Structural Breaks and Common Factors

In this paper, we survey the extant literature on structural breaks in the context of Dynamic Factor Models (DFMs). This literature mainly focus on stationary DFMs with the loadings and/or variances in the model having at least one break. Among the main conclusions is that structural breaks can inflate the number of factors determined using information criteria. However, in practice, large systems of economic/financial variables observed over large periods of time, are often non-stationary. Although there are results on the potential presence of spurious factors in non-stationary systems, as far as we are concerned, the extant literature has not dealt yet with the impact of structural breaks on non-stationary DFMs. By means of simulation experiments, we show that the number of factors determined using some popular information criteria, is inflated when a structural break in the loadings happens not too close to the starting or ending periods of the sample.