Previsão da produção industrial: indicadores antecedentes e modelos de série temporal

Abstract

The main purpose of this paper is to compare the leading indicator methodology with basic structural time series models, in order to predict the turning points of the general industrial production. index, 'which is the target serie chosen. Several methods of aggregation are used in order to obtain the weights of the series which are used to construct the leading indicator. The best weights are given by least square method. In order to improve the forecast property of the model, an error correction model was estimated. Another result, following Neftçi (1982), obtained in this paper is the probability of turning points. In order to derive this probability, the target series is assumed to be a stochastic process which switch regimes in a neighborhood of the turning points. These probabilities are computed, given a probability of false indication of turning point, for the observed and predicted target series.