32

calculation, i.e. to eliminate the numerous small earth tremors but to set the level well below the magnitude where significant losses occur. It is only earthquakes of magnitudes above 7 where losses are likely to occur to the bond, so the parametric trigger has no influence on the expected loss of the bond. If the event were deemed to qualify via the parametric trigger, the modeling service, RMS, would then use the fixed model to calculate estimated insured losses for the notional industry portfolio. If the Index Loss calculated exceeded certain dollar amounts then a loss payment would be triggered. The earthquake exposures of Hiscox Syndicate 33 are only relevant to the extent that the syndicate must have experienced losses of at least the amount paid under the reinsurance agreement with St Agatha Re.

Parametric triggers

70. A "parametric" trigger links recovery to the physical characteristics of the event that causes the losses (e.g., hurricane intensity, earthquake magnitude). Losses from the event may or may not match actual losses incurred but, since event parameters are quickly available, parametric triggers generate basis risk but no tail risk. Parametric structures are unlike other triggers. Clearly they add an increased risk of actual losses not matching recoveries. Basis risk tends to go up in these types of transactions therefore. Moreover, the modeling is very different because the probabilistic loss distributions are based exclusively on the physical parameters of the event. Whether quality underwriting or efficient claims management occurs after the event is irrelevant. Hence, unlike in the case of indemnity or index triggers, underwriting or claims practices need not be disclosed to investors. Lower disclosure needs also lessen the likelihood of potential litigation with investors. By the same token, rating agencies and investors scrutiny of the transaction is lower. Parameters tend to be more transparent and objective than indemnity or index calibrations. Hence, investors generally prefer this type of structure. This preference usually is reflected in slightly lower yields being needed to make the deal work.

71. In the Tokyo Disneyland transaction (discussed in paragraph 33), the payout is dependent solely upon the magnitude, location and depth of an earthquake, not on actual property damage. There are in fact two transactions, referred to as Concentric, Ltd. and Circle Maihama, Ltd.

33

Concentric, Ltd. provides Oriental Land (the owner of Tokyo Disneyland) with earthquake-contingent capital, while Circle Maihama, Ltd. provides it with earthquake-contingent financing. In both cases, there are three rings around a central point at the center of Tokyo Disneyland. The Inner Circle has a radius of 10km, the Inner Ring a radius of 50km, and the Outer Ring a radius of 75km. In order to trigger coverage, an earthquake with an epicenter within the Outer Ring and with a depth of less than or equal to 101km must occur. In the case of Circle Maihama, Ltd. the contingent financing is triggered if the magnitude of the earthquake is at least 6.5, 7.2 or 7.6 on the Japanese Meteorological Agency (JMA) scale for the inner circle, inner ring, and outer ring respectively. In the case of Concentric, Ltd. the principal payout is on a sliding scale depending on the JMA magnitude, and in which radius the epicenter lies. For the inner circle, the payout ranges from 25% at magnitude 6.5 to 100% at 7.5, for the inner ring it is 25% at 7.1 up to 100% at 7.7, while for the outer ring it is 25% at 7.6 up to 100% at 7.9.

Modeled loss triggers

72. A "modeled loss" trigger resembles both an index and a parametric trigger. The originating firm's portfolio is stored in a modeling firm's risk model. When the event occurs, the modeling firm calculates the modeled loss on the portfolio by using the physical parameters of the event. Hence, location and magnitude, for instance, determine the model's payout.

An illustration of modeled trigger transaction: The St. Agatha Re transaction. Hiscox Syndicate 33, one of the larger Lloyd's syndicates, recently entered into a catastrophe bond transaction designed to protect it against a major earthquake either in California or in the New Madrid region of the US. The bond secures up to US$33 million of property losses excluding liability over three years until April 15, 2005. The bonds were priced at 675 basis points over LIBOR and rated BB+ by Standard & Poor's. The deal uses a modeled loss index as the trigger, and the index is based on two industry models run by Risk Management Solutions (RMS) that measure insurance industry exposure in the two zones. The Qualifying Event trigger is parametric but the purpose of this is merely to set a realistic trigger for a loss calculation, i.e. to eliminate the numerous small earth tremors but

34

to set the level well below the magnitude where significant losses occur. It is only earthquakes of magnitudes above 7 where losses are likely to occur to the bond. So the parametric element of the trigger has no influence on the expected loss of the bond. If the event were deemed to qualify RMS would then use the fixed model to calculate estimated insured losses for the notional industry portfolio. If the Index Loss calculated exceeded certain dollar amounts then a loss payment would be triggered. The loss payment amount is on a predetermined sliding scale based on the Index Loss. The earthquake exposures of Hiscox Syndicate 33 are only relevant to the extent that the syndicate must have experienced losses of at least the amount paid under the reinsurance agreement with St Agatha Re.

73. While indemnity triggers provide the closest match between an originator's risk and its capital markets protection, non-indemnity triggers allow an originator to avoid detailed information disclosure in an offering memorandum. Because of heightened concern pertaining to the potential legal liability associated with erroneous disclosures in such a memorandum, some originators opt for a hybrid approach to securitizations. An originator enters into a traditional indemnitytriggered agreement with a transformer vehicle, which in turn transfers the risk to capital market investors by using an index-triggered securitization. The use of the transformer adds 1% to 1.5% to the cost of the transaction. The recent Western Capital transaction provides an example of this type of approach.

The role of the modeling agencies in the securitization process

74. Independent modeling is a crucial component to providing investors with confidence in the level of risk involved in the investment. Modeling firms provide an analysis of the risk pertaining to the securitization. A number of companies are recognized for their expertise in modeling. These include Risk Management Services Inc., EQE International Ltd., and Applied Insurance Research, Inc.

75. From a practical standpoint, it is extremely helpful to an originator to know that the major rating agencies have done extensive examinations and testing of these firms' models, and hence, a transaction can be brought to a successful closing more efficiently when one or more of these firms' models is employed.

76. The risk analysis results also become a major component of the analysis performed by the rating agencies. Moreover, the modeling firm also provides a number of the key ingredients for the ultimate offering circular for the transaction. Of utmost significance is the lossexceedance curve developed by the modeling firm. The following is an example of loss exceedance curve developed for the Residential Re transaction27:

27 See also Laurenzano, V. L. and Latza, W. D., Securitization of insurance risk. Insurance Securitization Educational Program of the National Association of Insurance Commissioners, San Francisco, December 4,