In Hagan's paper on valuing CMS swaps (Convexity Conundrums: Pricing CMS Swaps, Caps, and Floors), there is:

So the swap rate must also be a Martingale, and

$$E \big[ R_s(\tau) \big| \mathcal{F}_0 \big]=R_s(0) = R_s^0$$

To complete the pricing, one now has to invoke a mathematical model (Black’s model, Heston’s model, the SABR model, . . . ) for how $R_s(\tau)$ is distributed around its mean value $R_s^0$. In Black’s model, for example, the swap rate is distributed according to

$$R_s(\tau) = R_s(0)e^{\sigma x\sqrt{\tau}-\frac{1}{2}\sigma^2\tau}$$

where x is a normal variable with mean zero and unit variance. One completes the pricing by integrating to calculate the expected value.

And he stopped here. So I am trying to "complete the pricing by integrating", but I am not sure of what he actually meant. I think he just wants to get the swap rate $R_s(0)$, but I am not sure.

This integration would lead to:

$$ R_s(0) = E \big[ R_s(\tau) \big| \mathcal{F}_0 \big] = \int_{-\infty}^{+\infty} R_s(\tau) \frac{-\frac{x^2}{2}}{\sqrt{2\pi}}dx = \int_{-\infty}^{+\infty} R_s(0)e^{\sigma\sqrt{\tau}x-\frac{1}{2}\sigma^2\tau} \frac{-\frac{x^2}{2}}{\sqrt{2\pi}}dx $$

This does not make sens, since $R_s(0)$ is on both side of the equation. Maybe I am not looking at it the right way, and he meant something else. Or maybe my integration is wrong.

I am a bit lost here. Any help would be appreciated to understand how to "complete this pricing".

  • 1
    $\begingroup$ Since $R_s(\tau)$ is a stochastic process modeling a swap, he may be meaning that you need to use a model to describe this process with more depth. By integrating he means computing the expectancy of the process at time $\tau$ $\endgroup$ – Rerito Apr 23 '16 at 15:27

He meant to compute the swaption price given by (2.4a), that is, \begin{align*} V_{opt} = L_0 E\left((R_s(\tau)-R_{fix})^+ \mid \mathcal{F}_0 \right). \end{align*} Under the swap measure (i.e., with $L_t$ as the numeraire), the swap rate process $\{R_s(t), t \ge 0\}$ is a martingale, and is assumed to be of the form \begin{align*} R_s(\tau) = R_s^0 e^{\sigma X \sqrt{\tau} -\frac{1}{2}\sigma^2\tau }, \end{align*} where $X$ is a standard normal random variable. Then \begin{align*} V_{opt} &= L_0\int_{-\infty}^{\infty}\left(R_s^0 e^{\sigma x \sqrt{\tau} -\frac{1}{2}\sigma^2\tau }-R_{fix} \right)^+ \frac{1}{\sqrt{2\pi}}e^{-\frac{x^2}{2}}dx\\ &=L_0\big[R_s^0 N(d_1) - R_{fix} N(d_2) \big], \end{align*} where \begin{align*} d_1 = \frac{\ln \frac{R_s^0}{R_{fix}}+\frac{1}{2}\sigma^2 \tau}{\sigma \sqrt{\tau}}, \end{align*} and \begin{align*} d_2 = d_1 -\sigma \sqrt{\tau}. \end{align*}

  • $\begingroup$ Thank you, this makes sense. It's weird though that he does not say clearly that this is used for his next section. Also does it make Rerito's answer false, ie $R_s$ is not a martingale because Black's model has not a null mean? $\endgroup$ – MarinD Apr 25 '16 at 23:02
  • 1
    $\begingroup$ I don't think his answer is correct. $\endgroup$ – Gordon Apr 26 '16 at 2:37

Let's recall the definition of a Martingale first: it is a stochastic process $X(t)$ that has the following property: let $0 \leq t < T$ two real numbers. Let $\mathcal{F}_t$ be a filtration for the process $X$ at time $t$. We have then:

$$ \mathbb{E}[X(T)|\mathcal{F}_t] = X(t) $$

Now, if you use Black's model, you describe your asset price using a geometric Brownian motion. This is not a martingale. Let's study it a bit more:

$$ S(t) = S_0 \exp{(\sigma \sqrt{t} W(t) - \frac{1}{2}\sigma^2 t)} $$

If you consider the logarithm of this process, you'll establish that it follows a normal law of mean $\ln S_0 - \frac{1}{2}\sigma^2t$ and variance $\sigma^2 t$. By going a bit more into calculus you can establish it has the Markov property, but this is not a martingale.

  • $\begingroup$ That's right, but why would the author begin by stating that "the swap rate must also be a Martingale", and then try to model it with Black's dynamics which is not modeling a martingale process? This does not make any sense $\endgroup$ – MarinD Apr 23 '16 at 16:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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