Estimating the long memory parameter in nonstationary models: further Monte Carlo evidence

In this work we perform a Monte Carlo experiment to show and compare the performance of a variety of estimators of the long memory parameter d in case of nonstationary processes. Both parametric and semiparametric estimators are considered. Moreover they have been employed both on the original time series and on the first difference of the series. Results show that the Whittle estimator is the best performing and the strategy of preliminarily differentiate the series is worthy, but not for all the estimators.