We study a stochastic optimization problem under constraints in a general f
ramework including financial models with constrained portfolios, labor inco
me and large investor models and reinsurance models. We also impose America
n-type constraint on the state space process. General objective functions i
ncluding deterministic or random utility functions and shortfall risk loss
functions are considered We first prove existence and uniqueness result to
this optimization problem. In a second part, we develop a dual formulation
under minimal assumptions on the objective functions, which are the analogu
e of the asymptotic elasticity condition of Kramkov and Schachermayer (1999
). (C) 2001 Elsevier Science B.V. All rights reserved.