Catastrophe models and solvency IIMiscellaneous
Gabriela Chávez-López
Account Director – Europa / Asia
EQECAT
París – FRANCIA
The Association of British Insurers (ABI) recently released a guide to help insurance professionals with their catastrophe (CAT) model component of their Solvency II internal models1. This article presents a summary of the guide and the role CAT models play in Solvency II context.
The guide was written by industry practitioners in the United Kingdom and other European Union Member States. Even though it came out as a result of a Financial Services Authority (FSA)2 meeting, it is no way intended as a regulatory document. It offers technical guidance and suggestions for companies seeking to include a CAT model in their internal model under Solvency II.
The document covers the requirements to satisfy the model tests as set out in articles 120 to 126 of the Solvency II directive with regard to external models; it also includes a section on model selection and model change policy as set out in article 115 and a section on Outsourcing related to article 49 of the directive.
Working with CAT models requires addressing some very specific issues covered in the last section of the document under ‘Technical Considerations’, addressing multi-modelling approaches as well as how to deal with uncertainty in the models.
General Principles
The application of Solvency II essentially relates to good risk management. Senior Management has specific responsibilities to put in place processes ensuring that the internal model operates properly on a continuous basis and in line with the organisation’s overall risk appetite. Senior Management must also be able to demonstrate sufficient understanding of the organisation’s overall risk and tolerance towards risk exposure, as well as the reliability of the model with regard to proper and timely decision-making.
With regard to external CAT models, Senior Management should hold enough information to understand the strengths and weaknesses of the model being used whilst ensuring that adequate procedures are in place in order for the CAT model to be used and imbedded into the decision-making process.
The use of CAT models is a complex process and, due to the very nature of these models, complicated enough to be fully understood without adequate specialist knowledge. Senior Management is not required to understand in detail every single component of the model, but it needs to be knowledgeable enough to make pertinent decisions. Communication between departments should be paramount and part of the risk management process. This includes clear and regular documentation which can consist of correspondence, meeting notes, training and auditing reports.
Another aspect of risk management is to have the right people in the right place. The risk management team should be composed of personnel responsible for maintaining, running and evaluating the CAT models and thus fulfilling the function of providing information to the rest of the company.These specialists are required to evaluate regularly the pertinence of the company’s approach to CAT modelling and to document it, so that in the event staff members change the department or leave the company entirely, processes are not interrupted and day to day business activity maintained. Adequate process documentation ensures up-to-date information on any developments related to risk management and can act as a manual in helping other members of the staff or new joiners to be trained up to a desired level in any step of the process.
The use of third party service providers
Solvency II requires a full understanding of all components of an internal model, even when a company outsources CAT modelling to third party providers.
During the FSA workshop sessions, much was discussed about outsourcing and what it would mean under Solvency II as each company has different needs and uses different services. A company could use a broker for a complete service: from underwriting to CAT model use and interpretation of results. Alternatively it could use only a data management service company that simply cleans the exposure data used to run the model. The main issue is to have an outsourcing agreement that clearly states the outsourcing policy and what the relationship between each of the parties is. Again, good communication is important to transfer all necessary knowledge in order to understand all components of a CAT model.
However, ownership of the modelling process cannot be outsourced since the responsibility for understanding the model and key decisions on use and governance remain with the company.
Catastrophe model documentation
Companies have either a vendor CAT model or an internally developed CAT model. Alternatively, they could access the results of either type of model through a broker. Solvency II requires that documentation is consistent with the intended use of the model and that it is owned by the company.
Documentation should convey understanding of the CAT model capabilities and limitations, understanding of results, and evidence that it is used correctly. Thus, to prepare the required documentation, communication with CAT model vendors and brokers is essential.
An important issue regarding third party CAT model documentation, be it either vendor models or broker models, is intellectual property, which is subject to specific licensing agreements between the involved parties. Signed non-disclosure agreements are usually required for companies that do not have a license with the vendor model companies.
Use and management of catastrophe models data
As the old saying says “garbage in, garbage out”, the data quality used in the CAT model is an important part of the analysis process. They should represent the company’s risk profile as accurately as possible, so that the CAT model’s results reflect realistic estimations of the required capital. Data should be accurate, consistent and complete, with special attention to materiality and proportionality.
The limitations of the model will impact the data used for the overall analysis. In other words, not only does the data needto be accurate and complete but also appropriate to the model being used. For example, if a model is an aggregate model, detailed data will not give more precise results.
Where there is incomplete information, models will assign a default value to produce a reasonable approximation of the loss and users should be aware of the impact this may have on the final results.
Model selection and model change policy
One of the most difficult parts of CAT modelling is selecting the ‘right’ model and the ‘right’ method to evaluate the company’s CAT exposures. It will involve some research and tests on a particular peril model. This could be an onerous and lengthy process, but it is something all companies using a CAT model need to go through.
There are three main vendor models in the market as well as several in-house models developed by brokers or reinsurance companies. Each one of these models will have strengths, biases and limitations with regard to representing a firm’s view of the risk, and the company therefore needs to be in a position to be able to explain the motivation behind its selection.
Option and settings of catasprophe models
CAT models are built to be used by different companies with different needs. Although they may be ‘customized’ if they are in-house models, generally they are built to analyse any type of portfolio. They will cover gaps if necessary by assigning default values, but they will not be able to represent specific exposure without the capability of modifying settings and selecting options.
The selection of settings and options will depend on the company’s modelling approach. Since this selection will affect the final results, it should also be documented accordingly. An “option”relates to how the data is going to be modelled and manipulated to reflect the company’s overall risk exposure. For example, the user could use the generic code for structure type or select a specific structure type like “Concrete” for a particular risk. The user could further refine the exposure by using the height and age of the building, or just leaving it to the default values used by the model.
A setting refers to what the model can analyze, such as considering demand surge in the analysis, secondary perils, or alternative event catalogues. A company may choose to ignore these settings, to take them into consideration inside the model, or add their effect outside of the CAT model.
Catastrophe model validation
To be able to select a model and to understand if it represents the company’s view of the risk, the user needs first of all to understand how the model was created, make sure that the model is robust and consistent and that it accurately represents the hazard and the risk. The user needs to be aware of the model limitations and of any biases it may have. If the user finds that the model does not represent the risk as per the user’s view, he or she may choose to adapt the exposure and or the settings as discussed before. Consequently, the user will then need to validate the assumptions they have made to use the model by, firstly, understanding how the model has been validated by the modeller - be it a vendor modelling company or the company’s internal model development team – and secondly, once satisfied, validating that any adjustment or modification made still makes sense, representing the risk accurately, and that the model is robust and consistent.
Technical considerations
Last but not least, the issue of multiple models and the treatment of uncertainty are important issues to consider when using a CAT model.
The directive does not indicate if only one or multiple CAT models should be used. It requires that whatever method is used to make decisions, it is applied consistently through the whole process. The company has to make sure that any selected method used is well documented.
Uncertainty is one of the most challenging issues on the use of CAT models: “CAT models are constructed from the random variables which govern hazard event occurrence and severity, as well as the consequent loss potential.Only a few variables encountered in the study of natural hazards can be precisely determined through practical observation. As expected, the whole process involves a certain degree of uncertainty, which needs to be taken into account when evaluating the resulting estimations”.3
Some uncertainty can be reduced by using accurate input data, however not all uncertainty can be reduced and the user needs to understand the impact this may have on the overall estimation of the loss.
Conclusion
CAT models are sophisticated tools that use information coming from different sources and whose results are used by different people, departments and functions. Within an insurance company, they would be the underwriters, the analysts, the pricing actuaries, the accumulation control function, and the claims department. Thus, to have an adequate risk management process, Solvency II requires that everyone involved in the process have a clear understanding of the assumptions used to build the CAT models and their inherent uncertainties, in order to make sound business decisions. This understanding needs to be clearly communicated and should be easily audited, so that any change in the company’s personnel or organisation does not cause any disruption to the running of the business.
http://www.eqecat.com/Highlights
In December 2001, the Association of British Insurers (ABI) released a guide to help insurance professionals with their catastrophe (CAT) model component of their Solvency II internal models.
Senior Management must be able to demonstrate sufficient understanding of the organisation’s overall risk and tolerance towards risk exposure, as well as the reliability of the model with regard to proper and timely decision-making.
The use of CAT models is a complex process and, due to the very nature of these models, complicated enough to be fully understood without adequate specialist knowledge.
Ownership of the modelling process cannot be outsourced since the responsibility for understanding the model and key decisions on use and governance remain with the company.
Where there is incomplete information, models will assign a default value to produce a reasonable approximation of the loss and users should be aware of the impact this may have on the final results.
To have an adequate risk management process, Solvency II requires that everyone involved in the process have a clear understanding of the assumptions used to build the CAT models and their inherent uncertainties, in order to make sound business decisions.
1 ABI Association of British Insurers, London, December 2011, Industry Good Practice for Catastrophe Modelling – A guide to managing catastrophe models as part of an Internal Model under Solvency II
2 FSA Financial Services Authority, regulator of the financial services industry in the UK.
3 Chavez-Lopez G. (2010). Natural Catastrophe Loss Modeling: The value of knowing how little you know, 14 ECEE 2010, Ohrid, Macedonia.