Multiple Structural Breaks in Vector Error Correction Models

The analysis of structural breaks in vector error correction models is often confined to possible level shifts and trend breaks. In contrast, only rudimentary tools are available to deal with breaks in the cointegration matrix and the adjustment towards equilibrium. Particularly, the possibility of multiple structural breaks during long sampling periods is often ignored which can lead to inconsistently estimated coefficients. In this paper, we study a two-step estimator based on a penalized regression to determine the number of structural breaks, their timing, and estimate the model’s coefficients for each regime. We focus on two important cases, namely, (i) constant dynamics but changing long-run equilibria, and (ii) convergence to new long-run equilibria with new adjustment dynamics. We use simulations to investigate the finite sample performance of the two-step estimator and provide an empirical illustration using data on the term structure of interest rates.