM for forward-rate or observable-LIBOR-based frameworks when modeling a whole forward rate surface and complex calibration to market quotes is required.
Further reading and references
– Hull, J., & White, A. (1990). Pricing Interest-Rate-Derivative Securities. Review of Financial Studies, 3(4), 573–592.
– Vasicek, O. (1977). An equilibrium characterization of the term structure. Journal of Financial Economics.
– Cox, J., Ingersoll, J., & Ross, S. (1985). A Theory of the Term Structure of Interest Rates. Econometrica.
– Brigo, D., & Mercurio, F. (2006). Interest Rate Models — Theory and Practice: With Smile, Inflation and Credit. (Comprehensive text with derivations, implementation details and calibration examples.)
– Investopedia summary of Hull–White model
Concluding note
The Hull–White model balances tractability and flexibility: it exactly fits today’s yield curve, yields closed-form bond prices, and is easy to simulate. For vanilla derivative pricing and risk management it is a practical choice; for richer volatility and correlation structures you may need multi‑factor or more elaborate models. Always document calibration choices and check model output against market prices and business requirements.