The COVID Value-at-Risk (VaR) challenge - Assurance & Advisory Blog | Deloitte Australia has been saved
Limited functionality available
Given the increase in market volatility a sharp increase in Value-At-Risk (VaR) or the related tail risk metric Conditional VaR (CVaR) was largely expected due to the COVID-19 pandemic, but the magnitude of the increase for some markets has been beyond expectation. In this article we look at how best to deal with this volatility and cover model parameters, calibration and validation as well as VaR calculation and VaR explain.
Deloitte’s Jeff Reynolds, Chris Collins and Laurie Brown share actions to make sense of how to manage calculated Value-At-Risk (VaR) numbers versus management and board risk limits.
Given the increase in market volatility due to the COVID-19 pandemic a sharp increase in Value-At-Risk (VaR) or the related tail risk metric Conditional VaR (CVaR) was largely expected, but the magnitude of the increase for some markets has been beyond expectation.
In this article we look at how best to deal with this volatility and cover model parameters, calibration and validation as well as VaR calculation and ‘VaR explain’ analysis.
The first step to determining what the model parameters need to be is to understand the design strengths and weaknesses of your VaR model, and its use in measuring risk in the real world. Lookback period, decay rate and the VaR trading horizon are all important parameters that drive calculated VaR outcomes. In times of market shocks, a shorter lookback period with a faster decay rate means that recent market volatility significantly drives up the calculated VaR.
The VaR trading horizon should align to the trading strategy, so that hypothetical losses align to the normal turnover or performance period. A 1-day VaR may not align to a ‘buy and hold’ strategy, while a 1-year VaR would mean little to a ‘high turnover’ strategy that would not continue to hold the same positions through changing market conditions. Scaling VaR to extend to a longer time horizon can be a crude adjustment which assumes short-term volatility spikes are scaled and persistent over a long time period.
VaR models behave quite differently depending on the calculation type. This can be historical, parametric or Monte-Carlo simulation. How results are generated can be important, particularly where public data sources do not exist; for example with unlisted assets. Familiarity with the calculation methodology, inputs and the interpretation of outputs is important to be able to investigate, understand and explain VaR results to Risk Committees and Boards. Furthermore, as a tool that helps to model reality, the limitations of VaR are both important to understand and table with stakeholders. This is especially true in extreme events where attention is keenly focused on all risk measures.
As with any model, validation and back-testing are important to ensuring robustness during critical times. Furthermore, any VaR numbers need to be assessed in the context of the current market conditions. VaR models can underestimate risk prior to a crisis, coming out of a period of low volatility, and overestimate risk after a shock has occurred where the spike in volatility is captured in the model and projected in model outcomes. Industry participants need to consider their own assumptions about the persistence of extreme volatility spikes into the coming days and months.
Given these limitations, VaR will pose specific challenges for firms which have implemented a risk framework that includes limits on VaR. These industry participants face the challenge of having to change their asset class or specific asset mix in order to bring VaR levels down, or work with their governance authorities to endure temporarily breached limits with a clear understanding of how and why those breaches occurred, the likely duration of the breaches, and the nature of any further down-side risk.
Any limit for VaR that is ready to endure a one in 100 year event would also need to be multiple times more than expected ‘normal’ VaR levels. This would make the limit itself an extrapolation of stressed conditions with a large degree of uncertainty. An appropriately calibrated limit for VaR that results in breaches in very extreme conditions is expected, and it should result in the collective focus and attention of, and discussion and debate between, management, the line two risk team, and the board that supports proper risk management at this critical time.
Stressed VaR can serve as a useful complement, running current positions through historical scenarios and observing the calculated stressed VaR.
Beyond the model setup itself, transparency into the current VaR and its change over time is critical. ‘VaR explain’ type analysis deconstructs a change in VaR into the components of changes in positions, asset volatility, and correlation between assets. This type of analysis is critical to understanding what is driving VaR, which can otherwise be a difficult-to-decipher black box. Explaining the model in this way supports understanding and the decisions needed to be made in a crisis, addressing questions as to why the model is reporting these results and, importantly, what can we expect in terms of business impact and our capacity to endure further volatility.
Deloitte and organisations like ours are able to bring an independent and objective interpretation of VaR models as well as an important consideration of alternatives. We can also help review VaR model inputs, parameters and outputs. With the governance imperative for an independent review, Deloitte can also assist to review the place of VaR within your Risk Management Framework in the context of the business as a whole as well as bring our local, regional and global knowledge of other options for managing your risk.
Chris has 14 years of industry experience and has been with Deloitte for 2.5 years. He specialises in investment risk, counterparty risk, performance attribution, model risk, enterprise risk, investment governance and processes. Chris currently assists clients with their investment risk program, investment governance including manager selection and monitoring frameworks, and ensuring their performance and model risk functions align with best practice. Chris has industry experience working in investment risk / market risk functions in asset management, investment bank and hedge scheme industries. He was formerly the Investment Risk Manager at Nikko AM Australia with responsibility for building out and managing the market, counterparty and liquidity risk framework and shared oversight of Enterprise Risk Management. Prior to Nikko Asset Management, Chris worked as a Market Risk Analyst in the Equity Derivatives business at J.P. Morgan Australia.