Casualty catastrophe models, the younger siblings of models that help insurers and reinsurers underwrite property catastrophe exposures, need more development before they are widely embraced by reinsurers.

Developing models for casualty catastrophe exposures is more complicated than property exposures. For now, casualty cat models are better suited to primary underwriting operations with significant policyholder information than reinsurance, experts say.

“Casualty cat models are still in their infancy compared to property models,” said Andy Rapoport, a Minneapolis-based managing director and casualty modeling expert with Aon Benfield Analytics, in an e-mail.

David Wong, a director at PricewaterhouseCoopers L.L.P. in London, agreed and said casualty cat models have not evolved to the level of property models. Unlike property models offered by third parties that can be used by various insurers, catastrophe models have to be structured to specific insurer portfolios and often are developed in-house, he said.

“Property models have been developing since the late 1980s,” said Mr. Wong, who also is a member of the catastrophe modeling group at the London-based Institute of Actuaries. “Most of those being used on the casualty side are proprietary models.”

Interest in casualty cat models comes mostly from primary insurers with large exposures and potentially significant concentrations of risk, Mr. Rapoport said. “These companies have typically developed sophisticated approaches to managing property risk and are now trying to deploy the same technologies to understand their casualty exposures,” he said.

Guy Carpenter & Co. Ltd. will launch the second version of its casualty catastrophe model in the second quarter of this year, said David Lewin, London-based managing director of casualty for Europe for the reinsurance broker.

While Guy Carpenter's model theoretically could be used by reinsurers, it is better suited for insurers, Mr. Lewin said. “It requires a significant degree of granular detail, so it is more appropriate for insurance companies” that can provide necessary account-specific detail, he said.

“Reinsurers would have to have that same level of detail in their portfolios,” said Mr. Lewin. While some do, reinsurers generally do not collect the specific information that could be effectively used by the casualty model, he said.

“Reinsurers do not have the direct information on exposures that insurers have,” said Mr. Wong. “They are just not close enough to the exposure data and would have to rely on the insurers for that information.”

Steve Mildenhall, head of Aon Benfield Analytics Americas in Chicago, said casualty models are based on the same underlying paradigm as property catastrophe models, using hazard, vulnerability and potential losses as components of the modeling process.

“The hazard component produces a range of possible events that could cause a casualty catastrophe,” he said in an e-mail. “The vulnerability module describes how each hazard would affect different classes of business,” and “the loss component overlays an individual company's actual exposures and policy limits to convert vulnerabilities into a range of possible losses.”

Aon Benfield has developed casualty models for workers compensation, directors and officers liability, and other casualty exposures, said Mr. Mildenhall. The models are “in flux” with “considerable ongoing development and customization around each individual client application,” he said.

Whereas Guy Carpenter's first casualty cat model, launched in 2008, focused mainly on establishing relative measures of catastrophe risk within a portfolio to build a casualty catastrophe profile, the updated version will allow users to generate loss scenarios and calculate a distribution of loss severity, Mr. Lewin said.

“The scenario question is a very important part of what we are doing here,” he said. “We need to assess the potential impact on an insurance portfolio of large, single incidents as well as systemic events.”

Developing the model has been an exhaustive process of examining insurer portfolios to identify and connect all the risk components that are rolled into the modeling process, Mr. Lewin said.

Modeling casualty exposures differs from property because the kind of aggregate values that are considered in property risks are not available for casualty exposures, Mr. Lewin said.

Instead, a casualty model must consider a multitude of aspects, including the jurisdiction, trading and service relationships among different parties in determining the likelihood and potential impact of a catastrophe loss, Mr. Lewin said.

Historical data is useful in modeling the potential scope of casualty claims, he said. For example, the collapse of a terminal roof at Charles de Gaulle International Airport in 2004 involved more than 200 potentially responsible parties, Mr. Lewin said. Collecting such information “helps us gauge the realistic size and spread of such a disaster scenario,” he said.

While it is important to consider historical information in constructing models, a model has to look into the future, he said.

In a product liability claim, for example, the potential impact on the entire supply chain has to be understood. That's why a model must include information to allow projecting the impact of a claim according to complex relationships between multiple parties, Mr. Lewin said.

Zurich Financial Services Ltd. has used casualty catastrophe models since 2003, primarily on risks in the United States, the Zurich-based insurer confirmed. The models, provided by an unidentified third party, are used to assess and manage exposures within certain tolerance levels and to help plan reinsurance strategies, Zurich said.

“Most major insurers are using them,” Mr. Wong said of casualty models. “Compared to five years ago, we are seeing greater use.”

The models give insurers some guidance they can use in underwriting, but they have not made a significant impact on underwriting decisions, Mr. Wong said.

“Until there are industry-accepted casualty catastrophe models, these sort of insurer-developed models will have an impact on insurers themselves,” Mr. Wong said, but will not cause significant industrywide changes in underwriting behavior.