Some properties of the variance-optimal martingale measure for discontinuous semimartingales (Q2566718)

Let \(X= (X_t)\), \(0\leq t\leq T\) (defined on some complete filtered probability space \((\Omega,{\mathcal F},P,\) \(({\mathcal F}_t)_{0\leq t\leq T}))\), be an \(\mathbb{R}^d\)-valued RCLL special \({\mathcal S}^2\)-semimartingale which, in addition, is locally bounded. Let \(\Theta\) denote the space of all \(\mathbb{R}^d\)-valued predictable \(X\)-integrable processes \(\theta\) such that the stochastic integral \(G(\theta)= \int\theta_s \,dX_s\) is an \({\mathcal S}^2\)-semimartingale. A signed measure \(Q\) on \((\Omega,{\mathcal F})\) is called a signed \(\Theta\)-martingale measure provided \(Q(\Omega)= 1\), \(Q\ll P\), \(dQ/dP\in{\mathcal L}^2(P)\), and \(E[{dQ\over dP} G_T(\theta)]= 0\) for all \(\theta\in\Theta\). Let \(P_s(\Theta)\) denote the set of signed \(\Theta\)-martingale measures, and put \(P_e(\Theta)= \{Q\in P_s(\Theta)\mid Q\sim P\) and \(Q\) is a probability measure\} and let \[ {\mathcal D}= \Biggl\{D\in{\mathcal L}^2(P)\;\biggl|\;D={dQ\over dP}\text{ for some }Q\in P_s(\Theta)\Biggr\}. \] \(\widetilde P\in P_s(\Theta)\) is called the variance-optimal martingale measure (VOM) if \(\widetilde D:= d\widetilde P/dP:= \text{argmin}_{D\in{\mathcal D}} E[D^2]\). Let \(Z_t= E[\widetilde D\mid{\mathcal F}_t]\), \(0\leq t\leq T\). By definition the jump condition (JC) is satisfied if for some constant \(C> 0\), \(Z_-\leq CZ\). Finally, the author introduces Assumption 3.3. There exists some \(Q\in P_e(\Theta)\) satisfying the reverse Hölder inequality. (This means that for the density process \(Z^Q\) of \(Q\) there exists a constant \(C> 0\) such that, for every stopping time \(\sigma\leq T\), \(E[(Z^Q_T)^2\mid{\mathcal F}_\sigma]\leq C(Z^Q)^2\). The main results of the present paper are the following. Theorem 3.4. Under Assumption 3.3, \(Z\) satisfies the reverse Hölder inequality. Theorem 3.5. Under Assumption 3.3 and JC, the VOM \(P\) belongs to \(P_e(\Theta)\).