Introduction To Stress Testing

• Stress testing is a risk management activity that has become increasingly important since the 2007-2008 financial crisis. It involves evaluating the implications of extreme scenarios that are unlikely and yet plausible. Stress testing answers a key question for a financial institution – whether it has enough capital and liquid assets to survive various scenarios?
• Some stress tests are carried out because they are required by regulators, however, many financial institutions have adopted stress testing as an invaluable part of their internal risk management process. When stress tests are used in conjunction with VaR/ES analyses, they provide a more detailed picture of the risks facing a financial institution.
• An important advantage stress testing is that it can consider the impact of scenarios that are quite different from (and more severe than) the scenarios considered by VaR or ES. Bank regulators are increasingly moving towards basing market risk capital on stressed VaR and stressed ES.
• The stressed VaR and Stressed ES measures are based on how market variables behaved during a 12-month period that would be significantly stressful for a firm’s current portfolio.

Choosing Scenarios – Historical Scenarios

• The first step in choosing a stress-test scenario is to select a time horizon. While one-day or one-week scenarios are occasionally considered, scenarios lasting three months to two years are more common. The time horizon should be long enough for the full impact of the scenarios to be evaluated, and very long scenarios can be necessary in some situations.
• Scenarios are sometimes based on historical data, and it is assumed that all relevant variables will behave as they did in the past. It is assumed that Actual changes from the stressed period will recur for some variables, while proportional changes others.
• There are many historical scenarios that might be of concern to risk managers. For example, 2007-2008 US housing-related recession, which led to serious problems for many financial institutions, is an obvious one to use.
• Sometimes, a moderately adverse scenario from the past is made more extreme by multiplying the movements in all risk factors by a certain amount. For example, we could take what happened during a certain loss-making six-month period in the past and double (or triple) the movements in all relevant variables.
• When magnified in this way, the scenario might become a much more serious problem for a financial institution. However, this approach assumes there is a simple linear relationship between the movements in risk factors. This is not necessarily  the case, however, because correlations between risk factors tend to increase as economic conditions become more stressed.
• Sometimes, historical scenarios are based on what happened to all market risk factors over one day or one week. For example, the impact of a day like October 19, 1987 (when the S&P 500 fell by 22.3 standard deviations) could be assessed. If this is considered too extreme, a scenario could be created from the days around January 8, 1988 (when the S&P 500 fell by 6.8 standard deviations).
• These short-horizon stress tests can be supplements to stressed VaR and stressed ES calculations. While stressed VaR and stressed ES consider extreme movements during just one stressed period, short-horizon stress tests can pick big movements from many different stressed periods in the past.

Choosing Scenarios – Stress Key Variables

• One approach to scenario building is to assume that a large change takes place in one or more key variables. Changes that might be considered include
  • A 200-basis point increase in all interest rates
  • A 100% increase in all volatilities
  • A 25% decline in equity prices
  • A 4% increase in the unemployment rate
  • GDP declining by 2%

• Other changes could involve factors such as exchange rates, commodity prices, and default rates.
• For market risks, relatively small changes are measured in forms in the form of Greek letters such as delta, gamma, and vega. In the case of stress testing, however, the changes are so large that these measures cannot be used. Furthermore, whereas Greek letters quantify the risks arising from changes to a single market variable over a short period of time, stress testing often involves the interaction of several market variables over much longer periods.

Choosing Scenarios – Ad Hoc Stress Tests

• It is important for firms to develop scenarios reflecting current economic conditions, the particular exposures of the financial institution, and an up-to-date assessment of possible future adverse events. History never repeats itself exactly, and managerial judgement is necessary to either generate new scenarios or modify existing scenarios based on past data.
• Ad hoc stress tests could consider the impact of a change in government policy on a key issue affecting a financial institution or a Basel regulation that would require more capital to be raised in a short period of time. Adverse scenarios suggested by professional economists should be considered carefully.
• The boards, senior management, and economics groups within financial institutions are in a good position to use their understanding of markets, world politics, and current global uncertainties to develop adverse scenarios. One way of developing the scenarios is for a committee of senior management to engage in brain-storming sessions. Research suggests committees consisting of three to five members with different backgrounds work best. A key role of the committee should be to recommend actions that can be taken to mitigate unacceptable risks.l

Choosing Scenarios – Using The Results

• It is important that senior management recognizes the importance of stress testing and incorporates it into its decision making. Hence, involving senior management in building scenarios makes it more likely for the stress testing to be taken seriously and used for decision-making.
• It should be emphasized that the purpose of stress testing is not just to produce output answering, “What if?” questions. Senior management and the board should carefully evaluate stress-test findings and decide whether some form of risk mitigation is necessary. There is a natural human tendency for a decision-maker to base decisions on what he or she considers to be the most likely outcome and to regard alternatives to be so unlikely that they are not worth considering. Stress testing should be used by the board and senior management to ensure that this does not happen.

Regulatory Stress Testing

• Regulators in US, UK and EU require banks and insurance companies to carry out specified stress tests. In the US, for example, the Federal Reserve carries out a stress test of all banks with consolidated assets of over USD 50 billion. This is referred to as the Comprehensive Capital Analysis and Review (CCAR). Banks are required to consider four scenarios –
  • Baseline
  • Adverse
  • Severely adverse
  • An internal scenario

• The adverse and severely adverse scenarios describe hypothetical sets of events designed to assess the strength of banking organizations and their resilience. The baseline scenario is in line with average projections from surveys of economic forecasters. It does not represent the forecast of the Federal Reserve.
• Each scenario includes 28 variables (e.g., GDP and interest rates) which captures domestic and international economic activity. Along with it, the Board publishes a narrative that describes the general economic conditions and changes in the scenarios from previous year.
• Banks are required to submit a capital plan, documentation to justify the models they use, and the results of their stress tests. If they fail the stress test because their capital is insufficient, they are likely to be required to raise more capital and restrict the dividends they can pay until they have done so.
• Banks with consolidated assets between USD 10 billion and USD 50 billion are subject to the Dodd-Frank Act Stress Test (DFAST). The scenarios in DFAST are like those in CCAR. However, banks are not required to submit a capital plan.
• By choosing the scenarios, bank regulators can evaluate the ability of different banks to survive adverse conditions in a consistent way. But they make it clear that they also want to see scenarios developed by the banks themselves that reflect their particular vulnerabilities.

Model Building

• It should be possible to observe how most of the relevant risk factors behaved during the stressed period when building a scenario. The impact of the scenario on a firm’s performance can then be assessed in a fairly direct way.
• Scenarios constructed by stressing key variables (and ad hoc scenarios) typically specify movements in only a few key risk factors or economic variables. To complete the scenarios, it is necessary to construct a model to determine how a range of other variables can be expected to behave. The variables specified in the scenario definition are sometimes referred to as core variables, whereas the other variables are referred to as peripheral variables.
• One approach is to carry out an analysis (e.g., linear regression) relating the peripheral variables to the core variables. And data of stressed periods from the past are likely to be most useful in determining the relevant relationships. For example, in credit risk losses, data provided by rating agencies can be useful. The default rates can be related to economic variables, such as, the GDP growth rate and unemployment rate, to determine the overall default rates that can be expected in different scenarios. This can then be scaled up or down to estimate default rates for the different categories of loans on a financial institution’s books.

Model Building – Knock-On Effect

• A knock-on effect reflects the impact of how firms (particularly other financial institutions) respond to an adverse scenario. In responding to the adverse scenario, the companies often take actions exacerbating adverse conditions.
• If a scenario that might have constructed around the US housing price bubble in 2005-2006 is considered –
  • Banks were concerned about the creditworthiness of other banks and were reluctant to engage in interbank lending. This increased funding costs for banks.
  • There was a flight to quality where all risky assets were perceived to be less attractive. As a result, equity prices and corporate bond prices declined sharply. The decline in corporate bond prices meant credit spreads increased.

Reverse Stress Testing

• Reverse stress testing asks the question, “What combination of circumstances could lead to the failure of the financial institution?”
• One reverse stress-testing approach involves using historical scenarios. Under this approach, a financial institution looks at a series of adverse scenarios from the past and determine how much worse each scenario would have to be for the financial institution to fail. For example, it might conclude that a recession three times worse than the one in 2007-2008 would lead to failure. Ideally, a firm uses a more sophisticated model incorporating the tendency for correlations to increase as market conditions become more stressed.
• Analyzing all risk factors to find a plausible combination leading to firm failure is not usually feasible. Another approach is to define a handful of key factors (e.g., GDP growth rate) and construct a model relating all other relevant variables to them. Then scenario leading to failure is  searched iteratively over all factor combinations. Reverse stress testing can be an input to the work of a stress testing committee. The committee is likely to discard some of the scenarios generated by reverse stress testing as totally implausible while flagging others for further investigation.

Stress Testing Versus VaR/ES

• VaR and ES are based on the estimated loss distribution. VaR allows a financial institution to reach a conclusion in the form of the following – “We are X percent certain that our losses will not exceed the VaR level during time T.”
• In the case of ES, the conclusion is in the form of the following –“If our losses do exceed the VaR level during time T, the expected (i.e., average) loss will be the ES amount.”
• Essentially, while VaR reflects the level of loss that will not be exceeded on x number of days, it does not provide any insight into how bad the things can get in the small portion of time that falls outside the region, for which VaR is calculated. ES does solve this problem to a certain degree but it alone can not be relied on to provide enough information to protect the company in the extreme scenarios. This is where the need for stress testing becomes extremely high. 
• One major disadvantage of VaR and ES is that they are usually backward-looking. They assume the future will be like the past. Stress testing, however, is designed to be forward- looking and answer more general “What If?” questions. Unlike VaR and ES, stress testing does not provide a probability distribution for losses.
• The backward-looking VaR/ES analysis looks at a wide range of scenarios (either good or bad) that reflect history. On the other hand, stress testing looks at a relatively small number of scenarios (all bad for the organization).
• There are other differences between the stress-testing approach and VaR/ES analyses. In the case of market risk, the VaR/ES approach often has a short time horizon (perhaps only one day), whereas stress testing usually looks at a much longer period.
• The objective in stress testing is to obtain an enterprise-wide view of the risks facing a financial institution. Often, the scenarios are defined in terms of macroeconomic variables such as GDP growth rates and unemployment rates. The impact of these variables on all parts of the organization must be considered along with the interactions between different areas.

Stressed VaR And Stressed ES

• VaR and ES are calculated using data over from the preceding one to five years. Daily movements in risk factors during this period are used to calculate potential future movements.
• In stressed VaR and stressed ES, however, this data is gathered from particularly stressful periods. Stressed VaR are conditional on a repeat of a given stressed period and can be considered a form of historical stress testing.
• In contrast, stress testing usually has a longer time horizon. It seeks to answer the questions such as, “If the next year is a repeat of 2008, how would our organization survive?” or “If next year were like 2008 but twice as bad, how would we survive?” It does not consider what would happen during the worst T days of 2008. Rather, it considers the impact of the whole of 2008 being repeated.
• Traditional VaR measures are designed to quantify the full range of possible outcomes and can therefore be back-tested. However, it is not possible to back-test stressed VaR or the output from stress testing in this way because these measures focus on extreme outcomes, which we do not expect to observe with any particular frequency.

Importance Of Stressed Inputs

• It is crucial to understand the importance of stressed inputs and their role in the calculation of stressed VaR and stressed ES.
• Stressed inputs are essentially data from a certain period of time in the past, when the company was facing stressed conditions. A company is dependent on its external environment for various factors – the payments from major customers, the raw material price, the availability of cheap credit for overdraft, etc.
• A company might face problems in these areas, like if the major customers default on their payment, or if the price of raw material sky-rockets, or if banks withdraw lines of credit. The changes in external environment both temporary and permanent can have the capacity to threaten the very viability of the business. 
• Thus, it is important that the company ensures that they are in a position wherein if one or more adverse changes in external environment happens, the company can still survive.
• The way this can be done, is by taking the data of the performance of the company during a time when one or more of these adverse event happened and asking the question “whether the company can survive, if such a situation happened tomorrow?”.
• Companies can create scenarios wherein a few of these ‘stressors’ are taking place simultaneously and the expected performance of the company can then be judged.
• A good example for this can be that if an Indian company XYZ is into export of furniture to big hotels across the globe, and whose raw material is mainly supplied by Japanese companies, the major risk factors that the company is exposed to are –
  • The exchange rate of INR to USD rises sharply- a scenario where exported goods from India will suddenly cost way more in the international market.
  • The default of 3 out of the top 10 customers of the company – a scenario that the company’s management thinks is viable.
  • The sharp fall in the exchange rate of INR to Japanese Yen – since the company heavily imports from Japan, the raw material price would go up sharply.
• In the above example, the management of company XYZ might be concerned regarding the company’s liquidity and solvency in case one or more of the above events happen. Thus, to ensure that the company is in a position to withstand such events, the management might decide to take the data from the time periods in the past when such events did happen (maybe not at the scale as it is anticipated to happen in the future) and use the data in today’s context to see how vulnerable the company might get.
• Usually in stress-testing, the impact of more than one of the above factors are taken together – for example, the management might run a simulation of the company’s financial position if the exchange rate between the INR and USD rose sharply, while 4 out of 10 of the company’s biggest customers failed to make timely payments. 
• Logically, such events may have a certain degree of correlation – if the US economy has severe problems, the value of USD might fall while many major companies might go bankrupt, making the above scenario possible. Hence, the need for stress-testing is very real.

Advantages Of Stressed Risk Metrics

• The key advantages of using stressed risk metrics are –
  • They are more conservative, when compared to the typical VaR and ES analysis.
  • It is forward looking unlike VaR and ES that are backward looking. This ensures that we only take into consideration the relevant information during analysis.
  • It focuses on longer time periods as compared to VaR and ES.

Disadvantages Of Stressed Risk Metrics

• The key disadvantages of using stressed risk metrics are –
  • The stressed metrics are not good at responding to the current market condition.
  • Stress testing does not provide a probability distribution for losses.
  • Stress testing does not allow the effective backtesting of data due to its focus on extreme events.

Stress Testing Governance

• Governance is an important part of stress testing to determine the extent of the stress testing carried out by a financial institution. It also ensures that the assumptions underlying the tested scenarios have been carefully thought out, the results are prudently considered by senior management, and the actions based on the results are taken when appropriate.
• The governance structure within a financial institution is likely to depend on the legal, regulatory, and cultural norms within a country. Generally, there should be a separation of duties between the board of directors and senior management. However, some duties that are common to both senior management and the board are –
  • Senior management and the board should carefully consider whether the results of stress tests indicate that more capital should be held or that liquidity should be improved.
  • It is important for the board and the senior management to ensure stress testing covers all business lines and exposures. The same scenarios should be used across the whole financial institution, and the results should then be aggregated to provide an enterprise-wide view of the risks.

Governance – Responsibilities Of The Board Of Directors

• The board of directors has the responsibility to oversee the key strategies. It is also responsible for the firm’s risk appetite and risk culture.
• The board should define how stress testing is carried out. Specifically, it should determine the procedures used to create the scenarios as well as the way in which assumptions and models are used to evaluate them.
• Board members do not carry out stress testing themselves, but they should be sufficiently knowledgeable to ask penetrating questions. They should feel free to use their own experience and judgement to ask for changes in the assumptions underlying the scenarios (or even to ask for totally new scenarios to be considered).
• When key decisions to mitigate risks are required, the board should feel free to ask for other analyses to supplement the stress-testing results.

Governance – Responsibilities Of The Senior Management

• Senior management is responsible for ensuring the stress testing activities authorized by the board are carried out by competent employees as well as periodically reporting on those activities to the board.
• Senior management is also responsible for ensuring the organization is adhering to the policies and procedures.
• Senior management should ensure that the scenarios change as the economic environment changes and as new risks appear on the horizon. Stress testing should be an important part of the firm’s decision-making and risk-mitigation strategies.
• Senior management should have a deep understanding of how stress tests are carried out and should be in an even better position than the board to challenge key assumptions and models. If the nature of a scenario is not changed, then, at least, the severity should be changed.

Governance – Policies And Procedures

• A financial institution should have written policies and procedures for stress testing and ensure that they are adhered to. These policies and procedures should be clearly stated and comprehensive to ensure that different parts of the organization approach stress testing in the same way. The policies and procedures should –
  • Describe why stress testing is carried out
  • Explain stress-testing procedures to be followed throughout the company
  • Define the roles and responsibilities for those involved in stress testing
  • Define the frequency at which stress testing is to be performed
  • Explain the procedures to be used in building and selecting scenarios
  • Explain how independent reviews of the stress-testing function will be carried out
  • Provide clear documentation on stress testing to third parties as appropriate
  • Indicate how the results of stress testing are to be used and by whom
  • Be updated as stress-testing practices will change as market conditions change
  • Allow management to track how the results of stress tests change through time.
  • Document the operation of models and other software acquired from vendors or other third parties.

Governance – Documentation

• Documenting activities within a financial institution is often not a popular task. It is usually viewed as a less interesting and less creative activity than, say, building a model to investigate the impact of a recession where GDP declines.
• However, documentation is important as it ensures continuity if key employees leave and satisfies the needs of senior management, regulators, and other external parties.

Governance – Validation And Independent Review

• Stress-testing governance should include independent review procedures. The reviews themselves should be unbiased and provide assurance to the board that stress testing is being carried out in accordance with the firm’s policies and procedures.
• It is important that the reviewers of stress-testing procedures be independent of the employees conducting the stress test. The review should-
  • Cover the qualitative or judgmental aspects of a stress test,
  • Ensure that tests are based on sound theory,
  • Ensure that limitations and uncertainties are acknowledged, and
  • Monitor results on an ongoing basis.

• It is also important to ensure models acquired from vendors are subject to the same rigorous review as internal models.
• The validation of stress-testing models is more difficult than the validation of other models because stress testing deals with rare events. validation approaches are also difficult because of the limited amount of data available from previous stressed situations.
• Models describing the relationship between variables in normal market conditions may not describe how they behave in stressed market conditions. The independent review should ensure that these phenomena are incorporated into stress testing models.
• The independent review should reach conclusions on the conceptual soundness of the stress-testing approach. When there are doubts about the best model to use in a certain situation, the results from several models can be compared and the totality of stress-testing reports can make the attendant uncertainties clear.
• It is often more appropriate to provide a range of possible losses rather than a single estimate. Expert judgement should be used to ensure results are presented in a way that is most useful for decision-making.

Governance – Internal Audit

• The internal audit function has an important role to play in stress-testing governance. It should ensure that stress tests are carried out by employees with appropriate qualifications, that documentation is satisfactory, and that the models and procedures are independently validated.
• The internal audit function is not responsible for conducting the stress testing itself. Instead, it assesses the practices used across the whole financial institution to ensure they are consistent.
• The internal audit function should be able to find ways in which governance, controls, and responsibilities can be improved. It can then provide advice to senior management and the board on changes it considers to be desirable.

Basel Stress-Testing Principles

• The Basel Committee requires market risk calculations based on internal VaR and ES models to be accompanied by “rigorous and comprehensive” stress testing. Similarly, banks using the internal ratings-based approach in Basel II to determine credit risk capital are required to conduct stress tests to assess the robustness of their assumptions.
• After 2007-2008 crisis, the Basel Committee published stress-testing principles for banks and their supervisors in 2009. These principles emphasize the importance of stress testing in determining how much capital is necessary to absorb losses from large shocks.
• The principles note that stress testing plays an important role in –
  • Providing forward-looking assessments of risk
  • Overcoming the limitations of models and historical data
  • Supporting internal and external communications
  • Feeding into capital and liquidity planning procedures
  • Informing and setting of risk tolerance
  • Facilitating the development of risk mitigation or contingency plans across a range of stressed conditions
• The Basel Committee considers stress testing particularly important after long periods of benign conditions. The crisis showed that such conditions can lead to complacency and the underpricing of risk.
• In examining the shortcomings of the stress testing carried out prior to the 2007-2008 crisis, the Basel Committee reached several conclusions. They can be summarized as follows –
  • The involvement of the board and senior management is important. Top management and board members should be involved in setting stress-testing objectives, defining scenarios, discussing the results of stress tests, assessing potential actions, and decision-making. The banks that fared well in the financial crisis had a senior management that took an active interest in the development and operation of stress testing, with the results of stress testing serving as an input into strategic decision-making.
  • The stress-testing methodologies used at some banks did not enable exposures in different parts of the bank to be aggregated. Experts from different parts of the banks did not cooperate in producing an enterprise-wide risk view.
  • Some particular risks were not covered in sufficient detail in the scenarios. For example, risks relating to structured products, products awaiting securitization, imperfect hedging, counterparty credit risk were not fully considered. The impact of a stressed scenario on liquidity was underestimated.
  • The scenarios chosen in the stress tests were too mild and had durations that were too short. Also, the correlations between different risk types, products, and markets were underestimated. There was too much reliance on historical scenarios and the risks created by introduction of new products and the new positions taken by the banks were ignored.