Title : Estimating Korean Monthly GDP and Forecasting Korean GDP : Unobserved Component VECM and Gibbs Sampling Approach
Author : Ki-Ho Kim(BOK)
This paper suggests a Gibbs sampling estimation of an unobserved component cointegrated VAR Model to estimate monthly GDP with mixed frequency data.
We estimate a Canadian monthly GDP and compare the estimate and actual value of Canadian monthly GDP. The results show that the discrepancy between the estimate and actual value of Canadian monthly GDP is smaller or equivalent to the estimation results of Chow-Lin (1971), Denton (1971), Fernandez (1981), and Litterman (1983).
The suggested model can be used as a forecasting model for monthly, therefore quarterly GDP unlike the class of Chow-Lin. This paper ex post forecasts Korean monthly GDP(therefore quarterly GDP) in 6 month horizon and compares the forecast performance with other quarterly forecast models. Compared to Random walk, AR(1), AR(4), VAR(1), and VAR(4) models, the suggested model showed better forecast performance.