MOLDOVA GDP FORECASTING USING BAYESIAN MULTIVARIATE MODELS

PÂRȚACHI Ion, Academy of Economic Studies of Moldova

MIJA Simion, Academy of Economic Studies of Moldova

Abstract:
Building a multivariate GDP forecasting model based on relevant macroeconomic indicators selected through a proper selection process. This paper assesses whether alternative specifications of the Bayesian model can provide higher forecast accuracy compared to a standard VECM (Vector Error Correction Model). To achieve this, a Bayesian VAR (Vector Autoregressive) model is estimated using the Litterman precedent (1979). Compare the result based on the Bayesian VAR (Vector Autoregressive) model with the DFM (Dynamic Factor Model). The out-of-sample forecast performance of the models is then evaluated over a 5-year period (20 quarters), where model efficiency for a long forecast period is ascertained.

Keywords: GDP Forecast, Econometric Models, Bayesian VAR

JEL Classification: C10, C22, C38, C52

Volume: 76, Issue: 1

Pages: 85 - 93

Publication date: March, 2024

DOI: 10.56043/reveco-2024-0008

Download the article: http://economice.ulbsibiu.ro/revista.economica/archive/76108partachi&mija.pdf

”Cite