Testing for multiple structural breaks in multivariate long memory regression models

This paper focuses on the estimation and testing of multiple breaks that occur at unknown dates in multivariate long memory time series regression models, allowing for fractional cointegration. A likelihood-ratio based approach for estimating the breaks in the parameters and in the covariance of a system of long memory time series regressions is proposed. The limiting distributions as well as the consistency of the estimators are derived. Furthermore, a testing procedure to determine the unknown number of breaks is introduced which is based on iterative testing on the regression residuals. A Monte Carlo exercise shows the good finite sample properties of our novel approach, and empirical applications on inflation series of France and Germany and on benchmark government bonds of eight EMU countries illustrate the usefulness of the proposed procedures.