+44 (0)203 714 6957            info@lighthillrisknetwork.co.uk
Key Research Themes & Challenges
Research Themes for 2016 – 2017

Our research priority themes change annually upon consultation with our members and with the wider community. Click each topic heading below for details:

 

1. Developing global databases of standardised information. There are many potential benefits to global databases that standardise the formatting and presentation of risk information, such as a worldwide buildings database using unique building identification. The benefits would include considerable reduction in transaction costs, improved comparability and facilitation of electronic placement. Despite this, data in the industry is currently collected but neither consistently consolidated nor shared. What is the business case for industry data bases, and what are the obstacles to creating them when other sectors like aviation have gained so much?

 

2. Creating a business case for sharing anonymised claims data.
Despite some industry-wide data collection, we still see insurers baulk when asked to share data. Are they right or is sharing data is in everyone’s best interest? Research could consider whether the US experience of data sharing has been beneficial for the industry and how the UK open data initiative has worked out. What legal issues are there and what level of aggregation would be acceptable and still remain compliant? Is it possible to design algorithms that could be used in place of sharing data openly? What about independent data holders?

 

1. Unconscious bias in decision making. How can decision makers identify and reduce their cognitive biases? What tools could be developed, bearing in mind that senior executives may not acknowledge their own biases and probably would not want to admit them to others if they do. The research could also go back further to consider how unconscious bias may have affected the choice of subject and career, and appointments and promotions of senior executives. This is currently a concern in academia and education and the research may be transferable. Could some form of e-learning help? The best way would be to demonstrate the effect of changes on results. What are parallels with other industries such as engineering and aviation? “Many CEOs of insurance companies don’t think differently until not only is there blood in the streets, but also some of it is theirs.” Tom Bolt

 

2. Better methods of making decisions and reducing inertia. People often do not adopt new techniques even if those methods have demonstrated better results than the ones which they are using. Projects could be developed under this theme mostly based on synthetic research (summarising and communicating what can be done) and finding ways to stimulate change. They could consider:

• Behaviour: are there better methods available and if so, why are they not being used?
• Determining a small set of forecast skills scores and encouraging their adoption
• Bayesian methods in reserving and why has there not been more uptake

 

1. Practical tools for use of models in decision making. Models are becoming increasingly important in business applications like pricing and capital setting, in informing strategic decisions, and as tools for communication and learning, not just catastrophe modelling. Despite great advances in the modelling of insurance risks, very substantial uncertainty persists and different ways of using – or indeed not using – models produce different risks for an organisation.
A mechanical translation of model outputs to decisions creates vulnerability to potential modelling flaws, but excessive concerns about their limitations inhibits their effective use in identifying and optimally managing risks.
Practical tools could be designed for the use of models at two levels:
• Qualitative strategy tools to structure decision making at board level
• Quantitative decision tools to support business decisions by senior management

 

Both such tools should explicitly acknowledge (model) uncertainty and the “contradicting certainties” that this creates across stakeholders, and offer a structured way of counterbalancing conflicting perspectives in order to reach a decision.

 

2. Understanding and communicating uncertainty and probability. There is much evidence that people do not easily understand probability or risk. This deficit can affect many aspects of insurance, from the board looking at catastrophe models to communicating the rating of decisions with customers of risk and how this can influence them. “Catastrophe underwriters can be lucky for their whole career.” The literature includes work by Professor David Spiegelhalter at Cambridge and Gerd Gigerenzer, Director at the Max Planck Institute for Human Development on communicating risk probabilities in healthcare, including to health professionals.

 

3. Forensic analysis of decision-making over time. Long-term casualty insurance is arguably more exposed to behavioural risks than short-term lines of business, such as property, where there is faster feedback on the quality of the decision making. This could be a vehicle for understanding the process of decision making, including availability bias, representation risk, cognitive dissonance, group think and agency risk, considered in the light of results. There are potential regulatory aspects, too, for example the institutionalisation of solvency models shifting attention from long to short term, and the balance for the insurer between prudence, support for innovation and returns to shareholders. The cost-benefit analysis of policyholders ‘being safe’ and how this can be communicated.

 

4. Robustness of decision making: Models can be useful to explore different strategies but there is also a danger that the apparently optimal choice is overly dependent on the model being used. A project could consider how models can be used to produce ‘robust’ decisions which do not alter materially in the face of parameter, data or model uncertainty. The desired outcome would be ‘good practice’ based on synthetic research which could be used by modellers in the insurance industry when creating models for their business.

 

Global shipping movement patterns. Marine insurance is becoming uneconomic with too many players chasing finite premium. Marine (and aviation) are apparently no longer written as separate classes by many reinsurers, but in a general book of business for their diversification value. Yet the potential for catastrophic losses remains, as Superstorm Sandy (2012) and Hurricane Ike (2008) and the fire at Tianjin Port (August 2015) demonstrate.

 

Vessel size. The size of the vessel is rising to produce economies of scale for owners and operators but with an accumulation of risk for underwriters. Serious shipping casualties have become less common over the past ten years, but catastrophic losses are likely to be more expensive. Bigger ships = bigger wrecks and more lost cargo.

 

Terrain data. These increasingly large vessels are navigating oceans and seas often where the bathymetric data, which produces 3D images of sea beds, is inadequate or non-existent. Vast areas have never been surveyed or not re-surveyed for decades. There are implications for ship safety and for salvage.

 

Could the information available on specific vessel movements from IAS responders support the development of a global picture of all vessel movements over time? It could allow a view of risk by vessel type, size, ownership and shipping routes. It would also be interesting to look at how well identified and modelled are mobile risks in transit, especially in port areas.

 

1. Modelling the market. Although there are company models, there isn’t currently an insurance industry model that looks at the interactions of all the agents together. A whole market model could allow deeper understanding of cycles and systemic risk. Ideally, it would consider all major cash flows and processes, claims processes, underwriting rating, regulation, agency behaviour, reinsurance and management and policyholder options.
• Regulators could use the model to explore potential policy implications, such as testing the efficacy of VaR, TVaR and other risk measures
• It would enable testing of large-scale shock scenarios at industry level
• It would help identify sub-optimal processes and highlight good practice

 

2. The potato blight risk of using the same models. Industry dependence on a limited number of proprietary models could create systemic risk because we do not know the models’ weaknesses, either specific or generic. A serious flaw in those models could mean that a large part of the market may at the same time be under-pricing (or even be oblivious to) particular exposures.
There are further implications, both for the system and for individual organisations. For example, increasing confidence in the ability to quantify extreme risks has been associated with the entrance of new capital providers in the insurance market and the commodification of insurance products. The implications of such developments for the competitive position of reinsurers and for underwriting practices need to be explored.
Furthermore, the institutionalisation of models, especially through the Solvency II process, makes model changes and experimentation harder in practice. This reduces the ability of insurers to learn from their models and can stifle innovation. Understanding this issue requires examination of the ways that models are used as tools in practice within the specific context of insurance enterprises.

 

3. Model risks
What risks arise from the use or non-use of models and the different ways in which they are used? Can we develop a qualitative tool for model risk governance? Quantitative tools for reporting / analysing model risk?
• Limitation of models and how to deal with uncertainty within the risk management framework
• Practical application of models – we understand that they do not represent experience. How can their output be integrated with what academic research and loss experience tell us?
• How can you use the data produced by models for better decision-making?
• Decision-making in the context of uncertainty or conflicting certainties
• Flood modelling: why do so many losses fall outside the modelled flood zones?

 

1. Growing the value of insurance and bridging the insurance gap. General insurance is a low growth industry, but in many classes of business, especially catastrophe, insured losses are only a small proportion of economic losses. How, therefore, can the role of insurance can be increased in the financing of risks? The topic could be split into an analysis of personal lines and/or commercial lines of insurance and advanced or underdeveloped insurance markets and a look at demand and supply factors.
To what extent could simple products with parametric triggers and fixed pay-outs expand insurance purchase? What additional services could make insurance seem less of a grudge purchase for most consumers? Could a greater sense of value reduce the propensity to inflate claims?

 

2. Taming the cycle for the general good? The insurance industry is in a prolonged period of soft conditions with abundant capital and relatively few catastrophes. Opinions vary as to what level of catastrophe would have an impact sufficient to push up rates. Typically, however, when the market does harden, it does so quite sharply and policyholders face a material increase in premiums, at least for two or three years. If the disaster is sufficiently large, then there may be failures. The apparently inevitable cycle poses questions about its effect on the security of society and whether regulators can or should intervene, for example by raising barriers to entry. Would cycle modelling help?

 

1. The introduction of probabilistic regulation frameworks in the mid-1990s led to an increase in the use of stochastic models in the industry to mitigate the flows in deterministic methods. Now, however, there are concerns that regulatory processes are slowing changes in models compared to a more principles-based environment. Is there a danger that model development could stall and so stymie innovation? Questions to consider:

• How will the industry change with the institutionalisation of models?
• Have models become too tied to regulation, making firms less ‘nimble’?
• Have other industries suffered from the same problem?

• Are regulators aware of the risk and what are their views?

 

2. With Solvency II, is the use of the VaR metric driving everyone to behave in similar, blinkered way thus creating a systemic risk?

Technological change affects the insurance industry from a business model/process and risk perspective.

 

1. Capital distribution. Today insurance risk bearing capital is concentrated in a few places and the insurance contract is generally long-term and illiquid. Is it possible to distribute capital around the world in a more dispersed way that could deploy funds very quickly? Could it be crowd sourced? Or is this simply an electronic coffee house? How important is the principle of indemnity? It would be worthwhile looking at ILS and projects that use parametric triggers and unit-based payments for lessons.

 

2. Business processes:
• How much human interaction is still necessary or desirable?
• Future of fortuity – how would near certainty about liabilities change the business model?
• Expert decision making systems
• AI/algorithmic underwriting
• How might the product offering change over time?
• How can business processes improve?

 

Models create a great deal of data. How can we digest that and interpret it? A lot of decisions are made on gut feel. What about expert systems that can help us make decisions? A decision machine?

 

3. Emerging risks What is the optimum organisational design? What are called emerging risks frequently have a high degree of uncertainty; they are threats whose impact cannot yet be reasonably assessed. While risk is foreseeable and amenable to process, sanctions and compliance, managing uncertainty requires a different culture. People need to be able to take decisions to respond quickly to changing circumstances and sometimes they fail in doing so. Instead of wanting people to build up expertise in a specialist area, they should move around the business so knowledge is shared and there is a cross-function approach.

 

4. The changing risk environment – what are the implications for insurers? How can they balance prudence with support for innovation? A safe risk appetite with competitive shareholder returns?
• Artificial intelligence
• Autonomous systems (including the “internet of things”)
• Nanotechnology
• Synthetic biology
• Genetic testing
• Big data

 

Communicating the cost-benefit of policyholder safety. Are we swapping fewer small events for a small number of catastrophic ones? Will net utility improve?