最後，為檢驗實證結果之穩健性，本文採用分量向量自我迴歸（Quantile Vector Autoregression, QVAR）模型，以探討主要變數間在不同分位數下的動態關係。結果顯示，QVAR 與分量迴歸結果一致，金融變數皆是正向影響實質面經濟，且工業生產成長率對金融情勢指數與民間信用占比GDP 成長率不具內生性。衝擊反應分析進一步指出，金融情勢指數在較低分位時對工業生產成長率具更大的顯著影響。;This study examines the asymmetric effects of financial variables on Taiwan’s economy using a quantile regression framework. The results show that short-term financial easing supports industrial production growth but heightens long-term downside risks. To capture tail risk more precisely, the analysis incorporates a skewedt distribution into the model. The results show that Growth-at-Risk (GaR) remained stable before COVID-19 but increased significantly after, indicating greater sensitivity to adverse shocks. The analysis further integrates the skewed-t distribution into an entropy model to strengthen forward-looking risk assessment. Results show that downside entropy peaks in advance of economic recessions. Granger causality tests suggest it precedes and predicts future industrial production growth. To ensure robustness, a Quantile Vector Autoregression (QVAR) model analyzes the dynamic relationships across quantiles. The QVAR findings confirm those of the quantile regression, showing that financial variables positively affect the economy. Additionally, no evidence of endogeneity is found from industrial production growth to the Financial Conditions Index or credit-to-GDP. Impulse response analysis shows stronger effects of financial conditions at lower quantiles.
