Nonlinear and Heavy-Tailed Predictability in Transition-Energy Financial Markets
Nonlinear and Heavy-Tailed Predictability in Transition-Energy Financial Markets
Authors: Kpante Emmanuel Gnandi, Fredy Pokou, Jules Sadefo Kamdem Date: 2026-05-26 Paper ID: openalex:2605.26890
Summary
This paper addresses the limitations of conventional Gaussian-linear models in forecasting transition-energy financial markets, which are characterized by abrupt repricing and heavy-tailed volatility. The authors introduce a hybrid model that combines Student-t VAR with nonlinear recurrent neural networks to specifically filter out linear dynamics before learning residual nonlinear patterns. Extensive empirical testing on six sector-specific exchange-traded funds confirms that this approach improves forecast accuracy, especially during significant economic shocks. The results highlight the necessity of accounting for both heavy-tailed distributions and nonlinear dependencies when modeling modern energy transition assets.
Key Contributions
- Developed a hybrid forecasting framework that integrates Student-t Vector Autoregressions (VAR) with nonlinear recurrent residual learning to capture heavy-tailed dependencies.
- Demonstrated that transition-related financial returns contain residual nonlinear predictability that standard Gaussian-linear models fail to capture.
- Validated that the hybrid model significantly outperforms traditional VAR, standalone machine learning, and existing hybrid baselines, particularly during periods of high macro-financial stress.
Links
Metadata & Links
- url
- https://arxiv.org/abs/2605.26890
- paper_id
- 2605.26890
- paper_source
- openalex
- domain
- time-series
- tags
- time-seriesforecastingfinancial-econometrics
- architectures
-
- datasets
-
- concept_slugs
-
- dataset_slugs
-
- skill
- TimeSeriesSkill
- created_at
- 2026-05-29T06:28:58Z