Stochastic and robust control of nonlinear economic systems

This paper uses a robust policy concept that is an extension to nonlinear systems of the mean-variance approach for linear models. In the linear case, the approach reduces to the minimization of a convex combination of the mean and variance of the stochastic problem. The paper also discusses a method for computing expectations of nonlinear systems based on evaluating the bias from certainty equivalence using Monte Carlo simulations. This method computes robust strategies. If desired, these can be in the form of simple decision rules, as aggregate parametrisations of full feedback strategies. The parameters are determined to minimize a convex combination of the original objective function and a sensitivity (variance) expression. The latter can either be chosen as the sensitivity (variance) of the original objective function or the sensitivity (variance) of the individual endogenous variables.