This thesis presents a comprehensive set of techniques for solving, simulating, analysing and controlling large scale, nonlinear, econometric models that contain rational expectations of future dated variables. These expectations are generally treated as model consistent, whereby the expectation is set to the deterministic projection of the model. Solutions to such models are distinguished from those of conventional models by the fact that they are not recursive in time. The outcome for the current period depends on the expected outcome for future periods as well as past periods. This property means that all of the basic numerical procedures need to be altered. We consider the following topics: solution algorithms for the two—point boundary value problem; terminal conditions, uniqueness and stability; experimental design and stochastic simulation; model forms, solution modes and historical tracking; control methods including optimal control. We find that suitable procedures allow us to undertake all of the experiments usually conducted with conventional models. Each topic is illustrated by application to three large scale models of the United Kingdom economy which contain rational expectations terms. Only one of these models is constructed following the new-classical paradigm and hence their comparative properties revealed by our experiments are of some interest.