Bayesian forecasting and inference in latent structure for the Brazilian Industrial Production Index

Abstract

We consider the analysis of the Brazilian industrial production index (IPI) using statistical tools recently developed for time series. The main purpose is short-term forecasting and structural decomposition of the data through an autoregressive model that allows, but not imposes, nonstationary behavior. A very strong point of this model is that it incorporates all kinds of uncertainties by averaging forecasts across competing models, weighted by their posterior probabilities, in contrast with traditional analyses which assign probability one to a particular model. Additionally, the model considers innovation errors with heavy-tailed distributions and consequently accomodates for outlying observations. We interpret the results of the analysis in terms of its relation to the Brazilian economy.