A1059
Title: Heavy factor models
Authors: Jihyun Kim - Sung Kyun Kwan University (Korea, South) [presenting]
Nour Meddahi - Toulouse School of Economics (France)
Abstract: The purpose is to examine a factor model with heavy factors that may have unbounded variance. For this factor model, the validity of PCA-based identification and estimation methods is reassessed. Under unbounded variance, the usual L2-projection methods, such as PCA, are not well-defined at the population level. However, it is found that the factors and factor loadings remain well-identified using PCA, regardless of the boundedness of the factors' variance. Additionally, a structural interpretation of the heavy factor model is provided, which enables the determination of the number of factors as well as the number of structural factors under this interpretation. As an empirical application, the methods are applied to the FRED-MD dataset and discover that some macroeconomic factors exhibit significant fat-tailed behavior and/or high persistence.
