Estimation of high-dimensional factor models with multiple structural changes

Abstract

This paper considers a high-dimensional factor model with unknown number of breaks. A simple two-step procedure is proposed to determine the number of breaks and identify break dates. It can be shown that the estimator of the number of breaks is consistent and the distance between the estimated and actual break dates is stochastically bounded under some mild conditions. Monte Carlo simulations demonstrate that the proposed method has desired performance in finite samples. Two real data are analyzed for illustrations. Overall, the proposed method could be expected to be more straightforward to determine the number of breaks in the underlying factor model.