Singh: The 2005 hurricane season in the United States was the most costly since record-keeping began in 1851. Seven out of the ten most expensive hurricanes in US history occurred during the 14 months from August 2004 to October 2005.

This season was quiet so far, nevertheless, scientists still expect an active hurricane season in 2006 with 12 to 15 tropical storms, up to ten of which could become hurricanes.

However, overall activity is not necessarily a good predictor for expected losses. In 1992 hurricane Andrew was the first named storm in a quiet hurricane season, however, it did hit a high value concentration in a very vulnerable region.

Singh: We certainly hope not! Katrina ended up generating claims of about 30-40 billion euros. Despite being the largest insured loss event to date, Hurricane Katrina was by no means the worst case imaginable.

A "perfect storm", expected to hit the Caribbean and the US mainland with a probability of about once every 200 years, could generate insured losses well above 100 billion euros. However, at Allianz our pre-emptive exposure management ensures we can weather even such an event.

Singh: European storms are less intense than hurricanes, however, they are much larger in size. This can lead to accumulation of losses even across several countries, thus resulting in comparable loss potentials. However, compared to the US, buildings in Europe are usually better constructed.

In Asia we see the significant risks arising from earthquakes, typhoons and floods. Insurance penetration in these markets, however, is still lower than in the US or Europe. Therefore even in case of extreme economic losses we would not necessarily expect comparable insured loss amounts.

We do, however, have to keep in mind that the insurance coverages are rapidly catching up.

Singh: Yes and no. No single event can be directly attributed to a climate change trend, but the increased hurricane frequency in the last few years most likely reflects a global warming trend which is attributable to human greenhouse gas emissions.

On the other hand, the tropical sea surface temperature fluctuates mostly as a result of natural cycles; warmer phases correlating with more tropical cyclones. In the North Atlantic, this phenomenon is known as the "Atlantic Multidecadal Oscillation", which has demonstrated cycles lasting several decades. The latest upswing started around 1995.

Currently, the scientific community doesn't know to which extent the increased hurricane activity can be attributed to the natural cycle versus human greenhouse gas emissions. Nevertheless, both factors mean that the frequency and the intensity of tropical cyclones in the North Atlantic will continue at above-average levels for at least another 10 to 20 years.

In addition ongoing concentration of values along shorelines and more vulnerable technical infrastructure leads to ever increasing loss potentials.

Singh: Insurers are diversifying their risks more carefully - in 1992, several US insurers went bankrupt because the industry was virtually unprepared when Hurricane Andrew struck. This was a catalytic event for sophisticated catastrophe risk modeling and management.

Today, we underwrite more prudently in vulnerable regions and avoid excessive risk accumulation - both deductibles and rates have increased. The industry has also created incentives for home buyers and developers to promote disaster-resistant building structures

Singh: Traditionally, the insurance industry used deterministic risk models based on "probable maximum loss", or PML estimations: "How likely is a damage x, based on past experience?" Until the late 80s the US insurance industry did not anticipate an insured loss much greater than 1 billion US dollars. In 1989 Hurricane Hugo cost about 5 billion US dollars and become the basis for the new "PML". However, the insured loss caused by Andrew just three years later again turned out to be several times that much and could have been much higher had the storm made landfall a bit further north.

Since then, our knowledge has significantly improved and we can make use of more sophisticated catastrophe models. In our probabilistic models, thousands of realistic scenarios are developed, using so-called Monte Carlo simulation.

Singh: That actually really refers to the famous casino in Monaco – at the roulette table, each number has the same probability to be randomly picked by the ball. The Monte Carlo simulation technique basically also starts with equally distributed random numbers. By using distribution functions derived from historical data on all relevant parameters these simulations can then be transformed in events of different likelihood.

Singh: Models are by no means perfect. As Hurricane Katrina has shown, the industry had generally underestimated the US storm surge risk. Furthermore, in case of such huge catastrophes other factors than just wind speed and water height start to kick in and drive losses up. For example due to the evacuation of the city even mildly damaged buildings could not be repaired quickly leading to resultant losses such as mold.

Labor costs and prices for construction materials escalated and additionally buildings were looted as they were unattended. Such factors are hard to model and are not captured in detail.

Models always have to be used by experts who also understand their shortcomings.

For risk management also the question about the time horizon is important. Should we use the long term historical frequency or should we more reflect the current situation of elevated frequency in our models for evaluating the capital requirements? As previously stated, scientists are supporting the assumption that we will have increased frequency for at least the next decade or longer. This is now also reflected in our updated models.

We also learned that policymakers have a bigger influence on the effects of natural disasters than one would think - for instance, the design and maintenance of flood control systems or the enforcement of building codes impact vulnerability. Therefore the insurance industry supports strong building codes that must be rigorously enforced.

Singh: Yes, the CROs have strong cooperation. We regularly discuss risk issues within the Chief Risk Officer Forum, an industry network which counts 14 global insurance companies. As an outcome of this group, the Emerging Risk Initiative was launched in 2005 in order to raise awareness for new risks. In 2006 this initiative published a joint report on climate change and effects on US hurricanes which you can download from our webpage.

As with all content published on this site, these statements are subject to our Forward Looking Statement disclaimer, provided on the right.