Search

En este trabajo presentamos un nuevo procedimiento para seleccionar modelos econométricos. Está basado en un enfoque heurístico, empleando algoritmos genéticos, que permite explorar el universo de modelos disponibles a partir de un modelo general sin restricciones. Este proceso de búsqueda del modelo óptimo está guiado únicamente por el criterio de información de Schwarz, que actúa como la función pérdida de un algoritmo genético empleado para seleccionar el modelo óptimo. Este procedimiento muestra buen comportamiento en relación a otras metodologías alternativas. A modo de ejemplos de su utilidad se muestran tres problemas donde el algoritmo ha sido empleado con éxito: la selección de variables que explican el crecimiento económico, la predicción del fracaso empresarial y la formación de una cartera de pocos activos que siga el comportamiento de un índice bursátil como el IBEX35 Español. ABSTRACT In this paper we present a new procedure for selecting econometric models. It is based on a heuristic approach, called genetic algorithms, which permits us to explore the universe of available models starting from a general model without restrictions. This search process for the optimal model is guided only by the Schwarz Information Criterion, which acts as the lost function of the genetic algorithm employed for selecting the optimum. This procedure shows good performance with respect to other methodologies. As examples of its utility three problems where the algorithm was successfully employed are presented: the selection of variables that explain economic growth, the prediction of failure of firms and the construction of a portfolio with few actives that follows the behaviour of a stock exchange index such as the Spanish IBEX35.

In this paper, we propose a new methology for Index Tracking (IT) by means of cointegration which provides some significant improvements on that field. As the quality of the tracking portfolio (TP) depends highly on the stock selection procedure, we propose picking the stocks using a model selection technique based on optimizing the cointegration level of the TP and the benchmark index instead of selecting, as in previous papers the assets by ad hoc decisions. To illustrate an empirical application of these techniques we use daily closing prices in the Dow Jones Industrial Average (DJIA) index over two different periods; one period which goes from 1 January 1990 to 31 December 2001 previously used by other authors, and the bear and a turmoil period, which goes from January 2007 to May 2012, inside the current financial crisis. Using only five assets we are able to successfully track the DJIA index and our results improve the IT technique based on cointegration that chooses stocks with maximum capitalization level. We also have compared our results with a more traditional procedure based on correlation and again our results reveal superiority. The empirical illustration not only has been focused on the TP itself, but has also been extended to tracking the index with an added profitability of 5, 10, 15 or 20% and to long-short strategies, producing profitable results. [ABSTRACT FROM AUTHOR]