On simulation and optimization of macroeconometric models

This paper develops adapted methods for solivng and optimizing deterministic nonlinear discrete dynamic macroeconometric models. It gives computational results concerning a Belgian model of about 500 nonlinear equations per time period, which leads us to a large-scale minimization problem with several thousands of nonlinear constraints. The paper uses new optimization techniques based on partially known quasi-Newton Hessian matrix updates in a reduced-gradient framework. It also discusses different techniques for dealing with bound constraints arising form different model from different model formulations.