Estimating long memory: Scaling function vs Andrews and Guggenberger GPH

This paper compares the estimation methods of the long memory parameter. It focuses on the two specific cases of heavy tails or short memory. Monte Carlo simulations results show that the scaling function outperforms the modified GPH method [Andrews, D.W.K., Guggenberger, P., 2003, A bias-reduced log-periodogram regression estimator for the long-memory parameter. Econometrica 71 (2), 675-712].