We compare and evaluate the performance of four widely used numerical solution methods to dynamic rational expectations stochastic models, in the context of optimal and nonoptimal Pareto settings using a wide variety of statistical measures and two sample sizes. We find that: (i.) differences between methods do not necessarily increase with the complexity of the solved model (ii.) For all the example model economies we considered, a log-linear approximation behaves as well as a more complex to implernent finite element method. (iii.) Rejection of a particular solution method attending to the fulfilment of the rational expectation hypothesis is compatible with almost no differences between methods attending to other comparison criteria. (iv.) It is proper to consider 'large' sample sizes to check the properties of a particular solution.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados