Long Memory Factor Model: On Estimation of Factor Memories
61 Pages Posted: 22 Feb 2017 Last revised: 13 Feb 2018
Date Written: January 16, 2018
This paper considers the estimation of factor memories in the context of a high-dimensional factor model. Both factors and idiosyncratic error terms are potentially non-stationary fractional integrated processes. We propose a three-step procedure to estimate the latent factors. We then apply the fully-extended local Whittle (FELW) estimator of Abadir et al. (2007) to compute factor memories. This estimator is consistent and satisfies the same normal CLT, as if the factors are observed. Finite sample performance of the proposed procedure is evaluated in a simulation study. Finally, we apply the proposed estimator on a large dataset of macroeconomic variables.
Keywords: approximate factor models, principal components, long memory, fractional integration
JEL Classification: C14, C22, C38
Suggested Citation: Suggested Citation