Model selection via genetic algorithms illustrated with cross-country growth data

A new simple procedure is provided, which is used to select the regressors of the cross-country growth model regression, which shows good performance relative to other alternative methodologies when they are compared in a simulation environment.

Abstract

We provide a new simple procedure for selecting econometric models, which is used to select the regressors of the cross-country growth model regression. This procedure is based on a heuristic approach called genetic algorithms which are used to explore the universe of models made available by a General Unrestricted Model. This search process of the correct model is only guided by the Schwarz information criterion, which acts as the loss function of the genetic algorithm in order to rank the models. Our procedure shows good performance relative to other alternative methodologies when they are compared in a simulation environment.