- About Us
- Products & Solutions
- Industry Solutions
- Thought Leadership
- News Room
- Contact Us
|
By Andrew Robertson Stress testing has become an integral part of portfolio risk management to help eliminate potentially large losses from extreme events. This article explores the aspects and considerations needed to implement hypothetical stress testing as an effective risk management tool within an investment management organization. Portfolio Risk Management Perhaps the most important element of any risk management framework is buy-in from the top. Senior management within an organization needs to promote a 'risk culture' where the identification and measurement of risk becomes a central business driver. For senior management the risks need to be identified, measured and managed. Statistical models, such as Value at Risk, have long been used for risk management purposes. In general, VaR models assume normal market conditions. Under extreme market conditions, the underlying assumptions of these models break down. In these situations, stress testing has been used as a complementary technique to identify potential loss. Exhibit 1: VaR capturing risk as a plane, stress testing capturing risk as a point. Source: Yuko Kawai, Bank of Japan, 2006.
Hypothetical Model Building Blocks Hypothetical stress test models are comprised of four key building blocks1 - an extreme hypothetical scenario to define a set of exogenous risk drivers, a data generating process to map the exogenous factors to a set of exposure specific risk factors, an exposure, and finally, a risk measure. These building blocks are illustrated below. Exhibit 2: Stress test building blocks. Source: Drehmann 2008
If a stress test is to be used as a measurement instrument, the priority is model accuracy and forecast performance.1 In this case, a complex model should be created, containing many intricate parts and risk factors. Mechanisms such as feedback loops should be incorporated, liabilities considered and sophisticated quantitative models for co-movement employed. The type of portfolio also plays a fundamental role in the design of the stress test model. This requires an understanding of the way the fund manager makes money and to what risk types they are exposed. For example, the model design for a credit risk incorporating probabilities of default will be fundamentally different from a market risk stress test model. Scenario Development The purpose of a stress test is to explore the behavior of a portfolio under extreme conditions and make decisions based on their results. Developing a hypothetical stress scenario is a complex process because it requires not only envisioning an event that hasn't happened before, but it also needs to be able to persuade others the event is plausible. If a scenario is too extreme, it may be hard to justify to decision makers and could be dismissed as too unlikely to occur. If a scenario isn't extreme enough, it will provide limited information and may be disregarded as too insignificant to act on. Therefore it is important to find the right balance between severity and plausibility when developing a scenario. This is best achieved through the collaboration of different experts within an organization, such as risk managers, economists and fund managers. Using their expertise and experience, a state of the world can be formulated based on possible future events and then cross checked by senior management. Using this approach a scenario may not only be more realistic, but will also incorporate the culture and risk appetite of the organization. Exhibit 3: Iterative cycle of hypothetical scenario generation Source: J.P. Morgan
The most obvious starting point to develop a stress scenario is envisioning a future extreme world event. As well as having a degree of imagination, this calls on using economic fundamentals to create a forecast model. The extreme event is played through the model, with considerations made for arbitrage relationships between asset types and relationships between currencies. The laws of supply and demand are explored and challenged to produce a set of stress systematic risk factors such as global indices and major currency pairs. Another popular approach to building an extreme event is to consider the interrelationship of financial markets and the effect a liquidity shock may have on it. The technique starts with a 'thought experiment' on liquidity key drivers. Features to be considered are the interconnection of the financial system through advances in information technology and how this can amplify events through feedback in the financial system. The effects of market participants all managing risk in the same manner is another element that may need to be explored. An example of this technique might be an event causing a sharp increase in market risk and dealers exiting positions to avoid breaching trading limits. This contributes to further volatility and triggers action to be taken by other market participants resulting in herd behavior and further feedback in the system. The behavior spreads to other markets through the deterioration in liquidity and the inability to implement hedging strategies, thus causing further increases in volatility. As a result, a major market event unfolds. A third technique is to identify a maximum portfolio loss using mathematical optimization to create a hypothetical event scenario. The worst case scenario can be created by modifying the systematic risk factors within a stress test model under plausibility constraints, while ensuring the events are sensible due to the mathematical nature of the technique. Interpretation of results Hypothetical stress testing does not act as a crystal ball for future events. Stress testing is a risk management tool that requires both a quantitative and qualitative approach to establish the risk profile of a portfolio. The qualitative element of stress testing has led to skepticism in the past and the rationale results are of little use without associating a probability to an event. Yet, this is the nature and challenge inherent in risk management. Trying to assign a probability to a hypothetical stress test in isolation seems to be a dangerous exercise, which could lead to a false sense of security in numbers. One answer is not to look at single stress tests, but to apply a number of scenarios to a portfolio and use judgment to rank each on a probability of occurrence vs. impact of risk chart. Exhibit 4: Probability impact chart Source: J.P. Morgan, www.mindtools.com The process of interpreting the results of stress tests also aid psychological preparation to events. Bringing different members of an organization together to discuss the topic of risk helps raise decision and thought awareness within an organization. In turn, this helps formulate contingency plans and impact mitigation, lowering the chances of a shock and awe response to market events. Conclusion The business of risk management is a balancing act of risk against return. It sometimes feels like the playing field is not even, as small frequent opportunities to make profit are offset by large infrequent losses. The aim of hypothetical stress testing is to explore weaknesses in a portfolio under extreme circumstances to help mitigate these losses. But it is clear that there is no one-size-fits-all formula for stress testing in an organization. This is determined by the culture, the objective at hand and the overall risk appetite. A hypothetical stress testing program needs to have buy-in from all aspects of the organization. In some sense, stress testing is born of a state of mind, where ideas can be explored and where people and methodologies are open to be challenged. It is most important that senior management is involved in the design process, as ultimately they will be the ones making decisions on the back of any stress scenario. As with any practice, experience gains understanding and grows confidence in a method.
References: 1Drehmann 2008, 'Stress Tests: Objectives, Challenges and Modelling Choices' Riksbank Economic Review. |
