时间序列时域分解与混合频率序列协整分析

To make full use of the data information of economic time series observed at different frequencies, it is necessary to disaggregate the time series observed at low frequency into compatible higher frequency time series, or conduct econometric analysis with mixed-frequency economic data.. In this project, the high-frequency time series are modeled as unobserved components model, which is represented as state space model to be estimated, and the unobserved high-frequency observations are treated as missing values. Firstly, the temporal disaggregation by state space model and its relation with other disaggregation methods are studied to develop a unified analytical framework of temporal disaggregaton, which is used to clear the difference & relation among these disaggregaton methods and then achieve the dynamic expansion of Chow-Lin method. Secondly, the cointegration analysis of mixed-frequency time series is studied under the same model framework as temporal disaggregation, in which the impact of missing values on cointegration analysis is considered, and then the asymptotic distribution of test statistics and parameter estimators is derived. The last is empirical research, in which the real economic problems in China are analyzed based on the above theoretical research. . The research in this project will not only help to further perfect the theoretical system of temporal disaggregation, but also develop an approach to cointegration analysis with mixed-frequency data, which is important to solve the practical problems of serious data insufficiency in domestic macro-econometric analysis.

面对观测频率不同的经济序列，为保证数据信息的充分利用，需要通过时域分解将低频序列转化为高频序列，或直接使用混合频率的经济序列进行分析。本项目以未观测成分模型作为高频序列模型，将其表示为状态空间模型进行估计，未观测高频数据作为缺失值处理。在此基础上，首先研究基于状态空间模型的时域分解及其与其他分解方法的关系，将不同分解方法纳入统一分析框架，明确这些方法的分歧和联系，进而实现Chow-Lin分解的动态化扩展；其次，使用与时域分解统一的模型框架，研究混合频率序列的协整分析，考虑数据缺失的影响，推导协整检验统计量和参数估计量的渐近分布；最后将理论研究应用于分析中国现实经济问题。. 本项目的研究不仅有助于时域分解理论体系的进一步完善，还可以给出一种利用混合频率时间序列数据进行协整分析的方法，对解决国内宏观计量分析中样本数据严重缺乏的实际问题有重要帮助。

宏观计量分析中经常遇到不同频率的经济序列，这种情况下如何保证数据信息的充分利用，同时避免观测频率对结论稳健性的影响？本项目研究主要从三个方面展开：一是混频动态因子模型和混频VAR模型的估计，即利用高频数据和低频数据的关系，将高频模型表示为状态空间模型，在此基础上利用混频数据进行估计，进而估算不可观测的高频数据；二是分流量指标和存量指标，分别探讨了时域聚总和系统抽样对脉冲响应分析的影响，并给出相对稳健的处理方法；三是混频序列的单位根检验和协整分析，将Muller和Watson（2008，2013）关于序列低频波动性和协整关系的研究推广到混频情形，采取不同的权重，将混频的观测值序列转化为低频加权平均，在此基础上进行单位根和协整分析。本项目还将理论研究的结论应用于解决中国实际问题，探讨了一致指数的构建，以及通胀惯性和经济波动的持久性。. 本项目给出了宏观计量分析中常用的动态因子模型和VAR模型的混频估计方法，以及混频序列的单位根检验和协整分析方法，对于相关实证研究提供了必要的技术支持。此外，关于一致指数构建和高频数据估算的研究可以为统计部分和经济景气监测部门的相关工作提供一定的借鉴。