| 刘玲君,陈小亮,陈彦斌.货币政策有效性的影响因素再评估：基于可解释性机器学习方法[J].数量经济技术经济研究,2026,(1):27-52 | | 货币政策有效性的影响因素再评估：基于可解释性机器学习方法 | | Reassessment of Factors Affecting the Effectiveness of Monetary Policy: Based on Explainable Machine Learning Methods | | | | DOI： | | 中文关键词: 货币政策有效性 经济增长 经济结构 可解释性机器学习方法 | | 英文关键词: The Effectiveness of Monetary Policy Economic Growth Economic Structure Explainable Machine Learning Methods | | 基金项目: | | | 中文摘要: | | 实现党的二十届三中全会提出的“ 畅通货币政策传导机制 ”和2025 年政府工作报告提出的“进一步疏通货币政策传导渠道 ”重要目标，需要准确识别货币政策有效性的影响因素 。本文基于中国 2002~2022 年的月度数据，综合使用随机森林机器学习方法和 SHAP 值可解释性方法，测算并识别了近年来中国货币政策有效性的主要影响因素 。研究结果表明，第一，宏观经济结构失衡和经济增速放缓是 2008 年之后货币政策有效性下降的主要原因 。第二，虽然目前老龄化对货币政策有效性的影响尚不突出，但已呈现明显上升态势 。伴随着老龄化率的持续升高，其对货币政策有效性的冲击还将进一步加大 。有鉴于此，为了“ 畅通货币政策传导机制”，顺利实现 2025 年政府工作报告提出的“大力提振消费、提高投资效益，全方位扩大国内需求 ”等重要目标，不仅要完善金融市场和货币政策，还要做好两方面重要工作 。一是从改善经济增长态势和完善经济结构等视角入手，扫清货币政策传导面临的阻碍 。二是前瞻性防范老龄化对货币政策有效性的冲击。 | | 英文摘要: | | The key to achieving the important goals of “unblocking the transmission mechanism of monetary policy” proposed at the Third Plenary Session of the 20th Central Committee of the Communist Party of China and “further unblocking the transmission channels of monetary policy” proposed in the 2025 government work report lies in accurately identifying the influencing factors of the effectiveness of monetary policy. However, the existing literature has only analyzed whether the influence of a certain factor is significant and has not analyzed and ranked the importance of the influences of each factor. This is largely because existing research mainly uses traditional measurement methods, which can only focus on determining the impact of one or a few factors on the effectiveness of monetary policy. Therefore, it is unable to answer the major question of why the effectiveness of China’s monetary policy needs to be improved after the 2008 international financial crisis nor can it provide clear and targeted response plans. In addition, traditional measurement methods mainly examine the linear relationships among variables, and research on nonlinear relationships is relatively limited. However, the influence of various factors on the effectiveness of monetary policy may have complex nonlinear relationships.Compared with traditional empirical methods, the problem-solving ability of machine learning methods has been significantly enhanced. They not only conduct a more comprehensive examination of the influencing factors of monetary policy effectiveness but also deeply explore the nonlinear relationship between each influencing factor and the effectiveness of monetary policy. By combining machine learning methods with SHAP value interpretability methods, the influence of each factor on the effectiveness of monetary policy can be calculated, and the importance of each factor can be ranked. In view of this, based on the monthly data of China from 2002 to 2022, this study constructs an indicator system that includes a total of 17 indicators in 5 major categories-financial markets, macro policies, economic growth, economic structure, and aging. It comprehensively employs the random forest machine learning method and the SHAP value explainability method. The main influencing factors of the effectiveness of China’s monetary policy in recent years have been identified.The three main research conclusions of this study are as follows: First, the persistent economic structural imbalance and the slowdown in economic growth since 2008 are the main reasons for the need to improve the effectiveness of monetary policy. Second, the factors related to financial markets and macro policies that the academic and policy circles have been focusing on are important influencing factors for the effectiveness of monetary policy. However, their significance has considerably declined since 2008. Third, the impact of aging on the effectiveness of monetary policy is constantly increasing. Under the new national conditions where population aging and negative population growth are superimposed, the impact of aging on the effectiveness of monetary policy is expected to further rise.Compared with existing studies, this study mainly makes the following contributions to the literature. First, machine learning methods can rank the importance of different influencing factors, which helps us identify the main factors affecting the effectiveness of monetary policy. Based on this, this study uses the causal SHAP value method, which can alleviate the impact of endogeneity problems on the importance of influencing factors. It provides new causal evidence for the core conclusion that “enhancing the effectiveness of monetary policy should not only focus on the factors of the financial market itself but also attach importance to the impact of economic growth and economic structure.” Second, by combining machine learning methods with SHAP values method, a dynamic comparative analysis was conducted on the main influencing factors of the effectiveness of monetary policy in different periods, thereby answering the important question of what the main reasons are for the need to improve the effectiveness of monetary policy after 2008 and providing valuable decision-making references for implementing important policy arrangements such as “unblocking the transmission mechanism of monetary policy” proposed at the Third Plenary Session of the 20th Central Committee of the Communist Party of China. | | 查看全文 相关附件： 下载数据代码附录 |
| | |
