Wednesday, April 21, 2010

Reverse Stress Testing: What are the chances.....

Much has been written in this blog regarding stress testing and how it should be included as part of a complete risk analysis process in parallel with more standard exposure, tracking error, contribution analysis, etc. There has also been an amount of recognition that an increased effort is required in order to spread the understanding, and therefore responsibility for, risk exposure across areas of a firm outside merely the risk and performance teams.

This week's blog therefore aims to shed a little light on an area that sits alongside standard stress testing and potentially services that educational need: the concept of "reverse stress testing." Reverse stress testing is based on exactly the same principles as standard stress testing, but by approaching the analysis from a different, and some might say more intuitive, angle, conclusions can be drawn as to the risk exposure of a portfolio.

In standard stress testing, you consider an extreme event in terms of a shock to a valid market data series (e.g., S&P Financials drop 18%), and then use a model such as the conditional multivariate normal distribution to measure the impact that event would have on an existing portfolio. In reverse stress testing, you instead set limits on those impacts (for reasons such as mandate, internal compliance, etc.), model the implied shocks necessary to break those limits, and then consider the potential for that shock to actually happen. This process can leverage internal expertise in a manner that might not be considered through standard risk calculation.

Consider an example: The manager of the ABC Global Energy and Resources Fund has an imposed absolute monthly loss trigger of -10%, which might lead to a reallocation of mandate. He therefore looks to generate some reverse stress tests modelling impacts to several relevant measures generating the following scenarios:

The table shows three modelled events, the impacts of which be a 10% loss to the portfolio, alongside the standard deviation and implied probability of the event. The event that immediately stands out is the potential fall in the oil price of 20%, and, when we consider that these events are modelled over a one-month period, I believe the current climate allows a rapid disregarding of it. At the time of writing, the impact of the Greece de-rating is only just starting to feed through, and therefore the MSCI World event seems a lot more likely and is therefore the stress test I would allocate resources to in terms of exposure analysis.

This is only a simple example, but I hope that I have demonstrated that although the mechanics of the calculations are the same when it comes to stress testing, it is the analysis of the "reversed" tests that adds some extra colour to the above risk analysis and underlines how it is possible to further leverage indirect risk knowledge and experience and knowledge.

To receive new posts by e-mail, subscribe to this blog.

No comments:

Post a Comment