Solving stochastic optimal control problems with a quadratic objective function for nonlinear models has attracted increasing attention in recent econometric literature. However, algorithms that implement the solution in its full form are very complex and require a large amount of computing resources. Consequently, optimal control problems are usually solved either by neglecting the stochastic nature of econometric models (i.e., by means of standard deterministic control), or by using algorithms that give approximate solutions.
This paper explores the full stochastic approach, presenting an operational iterative procedure that implements the full stochastic optimal control solution. This is done by using stochastic simulations to compute the multiplier matrix that maps changes of the control variables into changes of the objectives and of the variances of the objectives.
