ESRI Discussion Paper Series No.357 Economic Structural Change and Business Cycle Monitoring within the Framework of PCA-DFM

From a practical aspect, this paper is concerned about the general question of how economic structural changes matter in business cycle monitoring. Recent works provide a theoretical answer within the framework of a principal component estimation of dynamic factor model: the structural changes as parameter shifts in dynamic factor model do not affect the cyclical composite indicator as an estimated common component of a canonical time series. Not only is the effect of instability averaged out in a principal component estimation to some extent, but spurious factors absorb the effect if its magnitude is larger. Because this proposition relies on an asymptotics and some thought-to-be general but unverifiable conditions, this paper sees its validity using Japanese monthly 330 time series variables spanning Feb. 1983 to Oct. 2018. In addition, recently proposed tests for structural change are applied to this dataset.