BSDES WITH WEAK TERMINAL CONDITION

We introduce a new class of backward stochastic differential equations in which the T-terminal value Y-T of the solution (Y, Z) is not fixed as a random variable, but only satisfies a weak constraint of the form E[Psi(Y-T)] >= m, for some (possibly random) nondecreasing map Psi and some threshold m. We name them BSDEs with weak terminal condition and obtain a representation of the minimal time t-values Y-t such that (Y, Z) is a supersolution of the BSDE with weak terminal condition. It provides a non-Markovian BSDE formulation of the PDE characterization obtained for Markovian stochastic target problems under controlled loss in Bouchard, Elie and Touzi [SIAM T. Control Optim. 48 (2009/10) 3123-3150]. We then study the main properties of this minimal value. In particular, we analyze its continuity and convexity with respect to the m-parameter appearing in the weak terminal condition, and show how it can be related to a dual optimal control problem in Meyer form. These last properties generalize to a non-Markovian framework previous results on quantile hedging and hedging under loss constraints obtained in Follmer and Leukert [Finance Stoch. 3 (1999) 251-273; Finance Stoch. 4 (2000) 117-146], and in Bouchard, Elie and Touzi (2009/10).