Paper: arXiv 2409.18970

Authors: Krishan Mohan Nagpal

Abstract

Value at Risk (VaR) and stress testing are two of the most widely used approaches in portfolio risk management to estimate potential market value losses under adverse market moves. VaR quantifies potential loss in value over a specified horizon (such as one day or ten days) at a desired confidence level (such as 95’th percentile). In scenario design and stress testing, the goal is to construct extreme market scenarios such as those involving severe recession or a specific event of concern (such as a rapid increase in rates or a geopolitical event), and quantify potential impact of such scenarios on the portfolio. The goal of this paper is to propose an approach for incorporating prevailing market conditions in stress scenario design and estimation of VaR so that they provide more accurate and realistic insights about portfolio risk over the near term. The proposed approach is based on historical data where historical observations of market changes are given more weight if a certain period in history is “more similar” to the prevailing market conditions. Clusters of market conditions are identified using a Machine Learning approach called Variational Inference (VI) where for each cluster future changes in portfolio value are similar. VI based algorithm uses optimization techniques to obtain analytical approximations of the posterior probability density of cluster assignments (market regimes) and probabilities of different outcomes for changes in portfolio value. Covid related volatile period around the year 2020 is used to illustrate the performance of the proposed approach and in particular show how VaR and stress scenarios adapt quickly to changing market conditions. Another advantage of the proposed approach is that classification of market conditions into clusters can provide useful insights about portfolio performance under different market conditions.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 6.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced statistical techniques like Variational Inference (VI) for regime clustering, demonstrating significant mathematical sophistication. While it uses a real-world dataset (COVID-2020) and focuses on practical portfolio risk management, the summary and excerpt lack specific backtest metrics, code, or implementation details, placing it between pure theory and full empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Incorporate prevailing market conditions<br>in stress testing & VaR"] --> B["Methodology: Variational Inference"]
  B --> C{"Data Input<br>Historical Market Data & Portfolio Positions"}
  C --> D["Compute: Clustering via VI<br>Identify Market Regimes"]
  D --> E{"Apply: Weight Historical Data<br>by similarity to current regime"}
  E --> F["Outputs: Adaptive VaR &<br>Stress Test Scenarios"]
  F --> G["Key Outcome<br>Rapid adaptation to changing<br>market conditions e.g., Covid 2020"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a VaR and stress‑testing approach that weights historical observations more heavily when past market conditions resemble the current regime, using variational inference to cluster market regimes.

Source sentence
The proposed approach is based on historical data where historical observations of market changes are given more weight if a certain period in history is "more similar" to the prevailing market conditions.

Assumes 3

  • Each day the market is assumed to be in one of K possible latent clusters. model
    Source sentence
    Each day it is assumed that the market is in one of K possible clusters (which are hidden/latent variables) which depend on market data xt ∈Rn which is known at time t.
  • The set of input variables xt used for clustering has already been identified. data
    Source sentence
    we will assume that we have already identified the choice of n input variables in xt that are most useful in differentiating clusters.
  • VaR is computed using the prior 250 daily risk‑factor observations. data
    Source sentence
    VaR estimate in all cases is based on the same historical data of prior 250 observations of daily risk factor changes.

Data

  • Covid‑related volatile period around 2020 market-prices unknown
    2020
    Source sentence
    Covid related volatile period around the year 2020 is used to illustrate the performance of the proposed approach

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • VI‑based VaR adapts quickly to changing market conditions and is usually more accurate than Historical Simulation or Gaussian‑based VaR. performance
    Source sentence
    VaR based on the proposed approach adapts quickly to chaanging market conditions and VaR estimates are usually more accurate than those computed using Historical Simulation or based on Gaussuain distribution of returns.
  • Both Historical Simulation and Gaussian approaches underestimate VaR in February and March 2020 relative to the VI method. performance
    Source sentence
    Both Historical Simulation and Gaussian approaches underestimate VaR in February and March relative to VI.
  • Gaussian‑parametrized VaR is lower than Historical Simulation VaR for the 97.5% confidence level because the Gaussian distribution does not capture heavy tails. statistical
    Source sentence
    Gaussian distribution based VaR is lower than Historical Simulation VaR as Gaussian distribution does not capture the heavy tails of the distribution.

Checked by others

Supported by

  • Comparative Evaluation of VaR Models: Historical Simulation, GARCH-Based Monte Carlo, and Filtered Historical Simulation (2025)
    On the question: Do Gaussian‑based VaR methods underestimate tail risk compared to non‑Gaussian approaches?
    Compare the findings

    This paper: Gaussian‑parametrized VaR is lower than Historical Simulation VaR for the 97.5% confidence level because the Gaussian distribution does not capture heavy tails.

    That paper: GARCH‑N underestimates tail risk because of its Gaussian assumption.

    Both find Gaussian‑based VaR is lower: A shows Gaussian VaR < Historical Simulation VaR at 97.5% confidence, and B shows GARCH‑N underestimates tail risk due to its Gaussian assumption.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.