关键词: 因子模型, 结构突变, 突变点, 累积平方和统计量

Abstract: In order to endogenously ascertain the structural changes of factor loading matrix in dynamic factor models, this paper obtains a cumulative sum of squares statistic for the estimated pseudo factor sequences by using the principal component estimation to test the structural changes of factor loading matrix. Furthermore, the iterative cumulative sum of squares algorithm is applied to detect the positions of the change points. It is found that the testing statistic proposed in this paper is robust as to the biased number of factors set. Moreover, there are good finite sample and asymptotic convergence properties in the statistic. In addition, an empirical analysis shows that there are common factors of structural changes in the logarithmic return sequence of the listed companies in the A-Shares of the Shanghai Stock Exchange.

Key words: Factor Model, Structural Change, Change Point, Cumulative Sum of Squares Statistic