4
$\begingroup$

Given $f\in C^2$ with compact support and $\tau$ be a stopping time with $E^x[\tau]<\infty$. Suppose $$Y_t=x+\int_0^t u(s,\omega)ds+\int_0^t v(s,\omega)dB_s(\omega), $$ with bounded $u,v$. Then by Ito formula with fixed $t$, I have

$$ f(Y_t)=f(Y_0)+\int_0^t (u\frac{\partial f}{\partial x}+\frac{1}{2}v^2\frac{\partial^2 f }{\partial^2 x})\,ds +\int_0^t v\frac{\partial f}{\partial x}\, dB_s, $$ My question is, how could I derive the following from the above identity that $$ E^x[f(Y_\tau)]=f(x)+E^x\bigg[\int_0^\tau (u\frac{\partial f}{\partial x}+\frac{1}{2}v^2\frac{\partial^2 f }{\partial^2 x})\,ds\bigg] +E^x\bigg[\int_0^\tau v\frac{\partial f}{\partial x}\, dB_s\bigg]. $$ In other words, how to change the deterministic $t$ into an integrable stopping time?

1 Answers 1

0

I think I got it. Please let me know if it has some problems.

For fixed integer $k$, we consider $$Z_t=x+\int_0^t 1_{\{s<\tau\}}u(s,\omega)ds+\int_0^t 1_{\{s<\tau\}}v(s,\omega)dB_s(\omega),t\in [0,k],$$ then Ito formula to $f(Z_t)$ implies $$ f(Z_k)=f(Y_{k\wedge \tau})=f(Y_0)+\int_0^{k\wedge \tau}(u\frac{\partial f}{\partial x}+\frac{1}{2}v^2\frac{\partial^2 f }{\partial^2 x})\,ds +\int_0^{k\wedge \tau} v\frac{\partial f}{\partial x}\, dB_s, $$ taking expectation and pass to limit (due to the dominant convergence, $\tau<\infty$ a.s., and the boundedness of the integrand), we get the desired result.

  • 1
    Looks almost good. 1. There should be $Y_{t\wedge \tau}$ in the first formula. 2.The phrase $\tau<\infty$ is irrelevant, but you may show that $\tau$ is (up to constant) is an integrable majorant for your expression. 3. As a side note, asking about downvotes is quite meaningless. – 2017-01-20
  • 0
    @zhoraster Thanks for your comment. I am not sure that I understand your second point. I show $\tau<\infty$ in order to pass the limit $k\to\infty$ and get the result $k\wedge \tau\to \tau$ a.s. . Are you saying this one is irrelevant? May you elaborate more on your second point? – 2017-01-20
  • 0
    Oh I see. Then it's relevant, of course. I thought it's not because you mentioned it near dominated convergence. – 2017-01-20
  • 0
    @zhoraster Thanks for your reply. – 2017-01-20