Multi-Unknown Breaks Estimates in Presence of Fractional White Noise Disturbance

Abstract

This paper analyzes the distribution of breaks estimators and finite sample property of the method based on different mean break and long memory situations based on Bai and Perron (1998) multi-unknown breaks method. As to the distribution of the multiple break estimates, the property of the estimates under the alternative hypothesis is better when d is close to-0.5 that the bias and variance are smaller and the distribution are smoother even better than the BP (1998) case. Further work can be made on the limit distribution of break number test statistics in the case of d (-0.5, 0), d (0, 0.5) and to the case of three and more mean, trend breaks.