Help with two proofs on the martingale properties of an American Put Option

125 Views Asked by At

Ok so I'm trying to prove two results for the standard American put option with pricing function: $$P^A(t,S(t))=\sup_{\tau\in (t,T)} \mathbb{E}_{\mathbb{Q}}\left [ e^{-r(\tau-t)}(K-S(\tau))^{+} | \mathcal{F}_t\right ]$$

Where $K$ is the strike, $T$ the expiry, and we are taking the supremum over all stopping times in the interval $(t,T)$. Here are the proofs I'm trying to do:

$\textbf{a)}\;$Denote by $\bar{P}^A(t,S(t))=e^{-rt}P^A(t,S(t))$ the discounted pricing function of the American put option. Show that $\bar{P}^A(t,S(t))$ is a supermartingale under the measure $\mathbb{Q}$.

$\textbf{b)}\;$ Denote $\tau^*(t)$ as the exercise time of the option after $t$, show that $\bar{P}^A(t,S(t))$ is a martingale under $\mathbb{Q}$ in the time interval $[t,\tau^*(t)]$, i.e., for any stopping time $t\leq \tau \leq \tau^*(t)$, $\mathbb{E}_{\mathbb{Q}}[\bar{P}^A(\tau,S(\tau))|\mathcal{F}_t]=\bar{P}^A(t,S(t))$

I believe I have a solid proof for $\textbf{a}$ however I am pretty stuck at $\textbf{b}$. Here's my proof for $\textbf{a}$.

Consider $t'<t$, then conditioning on this we have: $$\mathbb{E}_\mathbb{Q}[\bar{P}^A(t,S(t))|\mathcal{F}_{t'}]=\mathbb{E}_\mathbb{Q}\left [e^{-rt}\sup_{\tau\in (t,T)} \mathbb{E}_{\mathbb{Q}}\left [ e^{-r(\tau-t)}(K-S(\tau))^{+}\rvert \mathcal{F_{t}}\right ]\rvert \mathcal{F_{t'}}\right ]$$ $$\leq \sup_{\tau\in (t',T)}\mathbb{E}_\mathbb{Q}\left [ \mathbb{E}_{\mathbb{Q}}\left [ e^{-r\tau}(K-S(\tau))^{+}\rvert \mathcal{F_{t}}\right ]\rvert \mathcal{F_{t'}}\right ]$$ Where we used the expectation conditioned on $\mathcal{F_{t'}}$ to change the $\tau$ in the supremum and then took it out to create the inequality. Now by tower property we have $$=\sup_{\tau\in (t',T)}\mathbb{E}_\mathbb{Q}\left [ e^{-rt'}e^{-r(\tau-t')}(K-S(\tau))^{+}\rvert \mathcal{F_{t'}}\right ]$$ and hence taking $e^{-rt'}$ out the result follows.

Now for part $\textbf{b}$ I'm not even sure where to start. The question seems very ambiguous, is it asking to consider the same case but restricted to the interval $(t, \tau^*(t))$? Or is it asking to consider this interval but the pricing function is only evaluated at the stopping times between $t$ and $\tau^*(t)$? I feel like if it was the latter we could apply the optional stopping theorem, however that would imply that it is again a supermartingale. Any pointers to get me started would be greatly appreciated.