Author: Sehun Kim(Sogang University), Joonyoung Hur(Sogang University)

Economic agents base decisions on expectations of future fiscal policy changes. However, VAR models only use past and present data, making it difficult to capture these expectations and accurately identify government spending shocks. To address this caveat, this paper constructs a multivariate Bayesian (Large Bayesian) VAR model with 25 variables, including expectation variables of private agents, to analyze the impact of expansionary government spending on GDP. This model is estimated by using Bayesian methods incorporating Minnesota prior distributions, and the identification of impulse response is conducted through zero and sign restrictions. The analysis shows that while the nonfundamentalness problem arises in conventional smallscale VAR models, it does not occur in the multivariate VAR model specified in this study. The estimated government spending multiplier indicates that a one-won increase in government spending arises GDP by 1.45 won in the same period. However, responses after the following period are generally not statistically significant. These findings suggest that although the effect of government spending on GDP is short-term and lacks persistence, it remains effective for economic stimulation through its immediate impact on GDP.