INSTITUT Veolia Environnement

  • Home
  • Report n°3: Financial protection of critical infrastructure

Report n°3: Financial protection of critical infrastructure

The limitations of quantification models

The occurrence of particularly devastating and large scale natural disasters since the end of the 80s has encouraged a growing number of stakeholders to recognize that there is a need for developing new methods of quantification to measure the impact and likelihood of such disasters.

On the basis of mathematical analyses combining parameters characteristic of the natural phenomenon itself and information concerning the structures exposed to it (buildings, airports, stadiums, infrastructure networks, etc.) new models for risk quantification supply a priori assessments of potential damage (to life and limb or business interruption). These models were a great success in the 90s, particularly by offering insurance companies a measurement of their portfolio's exposure to these hazards. The market for quantification of catastrophic risk is presently dominated by three leading firms, AIR, EQECat et RMS.

In the year following the 9/ 11 attacks, the three firms developed new models to quantify the terrorist risk. However, these new instruments are of limited use.

In effect, whereas traditional quantification methods are based on objective risk evaluation criteria, evaluation of the probability - in the mathematical meaning of the word - of a terrorist attack is much more complex, and more often than not, impossible. The lack of available data and dynamic uncertainty are such that only by reliance on expert opinion can plausibility be factored into these quantification models.

However, experts also have their limitations. They all tend to specialize in a given subject, a particular terrorist group, or a specific type of weapon. Pioneer work undertaken by Linstone and Turoff in the 70s demonstrated that this could lead to an "illusion of expertise": each expert can only express himself through the prism of his own field of knowledge, which by definition is biased as compared to the entire subject, so that they may all be right about their own domain whereas the end result is faulty recommendations (Linstone et Turoff, 1975)(26).

Another bias of complete reliance on expert advice is well known. Their judgment is vastly influenced by recent events. For example, after the 9/11 attacks, the trend was to focus on the possibility of further attacks using commercial aircraft and to underestimate the possibility of attacks of another kind. At the other extreme, if several leaders of terrorist groups come to be arrested by the authorities, the tendency could be over-optimism and an under evaluation of the resolve of other terrorist groups.

The first generation of quantification models emerged two years ago and they are mainly used in the United States. They certainly provide a way of gaining better understanding of the exposure of a specific company or of a geographic area, but their real limitation is surely that they cannot offer a "probabilistic" approach of terrorist risk. In other words, these models can serve to determine potential risk but not expected risk (there is therefore no "scientific certainty"), which is the basis for pricing insurance coverage.

In this kind of context, it does seem difficult to rely on the traditional cornerstones of the insurance industry (actuarial computation, premiums indexed on the statistical risk incurred and on preventive efforts deployed by the policy holder, similarly modulated deductibles, etc.) A better course would be to use to advantage concepts based on global risk sharing methods based on a high degree of mutualization.

(26) It is worth noting that this discussion on the role of experts also exists in other environments whenever there is uncertainty on the entire set of possibilities; this is the case in particular for the debate on the principle of precaution; see Godard (1997).