Recursive identification, estimation and forecasting of nonstationary economic time series with applications to GNP international data

  1. García Ferrer, Antonio
  2. Hoyo Bernat, Juan Luis
  3. Novales Cinca, Alfonso
  4. Young, Peter C.
Journal:
Documento de trabajo: Serie Econometría

Year of publication: 1993

Issue: 11

Pages: 1-33

Type: Working paper

Abstract

In this paper we propose a recursive, unobserved components model, where parameter variation is characterized by a particular state space formulation. The choice about the characteristics that define each component are half way between an objetive function optimization strategy and subjective Bayesian approach: some parameter values need to be chosen, but these are reduced to a minimum, and some values are provided to aid in their choice. Annual real output data for nine countries are analyzed under both, the univariate and transfer function version of our unobserved componenets model, the latter using the money supply as a leading indicator. The performance of these models is compared with the forecasting results obtained in previous work with the same data set.