Author : Sunho Lee(Bank of Korea)

<Abstract>

This study constructs a Large Bayesian Global Vector Autoregression (LB-GVAR) model covering eight major economies including South Korea, the US, Japan, China, the Eurozone, Brazil, Russia, and India to endogenously analyze international shock transmissions. The model incorporates four features. First, by applying Bayesian regularization techniques, the number of variables included per country is expanded to a range of 10 to 20, thereby broadly encompassing real, price, financial, and climate variables. Second, an outlier correction procedure prevents distortions in parameter estimation from extreme events. Third, optimal hyperparameters are endogenously determined via marginal likelihood maximization. Fourth, a masking technique isolates direct impacts from indirect network spillovers. Scenario analysis of five macroeconomic and climate shocks reveals that US monetary tightening and financial uncertainty induce global economic slowdowns and disinflation. Conversely, energy and climate shocks create a policy trade-off with simultaneous growth contraction and inflation. Furthermore, masking results show that indirect network effects account for a substantial portion of the total impact, underscoring the necessity of explicitly modeling international transmission channels to rigorously evaluate macroeconomic spillovers.