Through alternative risk and model approaches - Takeaways from the Aventedge Cat Risk Management & Modelling Conference 2018
Dr. Matthew Jones
Aventedge Cat Risk Management and Modelling Conference, London, 7th March 2018
Simplitium hosted an interactive breakout session at the Cat Risk Management and Modelling Conference 2018. Any session like this is dependent on engagement and thankfully we had a lot of participation from a range of model developers, brokers, insurers and reinsurers which led to very interesting discussions. This article summarises the main takeaways from the session and highlights the barriers faced by insurers when trying to adopt alternative views of risk.
The subject of improving your internal view of risk is very close to me as I come from a scientific background and am well aware of the limitations of models. In my previous role as Global Head of Catastrophe Management for Zurich Insurance Group I very much championed the concept of improving an internal view of risk and not just using ‘out of the box’ model results.
I was also lucky enough to co-edit and co-author the book Natural Catastrophe Risk Management and Modelling, A Practitioner’s Guide together with Kirsten Mitchell-Wallace, John Hillier and Matthew Foote, where I wrote a chapter on the topic of ‘developing a view of risk’.
One of the things that frustrates me in the field of cat modelling is how limited we are in the scope of models we use. If you look at the discipline of climate science, which is also trying to estimate the impacts of something very uncertain, it is clear that multiple models are used. For example, in the IPCC 5th Assessment report over 30 models are used when projecting global average surface temperature out to 2100. Yet this usage of a wide range of models does not happen within cat modelling.
The latest ‘Catastrophe Risk in Bermuda’ report by the BMA shows that roughly half of firms that are regulated by the BMA use multiple vendor models, and half only use one vendor model – and this is a selection of highly cat exposed entities. The only cat modelling firms considered in the BMA report are RMS, AIR and EQECAT – which is in itself a comment on the limited model usage in the industry.
Although I’m frustrated with this being the state of our industry, I do completely understand it. For many operational and budgetary reasons, it is actually very hard to use alternative or multiple models, even just for the key perils. In the light of these challenges using a set of models from one vendor may well be the only pragmatic solution at this moment in time for many companies. However, I think it is important that we try and change this.
There is huge uncertainty in cat models. Much more than is normally communicated via the cat models themselves – more than is shown in the standard deviations we get in the event or year loss tables that the models output. Even if different models seem to be producing similar outputs at a high level, larger differences normally become evident as you drill into the results (which we could also see clearly during Federico Waisman’s model comparison session).
Using multiple models can help us understand the uncertainty range and stop us optimising our portfolio around one model. This gives us a better view of our risk and helps mitigate surprises when a catastrophe happens (the event itself never cares what the model results say).
In my view, anything that can be done to make it easier for firms to use multiple models will improve the industry.
During the breakout session we started by looking at some of the barriers to adopting alternative views of risk. There was a lot of feedback here and we probably could have spent the entire 40 minutes discussing this alone.
It was very clear that there are many barriers that prevent (re)insurers from adopting multiple or alternative models in the current landscape. Those mentioned in the discussion are shown in the table to the right. When asked to highlight the top three barriers the consensus was: Regulation, Fear of change and Cost.
After discussing the various barriers, we went on to discuss how the barriers could be lowered.
One of the initiatives discussed was the Oasis Loss Modelling Framework (Oasis LMF), which we hope everyone is aware of by now. Oasis LMF is a not for profit organisation created to lower barriers and provide the industry with a standardisation that leads to more models, lower costs and more transparency. It does so by delivering a set of open source software, including a set of C++ components ‘ktools’, a standardised framework for specifying models, and an engaged community.
Another initiative discussed was Simplitium’s ModExTM platform, which brings the Oasis LMF to market via a new, innovative and streamlined shared service cat modelling solution. ModEx lowers the barriers between (re)insurers and model vendors by delivering a centralised platform for multiple vendors (including JBA, Impact Forecasting, ARA, CATRisk Solutions and Combus). The web based platform delivers more choice and lower costs for (re)insurers while generating demand and encouraging model development from model vendors.
Quantifying the benefit of models and their costs was another area discussed in relation to lowering barriers, particularly the fear of change. It is important that we can communicate why using multiple models can benefit our organisation, whether it is by being more resilient to event ‘shocks’ or by designing an underwriting strategy that takes into account the confidence we have in account level pricing.
Closely linked to this is the transparency around vendor models, the time taken to validate models and the burden of regulation. Ironically, regulatory requirements seem to be a barrier that reduces the likelihood of a (re)insurer using more models, thereby leading to an increase in the systemic risk in the market – something which regulation is meant to reduce. Standardising validation, as far as possible, is something that could help a firm reduce their validation resource requirements, but the extent to which validation can be standardised is something that needs to be discussed with regulatory bodies.
Finally, we discussed which resources firms would find useful on the topic of improving our view of risk using alternative model or risk approaches:
- In terms of finding out what models are available, this page on the Oasis Hub is useful.
- A list of vendor models available on the ModEx platform can be found here.
- An excellent paper on model blending (Calder, Couper and Lo, 2012) is available here.
- The discussion on uncertainty and the chapter on view of risk in the following book could also be helpful.
We would appreciate your feedback on the aforementioned barriers and how you think these can be lowered. Please email me your comments and feedback (email@example.com).
For more information about how ModEx is working to make it easier for firms to adopt multiple views of risk, please visit the ModEx page.
About Dr Matthew Jones:
Matthew Jones is the founding Director of Cat Risk Intelligence, a UK based company providing catastrophe risk management consultancy to the (re)insurance industry.
In his previous role as Global Head of Catastrophe Management for Zurich Insurance Group, Matthew led the organisational change required to establish a global team with consistent processes to provide catastrophe knowledge, systems, models and services across Zurich’s general insurance lines of business. He worked for Zurich for fourteen years, including various roles in the actuarial pricing and catastrophe risk management fields.
Prior to this Matthew was a reinsurance pricing actuary for St Paul Re in London.
Matthew graduated from the UK’s University of Nottingham in 1993 with a degree in Physics. He then completed a PhD in Oceanography and Remote Sensing from University College London, whilst being based at the UK’s National Oceanography Centre in Southampton. He is a Fellow of the UK Institute of Actuaries and a co-author of ‘Natural Catastrophe Risk Management and Modelling: A Practitioner’s guide’.