Beneš condition for a discontinuous exponential martingale (Q357242)

Let \((\Omega,{\mathcal F},({\mathcal F}_t)_{t\geq 0},P)\) be a filtered probability space satisfying the usual conditions. Let \((\alpha(t))\) be an \(({\mathcal F}_t)\)-predictable stochastic process such that \(\int^t_0 \alpha^2(s)\,ds< \infty\), a.s. for all \(t>0\). Let \((B_t)\) be a Brownian motion w.r.t. \(({\mathcal F}_t)\). Then it is well known that \[ Z_t= \exp\Biggl(\int^t_0 \alpha(s)\,dB_s- 1/2 \int^t_0 \alpha^2(s)\,ds\Biggr),\quad t\geq 0,\tag{\(*\)} \] is a positive local martingale such that \(\operatorname{E}[Z_t]\leq 1\), \(t\geq 0\). Furthermore, \((Z_t)\) is a martingale if \(\operatorname{E}[Z_t]=1\), \(t\geq 0\). The author studies the situation arising when in \({*}\) \((B_t)\) is replaced by a purely discontinuous square integrable (càdlàg) martingale \((M_t)\) with \(M_0= 0\), having jumps \(\alpha(t)\Delta M_t>-1\), \(t\geq 0\), where \(\Delta M_t= M_t- M_{t-}\). Furthermore, \((M_t)\) is assumed to be a homogeneous process with independent increments which in particular means that the quadratic variation \((\langle M\rangle_t)\) of \(M\) satisfies \(\langle M\rangle_t= \lambda t\) for some positive constant \(\lambda\). Let \(\mu(dt,dz)\) denote the integer-valued random measure associated with the jump process \((\Delta M_t)\), i.e., for each \(t\geq 0\) and any measurable set \(\Gamma\subset\mathbb{R}_+\setminus\{0\}\), we have \[ \mu([0, t]\times\Gamma)= \sum_{0\leq s\leq t} 1_{\{\Delta M_s\in \Gamma\}}\quad (t\geq 0). \] The compensator \(\nu(dt,dz)\) of \(\mu(dt,dz)\) is of the form \(\nu(dt,dz)= dt\,K(dz)\), where \(K(dz)\) is \(\sigma\)-finite, and \[ \int_{\mathbb{R}_+}z^2 K(dz)= \lambda. \] The main result in the present paper says that \textit{V. E. Beneš} type conditions [SIAM J. Control 9, 446--472 (1971; Zbl 0203.47301; Zbl 0219.93029)] guarantee that \(\operatorname{E}[Z_t]= 1\), \(t\geq 0\). More precisely, the author proves the following Theorem. Assume \(\int_{\mathbb{R}_+} z^3K(dz)<\infty\). If, for any \(T>0\), \(\alpha^2(t)\leq \text{const.}[1+ \sup_{0\leq s\leq t} M^2_{-s}]\) holds for all \(0\leq t\leq T\), then \(\operatorname{E}[Z_T]= 1\). The proof is essentially different from Beneš's original proof.