Indicadores líderes, redes neuronales y predicción de corto plazo

• Autores: Kapsoli Salinas, Javier, Bencich Aguilar, Brigitt
• Formato: artículo
• Fecha de Publicación: 2004
• Institución: Pontificia Universidad Católica del Perú
• Repositorio: PUCP-Institucional
• Lenguaje: español
• OAI Identifier: oai:repositorio.pucp.edu.pe:20.500.14657/118091
• Enlace del recurso: http://revistas.pucp.edu.pe/index.php/economia/article/view/867/828; https://doi.org/10.18800/economia.200401.006
• Nivel de acceso: acceso abierto
• Materia: Economía; https://purl.org/pe-repo/ocde/ford#5.02.01

This paper presents a procedure for constructing a short-term predictor of the level of economic activity. To do so, the Baxter-King filter is used to decompose the monthly GDP series into its three components: seasonal, cyclical, and trend. The cyclical component is then estimated and forecasted using a set of leading variables that lead GDP. It is proposed that the relationships between these variables and the GDP cycle are determined through a nonlinear neural network model. The remaining components are estimated using standard econometric models. Finally, the three components are aggregated to obtain an indicator of future GDP developments. The resulting prediction demonstrates a reasonable level of reliability, making the proposed index a useful tool for decision-making given its ready availability relative to official statistics.