Stress Testing approaches at Pension Funds/Asset Management companies

Question

I am looking for resources on Stress Testing for non-banking institution, specifically for long term oriented Asset Management companies, Hedge Funds, Pension Funds, and other Investment companies. Although highly subjective and company-specific, there are similarities in this space and it would be great to know best practices.

For example, here is a excerpt from one of the CPPIB's reports on Stress Testing with no details:

Stress testing

To complement the suite of risk measures used to monitor CPP Investment Portfolio risk, CPP Investment Board further examines the potential impact of exceptional but plausible adverse market events. Stress testing considers the effect of various hypothetical scenarios. CPP Investment Board has a bottom-up approach for stress testing that considers the effect of either historical or predictive shocks across the entire portfolio. The focus is predominantly on those expected to be more extreme tail event losses for the CPP Investment Portfolio. Generally, the forecasted time line considered is one year in order to appropriately include the effect of private asset valuations, as well as initial public market shocks. These are reported to both senior management and the Board.

Unlike VaR measures, which have an implied probability because they are calculated at a specified confidence level, there is generally no implied probability that our stress test scenarios will occur. Instead, trigger points will be highlighted as a potential situation starts to evolve and then the directional trend gauged accordingly.

Thank you for any pointers and resources. I will try to enhance the question as I will have more specific details.

Update

I think we can break stress testing into three broad categories:

1) Historical Stress Testing - pretty much what @AlRacoon wrote below;

2) Deterministic Stress Testing - where you specify some deterministic scenarios like parallel shift interest rates by 100 bps, shift volatility by 10%, etc. and see what would the results be (basically individual correlated or uncorrelated shocks to other risk factors);

3) Predictive Stress Testing - where you identify major factors (let's say Trade War between US and Chine is imminent and thus formulate a hypotheses such as EM goes down by 20%, Oil goes down by 10%, Fed cuts rates by 50 bps, etc. Take the combination of these stress factors and propagate (link) to other risk factors using some type of regression and figure out the final P&L.

How do you specify a time window for a regression model? Do you rationalize like market conditions change frequently and in order to reflect the current market regime I am going to go ahead and take the last 2 years for my regression model? Or do you say like my investment horizon is long term and thus I want to use 10 year data for my regression model?

Also, how can you incorporate the forecasted timeline (6 months, a year, etc)?

Are there any source that offer comprehensive walkthrough?

Answer 1

Many long term investors use historical events and the market moves associated with such events to stress test their portfolios. For example, they use the dot-com bust, the latest "great recession", LTCM, Asian Crises, Black Monday, etc and any other dramatic events in history and see how their portfolios would have performed under those conditions. Of course this is limited to events that have occurred and not necessarily what might have occurred or will occur in the future. They augment these historical stress tests by creating stressed scenarios. By taking the largest market moves that would be adverse to their portfolio and using high correlations (either by assuming correlations are 1, or by making adverse moves in markets occur simultaneously) and assess the impact on their portfolios. Others impose a rationality check overlay on these scenarios to include what is likely to occur simultaneously. As you can see this is very much of an art. Many risk management systems incorporate such simulations as a part of their package.