Illustrating extraordinary shocks causing trend breaks

Structural breaks in a trending variable have been specified as changes in the drift parameter in the trend component, but extraordinary shocks causing these breaks have not been explicitly formulated. In this paper, the Hodrick–Prescott filter is extended by assuming two kinds of variance for the system noise driving the trend component: the larger one adopted in a point of time causing a trend break, and the smaller one adopted for remaining sequences. The number and location of structural breaks are determined by information criteria. In the proposed method, extraordinary shocks themselves can be illustrated. A Monte Carlo study shows the efficacy of the proposed model. Empirical results suggest that except for the UK, extraordinary shocks in quarterly time series of industrial production are detected for remaining six developed countries. Finally, it is shown that the proposed method considerably outperforms the other competing methods in correctly detecting business cycles.