By Steve Coates, Chief Underwriting Officer, at Pool Re
However there are still classes of insurance where end to end modelling is very difficult and one of these is terrorism. Effectively terrorism didn’t exist as a distinct class of insurance until 1993 when the effect of the IRA campaign in England forced reinsurers to exclude this cover from property treaties, which in turn would have forced UK insurers to exclude cover from commercial property policies. Fortunately the UK Government stepped in and provided the solvency guarantee for an industry mutual, Pool Re, to be created. Terrorism cover was provided by this reinsurance vehicle and UK plc could continue to buy terrorism cover, although they now had to pay a separate price.
Prior to 1993 commercial policies were silent on terrorism and cover was effectively included as part of the fire and explosion perils. Accordingly there was no historical loss data or rating tariffs so that the rates charged by Pool Re at the outset used a crude comparison with the fire rate. As the scheme paid claims through the 1990s these rates were moved up and down but there remained no real scientific basis for what was charged. In more recent times Pool Re has embarked on a modernisation programme and a key element of this is the adoption of modern modelling and actuarial techniques in our event modelling and pricing. However the models available to an underwriter of terrorism are very different to those in other perils, and in many ways that is why an element of government assistance remains necessary for terrorism in many countries.
The primary reasons for this are not only the lack of both historical event and future threat data, but the fundamental nature of terrorism itself. Modelling random events such as storm and earthquake is difficult enough but trying to predict and model something that is the product of a terrorist’s mind is near impossible. The effect of a storm hitting a certain place can be modelled but a terrorist theoretically has multiple attack vectors open to him; the bomb could be placed in different locations; be of different sizes or there could be two bombs. Moreover the terrorist can adapt his tactics to the activities of counter-terrorism agencies in preventing such attacks. Good security at one location means the terrorist seeks a substitute target that is not as well protected.
So how do terrorism underwriters approach this issue and what credibility can terrorism models have? In reality, as with some other low frequency, high severity events such as earthquakes, the best way forward is to invest in developing robust deterministic techniques that will outline the costs of a wide spectrum of event types should they happen and accept that any probabilistic view might be flawed. No-one predicted 9/11 and nor could they have since it had never happened before. The next big terrorist attack might be something equally unpredictable and thus outwit the contemplation of any current model. Equally the last big New Madrid earthquake was 1811-1812 and there have been more than 4000 earthquakes recorded since seismic equipment was installed there in 1974, happily most too small to be felt. We have no real idea when the next big one will come, but we can estimate its cost ($300bn).
Deterministic terrorism models have progressed significantly in the past decade. For Pool Re, a provider of significant aggregate capacity, significant investment is still required to vendor models if they are to achieve sufficient accuracy in urban areas. Flat radius modelling, which assumes a blast wave propagates out in a linear fashion, is flawed in densely built cities as blast waves are shielded and deflected by large and high rise buildings. The 3D model used by Pool Re gives very different modelled losses to models without the building layer applied, particularly in such places as Canary Wharf and the City. We must run a variety of loss scenarios and bomb locations to establish realistic disaster scenarios. We look at exposure density ie the exposure value in a postcode district, and the number of buildings in that postcode that might be considered a greater terrorist target than others.
The Pool Re scheme also covers CBRN (chemical, biological, radiological and nuclear) and given these are excluded perils in the property market and thus there are no credible models, we are building a model ourselves. This will be informed by threat and event analysis from ex-government consultants so we understand not only the blast and contamination elements, but also the wide area and access issues.
We are examining the credibility of applying probabilistic considerations but for now, the continuing rationale for schemes such as Pool Re is that the market does not think it can model robustly enough to warrant significant capital being allocated to terrorism.
As the threat changes and models develop, that may change. For now however models are a useful tool that can help us, but only if used wisely.
|
