Beyond Exposure Modeling

Catastrophe models are helpful, but they should be supplemented with a systematic, continuous and comprehensive approach to catastrophe risk management

In the face of frequent, potentially large and multi-line catastrophe losses, insurance companies need a systematic, continuous and comprehensive approach to catastrophe risk management. Catastrophe loss estimation models – software programs that simulate thousands of years of losses from potential future disasters – play an important role in this process, but they are evolving and have limitations that should be understood and overcome.

Due to the complexities involved in managing catastrophe risks, a diverse group of specialists is required to design, develop, and implement a process that meets a company’s objectives. At Aon Re, for example, the catastrophe risk management advisory team consists of catastrophe risk analysts, actuaries, meteorologists, seismologists, engineers, financial/accounting/tax professionals, rating agency and regulatory specialists, reinsurance structuring and market experts and a capital markets solution team. It is imperative that these experts work together as a team to provide cohesive risk management solutions and sound advice.

Catastrophe risk management is a continuous and recurring process. The goal of this process is to create an optimal risk portfolio that meets a company’s objectives. The process of assessing, financing, and reducing the risks is ongoing and roughly follows a one-year recurring cycle.

Within this cycle, each step has its own business objectives and is therefore a significant and comprehensive activity. Steps may include assessment through exposure modeling, rate-setting, reinsurance treaty design and placement or portfolio optimization. Insurers are assessing exposure accumulations and estimation of potential losses more frequently because they are now facing a growing or changing portfolio of risks, new developments in the industry and significant changes to catastrophe models.

The point is, catastrophe risk management is not just about consulting a computer model. It is a broader process requiring dedicated time, personnel, expertise and commitment to a formal process.

COMPREHENSIVE STRATEGY

Identifying

The need for a comprehensive approach to analyze catastrophe risk has never been clearer. In 2005, North America and the Caribbean experienced record number of catastrophes. Fifty-two natural catastrophes resulted in more than 3,431 deaths, 1.7-million damaged structures, and roughly US$68 billion in insured losses, according to Aon Impact Forecasting. Worldwide, nearly 6,000 natural and technological disasters between 1995 and 2004 resulted in an estimated cost of US$740 billion in damages and nearly 900,000 deaths, according to the 2005 World Disasters Report by the International Federation of Red Cross and Red Crescent Societies.

A comprehensive catastrophe risk management process starts with the identification of risks. Three things contribute to the probability of a catastrophe loss: peril or hazard, exposures to that hazard, and vulnerability or susceptibility of the exposure to that hazard. The identification phase is best accomplished by a thorough review of insurance policy forms to determine what is and is not covered by a policy. Also to be identified are catastrophe events that might trigger claims in areas where the company writes those policies.

During this stage, information regarding the historical frequency of catastrophic events and their potential severity of losses is collected to help prioritize assessment activities that follow. A company’s historical experience is a good starting point to identify critical geographic regions, lines of business, and perils.

One should not forget, however, the past does not always indicate the future. Each policy review should conclude with an answer to the question: “Are we missing anything that can lead to big surprises down the road?” One important benefit of this process is a clear understanding of what is and isn’t covered in the company’s policies. Another benefit is the opportunity for policy language improvements.

Available Tools

The next step is to consider what tools are available in ordet to model the insurance coverage and catastrophe perils that were identified in the previous step. Current inventory of catastrophe loss estimation models include hurricane, earthquake, fire following earthquake, landslide, liquefaction, tornado, hail, straight-line wind, winter storm, wildfire, flood and terrorist attacks. However, not all models cover all regions of the world: there are differences among the models used. Since catastrophe modeling is not an exact science and is subject to a lot of uncertainty, it is advisable to estimate potential losses from all available catastrophe models. Two key considerations in modeling catastrophe exposures are:

* Ensuring the accuracy and completeness of the input.

* Ensuring the accuracy and completeness of the output.

The clich “garbage-in, garbage-out” holds true for catastrophe modeling. Input data must be complete, current and accurate. Business units must verify and validate the policy data before they are sent off to risk analysts.

Catastrophe models are very sensitive to certain inputs. If the exact location of a risk, its insured values and limits or its structural and occupancy characteristics are inaccurate or missing, the models’ loss estimates can be way off the mark. A side from the obvious benefit of having complete and accurate information for loss estimation purposes, complete and accurate data also helps differentiate a company in its risk-financing transactions.

Interpreting Results

How a catastrophe risk analysis is carried out and what reports are generated from a model is essential to understanding the key drivers of potential losses. Typically, the key factors in successful exposure management include loss estimates by geographic sub-units (region, state, province, county, etc.), lines of business and products. Even so, potential loss estimates viewed by meaningful categorization of catastrophic events over entire areas of impact provide useful information about various loss scenarios and their severity. Viewed through this prism, such loss estimates are easier to comprehend.

In order to obtain meaningful results from models, users must first understand that catastrophe models are limited in scope. For example, not all plausible scenarios of damage are considered in a model, and models do not include all coverages in a policy. Model users should be prepared to make suitable adjustments.

Another example relates to hurricane models which do not include, among other things, loss-adjustment expenses, tornado-related damages, inland flooding, extended or contingent business interruption, pollution, food spoilage and debris removal. A seemingly insignificant add-on to a policy could result in a significant loss after a catastrophic event such as Katrina. For example, homeowners suffering little or no hurricane wind damage could still make millions of dollars’ worth of food spoilage claims due to power outage. If 20,000 homeowners each make a US$500 food spoilage claim, that adds up to US$10 million to the total claim amounts. Unfortunately, hurricane models available today do not predict the extent of power outage or food spoilage losses. In order to fill in such estimation gaps, adjustments to loss estimates are necessary.

CONCLUSION

Risk financing decisions, such as structuring and pricing of an excess of loss reinsurance treaty or a catastrophe bond, are increasingly relying on catastrophe loss estimation models. A rational reinsurance transaction will have answers to:

* How much reinsurance limit is required in light of the volatility that is transferred.

* What the expected price of that transaction is.

* Whether
or not the transaction is accretive or dilutive to the insurance company’s cost of capital.

Once a reinsurance program is in place, the cost of reinsurance can be allocated down to the individual policy level by using the catastrophe model output.

Commercially-sold catastrophe models have come a long way since their development in the late 1980s. They are now available as multi-purpose tools providing a basis for making risk management decisions for a wide range of catastrophe perils. Catastrophic events such as hurricane Katrina have provided opportunities to authenticate the accuracy of these models and determine how they should be used in light of their limitations. Model improvemets customization can address some shortcomings but a comprehensive and multi-disciplinary risk management approach is necessary to achieve a company’s risk management objectives.