Forecasting UK consumer price inflation with RaGNAR: Random generalised network autoregressive processes
File(s)
Published version
Author(s)
Nason, Guy
Palasciano, Henry
Type
Journal Article
Abstract
This article forecasts CPI inflation in the United Kingdom using Random Generalised Network Autoregressive (RaGNAR) Processes. More specifically, we fit Generalised
Network Autoregressive (GNAR) Processes to a large set of random networks generated according to the Erdős–Rényi–Gilbert model and select the best-performing networks each
month to compute out-of-sample forecasts. RaGNAR significantly outperforms traditional benchmark models across all horizons. Remarkably, RaGNAR also delivers materially moreaccurate predictions than the Bank of England’s four to six month inflation rate forecasts
published in their quarterly Monetary Policy Reports. Our results are remarkable not only for their accuracy, but also because of their speed, efficiency and simplicity compared to the Bank’s current forecasting processes. RaGNAR’s performance improvements manifest both in terms of their root mean squared error and mean absolute percentage error, which measure different, but crucial, aspects of the methods’ performance. GNAR processes demonstrably predict future changes to CPI inflation more accurately and quickly than the benchmark models, especially at medium- to long-term forecast horizons, which is of great importance to policymakers charged with setting interest rates. We find that the most robust forecasts are those which combine the predictions from multiple GNAR processes via the use of various model averaging techniques. By analysing the structure of the best-performing graphs, we are also able to identify the key components that influence
inflation rates during different periods.
Network Autoregressive (GNAR) Processes to a large set of random networks generated according to the Erdős–Rényi–Gilbert model and select the best-performing networks each
month to compute out-of-sample forecasts. RaGNAR significantly outperforms traditional benchmark models across all horizons. Remarkably, RaGNAR also delivers materially moreaccurate predictions than the Bank of England’s four to six month inflation rate forecasts
published in their quarterly Monetary Policy Reports. Our results are remarkable not only for their accuracy, but also because of their speed, efficiency and simplicity compared to the Bank’s current forecasting processes. RaGNAR’s performance improvements manifest both in terms of their root mean squared error and mean absolute percentage error, which measure different, but crucial, aspects of the methods’ performance. GNAR processes demonstrably predict future changes to CPI inflation more accurately and quickly than the benchmark models, especially at medium- to long-term forecast horizons, which is of great importance to policymakers charged with setting interest rates. We find that the most robust forecasts are those which combine the predictions from multiple GNAR processes via the use of various model averaging techniques. By analysing the structure of the best-performing graphs, we are also able to identify the key components that influence
inflation rates during different periods.
Date Issued
2026-01-01
Date Acceptance
2025-04-22
Citation
International Journal of Forecasting, 2026, 42 (1), pp.181-202
ISSN
0169-2070
Publisher
Elsevier
Start Page
181
End Page
202
Journal / Book Title
International Journal of Forecasting
Volume
42
Issue
1
Copyright Statement
© 2025 The Author(s). Published by Elsevier B.V. on behalf of International Institute of Forecasters. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
License URL
Identifier
10.1016/j.ijforecast.2025.04.005
Publication Status
Published
Date Publish Online
2025-05-17