| 141 | 0 | 13 |
| 下载次数 | 被引频次 | 阅读次数 |
精准预判经济风险与甄别预警信息,对短期政策调控与中长期高质量发展至关重要。因此,构建了一个融合多分布高维混频数据驱动的机器学习风险预测与预警信息甄别框架,基于多维度高维混频宏观经济数据,对中国经济增长风险进行实时预测与预警信息甄别。实证结果表明:第一,基于Sinh-Arcsinh分布及其与偏态t分布组合的MF-QRLSTM模型,通过高维混频数据的无损融合与变量间非线性动态关系的精准刻画,较传统预测方法显著提升了经济增长风险的实时预测精度。第二,金融状况对经济增长风险的边际预测贡献呈现明显的状态依赖性:经济平稳期其边际贡献较弱,而在经济冲击期则能提供更多的预测增量信息。第三,实体经济发展、国内外金融状况与宏观预期等是影响经济增长风险预测的关键因素,能够为经济增长风险提供重要的预警信息。本研究对助力构建新形势下经济增长风险监测预警体系具有重要的启示意义。
Abstract：Accurately forecasting economic risks and identifying early warning signals are crucial for short-term policy regulation as well as long-term high-quality development.This study develops a machine learning framework for risk forecasting and early warning signal identification,integrating multidistribution and high-dimensional mixed-frequency data,which enables real-time forecasting of economic growth risks and screening of early warning information for China.The empirical results show that,first,the proposed MF-QRLSTM model based on the Sinh-Aresinh distribution and its combination with the skew-t distribution significantly improves the real-time prediction accuracy of economic growth risks compared to traditional Ga R methods.This improvement is achieved through integration of highdimensional mixed-frequency data and precise characterization of nonlinear dynamic relationships among variables.Second,the marginal predictive contribution of financial conditions to economic growth risks exhibits clear state dependence:their contribution is relatively weak during periods of economic stability,but provides significantly more incremental predictive information during periods of economic shocks.Third,real economic development,domestic and international financial conditions,and macroeconomic expectations are key factors influencing the prediction of economic growth risks and can provide important early warning signals.The findings of this study offer valuable insights for building a risk monitoring and early warning system for economic growth under the new circumstances.
[1] ALTIG D,BARRERO J M,BLOOM N,et al.Surveying business uncertainty[J].Journal of econometrics,2022,231(1):282-303.
[2] 刘汉，刘营，卓杏轩.基于高维混频信息甄选的宏观经济总量实时预报与短期预测[J].经济学(季刊),2023,23(3):1112-1130.
[3] 王维国，于扬.基于混频回归类模型对中国季度GDP的预报方法研究[J].数量经济技术经济研究，2016,33(4):108-125.
[4] 刘汉，刘金全.中国宏观经济总量的实时预报与短期预测——基于混频数据预测模型的实证研究[J].经济研究，2011,46(3):4-17.
[5] 徐晓芹，赵文聚，仇东山，等.长短时记忆神经网络模型在兴济河洪水预报中的应用[J].济南大学学报(自然科学版),2025,39(6):843-851.
[6] 刘金全，张龙.中国混频金融状况指数的经济增长预测效果与检验[J].统计与信息论坛，2019,34(1):30-39.
[7] 韩阳.基于深度学习的中国宏观经济运行评估[J].数量经济技术经济研究，2023,40(3):189-212.
[8] 张蜀林，杨洋.基于混频数据的人民币汇率预测研究[J].商业研究，2016(12):65-72.
[9] ADRIAN T,BOYARCHENKO N,GIANNONE D.Vulnerable growth[J].American economic review,2019,109(4):1263-1289.
[10] 张晓晶，刘磊.宏观分析新范式下的金融风险与经济增长——兼论新型冠状病毒肺炎疫情冲击与在险增长[J].经济研究，2020,55(6):4-21.
[11] GU X,CHENG X,ZHU Z,et al.Economic policy uncertainty and China’s growth-at-risk[J].Economic analysis and policy,2021,70:452-467.
[12] LIU H,WANG L,ZHUO X.Unveiling the shadows:the effects of financial conditions on the tail risks of China’s macroeconomic activities[J].Economic analysis and policy,2025,85:1-14.
[13] HARTIGAN L,WRIGHT M.Monitoring financial conditions and downside risk to economic activity in Australia[J].Economic record,2023,99(325):253-287.
[14] 邓创，吴健，吴超.外部经济、金融不确定性与我国的宏观经济下行风险[J].统计研究，2022,39(6):36-51.
[15] 郑挺国，叶仕奇，范馨月.大数据下经济在险增长测度与风险探源研究[J].经济研究，2023,58(11):133-152.
[16] 隋建利，吕文强.从“脆弱性”到“韧性”:中国经济在险增长的脆弱性溯源与韧性解构[J].中国工业经济，2024(4):17-36.
[17] ADÄMMER P,PRÜSER J,SCHÜSSLER R A.Forecasting macroeconomic tail risk in real time:do textual data add value?[J].International journal of forecasting,2025,41(1):307-320.
[18] PLAGBORG-M∅LLER M,REICHLIN L,RICCO G,et al.When is growth at risk?[J].Brookings papers on economic activity,2020(1):167-229.
[19] PRÜSER J,HUBER F.Nonlinearities in macroeconomic tail risk through the lens of big data quantile regressions[J].Journal of applied econometrics,2024,39(2):269-291.
[20] SCHICK M.Real-time nowcasting growth-at-risk using the survey of professional forecasters[EB/OL].AWI discussion paper,No.750,2024.
[21] PFARRHOFER M.Modeling tail risks of inflation using unobserved component quantile regressions[J].Journal of economic dynamics and control,2022,143:104493.
[22] FERRARA L,MOGLIANI M,SAHUC J-G.High-frequency monitoring of growth-at-risk[J].International journal of forecasting,2022,38(2):582-595.
[23] CARRIERO A,CLARK T E,MARCELLINO M.Nowcasting tail risk to economic activity at a weekly frequency[J].Journal of applied econometrics,2022,37(5):843-866.
[24] HUBER F,KOOP G,ONORANTE L,et al.Nowcasting in a pandemic using non-parametric mixed frequency VARs[J].Journal of econometrics,2023,232(1):52-69.
[25] JONES M C,PEWSEY A.Sinh-arcsinh distributions[J].Biometrika,2009,96(4):761-780.
[26] 袁铭，温博慧，徐子钦.公众信心与中国宏观经济波动的非线性因果关系研究[J].统计与信息论坛，2018,33(5):58-66.
[27] BARNES E A,BARNES R J,DEMARIA M.Sinh-arcsinh-normal distributions to add uncertainty to neural network regression tasks:applications to tropical cyclone intensity forecasts[J].Environmental data science,2023,2:e15.
[28] AZZALINI A,CAPITANIO A.Distributions generated by perturbation of symmetry with emphasis on a multivariate skew t-distribution[J].Journal of the royal statistical society series a:statistics in society,2003,65:367-386.
[29] 于扬，王维国.混频数据回归模型的分析技术及其应用[J].统计与信息论坛，2015,30(12):22-30.
[30] GHYSELS E,HILL J B,MOTEGI K.Testing for Granger causality with mixed frequency data[J].Journal of econometrics,2016,192(1):207-230.
[31] LLOYD S,MANUEL E,PANCHEV K.Foreign vulnerabilities,domestic risks:the global drivers of GDP-at-risk[J].IMF economic review,2024,72(1):335-392.
[32] XU Q,XU M,JIANG C,et al.Mixed-frequency growth-at-risk with the MIDAS-QR method:evidence from China[J].Economic systems,2023,47(4):101131.
[33] DIEBOLD F X,MARIANO R S.Comparing predictive accuracy[J].Journal of business & economic statistics,1995,13(3):253-263.
[34] REICHLIN L,RICCO G,HASENZAGL T.Financial variables as predictors of real growth vulnerability[EB/OL].CEPR discussion paper,No.DP14322,2020.
[35] CLARK T E,HUBER F,KOOP G,et al.Forecasting US inflation using Bayesian nonparametric models[J].The annals of applied statistics,2024,18(2):1421-1444.
[36] 许启发，蒋翠侠，汤彬彬.基于LASSO-QVAR模型的中国金融机构尾部关联网络特征与系统性风险研究[J].统计与信息论坛，2025,40(3):56-72.
[37] 刘汉，王李俊，刘金全.三重压力下经济下行的识别与溯源[J].经济科学，2023(6):5-27.
(1)首先,鉴于前文SAS分布具有较好的经济增长风险预测性能,本文遵循这一方案;其次,中国金融类相关变量多以月度频率统计,故本文选择在混频数据模型中量化金融因素对中国经济增长风险的预测能力。
基本信息:
DOI：10.20207/j.cnki.1007-3116.20260519.001
中图分类号:F224;F124
引用信息:
[1]王李俊,刘汉.中国经济增长风险的实时预测与预警信息甄别[J].统计与信息论坛,2026,41(06):17-28.DOI:10.20207/j.cnki.1007-3116.20260519.001.
基金信息:
国家社会科学基金重大项目“大数据方法在宏观经济预测中的应用研究”（23&ZD075）
2026-05-20
2026-05-20
2026-05-20