Description

The aim of this paper is to address the problem of dimensionality in a large database for the Chilean economy. An alternative to deal with this problem is through Bayesian regression methods (shrinkage), and the other given by classical literature based on the use of principal components. In general, this work shows that the proposed Bayesian methods provide a good description of the Chilean data: economic activity indicator (IMACEC), inflation, and productive sectors such as commerce and industry, making forecasts many times with small predictive error and less volatile than other time series models. The main result is that these Bayesian methods are a good alternative for forecasting the most relevant series of the Chilean economy.