【专家简介】：郑挺国，厦门大学宏观经济研究中心副主任、邹至庄经济研究院和经济学院教授、博士生导师，厦门大学南强重点岗位教授，教育部青年长江学者、国家万人计划青年拔尖人才、福建省哲学社会科学领军人才、教育部新世纪优秀人才。研究领域为宏观经济与政策分析、宏观计量经济学、大数据方法与应用。先后在《经济研究》（13篇）《管理世界》《经济学（季刊）》、Journal of Econometrics、Journal of Business & Economic Statistics、Journal of Economic Dynamics & Control、International Journal of Forecasting等国内外重要学术期刊上共发表百余篇论文。主持国家社科基金重大项目1项、国家自然科学基金面上项目4项、教育部人文社科重点研究基地重大项目1项。
【报告摘要】：This paper proposes a novel time-varying parameter mixed-frequency dynamic factor model (TVP-MFDFM) for high-frequency economic condition tracking. By integrating time-varying factor loadings and conditional heteroscedasticity in idiosyncratic errors, the model effectively captures the dynamic relationships among economic variables across daily, weekly, monthly, and quarterly frequencies. To address the high-dimensionality issue in parameter estimation, the model further assumes that factor loadings follow multivariate random walks driven by a common variance parameter and that time-varying volatilities follow exponentially weighted moving average processes. The quasi-maximum likelihood method is developed for estimation, which can effectively handle non-Gaussian disturbances induced by a dynamic bilinear structure and missing values in mixed-frequency data. Empirical studies on U.S. and Chinese economic data are conducted. For the U.S., the model produces a robust daily economic index that aligns with established benchmarks such as the ADS index, but with reduced volatility. For China, the model accurately captures economic fluctuations since 1979, while real-time analysis confirms its timeliness and accuracy in tracking economic conditions. These results highlight TVP-MFDFM's potential for improving high-frequency economic monitoring and policymaking.
【报告时间】：2026年03月26日15:00-16:20
【报告地点】：位育楼417
