Stability of backward stochastic differential equations (Q1915848)

Let \(p \geq 1\), denote by \(S^p\) the space of semimartingales such that \(\sup_{0 \leq t \leq T} |X_t|\in L^p(P)\) and with \(H^p\) the space of the special semimartingales with canonical decomposition \(X = M + A\) (with \(A\) predictable), so that \(\left([M,M]^{1/2}_T + \int^T_0|dA_s|\right) \in L^p(P)\), where \((\Omega,{\mathcal F},P)\) is a complete probability space with filtration \(\{{\mathcal F}_t\}_{0 \leq t \leq T}\). Moreover, let \(V_t\) and \(V^n_t\) be solutions, respectively, of \[ V_t = E\left(\int^T_t g_s(V_s) dA_s + Y |{\mathcal F}_t\right), \] and \[ V^n_t = E\left( \int^T_t g^n_s(V^n_s) dA^n_s + Y^n|{\mathcal F}_t\right) , \quad \forall n \in \mathbb{N}, \] under the following hypotheses: (1) \(A^n\), \(A \in H^\infty\), which might be taken predictable, (2) \(Y^n\), \(Y \in L^1(P)\) are \({\mathcal F}_T\)-measurable random variables, (3) \(g^n\), \(g : [0,T] \times \Omega \times \mathbb{R} \to \mathbb{R}\) are cadlag, adapted and Lipschitz in \(\mathbb{R}\) with constant \(k\) for some \(k > 0\), uniformly with respect to \(s \in [0,T]\) and \(\omega \in \Omega\), (4) \(E\left(\int^T_0 |g^n_s (0)||dA^n_s|\right) < + \infty\), each \(n\) and \(g\) is bounded, say \(|g|< \gamma\). The author proves the following theorem: Let us assume that \(A^n\), \(A\), \(Y^n\), \(Y\), \(g^n\), \(g\) verify the foregoing hypotheses (1)--(4) and that (i) \(Y^n @>>n \to \infty> Y\) in \(L^1(P)\), (ii) \(A^n @>>n\to \infty> A\) in \(H^\infty\), (iii) \(g^n(V) @>>n \to \infty> g(V)\) in \(S^1\). Then \(V^n\) converges to \(V\) under the u.c.p. (i.e. the uniform convergence in probability) topology.