Indirect estimation of ARFIMA and VARFIMA models

Indirect estimation methods are proposed for estimating ARFIMA, as well as more complex VARFIMA models. A general framework for conducting indirect es timation of fractional models is developed that covers simulation methods, choice of auxiliary model and estimation algorithm. Special attention is gi ven to comparing the finite sampling properties of the indirect estimator w ith Sowell's (1992a) exact time domain maximum-likelihood estimator, the sp ectral maximum-likelihood estimator of Fox and Taqqu (1986) and the Geweke and Porter-Hudak (1983) spectral regression estimator. The indirect estimat or can be computationally faster than the exact time domain maximum-likelih ood estimator while generating similar small sample properties. The computa tional gains of the indirect estimator over maximum likelihood increase as the complexity of the data generating process increases. (C) 1999 Elsevier Science S.A. All rights reserved. JEL classification: C13; C22.