The importance of forecasting scenarios has become more important than ever. Daniel Finn, head of risk solutions at Conning, explains how Conning's GEMS stress scenarios software can help insurers
What are the key strengths of the GEMS Stress Test Scenarios?
Conning's GEMS Stress Test Scenarios have several key strengths. First, a great deal of care has been taken in putting the scenarios together to determine a clear, precise definition of each period to be used. For example, most people understand that the 2008 financial crisis was bad, but until one comes up with a specific definition of when it started, how long it lasted, and what it encompassed, it's impossible to put it into context for clients. Second, we've structured the scenarios in such a way they can be incorporated into any of our clients' models: they simply import a single file, and the model recreates the historical events. This not only makes it easier for our clients to maintain a single risk management platform, but also ensures they can be used across all of their different risk metrics, not just asset returns.
Do clients use these scenarios to augment or replace their existing risk management process?
Conning's clients are using the stress scenarios to augment their existing risk management processes, and there are two key areas where they have found them particularly useful. First, given the wide number of stochastic variables in a typical economic scenario generator (ESG), it can be difficult for companies to gauge the appropriateness of the resulting distributions. By incorporating historical events, however, companies can compare their results against a meaningful benchmark. Another area where we've seen companies struggle with their models is in understanding what the key drivers of change are from one run to the next. What were the initial conditions? What were the model assumptions? By instead using historical events, clients can strip out most of these differences and focus on what they really care about: how the changes they've made in their risk profile have impacted their projected results.
How are clients using your new Covid-19 scenarios?
As the financial markets started going off the rails in mid-March, Conning received a number of client inquiries about how our models would handle these "unprecedented" times. Typically, these requests were either in the form of "Did the GEMS software anticipate this level of market movement?" or "How will these results impact my company's results going forward?"
Our Covid-19 package was designed to help companies start addressing these questions. To help answer the first question, we included two of the most extreme events in our historical stress scenarios: the 2008 financial crisis and 1973–74 stagflation. When companies ran these scenarios, many were quite surprised to see there have been historic events worse than the present situation, although they did take much longer to develop. Second, by incorporating the first quarter's historical results into our annual model, we were able to help companies see the impact of the three major market changes upon our GEMS ESG projections: 1. Much lower treasury yields 2. Massively higher corporate spreads, and 3. Increased equity volatility.
What future services are you considering implementing to help insurers in this business?
Conning is constantly evolving our offering to help our clients. Two of the things we're currently working on are tail scenario selection and risk decomposition.
We see a lot of similarities between our stochastic models and cat modelling: both produce a wide range of plausible outcomes based on complex models, and when used by skilled practitioners, they can be a key component of a company's risk management platform, but they can also difficult for laymen to understand.
So, we think it will be useful to borrow another one of the cat modellers' key ideas: the 1-in-100 and 1-in-250 events. Specifically, our tail scenario selection method will let clients choose a single economic driver such as 1-in-250 drop in the equity markets or 1-in-100 increases in corporate spreads. This tool will identify scenarios with these characteristics from a massive stochastic set, possibly as many as 2,000,000 paths, that can be run through our models just like the historical events are.
For risk decomposition, we are increasingly seeing clients interested in factor-based asset allocation strategies, and we want to support these clients by giving them the ability to seamlessly decompose their simulated results.
