VIDEO BLOG

In this video blog Dan Schobel, Actuary for Numerix discusses the use cases for real world Economic Scenario Generation and addresses how insurance companies are using these scenarios to improve their risk management capabilities while gaining a new perspective on their business. He also addresses some of the common challenges that institutions face when implementing real world ESGs, including model selection and calibration. 

Weigh in and continue the conversation on Twitter @nxanalytics, LinkedIn, or in the comments section.

Transcript:

Jim Jockle (Host): Hi, welcome to Numerix video blog, I'm your host Jim Jockle. Joining me today, following up on his recent webinar on economic scenario generation, looking at real world challenges, is Dan Schobel, Actuary from the product management group here at Numerix. Dan, welcome.  

Daniel Schobel (Guest): Thank you, Jim.

Jockle: First of all, let’s just start. ESG. Real world ESG. What is it?

Schobel: Real world ESG is a tool that insurers and other organizations, such as asset managers, pension funds, P&C insurers, Life insurers, and many other organizations will utilize to get a better understanding of the risk that they face. Really a Real World ESG is intended to make probabilistic statements. So gaining comfort around an organizations solvency position for example.

Jockle: So, you bring up solvency and now we’re at stages of rollout of Solvency II. How does real world ESG play into the regulatory requirements?   

Schobel: Yes, so because Solvency II is a framework that is based on using essentially a probabilistic view of a company’s capital position to understand whether or not they would be solvent over the next year. Naturally, this leads to using real world scenarios to measure the economic risk portion of Solvency II. Now solvency II encompasses more than just economic risk, it’s also non-market risk like mortality, and longevity, as well as operational risks but we’ll just focus on the economic risk portion which would incorporate like interest rate risk, equity risk, exchange rate risk, credit risk; these kinds of risks would be captured inside of that component.  

Jockle:  So now, clearly quantitatively challenging and in many ways we’re definitely seeing more and more introductions of different models in product design, the introduction of real world modeling into economic scenario generation; clearly not something that is very easy to do. Perhaps you can give us a little insight that companies should be thinking about when looking to do different or looking at different ESG type solutions.

Schobel: Oh yes, absolutely. And it depends very much on what their use case is because many different real world use cases will often be looking over different time horizons, they may be sensitive to different risk factors and they’ll need to very carefully choose what sorts of models they’re going to use inside of their real world ESG for their given use case because if you try to have just a single standardized real world calibration to use across all use cases and perhaps all lines of businesses within an insurance organization it may accurately capture the risks associated with some lines but it may not accurately capture risks for other lines. Some liabilities are naturally short duration others are long duration and depending on how those real world scenarios have been calibrated or tailored the scenarios may be more appropriate at one time horizon vs another.

Jockle: And there was one of the topics as related to your webinar was some of the challenges specifically around calibration, is that correct?

Schobel: That’s right, yes. And real world calibration is, by its very nature, very subjective. Because ultimately when recalibrating all these real world models we’re attempting to estimate risk premiums, but you can’t really know what the true risk premium on any asset class actually is. We could go out to ten different economists, and we would probably get back ten completely different answers for the risk premium on some given asset like the S&P 500 index versus government bonds for example. Whereas in the risk neutral framework everything is just about reproducing observable market prices. So when we calibrate the risk neutral model there a very objective and clear way to validate whether or not we have a good calibration, but for real world its objective because my view of what the appropriate risk premium is could be very different from your view or someone else’s view.

Jockle: So what are some of the best practices in normalizing an individual’s perspective?

Schobel: So what individuals should take into consideration when they want to consider views that they’re going to incorporate into their real world calibration is the nature of their use case, what are their most sensitive protection horizons, what are their most sensitive risks, and be sure that the scenarios that they’re calibrating and the resulting distributions of those scenarios are actually going to capture real risks that they would face. They want to avoid situations where their own internal views may lead to some risks being hidden or not considered. An example of this would be if they were assuming steady or even a rapid rise in interest rates over the next few years. So all their scenarios would generally tend to incorporate that behavior into them and so basically all your paths in that real would set would have all these steadily rising rates and you would completely omit the risk of having continued, prolonged periods of low interest rates as we have today.

Jockle: So you know you bring up a very interesting thought. In the sense of interest rate risk. You know obviously we’ve has long durations of negative rates, of low interest rate environments, and, you know, rates tend to move in step over time. But, take into consideration something like volatility, where you have, especially now, you’re getting much more spikes into the marketplace. How are insurers thinking about volatility in this perspective?

Schobel: Yes so, more insurers are starting to take a much harder look at volatility in their real-world scenarios and starting to get a better understanding of how even something as simple as interest rate volatility can have quite a large impact on their results because particularly for many real world use cases we are often looking deep in the tail, you know? Like the 99th percentile or a deep tail, conditional tail expectation. And, these kinds of measures are going to be very sensitive to basically the most extreme scenarios in the real world set and so your assumptions around volatility, whether it’s for interest rates or even other asset classes like equity or FX, these are going to really drive how those detailed percentiles are paving.

Jockle: So, from that perspective, you know, you’re running hundreds of passes, hundreds of scenarios and going into downstream impacting, you know obviously your valuations and your long term projections for capital. What is a best practice in terms of your thinking to, you know, obviously you want to manage your volatility but you know, having a longer-term perspective, how do you set bands around the spikes in the market to make sure you’re running a stable business?

Schobel: Yea so there’s a few ways that users can do this. Often a good place to start is actually to first look back over recent or relevant historical periods to see what the most extreme level of volatility has been in the historical record. That’s often a good starting place as perhaps setting an upper bound on how much volatility you would want to observe in your scenarios but you may want to take an even more conservative view and consider, well what if we end up in a market environment that’s actually even worse than anything we’ve ever seen historically. So for example, something worse than the 2008 financial crisis, worse than the Great Depression.

Jockle: Texas?

Schobel: [Laughing] Yes, sure. So if we’re going to be in an environment which is even worse than anything historically observed, then we can pad our volatility assumptions and actually make them even more extreme with the understanding that such a view would drive up like capital requirements, for example, in an economic capital exercise.  
Jockle: Well Dan, I want to thank you so much for joining us and of course if you want to catch the replay/on-demand version of Dan’s webinar, it’s available on numerix.com. And we absolutely want to talk about the topics that you’re interested in, so please follow us along on LinkedIn or on Twitter @nxanalytics. Thank you so much, and thank you Dan.

Schobel: Thank you, Jim.