Model Management
Regulators to require US, Bermuda and EU markets to validate their reliance on catastrophe models.
Months after Hurricane Sandy, commentary has emerged about the reliability and proper use of catastrophe models by insurers and reinsurers.
These discussions raise interesting questions regarding the emphasis placed on cat models, as well as the assessment of the models' validity for purposes of regulation and solvency supervision under the evolving regulatory environments in the United States and the European Union.
With the advent of regulatory reform and modernization, cat modeling is a central piece of enterprise risk management and Own Risk and Solvency Assessment for companies insuring and reinsuring property catastrophe risks in the United States, the EU and elsewhere. This holds true for the Bermuda insurance and reinsurance market due to its emphasis on reinsurance of U.S. property catastrophe risks in particular, and its connection to both the United States and EU markets.
Criticism of cat modeling is not new. In the wake of hurricanes Katrina, Rita and Wilma in 2005, a furor was raised over cat models' underpredicting of the potential force and effect of these storms, particularly with regard to storm surge. Additionally, it was emphasized then--as it has been more recently--that certain elements of windstorms and other natural catastrophes, such as business interruption, disruption of ingress/egress and interruptions caused by the acts of civil authorities, are difficult to account for in cat models.
Cat modeling did not exist three decades ago. Today, it is an essential piece of all property insurers' or reinsurers' risk evaluation. Cat modeling is used by insurers to identify the potential exposures from natural and man-made catastrophes such as windstorm, earthquake, flood and other occurrences, with an eye toward identifying the amount of reinsurance that will be needed to cover a portfolio.
A key component to using natural catastrophe models effectively is recognizing that they are not perfect. These models do not, and cannot, predict with certainty the losses that will be incurred from an event that has not happened yet.
Of course, cat models are subject to change, which carries the potential for modifications in the values generated, and in turn may impact an insurer's or reinsurer's evaluation of exposure.
Such changes to cat models may seriously affect a company's risk assessment and require that underwriting practices, capital requirement and reinsurance protections be re-evaluated.
For example, when RMS released a new version of its U.S. windstorm model in August of 2011, many licensees were incensed over the increased volatility in the model. The new model significantly increased the potential losses from U.S. hurricane risk based upon information from more recent hurricanes, including Katrina, Rita and Wilma, which resulted in a significant increase in the projected losses in some areas.
Climate change also raises questions and challenges for cat modeling. Whereas cat modeling is generally based upon the notion that future losses may be predicted based upon historical losses, climate change involves the potential for conditions that have not been observed or recorded.
Moreover, climate change is anticipated to be an ongoing issue, and is expected to cause more numerous and intense windstorm events, higher sea levels and other extreme weather events. Therefore, it seems necessary for cat models to evolve as advances are made in the science of climate change and its likely impact on catastrophe events including windstorms, floods, droughts, winter storms and even (according to some theorists) earthquakes.
Cat modeling is an essential part of property catastrophe reinsurance risk assessment and pricing. Although it is generally accepted that cat models do not predict loss outcomes with complete certainty, the market continues to do a proficient job of identifying and underwriting risks, and ensuring that there is adequate capital to respond to significant losses.
As reported by the Association of Bermuda Insurers and Reinsurers, in its comments regarding the Federal Insurance Office's report on the global reinsurance market, global reinsurance capital was $470 billion in 2010 and dipped only 5% in 2011 to $445 billion, despite record global catastrophe losses of $105 billion that year. 2011 included the Japan and New Zealand earthquakes, an active Atlantic hurricane season, severe tornadoes in the U.S. and unanticipated events such as flooding in Thailand.
While 2012 did not see the same level of cat activity as 2011, it is worth noting that total global reinsurance capital reportedly increased to a record $505 billion in 2012, according to the latest aggregate study released by Aon Benfield.
Bermuda's reinsurers, in particular, have posted profits for calendar year 2012, according to a recent report from Property Casualty Insurers Association of America.
As such, despite conjecture about the limitations of cat modeling, the industry has used this tool to respond effectively to losses while maintaining, and increasing, its capacity to respond to future events.
Bermuda and Cat Reinsurance
The Bermuda insurance and reinsurance market is principally involved in high-layer excess liability insurance coverage and property catastrophe insurance and reinsurance, and the market is generally perceived as having a risk appetite that is not averse to volatility. According to statistics reported by ABIR to the FIO, Bermuda insurers and reinsurers have licensed entities in 61 jurisdictions, 142 domiciled entities in 18 U.S. states, with 34,000 employees globally, 15,000 of which are in the United States.
Bermuda insurers and reinsurers covered 37% of the reported losses from Europe's windstorm Xynthia in 2010, 38% of the 2010 Chilean earthquake, 51% of the Christchurch, New Zealand earthquake, and 29% of the internationally reinsured losses from the 2011 earthquake in Japan. According to ABIR, Bermuda reinsurers paid more than $22 billion to U.S. cedants for the 2004 and 2005 hurricanes, including Katrina, and Bermuda reinsurers expect to pay out at least $3 billion to cover claims arising from Sandy.
Use of cat models is critical to a reinsurance market like Bermuda's that focuses on property catastrophe claims. According to data from the Bermuda Monetary Authority, the Bermuda market relies heavily upon licensed cat modeling applications from third-party vendors such as AIR Worldwide Corp., RMS and Eqecat Inc.
Regulatory Impact
Current regulatory reform initiatives in the United States and European Union include requirements pertaining to the vetting and validation of cat models. The International Association of Insurance Supervisors' Insurance Core Principles include a requirement that companies conduct an Own Risk and Solvency Assessment, which has been incorporated under Pillar 2 of the Solvency II regime.
Under Solvency II, insurers and reinsurers will be required to carry out and document model validation to ensure accuracy and predictability. This applies to both internal models and external models licensed from third-party vendors.
Model validation is essentially a process of understanding whether the external cat model provides a valid representation of the risk for a particular portfolio. Solvency II will require that insurers and reinsurers demonstrate that there is sufficient understanding about a particular cat model to decide that it is appropriate for use with a given portfolio, and also that the reinsurer provides documentation showing that this process has been followed.
In 2011, the Association of British Insurers generated its Industry Good Practice for Catastrophe Modeling, developed by industry experts with input from the UK's Financial Services Authority and designed to provide guidance to the insurance and reinsurance market for the application of Solvency II to cat modeling.
The ORSA model developed by the NAIC is similar in many key respects, and is viewed by some as the U.S. "response" to the EU model. In the NAIC's version of ORSA, companies may be required to provide information on models and model validation processes.
Under Bermuda's regulatory scheme, Class 4, Class 3B and Class 3A insurers are likewise required to submit ORSA filings.
According to the Catastrophe Risk Return Guidelines produced by the BMA, Class 4, Class 3B, and Class 3A insurers must file a Catastrophe Risk Return annually, which should detail the extent of a company's reliance on cat models, the actions taken to mitigate "model risk" (the risk that the model underestimates potential losses from a cat event) and a description of the analytics in place to quantify exposure to vendor models. It also should allow the BMA to assess the reasonableness of inputs into the cat component of the regulatory capital requirement.
Under each of these regulatory schemes, insurers and reinsurers will be expected to assess the potential losses from catastrophe risks themselves, but inherent in that assessment are the validity of the models used and the potential risk of model inaccuracy.
Given the underwriting focus of the Bermuda reinsurance market, the United States and EU regulatory changes, as they relate to the property catastrophe business, necessarily would seem to be of importance to the Bermuda reinsurance market, which will be impacted by the regulatory initiatives and developments in the U.S. and the EU as well as under Bermuda's own regulatory scheme.
The latest round of commentary on the accuracy and dependability of cat models and the need to ensure that they are given the appropriate weight by underwriters is not saying anything that the international market has not heard before. Certainly insurers and reinsurers must consider how the information generated by models may be supplemented to account for nuances in the risks to be covered and, where possible, obtain information not accounted for by the models. Lastly, the models need to be "stress tested" so that factors that could potentially affect a model's accuracy are identified and disclosed.
What is new, however, is the incorporation of these principles into regulatory and supervisory schemes--that these concepts are required pieces of the ERM/ORSA of companies that use these tools in the evaluation of their loss exposure. Regardless of outstanding market performance and continued increases in capital surplus in the reinsurance market, companies operating in the United States, EU and Bermuda will be required to demonstrate to regulators and supervisors the extent of their reliance upon cat models, both internal and external, and that the models have been validated.
By Michael J. Kurtis and Nicholas C. Bacon
(Contributors: Michael J. Kurtis is a partner and Nicholas C. Bacon is director, International Markets at Nelson Levine de Luca & Hamilton. They may be reached at :mkurtis@nldhlaw.com or nbacon@nldhlaw.com)
