| 讲座简介： | The time-varying parameter dynamic factor model (TVP-DFM) has received increasing attention in recent years due to its consideration of the changing economic environment and structure. However, its estimation and applications are still very limited. This paper proposes an efficient joint quasi-maximum likelihood estimation approach to estimate TVP-DFM. The joint estimation approach represents TVP-DFM as a linear non-Gaussian state space model and then exploits quasi-maximum likelihood estimation to estimate unknown parameters of the model, as well as common factors and time-varying factor loadings. Other two related estimation approaches, namely dual estimation and approximate estimation, are also introduced. In the simulation study, the finite-sample performance and in-sample fitting performance of the proposed estimators are further compared. Moreover, to adapt to high-dimensional modeling and estimation of large-scale economic variables, we propose a parsimonious TVP-DFM and confirm its applicability and accuracy in high-dimensional scenarios. Finally, we use basic TVP-DFM and parsimonious TVP-DFM to empirically measure China's economic conditions from low and high dimensions, as well as from a real-time data-rich environment. |
