3
$\begingroup$

Is there a closed formula to approximate the quantiles of swaprates in the 1-factor Hull White model?

Background

The Hull-White is a Gaussian model for the short rate. Its mean and covariance function can be explicitly given in terms of calibration input, i.e. the initial yield curve and estimates for mean reversion strength and volatility. In this model bond prices $B(t,T)$ are log-normal. But this means that swap rates, defined as $$S_{i,j}(t)=\frac{B(t,T_i) - B(t,T_j)}{\sum_{k=i+1}^j B(t,T_k)} $$ have no "simple" distribution.

Details

  • Do the swap rates follow a distribution, whose properties have been analysed and described somewhere?
  • Is there a way to estimate quantiles of the swap rate i.e. values s(p) such that the probability $P(S\leq s(p)) = p.$

I guess there will only be approximation possible. Of course, I can always simulate and take empirical quantiles. But a closed formula would be convenient. I am fine with approximation errors in the ballpark of estimates by simulation with sample size a few 100'000.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

No closed form formula but the $B(t, T)$ and thus $S(t)$ are functions of the spot rate $r_t$, and $r_t$ has a Gaussian distribution (details in e.g. "brigo mercurio interest rate models theory and practice") so building the distribution of $S(t, r_t)$ is straightforward.

Improve this answer
$\endgroup$
4
  • $\begingroup$ What do you mean by "building the distribution"? $\endgroup$
    – g g
    Commented Dec 11, 2019 at 13:42
  • $\begingroup$ Assuming $S(r)$ is an increasing function of $r$, $P(S(r_t) < s) = P(r_t < S^{-1}(s))$ $\endgroup$
    – Antoine Conze
    Commented Dec 11, 2019 at 13:45
  • 1
    $\begingroup$ You claim $S(r)$ is always an increasing function of $r$ in the Hull-White model? How would you show that? $\endgroup$
    – g g
    Commented Dec 11, 2019 at 15:06
  • 1
    $\begingroup$ I haven't made any such claim. And usually $S(r)$ is not monotonic on $]-\infty, +\infty[$. However it generally is on $+/-$ 5 stdev around $E[r_t]$ which is generally sufficient for e.g. pricing options on $S$. You should look into european swaptions pricing with the HW model as it is related to your question. $\endgroup$
    – Antoine Conze
    Commented Dec 12, 2019 at 7:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.