Modelling the unthinkable

Insurance special – Terrorism insurance

insurancespecial-chart-gif

In November, the US government passed a law requiring all insurers to underwrite terrorism risk, forcing them to find a way to price this exposure. Could terrorism risk models be the answer? By Paul Lyon

When Boston-based catastrophe modelling company AIR Worldwide placed Washington DC in the top tier of its three-tier terrorism risk assessment at the end of last year, Lawrence Mirel, the District of Columbia’s chief insurance regulator, complained. Attempts to model terrorism risk are inherently unreliable, he argued, since they rely on insufficient data that purports to quantify human behaviour.

Unsurprisingly, modelling companies say this isn’t the case, arguing that it is in fact possible to measure the probability of a terrorist attack in the same way that they forecast the probability of an earthquake. Insurance companies have used natural catastrophe models for years. Now it looks like some are set to use terrorism risk models in much the same way.

Following the terrorist attacks of September 11, 2001, the majority of US insurers stepped back from providing terrorism coverage. But with the Senate’s passage of the Terrorism Risk Insurance Act (TRIA) in November 2002, all commercial property and casualty insurers are required to cover losses due to international terrorist activity within the US, forcing insurers to make difficult pricing decisions regarding terrorism risk. And modelling companies are consequently bracing themselves for an upsurge in business.

Besides AIR, there are two major terrorism-modelling companies in the US vying for this business. Both Eqecat and Risk Management Solutions (RMS) (both based in the Bay area of San Francisco), like AIR, have noted an upswing of interest in terrorism risk models since the passage of the TRIA.

The bill provides Federal reinsurance to insurance companies for losses over certain thresholds caused by foreign terrorist acts certified by the US Treasury secretary. It expires at the end of 2004, unless extended by the Treasury secretary – which insurers think is likely. It covers losses up to $100 billion a year and voids all foreign terrorism exclusions in commercial property and liability insurance policies. Under the legislation, insurers are liable for 7%, 10% and 15% of their annual commercial property and casualty premiums for the first, second and third years, respectively, before being eligible for federal assistance. Once the deductible level is reached, the federal government will cover 90% of losses up to the annual programme limit of $100 billion, leaving insurers with the remaining 10%.

The TRIA didn’t go as far as some insurers were hoping. Prominent industry figures, including Dean O’Hare, chief executive of Chubb Corporation, argued that the Bush administration should replicate Pool Re, the mutual created by the UK government to reinsure terrorism risks after the Irish Republican Army’s bombing of the City of London caused commercial insurers to withdraw cover. Some lament that the US government didn’t create a similar mutual because the 10% of any claim for which they are liable under the US system could be a sizeable amount.

The ability to model risk has therefore become important – even if the US government is providing a backstop of sorts. Enter RMS, AIR and Eqecat. The models, which debuted before the TRIA was passed, take different approaches to estimating exposures and concentrations of risk – but all attempt to provide a framework that can be used to price terrorism risk. RMS uses game theory to determine the likelihood, and targets, of terrorist attacks, taking into account 16 attack modes on 1,500 US targets, ranging from conventional explosives and improvised weapons to a range of chemical, biological, radiological and nuclear scenarios. Developed in the 1950s by the Rand Corporation – the strategic think-tank – game theory research involves studies of the interactions between groups of people attempting to determine the optimum strategy for dealing with a given situation or confrontation. In addition to the probability of various attacks, the model estimates the insured losses stemming from property damage, business interruption and human casualties.

But any model, however logically designed, must ultimately involve some expert judgement, says Peter Ulrich, managing director of enterprise risk at RMS in Newark, California. “Accordingly, RMS sought input from a team of advisers, including representatives from the Centre for the Study of Terrorism and Political Violence, Jane’s Consultancy and Rand Public Safety and Justice, as well as international relations professors from the University of St Andrews, Scotland,” Ulrich says. “Without this knowledge databank at our fingertips, the model would have been redundant.”

AIR was the first on the market with a commercial terrorism risk model, in September 2002. Like RMS, it employed a team of counter-terrorism specialists with experience at government agencies such as the FBI, CIA and the Department of Defence to develop the model. But, unlike RMS’s use of game theory, AIR estimated the frequency and severity of attacks using the Delphi method, a group decision process about the likelihood that certain events will occur, developed by the Rand Corporation at the start of the Cold War. The Delphi method has since been used to generate forecasts in a variety of areas, including intercontinental nuclear warfare and technological change.

AIR has sold the model to at least six insurance companies, and says it has had more than 100 enquiries. The model estimates the financial impact of insured property and workers’ compensation losses from potential terrorist attacks in the US. It analyses threats posed by domestic extremists and international and state-sponsored terrorist organisations, as well as loosely affiliated networks, examining a range of potential conventional attacks, including air crash and bomb blast, and their impact on insurers’ and reinsurers’ books. AIR is also examining non-conventional weapons damage, including chemical, biological, radiological and nuclear, and plans to incorporate these threats into the model by early 2004, says Jack Sequist, product manager for AIR’s terrorism model.

Although most examples of terrorism risk modelling revolve around property exposures, AIR has also been successful in selling its solution to a life insurer. At the end of January, US-based insurer UnumProvident Corporation asked AIR to prepare terrorism and earthquake catastrophe loss analyses on its life insurance portfolio. Using its terrorism loss estimation model and US earthquake model, AIR developed a probabilistic loss and concentration of exposure analysis for three insurance lines, including life, accidental death and dismemberment, and disability.

Oakland-based Eqecat decided to incorporate both the Delphi method and game theory into its terrorism risk model. Robert Healy, Eqecat’s senior vice-president, claims Eqecat has a far wider-reaching model than its two competitors, with a fully probabilistic stochastic set of 15 million events covering all 50 states plus the District of Columbia – having around 150,000 identified target sites with high probability of attack. The Florida-based National Council on Compensation Insurance, a shared services organisation for the workers’ compensation industry, has already selected Eqecat’s product, as have other undisclosed insurance companies.

Prices for the three models can range anywhere from the tens of thousands up to $750,000 – according to the number of policies involved. Typically, insurance companies model client exposures on request, says Sean Mooney, chief economist for Guy Carpenter, the New York-based reinsurance division of financial services company Marsh & McLennan. But companies such as his, which has extensively researched the benefits of using game theory, are also modelling terrorism risk themselves – thereby providing a source of competition for the modellers.

In November, for example, Aon Corporation’s catastrophe modelling affiliate, Impact Forecasting, completed development of its property and workers’ compensation terrorism exposure analysis tools for US insurers, large employers and property owners. Aon says it used street address and zip code information to assess aggregations of property and workers’ compensation exposures and their relative proximity to a database of around 5,200 high-visibility potential targets and more than 1 million less visible but important landmarks or facilities.

Swiss Re has also developed its own internal risk model. Andreas Schraft, Swiss Re’s New York-based product manager for property reinsurance, says Swiss Re started to develop its model shortly after September 11, but will not say whether the reinsurer employs game theory or not. He does, however, admit to having an interest in the work of the three main modelling firms, having combined Swiss Re’s own model with that of one of the external firms. “We are conservative enough with our pricing to be confident in our model’s ability,” says Schraft. “But unless we have an event that allows us to test the models there will always be an underlying scepticism. However, there was scepticism about earthquake modelling when that first began, but recent tragedies, such as the earthquakes in Mexico, have proved that the model was capable of producing accurate loss analyses.”

Pitfalls

Rating agencies have mixed feelings about terrorism risk models. Steven Dreyer, New York-based managing director of Standard & Poor’s (S&P) insurance ratings, says the adoption of terrorism risk by commercial property and casualty insurers in the US casts a shadow on ratings prospects, compounding S&P’s negative outlook on the industry. Indeed, the TRIA creates numerous pitfalls and grey areas for insurers to contend with, Dreyer adds.

“The legislation amounts to a command from government for insurers to get back into the pool of terror risk,” Dreyer says. “But while terrorism risk models provide a good starting point for assessing risk, claims for the effectiveness of terrorism pricing models are wildly exaggerated. Bathed in an aura of invincibility by such phrases as ‘fully probabilitistic’, they are at best a blunt instrument that could nevertheless lull insurers into a false sense of security.”

Indeed, there is a worry among rating agencies, and insurance professionals alike, that insurers will never know how to price terrorism risk and could use government backing as a crutch for taking on exposure irresponsibly. “And if insurers are pricing something they do not fully understand, they will tend to undercut each other,” Dreyer adds. “If they are prudent, insurance companies will continue to price terrorism coverage at high levels because, unlike hurricane models, these new terrorism models don’t provide historical information, meaning that it is impossible to know what the right levels of pricing are.”

But as Moody’s analysts note, terrorism risk insurance must be made affordable if the TRIA is to be deemed a success. So far, Moody’s estimates that insurers are adding premiums of up to 75% for TRIA-mandated coverage. Whether the availability of terrorism risk models will help readdress this trend remains to be seen.

And as one rating agency analyst notes: “The cheapest and perhaps most sensible way for insurance companies to reduce concentrations of terrorism risk is to make sure they limit their exposures in vulnerable areas. For a price models can tell you which these areas are. But common sense will also provide similar results.”

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact info@risk.net to find out more.

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here