Large catastrophes often reveal unexpected correlations and loss amplifications. Nick Thorpe examines whether these correlations of risk could catch out insurers, reinsurers and their investors

The events of September 11 and Hurricane Katrina served as a wake-up call for insurers and reinsurers around the world. Both generated huge losses across portfolios and caused numerous unusual correlations that had not surfaced during the risk modelling process.

“The World Trade Center loss highlighted that you can have things as diverse as aviation and marine, personal accident and property, all having extreme losses,” says Simon Margetts, senior manager of research and development in insurance and actuarial advisory services at Ernst & Young LLP. “Normally you wouldn’t expect these lines to have any common risk drivers.” Similarly, while the strong winds of Hurricane Katrina caused substantial property damage, it was the associated storm surges and flooding that caused significant losses across lines of business. This contributed to losses in lines as diverse as cargo, inland marine and onshore energy, according to a Risk Management Solutions (RMS) report written shortly after the hurricane hit New Orleans.

However, what sets a megacatastrophe apart from a regular catastrophe is its knock-on affect on the wider economy. Hurricane Katrina, for example, affected petroleum supplies in the Gulf of Mexico and hit prices of crude oil and natural gas globally. According to the RMS report, “the shortage of refined petroleum even led to increases in the cost of gasoline worldwide as producers sought to divert supplies.”

As capital market investors continue to demonstrate their penchant for investing in catastrophe risk (through vehicles such as catastrophe bonds, sidecars and industry loss warranties), there are signs that risk correlations could affect their portfolios too, linking their equity investments and insurance investments (despite the fact capital market investors generally believe insurance to be uncorrelated to their other investments).

Catastrophe modellers warn that we should expect more frequent and more intense hurricanes like Katrina in the near-term whilst at the same time insured values in coastal “peak zones” are going up every year. Given this increasing risk of major losses – both economic and insurance losses – are the hedge funds, private equity firms and institutional banks investing in the industry risking a “double hit” to their portfolios? If they are, how could this affect the insurance industry’s newest source of capital?

Off the scale

RMS defines a megacatastrophe as being “characterised by damage on a massive scale that gives rise to non-linear loss amplification, correlation and feedback”.

Certainly, recent high-profile disasters such as Katrina have served to bring “megacats” to the front of insurance and reinsurance director’s minds as they seek ways of cushioning the blow from future events. The Katrina, Rita, Wilma combination caught many insurers and reinsurers off guard, exposing deficiencies in cat models and business pricing that rocked the industry to its core. Katrina alone is now estimated to have cost the property casualty insurance industry $41bn from around 1.7 million claims (see page 31).

Traditionally, insurance companies have dealt with this risk by examining “correlation coefficients” which uses statistics to anticipate interaction between portfolios and lines of business. However this process is limited to indicating the likely average correlations between lines and allows for no variation on the magnitude of an event.

According to Dr Jayanta Guin, senior vice president of research & modelling, AIR Worldwide, there is now a much greater chance of major losses through megacatastrophes. “In the US, we estimate that in any given year there is a 1% chance that the industry will experience a catastrophe loss of over $100bn. With current exposures and increases in insured values, we expect this estimate to double every ten years, so by 2017 we would expect that figure to be more than $200bn.”

Hurricane Katrina was a notable megacatastrophe due to a number of characteristics that caused losses to be amplified outside of the normal spectrum of catastrophe effects. According to RMS, “In super catastrophes like Katrina losses become nonlinear, ie the scale of the event itself causes losses to increase further.”

Katrina exposed unusual losses that were historically treated as “rare exceptions” by RMS, causing the modellers to reconsider knock-on events following catastrophes, such as fires following earthquakes and demand surge. Whereas in the past these processes were surplus to the core model, Katrina and other megacats have seen modellers integrating these additional losses into the methodology. “Companies need to focus on the high quality collection of information and data, and the anticipation of events that could happen,” agrees Robert Muir-Wood, chief research officer at RMS.

Wider impacts

This loss amplification is at the heart of the potential risks facing new capital market investors, which are currently even more enamoured with insurance following massive returns in 2006 (thanks to a benign season and sky-high property catastrophe rates). Since Katrina, $35bn has been raised, most of it through the capital markets, according to PricewaterhouseCoopers. This new source of capital is attracted by the potentially lucrative returns the market can offer, particularly in the high risk/high return catastrophe business. These investors have replenished and transformed the industry, demanding higher levels of transparency and tighter risk management as their understanding grows.

“The size of the earthquake caused the Japanese stock market to tumble, losing over a thousand points in one day. It was this financial meltdown that caused the collapse of Barings Bank as trader Nick Leeson speculated vast sums on Japanese derivatives

But will they stay if made to pay out for major catastrophe losses, which hit more than just their insurance investments? Equity cycles and insurance cycles are correlated, warns Ernst & Young’s Margetts, with the amount of capital entering an insurance market correlated to the availability of capital via what’s happening elsewhere, which in turn affects the returns on assets. While insurance represents just a drop in the ocean for the global capital markets, it is growing and “the correlations between investment portfolios and insurance portfolios are very important,” adds Muir-Wood. “Sooner later the two will collide.”

Margetts advises naive investors to be aware. “I think it is quite difficult to model or quantify these types of risk without an intimate knowledge of the insurance industry and I think it is something the investors need to keep an eye on.”

While few catastrophe bonds have so far been triggered, 2007 promises an active hurricane season and earthquakes remain an ever-present risk. It might not be long before the opportunities turn into expenses for new investors. Although seasoned insurance investors know the drill, more opportunistic capital could be surprised by a big loss this year. And with their notoriously fickle nature, hedge funds in particular could theoretically pull their money out very quickly, leaving the insurance and reinsurance industry high and dry.

A global future

AIR’s Guin points to the globalisation of the world economy as one of the reasons behind the growing correlation between insurance and the capital markets. “The world economies are becoming ever more integrated,” he confirms. “There is a much higher level of dependence across regions and that is reflected in the economic and financial impacts from catastrophes. Today we worry about what would happen in the Tokyo financial markets if a catastrophe hit but this was not the case a few decades ago.”

Guin is referring to the theory that should a major earthquake occur under Tokyo, one of the most important financial centres in the world, it could cause equity prices to tumble on exchanges around the world, possibly triggering a global recession. In this situation, not only would there be insurance losses to deal with, investors would also face a downturn in their equity investments.

A historical precedent for such correlations was set with the Great Hanshin Earthquake disaster, or Kobe earthquake, in Japan on 17 January 1995. The quake caused an estimated $200bn in damage, although actual insured losses were much lower due to a mere 3% of property being insured. However the size of the earthquake caused the Japanese stock market to tumble, losing over a thousand points in one day. It was this financial meltdown that caused the collapse of Barings Bank as trader Nick Leeson speculated vast sums on Japanese derivatives.

The World Trade Center attacks, meanwhile, allowed insurers to extrapolate what future terrorist attacks could mean for major economic hubs like Tokyo or New York, especially in the arena of nuclear, biological, chemical and radioactive. It exposed hitherto unseen correlations between various lines of business and, in addition, losses on the global stock markets.

Another cause for concern for insurers and their investors is the current threat of a pandemic event such as bird flu. Margetts predicts a pandemic “would be as significant as a dirty nuclear bomb in central London, although there is more possibility of mitigation with a pandemic, via the application of medical science”. Affects could range from impacts on payment protection plans to business interruption to a complete meltdown of financial structures around the world. But thanks to high profile threats, such as bird flu, insurance companies are already considering the impact and closely monitoring the situation.

Changing times

These low probability/high consequence scenarios, and 9/11 in particular, transformed underwriting says Muir-Wood. “People still try to discover the risks involved with every underwriting decision but now the location of the risk has become very important. There has been a big transformation in how underwriters track locational information.” And as megacatastrophes expose more and more correlations across capital market portfolios, so they continue to demand more detailed risk information.

With global financial structures becoming more complex, the likelihood of failure rises, threatening to trigger cat bonds and hit all four corners of an investor’s portfolio. Unless correlations are thoroughly evaluated, and products become more sophisticated (which can already be seen in some of the newer cat bonds), it can become easy to miss the possibility of the combined impact of multiple losses. “Since Hurricane Katrina, the volume of capital market transactions has significantly increased,” adds Guin. “Although there is certainly more room to develop, with the growth in alternative risk transfers the correlations of risk in a portfolio of these instruments will likely continue to increase.”

Nick Thorpe is associate editor of Global Reinsurance.

Characteristics of a megacatastrophe


• Non-linear losses caused by the extreme scale of the event – eg As buildings disintegrate under wind or flood loads, they create debris that increases the damage to buildings;
• Demand surge – Megacatastrophes cause unprecedented damage and interruption, placing unusually high demand on insurers and service suppliers;
• Delays – Larger scale losses take longer to repair and resolve, increasing the likelihood of properties sustaining further damage or deteriorating;
• Claims exaggeration – With larger numbers of claims, insurers are unable to monitor them as rigorously, leading to the potential for fraud and claims exaggeration;
• Increased correlations of loss – There is increased correlation in loss outcomes across different locations, lines of business, and insurance coverages;
• Pollution – Pollution clean up greatly increases the costs of repair and lengthens the time it will take to start repairs and re-occupy undamaged properties; and
• Evacuation and consequent losses – The mandatory evacuation of the population of New Orleans after the flooding of the city became a major source of economic and physical damages.