Chapter 9

Continuous semimartingales

[7/0%]
LaTeX
Problem 9.1

Let WtW_{t} be a Brownian motion started at 1 and T0=inf⁡{t>0:Wt=0}T_{0}=\inf \left\{ t>0: W_{t}=0\right\}. Is Mt=Wt∧T0M_{t}=W_{t \wedge T_{0}} a square integrable martingale? A locally square integrable martingale? A uniformly integrable martingale? A martingale? A local martingale? A semimartingale?

?
Problem 9.2

Prove that if MM is a submartingale such that the paths of MM are continuous, sup t∣Mt∣{ }_{t}\left|M_{t}\right| is integrable, and S≤TS \leq T are finite stopping times, then E[MT∣FS]≥MS\mathbb {E}\left[M_{T} \mid \mathcal{F}_{S}\right] \geq M_{S}. Note that the last part of the proof of Proposition 9.3 breaks down here.

?
Problem 9.3

Suppose MtM_{t} is a local martingale with continuous paths. Show that if N>0,TN=inf⁡{tN>0, T_{N}=\inf \{ t : ∣Mt∣≥N}\left.\left|M_{t}\right| \geq N\right\}, and MtN=Mt∧TNM_{t}^{N}=M_{t \wedge T_{N}}, then MNM^{N} is a uniformly integrable martingale.

?
Problem 9.4

Suppose Wt1W_{t}^{1} and Wt2W_{t}^{2} are two independent Brownian motions, t0>0t_{0}>0, and Mti=Wt∧t0,i=1,2M_{t}^{i}=W_{t \wedge t_{0}}, i=1,2. Show ⟨M1,M2⟩t=0\left\langle M^{1}, M^{2}\right\rangle_{t}=0.

?
Problem 9.5

Show that the norm defined in (9.10) is complete.

?
Problem 9.6

Let ZtZ_{t} be a bounded supermartingale with continuous paths that is constant from some time t0t_{0} on. Show that for each tt

E[∫t∞E[Zs−Zs+h∣Fs]ds∣Ft]=∫t∞E[Zs−Zs+h∣Ft]ds, a.s.  \mathbb {E}\left[\int _{t}^{\infty } \mathbb {E}\left[Z_{s}-Z_{s+h} \mid \mathcal{F}_{s}\right] d s \mid \mathcal{F}_{t}\right]=\int _{t}^{\infty } \mathbb {E}\left[Z_{s}-Z_{s+h} \mid \mathcal{F}_{t}\right] d s, \quad \text{ a.s. }
?
Problem 9.7

We mentioned that one can prove the existence of ⟨M⟩\left\langle M\right\rangle without using the Doob-Meyer theorem. Here is how that argument starts. Let MM be a bounded continuous martingale and for each nn, define

In(t)=∑i=0[t2n](M(i+1)/2n−Mi/2n)2. I_{n}(t)=\sum _{i=0}^{\left[t 2^{n}\right]}\left(M_{(i+1) / 2^{n}}-M_{i / 2^{n}}\right)^{2} .

Here [x][x] is the integer part of xx. Prove that for each t>0t>0, E[∣In(t)−Im(t)∣2]→0\mathbb {E}\left[\left|I_{n}(t)-I_{m}(t)\right|^{2}\right] \rightarrow 0 as n,m→∞n, m \rightarrow \infty. One can then define ⟨M⟩t\left\langle M\right\rangle_{t} as the L2L^{2} limit of In(t)I_{n}(t).

Hint: If n>mn>m, note that

M(i+1)/2m−Mi/2m=∑j=2n−mi2n−m(i+1)−1(M(j+1)/2n−Mj/2n). M_{(i+1) / 2^{m}}-M_{i / 2^{m}}=\sum _{j=2^{n-m} i}^{2^{n-m}(i+1)-1}\left(M_{(j+1) / 2^{n}}-M_{j / 2^{n}}\right) .
?