Picturing the ideal risk model

In part two of this InsuranceERM/Towers Watson roundtable, modelling experts discuss what they want from risk models and the potential drawbacks to innovative approaches

The first part of this roundtable is available here

Chris Cundy: The models that we have been describing, with linkages to the real economy, suggest a more complex model. Isn't simplicity desirable, or is greater complexity unavoidable?

Karina Lo Dico: A good example is the defaults on US AAA-rated mortgage-backed securities. Most people simply accepted that was the risk level. If you had been watching mortgage defaults and the types of products being sold, you could have made a better assessment of the credit risk.

Joel Fox: It is about understanding and using the right drivers in your models. That might lead to more complex models but it might also lead to some simplification of assumptions, recognising we do not know all the answers, but let's model what we can.

Kevin Borrett: It may be utopia, but I would like models that themselves limit and reduce model risk. Otherwise, there is a danger that one builds up a process of thinking driven by the model and it becomes too linear. We need a sanity check within models that prompts the user to actually think about the outputs in a more holistic way.

Joel Fox: This is the danger of these models: you can get very easily caught up in the detail and what that particular scenario is telling you without stepping back to think, 'actually, is that even the right scenario to be thinking about here?'

John Rowland: Which is the danger, I think, with the abundance of the technology now. It is possible to build more complex models, it is possible to build more granular models and in doing so forget the need to link them to reality and understand the risks associated.

Take operational risk modelling; people went down the road of trying to model it and ended up with something that looked complicated and was therefore believed to be right, but it was so dependent on the input assumptions that they had to backtrack. So rather than modelling operations using a complex Monte Carlo model, a lot of companies went back to a much simpler scenario-based approach. That was a more trustworthy approach, because it was more transparent that the answer was a direct function of the inputs, which were expert judgement-based assumptions.

Karina Lo Dico: The model we used to have for operational risk was a complex Monte Carlo model, but what was helpful was the sessions we held to set the inputs. It made people really think about the risk.

Costs of technology

Chris Cundy: There seems to be a growing emphasis towards cost saving, rather than regulatory compliance, in investment cases for new technology. What kinds of opportunities for spending remain?

John Rowland: Solvency II is driving a timetable of solvency reporting four times each year with, in the end state, maybe two weeks to produce the numbers, so there is a huge trend around cost reduction through automation, and companies that talk to us have very clear objectives. The costs here are directly linked to the large numbers of people, large numbers of spreadsheets and operational risk around processes.

Karina Lo Dico: Within a small team it is easy to do the maths to show how we can reap the benefits of process automation, but I can imagine in larger organisations with more complexities, cost-benefit analysis is less easy to do.

Tim Thornham: Clearly there has been a large investment in the past five years in Solvency II systems and there are ongoing maintenance costs. Within an organisation like Aviva, certain elements have been centralised and can be deployed across multiple business units. Doing that should drive costs down. And centralising the maintenance budget makes investment decisions simpler. The conversation changes from 'do you invest your money?' to 'where do you invest your money?', because you know these are key systems that need to be maintained as the business and risks we are exposed to evolve.

Downsides to new technologies

Chris Cundy: We have talked about the challenges in developing and integrating technologies, but are there other factors preventing you from adopting new modelling technologies? Security issues, for example?

Joel Fox: We have certainly seen many people not taking up cloud computing earlier because of anxiety about security. We are starting to see that change, but I am curious as to what extent others in the industry see the nervousness evaporating?

Kevin Borrett: We do not see the use of cloud as a barrier to new technologies, which in some respects might even improve industrialisation and robustness of infrastructure.

Karina Lo Dico: Having modelling platforms in the cloud takes away the time an actuary spends on setting up, upgrading and the like, so it does seem like an easy win. There is a cultural change associated with that, but it does seem to take risk out of the process without a lot of management time required.

Kevin Borrett: It is also helpful in your dialogue with regulators, where you can point to increased resilience within core and critical systems.

Tim Thornham: Our 'digital first' strategy at Aviva means we are committed to adapting to new technology. We are currently in the process of transitioning our risk models to cloud-based solutions. That clarity of focus enables people, including IT security, to get on with solving the problem.

John Rowland: Our most recent product was launched as cloud-only, which we would not have done if we did not think that was a viable method of delivering it. We believe the trend will be towards software that is a service, rather than licensed traditionally.

Joel Fox: There is a cost benefit for companies using the software and for the developers of software, because it is easier to develop and deploy solutions in the cloud as 'Software as a Service', as opposed to traditional on-premise solutions. The cloud and associated technologies also enables faster release cycles and clients can adopt new technology quicker.

The ideal risk model

Chris Cundy: In September 2015 we surveyed 15 CROs, chief actuaries and risk managers and we came up with a definition of the holy grail of risk and capital modelling systems. It reads:

It is a single, accurate risk model that responds quickly and easily to changing economic and demographic scenarios, producing meaningful and validated results in real time.

Do you agree with this?

Joel Fox: We have challenged the question of whether it should be a single model already as we see value in analysing risk from different perspectives, but is that because we do not think any of the single models is capable of that, with all the appropriate linkages to the real world? If we did have that, maybe we could convince ourselves that a single model is utopia.

John Rowland: I would add 'well-understood'. The utopia, for me, is a useful model and a useful model is one that you understand its strengths and weaknesses.

Kevin Borrett: Yes, but I would also add 'easy to use'.

Blue-sky scenario

Chris Cundy: Can you imagine if all the technology constraints went away? What would you be doing differently in terms of managing your risk and your business?

Karina Lo Dico: When you have a portfolio of risks, it is very much like holding a portfolio of stocks and there seems to be so much analysis in investment management about volatility and the driving factors. I would like to get to a point where you can understand your portfolio of risks in a similar manner; to understand where you do not have exposure to a particular area and how you get exposure. For that, you need to understand how much capital you must hold for each policy, what risks they are exposed to today, what is systematic, what is unsystematic and the like. We are miles away from that.

John Rowland: We were asked by an insurance company CEO, who had a banking background, to build a solution that enabled him to monitor the performance of his portfolio of businesses actively, almost on a daily basis. While it's not been decided whether it will get deployed, the technology is there to implement such a capability.

Tim Thornham: Capital is considered a scarce resource and so if you want it, you need to show senior management you are worth it. Having sufficiently granular metrics supports senior management in capital and risk-based decision making at the portfolio level.

Kevin Borrett: If we ever got to the point where models dealt with the business-as-usual, then it would be the opportunity to stand back and exercise some real blue-sky thinking about what the new opportunities are.

Karina Lo Dico: If you can create transparency about how you price your risks and commoditise these products, anyone can come into the market.

Tim Thornham: Yes, if technology constraints were removed, then data becomes a key differentiator and there is a threat from the likes of Facebook and Google, who have global reach and have built their businesses from expert use of data. Insurers will need to adapt quickly to technology advances, whilst continuing to focus on their unique customer proposition and delivering a valued service.

The first part of this roundtable is available here