Algoritmos Genéticos En La Estimación De Un Modelo Macroeconométrico Para Ecuador

Abstract

This paper presents the articulation of genetic algorithms with traditional least squares techniques for the estimation of a simultaneous equation macroeconometric model for Ecuador, whose application would allow us the monitoring of the relationships between macroeconomic and social variables (particularly, poverty evolution) in a very flexible too! for the simulation of several expectational hypothesis. Due to the rigid structure in most of the model specification --there is a theoretical precondition for numerous coefficients- we need a net of constraints on the parameters, and that increases the complexity of the estimation procedure. A common practice in the literature is to bypass this problem using parameter callibration. By the contrary, the option here takes the risk uf an econometric constrained estimation, assuming as given the parameter space for certain coefficients and rigorously searching for the estimates that best fit the data.