The process of model development must involve the user, capturing his perceptions of their business world, for integrated risk modelling to become a true business instrument, argues Oliver Peterken.

Positioned above asset liability modelling and statistical techniques, integrated risk modelling (IRM) is portrayed as the ultimate evolution of modelling for insurance and reinsurance businesses. Consequently, most firms of consulting actuaries, large reinsurers and reinsurance brokers, as well as various consultancies, now offer risk-return modelling services.

However, a noticeable feature of many IRM models is their lack of transparency and there are frequent criticisms from users as to the lack of ability to evolve and expand.

This, as well as their failure to account for poor strategic decisions and unforeseen losses, may contribute to their limited, if increasing, use.During the past two years there has been a distinct change in tone of the proponents of IRM who suggest that they are no longer a specialist tool for tackling unusual and isolated problems, but methodology for tackling the strategic issues facing an insurer.

The growth in support for the use of IRM models is explained by both the increasing pressures under which insurers find themselves from shareholders and analysts, and developments by their banking competitors where risk modelling techniques have quickly become the norm and are integrated into every-day trading and risk management activities.

This raises the following questions:
• How widespread is the use of such techniques, how are they used and what types of decision are they supporting?
• How “totally integrated” are these models in terms of the risks facing an insurer they encompass?
• How good are these models when compared to accepted standards for model development?
• How well do such models tackle the important issues facing insurance company chief executives?
• What is needed for further development into the next generation of models?

The use of IRM models falls under the classification of “decision support” in systems terms, that is providing information to management to improve the effectiveness of their resource allocation decisions. The types of decisions IRM models have assisted include asset mix strategies, capital allocation, cost of risk capital, insurance risk control, acquisition/divestment, price setting and risk transfer. It can, therefore, be said that IRM models have the potential to support the most important decisions made by insurance company senior executives.

However, while IRM models can achieve all these things, they are unable to take into account the effects of a poor strategic decision, or the effects of an unforeseen loss.

In addition, IRM models rarely tackle operational, reinsurance and regulatory risks which suggests that “total integrated risk” or “enterprise risk” models fall short of encompassing the totality of insurance enterprises' risk. What IRM models do, however, is model the risks to assets and liabilities which form the basis of every day risk to an insurance entity.

From a user's perspective, a good model needs to be flexible to adapt rapidly to changing requirements as a business' situation changes and believable, by which is meant that the model can be independently audited to verify that it functions as specified and that a user can understand the workings and is prepared to use the model's output to support critical business decisions. Further important characteristics are transparency, evolvability, parsimony, consistency, scalability.

Transparency enables users to see how their business is being modelled and thus gains their acceptance of the model's results. Evolvability enables a model to grow in complexity and is required because an effective development strategy is to build a simpler version to which more complexity is added over time. Parsimony prevents theories becoming too complex and, all things being equal, values fewer assumptions over more. Consistency means that the model explains logically the relationships between variables. Whereas evolvability refers to the expansion of the logical structure of the model, scalability applies to the model's capacity to capture more scope and depth of the “real world”. For example, a scaleable model can be readily extended to include more product lines and more sources of risk.

If IRM models are to achieve the status of being instruments, as opposed to technical tools, for insurance executives, they must become more acceptable to the users. This can only be done by making models more transparent.

At their core must be a process of model development which involves the user, capturing their perceptions of their business world. The new generation of programming tools based on object-orientated languages, which capture the nature of relationships between items, offers exciting possibilities for moving IRM models to the centre stage of insurance company management.

Oliver Peterken is managing director of the consulting practice at Willis.