Skip to content
All library documents

Deriving the Hull–White Short-Rate Distribution from the Forward Curve

Article Quant Q&A · Author: SmallChess

Summary

The document explains how two expressions for the conditional short-rate distribution in the Hull–White model are related. Starting from the model’s stochastic differential equation, it gives the conditional mean and variance when the drift parameter is constant, then derives the general time-varying-drift case. The variance depends on the mean-reversion and volatility parameters, while the conditional mean includes an integral of the drift function.

To connect that integral to observable market data, the derivation uses the model’s zero-coupon bond pricing formula and differentiates the initial bond-price curve to obtain the instantaneous forward rate. This yields a forward-curve-based expression for the shift term and, in turn, the conditional mean written in terms of a function built from the forward curve. The document supplies an algebraic derivation rather than empirical tests. Its formulas assume the stated one-factor Hull–White setup under the risk-neutral measure; the derivation does not discuss calibration choices or extensions to other models.

Key ideas

  • The Hull–White short rate has a conditional normal distribution with a mean-reverting component and a variance determined by mean reversion and volatility.
  • For a time-varying drift, the conditional mean includes an integral over the drift function.
  • The initial zero-coupon bond curve connects that drift integral to the instantaneous forward curve.
  • The resulting forward-curve expression reconciles the two forms of the conditional mean.

Tags

Full text
# Hull-White formula on wikipedia, correct?


# Hull-White formula on wikipedia, correct?












The distribution for the short rate in Hull-White model on Wikipedia is:

But the same equation in Damiano's `Interest Rate Models - Theory and Practice` is:

Q: I don't see how the formulas for the expectation are related. The formula in the book has instantaneous forward curve, which is nowhere in Wikipedia.

## Answer by Gordon (score 13, accepted)

https://quant.stackexchange.com/a/31852

For the Hull-White model, where \begin{align*} dr_t = (\theta(t)-a r_t)dt+ \sigma dW_t, \end{align*} under the risk-neutral measure, we have that, for $t\ge s \ge 0$, \begin{align*} r_t = e^{-a(t-s)} r_s + \int_s^t \theta(u)e^{-a(t-u)} du + \int_s^t \sigma e^{-a(t-u)} dW_u. \end{align*} Then, if $\theta$ is a constant, \begin{align*} r_t \mid r_s &\sim N\left(e^{-a(t-s)} r_s + \int_s^t \theta e^{-a(t-u)} du\Big), \, \frac{\sigma^2}{2a}\Big(1-e^{-2a(t-s)}\Big)\right) \\ &\sim N\left(e^{-a(t-s)} r_s + \frac{\theta}{a} \Big(1-e^{-a(t-s)}\Big), \, \frac{\sigma^2}{2a}\Big(1-e^{-2a(t-s)}\Big)\right). \end{align*}

For the general case (see this question), the price of a zero-coupon bond price is given by \begin{align*} P(t, T) &= A(t, T) e^{-B(t, T)\, r_t}, \end{align*} where \begin{align*} B(t, T) = \frac{1}{a}\Big(1-e^{-a(T-t)} \Big), \end{align*} and \begin{align*} A(t, T) &= \exp\left(- \int_t^T \theta(u) B(u, T) du -\frac{\sigma^2}{2a^2}\big(B(t, T) -T+t\big)-\frac{\sigma^2}{4a}B(t, T)^2\right). \end{align*} Given the initial bond price curve, note that \begin{align*} \ln P(0, T) = \ln A(0, T) - B(0, T)\, r_0. \end{align*} Then \begin{align*} f(0, T) &= -\frac{\partial \ln P(0, T)}{\partial T}\\ &= \int_0^T \theta(u) \frac{\partial B(u, T)}{\partial T} du + \frac{\sigma^2}{2a^2}\Big(\frac{\partial B(0, T)}{\partial T} -1\Big)+ \frac{\sigma^2}{2a}B(0, T) \frac{\partial B(0, T)}{\partial T}\\ &\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\ + \frac{\partial B(0, T)}{\partial T} r_0\\ &=\int_0^T \theta(u) e^{-a(T-u)} du+ \frac{\sigma^2}{2a^2}\Big(e^{-a T} -1\Big)+ \frac{\sigma^2}{2a^2}\Big(e^{-a T}-e^{-2a T} \Big) + e^{-a T} r_0\\ &=\int_0^T \theta(u) e^{-a(T-u)} du - \frac{\sigma^2}{2a^2}\Big(e^{-a T} -1\Big)^2 + e^{-a T} r_0. \end{align*} That is, \begin{align*} \int_0^T \theta(u) e^{-a(T-u)} du &= f(0, T) + \frac{\sigma^2}{2a^2}\Big(e^{-a T} -1\Big)^2-e^{-a T} r_0\\ &=\alpha(T)-e^{-a T} r_0, \tag{1} \end{align*} where \begin{align*} \alpha(T) = f(0, T) + \frac{\sigma^2}{2a^2}\Big(e^{-a T} -1\Big)^2. \end{align*} Moreover, from $(1)$, \begin{align*} \int_0^T \theta(u) e^{au} du= e^{aT}\alpha(T)-r_0. \end{align*} Then \begin{align*} \int_s^t \theta(u) e^{-a(t-u)} du &= e^{-a t} \int_s^t \theta(u) e^{a u} du\\ &= e^{-a t}\left(e^{at}\alpha(t)-e^{as} \alpha(s) \right)\\ &=\alpha(t) - e^{-a(t-s)}\alpha(s). \end{align*} Therefore, \begin{align*} r_t \mid r_s &\sim N\left(e^{-a(t-s)} r_s + \int_s^t \theta(u) e^{-a(t-u)} du\Big), \, \frac{\sigma^2}{2a}\Big(1-e^{-2a(t-s)}\Big)\right) \\ &\sim N\left(e^{-a(t-s)} r_s + \alpha(t) - e^{-a(t-s)}\alpha(s), \, \frac{\sigma^2}{2a}\Big(1-e^{-2a(t-s)}\Big)\right). \end{align*}

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.