Files and links (1)
CC BY-NC-ND V4.0, Open Access
Dissertation
Analysis of the Predictive Performance of Value at Risk and Expected Shortfall Estimations in Times of Financial Crisis
University of the Sunshine Coast, Queensland
Doctor of Philosophy, University of the Sunshine Coast, Queensland
2026
DOI:
https://doi.org/10.25907/01004
Appears in COVID-19 Research
Abstract
Measuring and mitigating market risk is essential to the stability of financial institutions, particularly large banks and investment firms. While Value at Risk (VaR) remains the most widely used metric for assessing and managing these risks, it has shown notable weaknesses during past financial crises. In this study, the predictive performance of a range of VaR and Expected Shortfall estimation models including parametric (Normal, t-distribution, GARCH-based), non-parametric (Historical Simulation), and semi-parametric (Extreme Value Theory, Volatility-Weighted Simulation, Monte Carlo) across the G7 stock market indices of Canada, France, Germany, Italy, Japan, the UK, and the US is evaluated. The analysis spans over the period 2005–2022, with a focus on three selected sub-periods: the Global Financial Crisis of 2007–2009, the COVID-19 Recession of 2020 and a non-crisis period 2010-2019. Multiple backtesting procedures are applied to evaluate the predictive performance at the 97.5%, 99% and 99,9% confidence levels. These include the Kupiec and Christoffersen tests for VaR, multinomial backtest and unconditional and exceedance-based tests for Expected Shortfall (ES). The results reveal the limitations of VaR in crises conditions, particularly in capturing fat-tailed loss distributions. Dynamic volatility modeling, such as GARCH with t-distributions, significantly enhances VaR forecasts over more straightforward, static assumptions. Moreover, the findings confirm that ES provides more robust tail-risk coverage than VaR at the 97.5% and 99% confidence levels, aligning with emerging regulatory preferences for ES within the Basel Committee’s updated frameworks. Beyond quantifying backtesting performance, this research highlights how the COVID-19 Recession, with its abrupt and unprecedented economic shutdowns that led to rapid shifts in market sentiments, present unique challenges for risk estimation relative to the Global Financial Crisis, emphasizing that crisis-specific factors can undermine even sophisticated models. The findings offer practical guidance for regulators, financial institutions, and market participants by supporting more adaptive risk management, reinforcing the regulatory shift toward ES, and improving the resilience of capital markets during crises.
Details
- Title
- Analysis of the Predictive Performance of Value at Risk and Expected Shortfall Estimations in Times of Financial Crisis
- Authors
- Alexander Marx - University of the Sunshine Coast, Queensland, School of Business and Creative Industries
- Contributors
- Sajid Anwar (Principal Supervisor) - University of the Sunshine Coast, Queensland, Indigenous and Transcultural Research CentreHammad Siddiqi (Co-Supervisor) - University of the Sunshine Coast, Queensland, School of Business and Creative IndustriesRaffaella Belloni (Co-Supervisor) - University of the Sunshine Coast, Queensland, School of Business and Creative Industries
- Awarding institution
- University of the Sunshine Coast, Queensland
- Degree awarded
- Doctor of Philosophy
- Publisher
- University of the Sunshine Coast, Queensland
- DOI
- 10.25907/01004
- Organisation Unit
- School of Business and Creative Industries
- Language
- English
- Record Identifier
- 991203251002621
- Output Type
- Dissertation
Metrics
37 File views/ downloads
59 Record Views