AIG 2015 Annual Report Download - page 184

ITEM 7 / ENTERPRISE RISK MANAGEMENT

184
relative strengths and weaknesses of vendor models, and make adjustments to modeled losses to account for loss adjustment
expenses, model biases, data quality and non-modeled risks.
We perform post-catastrophe event studies to identify model weaknesses, underwriting gaps, and improvement opportunities.
Lessons learned from post-catastrophe event studies are incorporated into the modeling and underwriting processes of risk
pricing and selection. The majority of policies exposed to catastrophic risks are one-year contracts which allow us to adjust our
underwriting guidelines, pricing and exposure accumulation in a relatively short period.
We recognize that climate change has implications for insurance industry exposure to natural catastrophe risk. With multiple
levels of risk management processes in place, we actively analyze the latest climate science and policy to anticipate potential
changes to our risk profile, pricing models and strategic planning. For example, we continually consider changes in climate and
weather patterns as an integral part of the underwriting process. In addition, we are committed to providing innovative
insurance products and services to help our clients be proactive against the threat of climate change, including expanding
natural disaster resilience, promoting adaptation, and reducing greenhouse gas emissions. Our internal product development,
underwriting, modeling, and sustainability practices will continue to adapt to and evolve with the developing risk exposures
attributed to climate change.
Our natural catastrophe exposure is primarily driven by the U.S. and Japan, though our overall exposure is diversified across
multiple countries. For example, we have exposures to additional perils such as European windstorms and floods. Within the
U.S., we have significant hurricane exposure in Florida, the Gulf of Mexico, Northeast U.S. and mid-Atlantic regions. Events
impacting the Northeast U.S. and the mid-Atlantic may result in a higher share of industry losses than other regions primarily
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 in a year, respectively.
The following table presents an overview of OEP modeled losses for top perils and countries.
At December 31, 2015 Net of 2016 Net of 2016 Percent of Total
(in millions) Gross Reinsurance Reinsurance, After Tax Shareholder Equity
Exposures:
U.S. Hurricane (1-in-100)(a) $ 5,332 $ 3,144 $ 2,044 2.3%
U.S. Earthquake (1-in-250)(b) 7,484 3,827 2,487 2.8
Japanese Wind (1-in-100) 936 523 340 0.4
Japanese Earthquake (1-in-250)(c) $ 993 $ 601 $ 391 0.4%
(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, 2015, for both U.S. and Japan exposures.
The catastrophe reinsurance program is as of January 1, 2016.
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