Semiparametric Estimation of Fractional Integration: An Evaluation of Local Whittle Methods

TL;DR

This paper evaluates various local Whittle estimators for fractional integration, providing guidance on their performance under different data conditions through simulations and empirical analysis.

Contribution

It compares multiple estimators, extends previous studies with new experiments, and offers practical recommendations for their application in real-world data analysis.

Findings

Performance varies with short-run dynamics and structural breaks

Guidance on estimator selection and bandwidth tuning

Diagnostic value of estimator disagreements

Abstract

Fractionally integrated time series, exhibiting long memory with slowly decaying autocorrelations, are frequently encountered in economics, finance, and related fields. Since the seminal work of Robinson (1995), a variety of semiparametric local Whittle estimators have been proposed for estimating the memory parameter $d$. However, applied researchers must determine which estimator to use, and under what conditions. This paper compares several local Whittle estimators, first replicating key findings from the literature and then extending these with new Monte Carlo experiments and in-depth empirical studies. We compare how each estimator performs in the presence of short-run dynamics, unknown means, time trends, and structural breaks, and discuss how to interpret potentially conflicting results with real datasets. Based on the findings, we offer guidance to practitioners on estimator choice, bandwidth selection, and the potential diagnostic value of disagreements between estimators.