FEE

Methodology for estimating brazilian quarterly GDP series employing cointegration and kalman filter procedures

Authors
  • Luiz Fernando Cerqueira

Keywords:
Estatística - Modelos matemáticos, Kalman, filtro, Produto interno bruto, Contas Nacionais
Abstract
In this paper we present a methodology to estimate the Brazilian GDP's quarterly series from 1960 to 1996. Firstly, the suggested methodology applies Engle-Granger two-step cointegration procedure, involving annual GDP data and some GDP correlated variables. The estimated coefficient of the long-run relationship is used to create a quarterly equation with the GDP and its covariates. From this equation the first quarterly GDP estimate is obtained. The following step consists of improving this estimate with the state space framework, which use the Kalman filter. We estimated univariate models with missing values and benchmarking process, as well as a multivariate SURE with explanatory variables. The estimated model shows appropriate goodness of fit and the diagnostic tests for standardized innovations suggest the assumptions underlying the Gaussian linear model are valid for the data. The paper's major contribution is to present an approach for estimating the quarterly GDP series and developing a methodological framework for its forecasting.
Section
Artigos
License
A remessa dos artigos à revista implica a cessão dos direitos autorais à FEE.