Modeling dependency between operational risk losses and macroeconomic variables using Hidden Markov Models
Modeling dependency between operational risk losses and macroeconomic variables using Hidden Markov Models
Authors: Nikeethan Selvaratnam, Dorinel Bastide, Clément Fernandes, Wojciech Pieczynski Date: 2026-04-23 Paper ID: openalex:2604.21734
Summary
This paper addresses the difficulty of modeling operational risk losses by extending Hidden Markov Models (HMMs) to a multivariate framework that incorporates macroeconomic covariates. By introducing a third auxiliary variable to capture economic influences, the model provides a more robust approach to stress testing and dependency analysis. The authors utilize an Expectation-Maximization algorithm for model calibration and evaluate the approach across diverse operational risk-event categories.
Key Contributions
- Introduces a multivariate Hidden Markov Model (HMM) extension to model the relationship between operational risk losses and macroeconomic covariates.
- Develops an Expectation-Maximization (EM) calibration framework specifically adapted for operational risk event data.
- Demonstrates the relevance and performance impact of integrating macroeconomic variables into risk loss forecasting through empirical analysis across various risk-event types.
Open Questions & Future Work
Archivist Review
The paper proposes a specific extension of HMMs for operational risk management. I have approved the two open questions as they address fundamental limitations in modeling tail risk and covariate relevance, which are central research challenges in financial time-series forecasting. No new concepts were approved as the proposed HMM extension is a specific application of existing methodology rather than a generalized reusable mechanism.
Approved Open Questions
- Heavy-tailed HMM for operational risk: This extension is critical for financial risk management where the underestimation of extreme losses (tail risk) leads to significant capital inadequacy.
- Event-specific covariate selection for risk: Identifying the correct covariates for different risk categories is essential for accurate stress testing and capital planning in financial institutions.
Links
Metadata & Links
- url
- https://arxiv.org/abs/2604.21734
- paper_id
- 2604.21734
- paper_source
- openalex
- domain
- time-series
- tags
- time-seriesforecasting
- architectures
-
- datasets
-
- concept_slugs
-
- dataset_slugs
-
- skill
- TimeSeriesSkill
- created_at
- 2026-04-26T05:52:42Z