AIG 2014 Annual Report Download - page 189

ITEM 7 / ENTERPRISE RISK MANAGEMENT

172
due to our relative share of exposure in those regions. Within the U.S., we have significant earthquake exposure in California,
the Pacific Northwest and New Madrid regions. Earthquakes impacting the Pacific Northwest and New Madrid regions may
result in a higher share of industry losses than other regions primarily due to our relative share of exposure in these regions.
The estimates below are the Occurrence Exceedance Probability (OEP) losses, which reflect losses that may occur in any
single event due to the defined peril. The 1-in-100 and 1-in-250 PMLs are the probable maximum losses from a single natural
catastrophe event with probability of 1 percent and 0.4 percent, respectively.
The following table presents an overview of modeled losses (OEP) for top perils and countries.
At December 31, 2014 Net of 2015 Net of 2015 Percent of Total
(in millions) Gross Reinsurance Reinsurance, After Tax Shareholder Equity
Exposures:
U.S. Hurricane (1-in-100)(a) $ 4,887 $ 2,801 $ 1,821 1.7%
U.S. Earthquake (1-in-250)(b) 7,080 3,522 2,289 2.1
Japanese Wind (1-in-100) 1,279 1,225 796 0.7
Japanese Earthquake (1-in-250)(c) $ 1,159 $ 899 $ 584 0.5%
(a) The U.S. hurricane amount includes losses to Property from hurricane hazards of wind and storm surge.
(b) U.S. earthquake loss estimates represent exposure to Property, Workers’ Compensation (U.S.) and A&H business lines.
(c) Japan Earthquake represents exposure to Property and A&H business lines.
The OEP estimates provided above reflect our in-force portfolios at September 30, 2014, for U.S. exposures, and at June 30,
2014 for Japan exposures. The catastrophe reinsurance program is as of January 1, 2015.
As noted above, AIG, along with other non-life insurance and reinsurance companies, utilizes industry-recognized catastrophe
models and apply their proprietary modeling processes and assumptions to arrive at loss estimates. The use of different
methodologies and assumptions could materially change the projected losses. Since there is no industry standard for
assumptions and preparation of insured data for use in these models, modeled losses may not be comparable to estimates
made by other companies.
Also, the modeled results are based on the assumption that all reinsurers fulfill their obligations to us under the terms of the
reinsurance arrangements and all catastrophe bonds attach and pay as modeled. However, reinsurance recoverable may not
be fully collectible. In particular, the use of catastrophe bonds may not provide commensurate levels of protection compared to
traditional reinsurance transactions. Therefore, these estimates are inherently uncertain and may not accurately reflect our
exposure to these events.
Our 2015 catastrophe reinsurance program includes coverage for natural catastrophes and some coverage for terrorism
events. It consists of a large North American occurrence cover (without reinstatement) to protect against a large U.S. loss, and
a worldwide aggregate cover to protect against multiple, potentially smaller, losses. The attachment point for this reinsurance
program is at $3 billion.
Actual results in any period are likely to vary, perhaps materially, from the modeled scenarios. The occurrence of one or more
severe events could have a material adverse effect on our financial condition, results of operations and liquidity. See also Item
1A. Risk Factors — Reserves and Exposures for additional information.
Terrorism Risk
We actively monitor terrorism risk and manage exposures to losses from terrorist attacks. We have set risk limits based on
modeled losses from certain terrorism attack scenarios. Terrorism risks are modeled using a third-party vendor model and
various terrorism attack modes and scenarios. Adjustments are made to account for vendor model gaps and the nature of the
Non-Life Insurance Companies exposures. Examples of modeled scenarios are conventional bombs of different sizes, anthrax
attacks and nuclear attacks.
Our largest terrorism exposures are in New York City, and estimated losses are largely driven by the Property and Workers’
Compensation lines of business. At our largest exposure location, modeled losses for a five-ton bomb attack net of the