The payoff of a plain vanilla swap with respect to measure $Q$ is : $$V_{\mathrm{swap}}(t) = \beta(t) \sum_{n=0}^{N-1} \tau_n E_t^Q \left( \dfrac{1}{\beta(T_{n+1})} ( L(T_n, T_n, T_{n+1}) - k) \right)$$

How do we get : $$V_{\mathrm{swap}}(t) = \beta(t) \sum_{n=0}^{N-1} \tau_n E_t^Q \left( \dfrac{1}{\beta(T_{n})} ( L(T_n, T_n, T_{n+1}) - k) \, P(T_n, T_{n+1})\right) $$


$$ P(T_n, T_{n+1}) = E^Q_t \left(\dfrac{1}{\beta(T_n, T_{n+1}}\right) $$


Note that the money-market numeraire is defined in terms of the short-term interest rate as

$$\beta(t) = \exp\left(\int_0^tr(s)ds\right),$$

and the forward price of a zero-coupon bond issued at the future date $T_n$ and maturing at $T_{n+1}$ is the risk-neutral conditional expectation of the discounted payoff:

$$P(T_n, T_{n+1}) = E^Q \left(\left.\exp\left(-\int_{T_n}^{T_{n+1}}r(s)ds\right)\right|\mathcal{F_n}\right)=E^Q \left(\left.\frac{\beta(T_n)}{\beta(T_{n+1})}\right|\mathcal{F_n}\right).$$

Using the tower property of conditional expectation $$E_t^Q \left( \dfrac{1}{\beta(T_{n+1})} ( L(T_n, T_n, T_{n+1}) - k) \right)\\=E_t^Q \left[E^Q\left( \left.\dfrac{1}{\beta(T_{n+1})} ( L(T_n, T_n, T_{n+1}) - k)\right|\mathcal{F}_n\right) \right]\\=E_t^Q \left[E^Q\left( \left.\frac1{\beta(T_n)}\dfrac{\beta(T_n)}{\beta(T_{n+1})} ( L(T_n, T_n, T_{n+1}) - k)\right|\mathcal{F}_n\right) \right]$$

Taking what is known at time $T_n$ out of the inner conditional expectation we get

$$E_t^Q \left( \dfrac{1}{\beta(T_{n+1})} ( L(T_n, T_n, T_{n+1}) - k) \right)\\=E_t^Q \left[\frac1{\beta(T_n)}( L(T_n, T_n, T_{n+1}) - k)E^Q\left( \left.\dfrac{\beta(T_n)}{\beta(T_{n+1})} \right|\mathcal{F}_n\right) \right]\\=E_t^Q \left( \dfrac{1}{\beta(T_{n})} ( L(T_n, T_n, T_{n+1}) - k) \, P(T_n, T_{n+1})\right)$$

  • $\begingroup$ Awesome! Thanks a lot. $\endgroup$ – ZeroCool Aug 15 '14 at 14:22
  • $\begingroup$ @ZeroCool: You're welcome. I edited the answer with a correction for the money market as the exponential of the integral of the short-rate process. $\endgroup$ – RRL Aug 15 '14 at 14:30

In case any one is interested I figured out a short answer :

Let $M(T_{n+1}) = L(T_n, T_n, T_{n+1})-k$

Given $M(T{n+1}$ is $F_{T_{n+1}}-measurable$ the payoff at the end of each period becomes :

$$ E \left( E \left( \dfrac{M(T_{n+1})}{\beta(T_n) \beta(T_n, T_{n+1})} \; | \; \mathcal{F}_{T_{n+1}}\right) | \; \mathcal{F_t}\right) $$

$$ = E \left( \dfrac{M(T_{n+1})}{\beta(T_n)} E \left( \dfrac{1}{\beta(T_n, T_{n+1})} \; | \; \mathcal{F}_{T_{n+1}}\right) | \; \mathcal{F_t}\right) $$

$$ = E \left( \dfrac{M(T_{n+1})}{\beta(T_n)} P(T_n, T_{n+1}) | \; \mathcal{F_t}\right) $$



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.