University of Leicester
Browse

Inflation forecasting with rolling windows: An appraisal

Version 2 2025-02-04, 15:55
Version 1 2024-01-30, 16:22
journal contribution
posted on 2025-02-04, 15:55 authored by Stephen G Hall, George S Tavlas, Yongli Wang, Deborah Gefang

We examine the performance of rolling windows procedures in forecasting inflation. We implement rolling windows augmented Dickey–Fuller (ADF) tests and then conduct a set of Monte Carlo experiments under stylized forms of structural breaks. We find that as long as the nature of inflation is either stationary or non‐stationary, popular varying‐length window techniques provide little advantage in forecasting over a conventional fixed‐length window approach. However, we also find that varying‐length window techniques tend to outperform the fixed‐length window method under conditions involving a change in the inflation process from stationary to non‐stationary, and vice versa. Finally, we investigate methods that can provide early warnings of structural breaks, a situation for which the available rolling windows procedures are not well suited.

History

Author affiliation

School of Business, University of Leicester

Version

  • AM (Accepted Manuscript)

Published in

Journal of Forecasting

Volume

43

Issue

4

Pagination

827-851

Publisher

Wiley

issn

0277-6693

eissn

1099-131X

Copyright date

2024

Available date

2026-01-14

Language

en

Usage metrics

    University of Leicester Publications

    Categories

    No categories selected

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC