Online Program Home
  My Program

Abstract Details

Activity Number: 580 - Time Series and Factor Models
Type: Contributed
Date/Time: Wednesday, August 2, 2017 : 2:00 PM to 3:50 PM
Sponsor: Business and Economic Statistics Section
Abstract #323581 View Presentation
Title: Efficient Estimation and Inference in Factor Models with Nonstationary Common And/Or Idiosyncratic Components
Author(s): Haiqing Zhao* and Mohitosh Kejriwal
Companies: Purdue University and Purdue University
Keywords: Factor models ; Common trends ; Maximum likelihood ; Generalized least squares ; Principal components

This paper studies methods for efficient estimation and inference in factor models where the common factors and/or the idiosyncratic components are potentially nonstationary. Our paper explores the possibility of improving the Bai and Ng (2004) PANIC procedure by proposing a modification that accounts for serial correlation and heteroskedasticity in the error component when estimating the model parameters. We employ the Iterated Principal Components [IPC] estimator from Breitung and Tenhofen (2011) and the Iterated Maximum Likelihood [IML] estimator from Bai and Li (2016), which have been developed in the stationary framework, and show how to extend them to the nonstationary case. Our simulations indicate that with heteroskedastic errors, the IML approach dominates both in terms of factor estimation and hypothesis testing. Further, with serial correlated errors, the IPC and IML approaches offer comparable advantages over the original PANIC approach. Finally, we illustrate the relative performance of the various approaches in a macroeconomic forecasting application.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

Copyright © American Statistical Association