We examine the least-squares estimator of change point for nonstationary I(d) data with 0.5 < d < 1.5. We show that this estimator fails to locate the change point consistently when a change occurs and that it would suggest a spurious change even when there is none.