Will Europe experience its own Katrina? European insurers need to prepare for winter storm losses far beyond those experienced to date, warn Yorn Tatge and Peter Dailey.
Extratropical cyclones are the major driver of weather-related insured losses in Europe. These storms are strongest between October and March and can bring damaging winds and heavy precipitation across northwestern Europe. The cyclones annually cause an average of EUR1.5bn in insured losses, but extreme storms can cause far higher levels of loss. For example, Winter Storm Daria alone cost insurers EUR6bn in 1990. And the trio of storms in December 1999 - Lothar, Anatol, and Martin - together tallied EUR12.3bn in insured losses at the time.
Many within the industry believe Winter Storm Lothar, for example, provides a good benchmark as a European worst-case scenario and that the frequency of such large loss events is too low to be of great concern. This view stems from basic factors in our nature. As humans, we tend to discount the possibility of low frequency events when we have not experienced something similar. Even when we recognise the possibility of a low frequency scenario, we tend to envisage only a subset of all possible scenarios, leading us to underestimate its probability. Lastly, we generally discount data that contradicts our existing risk views. Thus, Katrina's $41bn insured loss caught most of the industry off guard.
Recipe for extremity
There are three ingredients that come together to make an extremely damaging winter storm: size, strength, and location. First, the storm must be large. Daria, which stretched over six countries, is an example of such a large European winter storm. Given current property exposures, were Daria to reoccur today with the same footprint, AIR estimates that insured losses would be in excess of EUR10bn.
Second, the storm must be intense. In December 1986 a very intense European extratropical cyclone with the lowest recorded central pressure in the Northern Hemisphere struck Greenland and Iceland. Few know of this storm, however, because it failed to reach the population centres of Europe. This brings us to the third element: location. To result in extremely large losses, the storm's footprint must coincide with a high density of insured properties. For European winter storms, this correlates to a corridor running from the UK and northern France, across the Benelux countries, into Denmark and northern Germany.
Europe's largest historical losses have resulted from storms that met only one or two of these criteria. At some point, the continent will experience a storm where all three coincide. A storm as large as Daria, with an intensity similar to that of the 1986 storm, and which follows a path over Europe's highest density of insured properties is a meteorological possibility. AIR's analysis shows that such a storm would be capable of causing insured losses in excess of EUR40bn.
The probability of such a storm is not as remote as one might think. For example, the annual loss exceedance probability of a storm causing a EUR10bn insured loss or greater is about 7%, according to the AIR extratropical cyclone model for Europe. Examining the next ten years, the probability of such a loss grows to more than 50%. For the EUR40bn loss scenario described above the annual probability is about a half a percent, rising to almost 5% over ten years. Clearly, companies cannot dismiss losses of this magnitude or likelihood.
Climate change in the mix
The ferocity of the 2004 and 2005 hurricane seasons, and the possible ties between elevated sea surface temperatures and hurricane frequency, have caused many to consider whether climate change holds any consequences for European winter storms.
In 2005, the Association of British Insurers (ABI) conducted a study aimed at probing this question. While acknowledging the great uncertainty surrounding the potential influence of climate on extratropical cyclone activity, ABI projected that a doubling of atmospheric carbon dioxide levels by the 2080s could potentially lead to a 20% increase in the frequency of the most intense European winter storms (those with wind speeds in the 95th percentile or above). Additionally, the frequency at which such storms track further into Western and Central Europe could potentially rise by 10%.
A sensitivity analysis performed by AIR at the behest of ABI indicated that a 20% increase in extreme storm frequency could translate into a 35% rise in average annual insured winter storm losses. Insurers could also face a 5% increase in the insured losses caused by the most extreme storms, those with exceedance probabilities in the range of 0.5%-1%.
It should be noted that the ABI study did not consider impacts on less intense storms due to limitations in the data on the effects of climate change on more "garden-variety" extratropical cyclones. Because storms of lower intensity occur more frequently, the full impact of climate change on European winter storm losses could potentially be far greater. Also, the ABI focused solely on wind-related losses. The costs of flooding associated with storm surge and precipitation were not considered, even though evidence suggests that they too could increase with climate change.
Managing the risk
To fully realise the value offered by detailed catastrophe risk models, the analysts running the models and the underwriters making risk management decisions based on the output must have a thorough knowledge of how the models work and how to interpret their output (See table 1). European insurers need to prepare for winter storm losses far beyond those experienced to date. The purpose of catastrophe modelling is to provide insurers with a realistic and robust distribution of potential large losses before they occur, not to mimic market prices or the industry's perception of the risk. A EUR40bn winter storm loss will occur in Europe. It is only a matter of when.
For more details of the ABI study, "The financial risks of climate change", go to www.abi.orguk/climatechange.
AIR warns of EUR40bn European windstorms
Re/insurers must do more about climate change
European reinsurers hoping for the best
TABLE 1: BEST PRACTICE FOR EFFECTIVE CATASTROPHE RISK MANAGEMENT
Capture detailed exposure data - High resolution catastrophe models require high resolution data to produce accurate loss estimates. However, European insurers do not generally gather detailed property data. At a minimum, companies need to collect address-level - not aggregate - data on their portfolios. Capturing building characteristics would improve accuracy even further.
Ensure high data quality - The single largest driver of loss estimate accuracy is data quality. AIR Worldwide undertook an exposure data survey of US insurers in 2005 and found significant gaps across the industry in the quality of the data supplied for catastrophe modelling. Models will inevitably underestimate potential losses if key pieces of exposure information (eg replacement value, occupancy class, construction, location) are incomplete or incorrect. For example, if the replacement values for a portfolio's exposures are underestimated by 50%, then losses will be underestimated by at least that much.
Benchmark results for reasonability - A lack of reasonability checks can also lead to gross underestimation of losses. Company loss estimates should be benchmarked against industry loss distributions. An insurer with a 1% market share, for instance, should expect something on the order of 1% of industry losses. Significantly lower losses would suggest the need for further scrutiny of the exposure data.
Consider loss exceedance probabilities rather than return periods - Instead of thinking about catastrophe modelling results in terms of return periods (eg 100-year return), think of exceedance probabilities (eg a 1% annual probability). Placing losses in this context will help avoid the misleading notion that multiple low probability losses cannot happen in back-to-back years.
Run analyses regionally, not separately by country - Insurers who estimate loss probabilities by country will invariably underestimate the risk to their portfolio as a whole. Lower probability risks from individual nations push up the loss levels for the region as a whole. Since extreme windstorms tend to impact multiple countries, such perils should be analysed regionally, not separately by country.
Manage to a loss exceedance probability, not to an event - It is not the risk of individual events that companies need to manage, but rather the level of loss that will cause financial impairment. This could come from single event losses, multiple losses within a year, aggregation of losses across perils, or multiple years of catastrophe experience. Companies should focus on the aggregation of risk across perils and over a planning horizon.
Manage exposure portfolios using the full loss distribution - Companies must look beyond the single loss level or 1% exceedance probability; it is outside this range that the risks of extreme winter storms exist. Insurers should go so far as to investigate their tail-value-at-risk (TVAR), their modelled TVAR-to-surplus ratios for various loss probabilities, and the regions and perils driving their large loss scenarios. Solvency II regulation is pushing insurance companies in Europe to look at their catastrophe risk in this more discerning way.
Judge models based on their scientific merits, not on whether they agree with market prices - Prices are set by market conditions and often do not adequately reflect the underlying risk. Catastrophe models are not intended to match or justify current market prices, but to help companies prepare for large losses before they occur. Therefore, when current market prices are not in sync with a scientifically credible, peer-reviewed catastrophe model, companies should use the model to get a credible assessment of their risk, including such activities as optimising their risk capital, determining how much reinsurance to buy, and how much capacity to allocate to a region.