The Bank of Korea has since 2007 been carrying out economic forecasting and policy effect analyses based upon a structural model that emphasizes the theoretical perspective of the Dynamic Stochastic General Equilibrium (DSGE) models, while complementing its power in explaining reality by taking a multi-model approach through time-series models such as VAR models.

 

    With the economic environment and behaviors of economic agents having changed greatly of late, there has been a growing need to reorganize the constituent models of the macroeconomic modelling system. As economic integration has progressed, the interconnectedness both between the financial and the real sectors and between countries has been heightened. Also changes in the economic environment including demographic aging and mounting uncertainty have modified the behaviors of economic agents. If the existing models continue to be used despite such changes in circumstances, this is likely to cause problems such as a reduction in their predictive power regarding the economy and inaccurate economic analyses.

    The Bank of Korea, impelled by the recognition of these problems, has started to reorganize its economic forecast modelling system since 2012 and is currently developing a new policy simulation model. The VAR model block constructed in this connection forms part of the work of improvement, and will act to heighten the power to explain reality of the BOK macroeconomic modelling system.

 

    The re-constructed VAR model block is divided into two models : a GVAR model and a BVAR model. The GVAR model includes macroeconomic variables of more than 30 major countries around the world and is expected to complement the BOKGM currently used by the Bank of Korea to analyze the global economy. While the BOKGM is a global model embracing six major countries around the world, the GVAR model has a greater power to explain reality, being based on a time-series model. Therefore, the two models can be employed in a mutually complementary manner. In addition, the BVAR model, which estimates a one-country model based upon a Bayesian methodology, has strengthened financial and real sector interconnectedness within the model and, as a result, its short-term economic forecasting performance has turned out to be satisfactory. The BVAR model is seen to find its main use in appraising the BOKDSGE model, whose reconstruction is currently underway, and for short-term economic forecasting involving a forecast combination.