기준금리 변동이 지역별 아파트 가격에 미치는 비대칭적 영향 - 비선형 자기회귀시차분포와 XGBoost-SHAP 모형을 활용한 분석 -

초록

This study examines the asymmetric long-run effects of the Bank of Korea’s base rate on regional apartment prices by combining a nonlinear autoregressive distributed lag (NARDL) model with the eXtreme Gradient Boosting (XGBoost)-Shapley additive explanation (SHAP) framework across 17 Korean provinces. The price-stimulating effect of rate cuts significantly exceeds the contractionary effect of rate hikes in several provinces concentrated in the capital region, with pronounced regional heterogeneity in the strength of this asymmetry; outer-capital provinces respond selectively to easing while barely reacting to tightening. In some provinces, structural industrial decline appears to weaken the standard transmission. Block-bootstrap tests confirm that the SHAP importance of the NARDL residual is statistically significant; however, the combined model does not improve short-horizon forecast accuracy over the standalone NARDL. Therefore, the contribution of the design lies in integrating structural interpretation with nonlinear attribution, rather than in forecast improvement. These findings provide grounds for state-contingent and regionally differentiated macroprudential policies.

키워드