Identification of Time-Varying Factor Models
研究了在因子和载荷均可随时间变化时，如何在不进行旋转的情况下渐近估计它们，并比较了不同识别限制对预测美国宏观经济变量的影响。
The emergence of large datasets with long time spans has cast doubt on the assumption of constant loadings in conventional factor models. Being a potential solution, the time-varying factor model (TVFM) has attracted enormous interest in the literature. However, TVFM also suffers from the well-known problem of nonidentifiability. This article considers the situations under which both the factors and factor loadings can be estimated without rotations asymptotically. Asymptotic distributions of the proposed estimators are derived. Theoretical findings are supported by simulations. Finally, we evaluate the forecasting performance of the estimated factors subject to different identification restrictions using an extensive dataset of the U.S. macroeconomic variables. Substantial differences are found among the choices of identification restrictions.