Risk and capital modellers share lessons from the LDI crisis, the shift to private assets and board interactions

In part two of this InsuranceERM and SS&C Algorithmics roundtable, chief risk officers and risk experts discuss how the liability-driven investment (LDI) crisis and moves to invest in private assets are influencing their approach to risk and capital modelling – and what the board has to say about scenario testing

Q: How did the LDI crisis affect your approach to liquidity stress testing and modelling?

David Harrison: There were a few lessons, in particular the severity of the stress. The main benefit really was the data on that level of stress, which we hadn't had before in history.

David Chen: We also learned about market liquidity in that scenario. This is invaluable tail event data that we use to form our expert judgement on available liquidity and haircut assumptions.

Guy Barton: The hard bit is the assumptions around how markets and government will react. You should always be questioning the assumptions you make based on the past.

Keith Davies: What's the implication of the Bank of England's new lending window? Does it create moral hazard?

Guy Barton: It's a sensible thing to introduce, but as a company you have to be very careful about assuming how much support you may receive.

Patrick van Beek: In our models we had quite strong correlations between UK and overseas risk-free rates. We saw in the LDI crisis that overseas rates moved way less than the sterling rates. That's not something that had been captured very well.

Guy Barton: It shows again that modelling using past data is useful, but it's got to be supplemented by stress testing.

David Chen: The PRA [Prudential Regulation Authority] is going through a consultation on standardisation of liquidity risk reporting across insurance firms. From there, we expect them to set a liquidity risk management framework, standardise stress requirements and possibly create an Operational Liquidity Requirement (similar to the concept of the SCR for solvency), as well as set out what's 'liquid' and what's not.

Whilst the intent is well understood, there is general nervousness around the risk of leading to an unwanted outcome of driving procyclical trading behaviour, which could exacerbate market volatility in a time of stress. We as an industry should collectively highlight such risk to the regulator.

Q: Insurers are investing more into private assets. What are the challenges around modelling these, compared with public assets?

Chris Craig: The amount of data on private assets, especially if we try to structure them in different ways to get 'highly predictable' cash flows, could become quite limited. We can end up in a world where we're making a lot of expert judgements based on limited data which isn't really representative of the way UK insurers want to invest. There's a need for a lot of challenge to ensure we're happy with the modelling, or the basis for the modelling, and what data is taken into that.

Ravin Jagatiya: Private assets have challenges along the whole chain which includes: sourcing, valuation, credit rating and capital modelling - each comes with a set of challenges that have to be overcome.

There's perhaps a philosophical question about whether you build a sophisticated granular internal model reflecting the underlying risk drivers, or a simpler model but have a more sophisticated validation model.

When investing in these assets it is worth noting that, as an industry, we have not had this level of exposure, and no one has seen what new risks might come about. This makes stress and scenario testing more important for these types of assets.

William Diffey: Smaller insurers investing in private assets via funds would be very reliant on the fund manager to provide information on risk and ratings. I question whether smaller firms would have the expertise required to invest in private assets.

Guy Barton: When problems emerge in a market, quite often it's because people got carried away in thinking it is the best thing since sliced bread and money starts piling in. You can't just rely on how things were a few years ago – you've got to keep an eye on the market and stay on top of what's happening.

David Chen: Solvency UK and the MA attestation is trying to push people in the right direction, from a risk perspective. The requirement for spread decomposition is making people ask questions like 'why are we getting that much spread?'. Maybe that will trickle down to modelling.

Ravin Jagatiya: If you can decompose the base spread, it makes capital modelling a lot easier as the stressing of each component can be considered in the internal model. However, it is worth noting that there may still be risks that only materialise under stress and as such are absent in the base spread decomposition i.e. the 'unknown unknowns'.

Q: How does the board of directors get involved in setting stress-testing scenarios and challenging the results?

Guy Barton: I don't find it challenging to get the board engaged with setting scenarios, as they have insights about the threats that should be considered. The key is to then to have a good discussion on what the results are telling you.

Keith Davies: One good challenge we got was if you ran the same scenarios the following year, would you get a different severity? i.e. have you taken actions to mitigate? Some of it was quantifiable, for example, the impact on capital. A lot was around operational and reputational risk.

David Chen: The elephant in the room for life insurers is the 2025 Life Insurance Stress Test the PRA is running. We're all trying to establish the benchmark for financial resilience for that.

Patrick van Beek: The results for each firm will be public, so there's a question around the solvency coverage ratio you're happy to show you're going down, to under stress.

Guy Barton: The worry is whether it's going to be consistent. Are we confident enough firms are going to apply it in the same way, and is it going to be a fair representation?

William Diffey: Boards are very interested in stagflation scenarios, and geopolitical risks and supply chains. Recovery and resolution planning has been a big push.

The actual model framework is catching up for demands from stress tests.

Ravin Jagatiya: You have to ensure the board has the right level of training to understand the SST result and limitations, as the underlying models are very complex in nature. This should ensure that they can provide valuable insight and challenge.

Leo Armer: Are the board discussions done as a briefing, or do you give them access to an interactive interface?

Chris Craig: It's more of a briefing to seek feedback. Currently, there's a real focus on scenario modelling of operational risks. That good – it's much more engaging and leads to a richer conversation.

Christopher Cundy: Leo, have you seen examples of boards using interactive user interfaces?

Leo Armer: Everyone is looking at it at the moment, especially with the potential for generative AI to help explain potential changes that have happened. We've done some work using a ChatGPT-like tool to help with comparisons of data. It saves you analysis time and allows you to focus on the outcomes, rather than describing it. Every user interface that people are looking to invest in now has to have this technology.

Q: Are you still looking to invest in speeding up your model runs, or are there other aspects of the modelling process which you see limiting your modelling capability?

David Harrison: We're at the end of the process to put our internal model on a new platform, and we've seen very significant speed improvements and cost reductions.

Leo Armer: We've built a high-speed simulator at SS&C called HiPER that was recently ported onto polynomials and callable bonds. It can also be used for nested Monte Carlo simulations for example, and we've seen 500 times performance gains, which is just unheard of. When you start to get that level of performance improvement, it completely changes business processes. Something that was previously run overnight, in a batch, can now be repeated several times per day – giving the business greater insight to results.

Part one of this roundtable is available here.