跨境资本异常波动的风险预警:基于机器学习视角  

Abnormal Fluctuation Risk Warning of Capital Flows:Based on Machine Learning Perspective

作　　者：谭小芬 王欣康 张碧琼[2] Tan Xiaofen;Wang Xinkang;Zhang Biqiong

机构地区：[1]北京航空航天大学经济管理学院,北京100191 [2]中央财经大学金融学院,北京100081

出　　处：《复印报刊资料（统计与精算）》2023年第6期93-108,共16页STATISTICS AND ACTUARIAL SCIENCE

基　　金：国家社会科学基金重大项目“负利率时代金融系统性风险的识别和防范研究”(20&ZD101);国家自然科学基金项目“汇率市场变化、跨境资本流动与金融风险防范”(71850005)。

摘　　要：基于“在险资本流动”的宏观研究新范式,运用预测分位数回归与稳定分布拟合方法,识别出中国的跨境资本异常波动风险。基于包含371个指标的高维数据集,通过构建的Lasso-PCA双重筛选机器学习方法,识别出不同类型跨境资本异动风险的关键预警因子。结果显示:(1)构建的跨境资本异动风险指标能较好地衡量中国所面临的资本流动尾部风险;(2)国内外长期利差扩大是债券资本涌入风险的关键预警因子,美国货币政策收紧是债券资本撤离风险的关键预警因子,全球风险偏好降低是股票资本涌入风险的关键预警因子,而美元名义有效汇率上升是股票资本撤离风险的关键预警因子;(3)2022年以来,美联储持续加息与美元走强会增加中国的跨境资本撤离风险。在金融市场双向开放进程加快的背景下,为中国防范化解外部冲击风险提供了重要参考依据。Extreme capital flows can significantly affect a country's economic and financial conditions and pose challenges to policymaking of authorities.In particular,when the capital flow trend reverses,or an extreme capital flow event occurs,it will have a non-negligible impact on a country.As the world's largest emerging market country,China has long faced large-scale inflows and outflows of international capital.Compared with other emerging economies,China is unique in terms of economic size,financial system,exchange rate regime,and capital account openness.The events that cause abnormal cross-border capital flows in China are different from those in other emerging economies in terms of frequency of occurrence,event characteristics,driving factors,and economic consequences.It can be expected that with the two-way opening of China's financial market,the pressure on China to deal with the large-scale two-way cross-border capital flows will further increase.Based on the new macro research paradigm of capital flow at risk(CFaR),this paper uses the methods of(forecast)quantile regression and stable distribution fitting to identify the risk of China's abnormal cross-border capital flows.Based on a high-dimensional data set containing 371 indicators,the Lasso-PCA double-screening machine learning method is constructed to identify the key early warning factors of different types of risks of abnormal cross-border capital flows.The research results are as follows.(1)The risk indicators of abnormal cross-border capital flows constructed in this paper can better measure the tail risk of capital flow faced by China.(2)The expansion of long-term interest rate spreads is the key early warning factor of the inflow risk of bonds,the tightening of monetary policy in the United States is the key early warning factor of the outflow risk of bonds,the reduction in global risk appetite is the key early warning factor of the inflow risk of equities,and the rise in the nominal effective exchange rate of the US dollar is the key early warning

关 键 词：跨境资本流动 异常波动风险 涌入风险 撤离风险 风险预警 机器学习 Lasso-PCA双重筛选模型 

分 类 号：F832.6[经济管理—金融学]