Discrete-time approximation for continuously and discretely reflected BSDEs (Q2518617)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Discrete-time approximation for continuously and discretely reflected BSDEs |
scientific article |
Statements
Discrete-time approximation for continuously and discretely reflected BSDEs (English)
0 references
16 January 2009
0 references
The authors of the present paper investigate a discrete-time approximation of the solution \((Y,Z,K)\) of a reflected backward stochastic differential equation (BSDE) in a Markovian setting. Given the solution \(X\) of the associated forward diffusion equation with diffusion coefficient \(\sigma\) they suppose that the reflecting lower obstacle is of the form \(h(X)\). Given a partition \(\pi=\{0=t_0<\cdots<t_N=T\}\) and the Euler scheme \(X^\pi\) of \(X\) they consider the approximation scheme \(\overline{Y}_T^\pi= g(X_T^\pi)\), \(\overline{Z}^\pi_{t_i}= (t_{i+1}-t_{i})^{-1}E [\overline{Y}^\pi_{t_{i+1}}(W_{t_{i+1}}-W_{t_i}) |{\mathcal F}_{t_i}^W]\), \(\widetilde{Y}^\pi_{t_i}=E [\overline{Y}^\pi_{t_{i+1}}|{\mathcal F}_{t_i}^W]+ (t_{i+1}-t_{i})f(X^\pi_{t_i},\widetilde{Y}^\pi_{t_i},\overline{Z}^\pi_{t_i}),\) \(\overline{Y}^\pi_{t_{i}}=\widetilde{Y}^\pi_{t_i}\vee h(X^\pi_{t_i}),\) \(i\leq N-1\). In the case of BSDEs without reflection such approximations were studied by \textit{B. Bouchard} and \textit{N. Touzi} [Stochastic Processes Appl. 111, No. 2, 175--206 (2004; Zbl 1071.60059)], \textit{J. Zhang} [Ann. Appl. Probab. 14, No.~1, 459--488 (2004; Zbl 1056.60067)], and for BSDEs with jumps by \textit{B. Bouchard} and \textit{R. Elie} [Stochastic Processes Appl. 118, No. 1, 53--75 (2008; Zbl 1136.60048)] and \textit{J.-L. Lemor, E. Gobet} and \textit{X. Warin} [Bernoulli 12, No. 5, 889--916 (2006; Zbl 1136.60351)]. In all these cases it has turned out that the error of the approximation is controlled by \(|\pi|^{1/2}+E [\int_0^T| Z_t-\overline{Z}_t|^2\,dt]^{1/2}\) where \(\overline{Z}_t=(t_{i+1}-t_{i})^{-1}E [\int_{t_i}^{t_{i+1}}Z_s\,ds|{\mathcal F}_{t_i}^W]\), \(t\in[t_i,t_{i+1})\). \textit{J. Zhang} [Some fine properties of backward stochastic differential equations. Ph. D. Thesis, Purdue University, (2001)] showed that the term of the error is bounded by \(C|\pi|^{1/2}\). In the case with reflection the estimate of this term becomes more difficult, except the case of \(f\) independent of \(Z\) [\textit{V. Bally} et al., Bernoulli 9, No. 6, 1003--1049 (2003; Zbl 1042.60021); Bouchard and Touzi, loc. cit.]. The case of \(f\) depending on \(Z\) was studied by \textit{J. Ma} and \textit{J. Zhang} [Stochastic Processes Appl. 115, No.~4, 539--569 (2005; Zbl 1076.60049)] and the estimate \(C|\pi|^{1/4}\) was obtained under smoothness assumptions on the coefficients and a strict ellipticity assumption on \(\sigma\). The authors of the present paper improve this latter work considerably by removing the ellipticity assumption on \(\sigma\).
0 references
Backward stochastic differential equation
0 references
reflection backward stochastic differential equation
0 references
discrete time approximation schemes
0 references
regularity
0 references
0 references