Stress-Testing & Scenario Analysis

Stress-Testing

Today, stress-testing is a critical part of any strategic planning. Traditional stress-testing methodology uses historical information and events, and ascertains what would happen now if we faced the same circumstances again. While this approach has some validity the problem is that history does not appear to repeat itself in an identical manner. In the case of historical stress test analysis you really do not know whether the historical parameters selected will be extreme enough. After all, financial markets have less than 100 years of more timely financial data to work from.

CheckRisk has developed a suite of more sophisticated, forward looking stress testing approaches. We go beyond what a software provider can do for you. We can stress-test your portfolios against multiple criteria, using new scenarios weighted to more recent asset class information. We can assess what effects a single extreme event, or the impact of a sequence of events may have on your organisation or investment fund.

Our view is that stress-testing should be about identifying risks that have yet to occur as opposed to using historical models and assuming that history repeats itself. Stress-testing can be used to build risk resilience into organisations or to help understand where vulnerabilities exist. A further advantage of the CheckRisk stress testing approach is that implications of a proposed course of action can be analysed prior to their execution. Being forewarned is to be forearmed.

From the non-parametric to the parametric, and correlations to causality CheckRisk can offer a wide array of stress-testing methods that will allow you simulate different scenarios. Our models let you programme any economic information or any security, equity, bond or commodity index you like.

Scenario Analysis

Scenario analysis differs from standard stress-testing in the sense that it is more about thinking of several more plausible ‘what if?’ situations. This is useful when thinking about the different outcomes we may face in any given year, from elections to interest rate rises and other risk factors.

The human mind is very good at thinking in terms of causal chains, and stories that are easy to understand. Pattern recognition is one of the basic human thought processes and often leads to mis-identification of risk. Scenario analysis is therefore especially useful for boards or other non-expert executives, who may not find the standard statistical methods as intuitive.

CheckRisk believes that the process of thinking about these plausible situations is as valuable as the results.

Example: Bayesian Network

Coherent Scenario Blue Red2(1)

Above is one such example of scenario analysis, and shows a visual representation of a Bayesian network. We selected several events that could impact markets over a given time period and attached probabilities of these events occuring along with the relevant profit and losses to a mixed portfolio of equity and bonds. The size of the node represents the impact on the portfolio should the event materialize. The grouping colours have been identified by the event conditionality. In other words, the algorithm identifies that a Chinese devaluation and the oil price moving to $20 are more dependent on each other and are therefore in blue.

The probabilities themselves have been input according to external models which we have verified where possible, and have been subjectively identified in those instances where it is not possible. CheckRisk considers probabilities in terms of possible outcomes which can then be weighted. Clients may also choose to input specific probability outcomes and CheckRisk will ensure that the model then reflects these choices.