2
$\begingroup$

Let $B$ be a standard Brownian motion and let $\alpha, \beta > 0$. Let \begin{align} \tau = \inf\{t \geq 0 : B_t = \alpha \ \ \text{or}\ \ B_t=-\beta\}. \end{align} It can be shown by defining independent events $A_j = \{|B_{j+1}-B_{j}| \geq \alpha + \beta\}$ for integers $0 \leq j \leq n-1$ and $\epsilon = \mathbb{P}(A_j) = \mathbb{P}(A_0) $, that there exists an $\epsilon \in (0,1)$ such that $\mathbb{P}(\tau \geq n) \leq (1-\epsilon)^{n}$.

My question is about showing that for all $p \geq 1$, $\mathbb{E}[\tau^p] < \infty$. I do not understand why \begin{align} \mathbb{E}[\tau^p] = \int_0^\infty p t^{p-1} \mathbb{P}(\tau \geq t ) dt. \end{align} And subsequently why is \begin{align} \int_0^\infty p t^{p-1} \mathbb{P}(\tau \geq t ) dt = \sum_{n=0}^\infty \int_n^{n+1} p t^{p-1} \mathbb{P}(\tau \geq t ) dt ? \end{align}

$\endgroup$
1
$\begingroup$

Note that $P(\tau\geq t)=E(I_{\{\tau\geq t\}})$ and $$\int_{0}^{+\infty}pt^{p-1}P(\tau\geq t)dt=\int_{0}^{+\infty}pt^{p-1}E(I_{\{\tau\geq t\}})dt\overset{!}{=} E\bigg(\int_{0}^{+\infty}pt^{p-1}I_{\{\tau\geq t\}}dt\bigg)=E\bigg(\int_{0}^{\tau}pt^{p-1}dt\bigg)=E(\tau^p),$$ where in ! we used the Fubini - Tonelli theorem. For the second part just note that $$\displaystyle \int_{0}^{\infty}g(t)dt=\sum_{n=0}^{+\infty}\int_{n}^{n+1}g(t)dt$$ for an integrable function $g$ on $[0,+\infty)$ (in the generalized sense).

EDIT. $$I_{\{\tau\geq t\}}=\left \{\begin {array}{ll} 1,~t\leq \tau\\ 0,~t> \tau\\ \end{array} \right. $$ and therefore $$\int_{0}^{+\infty}pt^{p-1}I_{\{\tau\geq t\}}dt=\int_{0}^{\tau}pt^{p-1}~dt + \int_{\tau}^{+\infty}0~dt= \int_{0}^{\tau}pt^{p-1}dt.$$

$\endgroup$
  • $\begingroup$ Why is integrating $pt^{p-1}\mathbb{1}_{\{\tau \geq t\}}$ over $[0,+\infty)$ equal to integrating $pt^{p-1}$ over $[0,\tau]$? $\endgroup$ – iJup Mar 21 '16 at 20:21
3
$\begingroup$

To derive the expression for $\mathbb{E}[\tau^p]$, note that by definition $$\mathbb{E}[\tau^p] = \int_{0}^{\infty} t^p f_\tau(t)dt,$$ where $f_{\tau}$ is the density function of $\tau$. Applying integration by parts we obtain \begin{align} \mathbb{E}[\tau^p] &= \int_{0}^{\infty} -t^p d[1-F_{\tau}] \\ &= \underbrace{\left[-t^p (1-F_{\tau}(t))\right]_{0}^{\infty}}_{*} + \int_{0}^{\infty} pt^{p-1}[1-F_{\tau}(t)]dt, \end{align} where $F_{\tau}$ is the distribution function of $\tau$. Now, $[1-F_{\tau}(t)] = \mathbb{P}(\tau \geq t)$. Furthermore it can be shown that $1-F_{\tau}(t) \to 0$ if $t \to \infty$ (follows by Markov inequality). Hence the first term $*$ vanishes and you obtain the expression.

$\endgroup$

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.