Aggregate versus Disaggregated Geopolitical Risks and Machine Learning Forecasting of Oil Market Volatility: Evidence from AI-Based Indexes

Onur Polat (), Goodness Aye (), Rangan Gupta () and Samrat Goswami ()
Additional contact information
Onur Polat: Hacettepe University, Informatics Institute 06800 Beytepe, Ankara, Türkiye
Goodness Aye: Department of Economics, University of Pretoria, Private Bag X20, Hatfield 0028, South Africa; Department of Economics, Landmark University, Omu-Aran, Kwara State, Nigeria
Rangan Gupta: Department of Economics, University of Pretoria, Private Bag X20, Hatfield 0028, South Africa
Samrat Goswami: Department of Rural Studies, Tripura University, Agartala 799022, Tripura, India

No 202636, Working Papers from University of Pretoria, Department of Economics

Abstract: This study evaluates the out-of-sample predictive power of disaggregated versus aggregate geopolitical risks (GPR) for crude oil market volatility. Leveraging the AI-GPR dataset of Iacoviello and Tong (2026), we forecast monthly WTI realized volatility over the 1960:01--2026:09 period across 1- to 24- month horizons using autoregressive and machine learning benchmarks. While the predictive content of the aggregate GPR index is largely localized to the 1990s, oil-specific and event-type risk components provide persistent incremental information through the post-2000 period. When allowed to weight the eight underlying event types endogenously, the model significantly outperforms the headline index, reducing root mean squared forecast errors by 2.1 percent to 7.5 percent at horizons up to 12 months. Conversely, geographic disaggregations (country-level and bilateral) yield negligible forecasting value. In addition, superior performance from random forests reflects intrinsic non-linear volatility dynamics rather than geopolitical risk extraction.

Keywords: Geopolitical risks; Oil-market volatility; Realized volatility; Disaggregation; Forecasting; Machine learning (search for similar items in EconPapers)
JEL-codes: C22 C53 F51 Q41 Q47 (search for similar items in EconPapers)
Pages: 19 pages
Date: 2026-10
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:pre:wpaper:202636

Access Statistics for this paper

More papers in Working Papers from University of Pretoria, Department of Economics Contact information at EDIRC.
Bibliographic data for series maintained by Rangan Gupta ().