\(L^{2}\)-approximating pricing under restricted information (Q985719): Difference between revisions

The aim of the paper is to get an approximate price of an assets, according to minimize a variance criterium. The assets price is \[ S_t= S_0+ M_t+ \int^t_0 \lambda_u d\langle M\rangle_u,\;t\leq T, \] on a probability space \((\Omega,({\mathcal F}_t)_{0\leq t\leq T}, P)\), but only a subfiltration \(({\mathcal G}_t)_{0\leq t\leq T}\) is observed. Hypotheses are that \(\lambda\) and \(\langle M\rangle\) are \({\mathcal G}\)-predictable, \({\mathcal G}\)-martingales are local \({\mathcal F}\)-martingales, and there exists an equivalent probability measure \(Q= ZP\) such that \(S\) is a local \((Q,{\mathcal F})\)-martingale. Moreover there exists a constant \(C\) such that for any \({\mathcal F}\)-stopping time \(\tau\), \(E[{Z^2_T\over Z^2_\tau}/{\mathcal F}_r]\leq C\) (namely reverse Hölder condition). Under these hypotheses, given \(H\in L^2({\mathcal F}_T,P)\), the main result give an explicit expression of an optimal portfolio \(\pi^*\) which minimizes the application \[ x\mapsto E\Biggl[\Biggl(x+ \int^T_0 \pi_u dS_u- H\Biggr)^2\Biggr]. \] This portfolio is linked to a solution \((Y, \psi, N)\) of \[ dY_t= \pi^*_t\lambda_t d\langle M\rangle_t+ \psi_t dM_t+ dN_t,\quad Y_T= 0, \langle M,N\rangle= 0. \] Section 4 is devoted to the link with the authors previous results [SIAM J. Control Optim. 47, No. 5, 2381--2409 (2008; Zbl 1173.91393)]. Finally the main result is applied to a market with two assets, \((S,\eta)\), only price \(S\) being observed: an optimal portfolio is exhibited.