Skip to main navigation Skip to search Skip to main content

Robustness and spurious long memory: evidence from the generalized autoregressive score models

Research output: Contribution to journal › Article › peer-review

Abstract

This paper employs the generalized autoregressive score (GAS) models to study the long memory and regime switching in the second comment. Via systematically constructed simulation studies, we firstly demonstrate the desirable robustness of both the long memory GAS (LMGAS) and Markov switching GAS (MS-GAS) models against outliers. Despite this, the LMGAS model still produces significant and spurious long memory, when the pure regime-switching processes are fitted. An illustrative cause of such spuriousness is further provided, which suggests an effective solution needs to accommodate both long memory and regime switching. Motivated from this, we propose an MS-LMGAS model and provide comprehensive simulation evidence on its effectiveness to resolve the spurious long memory and robustness against outliers. An empirical study of the West Texas Intermediate crude oil spot returns is then conducted, which contain historical-writing movements deemed as large outliers. Our empirical findings support the superiority of the new model over both the existing GAS and GARCH models. It is further verified that significant long memory only exists in the high-volatility state. Important financial implications are described to improve the risk management operations in practice.
Original languageEnglish
Pages (from-to)653-685
Number of pages33
JournalAnnals of Operations Research
Volume352
Issue number3
Early online date5 Jul 2023
DOIs
Publication statusPublished - Sept 2025

Keywords

  • Generalized autoregressive score
  • Long memory
  • Robustness
  • Regime switching
  • Crude oil

Fingerprint

Dive into the research topics of 'Robustness and spurious long memory: evidence from the generalized autoregressive score models'. Together they form a unique fingerprint.

Cite this