Guidance paper on the use of internal models for risk and capital management purposes by insurers October 1, 2008 Stuart Wason Chair, IAA Solvency Sub-Committee
Agenda Introduction Global need for guidance Guidance activity Black boxes, black swans and Rumsfeld Quick comparison of IAIS & IAA guidance Key features of internal models (material drawn in part from current IAA draft) 1
Introduction Global need for guidance The use of internal models for insurer risk assessment and capital management is increasing due to the: emergence of comprehensive insurer risk management practices, widespread use of economic capital for risk and capital management, and development of increasingly sophisticated risk-based insurance regulatory capital requirements. 2
Introduction Global need for guidance Company-specific internal models are also being used for other purposes within insurance companies, including: valuation of insurance obligations, financial condition analysis (known in some jurisdictions as Dynamic Solvency Testing, Dynamic Capital Adequacy Testing or Dynamic Financial Analysis), analysis of asset/liability mismatch and the refinement of investment policy, analysis of market risk in certain investment products with guaranteed values (known as segregated funds, variable annuities or maturity guarantees), pricing of insurance products, and pricing of reinsurance programs. 3
Introduction Guidance activity IAIS Standard & Guidance Paper March 2008 drafts The use of internal models for risk and capital management by insurers IAA Solvency Sub-Committee has developed an April 2, 2008 draft and has been working in concert with IAIS for the last year to align the material in both IAA and IAIS guidance papers Groupe Consultatif has formed a working group to assist the CEIOPS Internal Models group Other country specific initiatives (US, Canada, Australia etc.) 4
Introduction Black boxes, black swans and Rumsfeld The IAA paper is written to address the concern that internal models are black boxes which are difficult to understand and on which to place reliance. The IAA paper provides guidance for those who are constructing and using internal models to assess and manage risk and capital within insurance enterprises. Internal models are frequently associated with regulatory capital requirements, but they have wider application in connection with other measures of required capital and enterprise risk management (ERM). By understanding the possibilities and limitations of internal models we seek to minimize black swan events or the unknown unknowns. 5
Quick Comparison of IAIS & IAA Guidance Focus of guidance IAIS The use of internal models for regulatory capital by insurers IAA The use of internal models for risk and capital management by insurers Internal model definition IAIS - Risk measurement system developed by an insurer to analyse its overall risk position, to quantify risks and determine EC to meet those risks IAA - Mathematical model of an insurer s operations to analyse its overall risk position, to quantify risks and determine the capital to meet those risks 6
Quick Comparison of IAIS & IAA Guidance Contents IAIS 17 requirements covering general provisions, initial and on-going validation & supervisory approval, supervisory reporting and public disclosure IAA 8 sections 8Introduction 8 Model Fundamentals 8Design Considerations 8 Model Construction 8Controls 8 Governance 8Communication 8Supervisory Approvals 7
Quick Comparison of IAIS & IAA Guidance Tests IAIS 8Statistical quality test (appropriateness; addresses risks) 8Calibration test (satisfies modeling criteria of supervisor) 8 Use test and governance (embedded in RM; governance) 8On-going validation (on-going fit for purpose; change controls) IAA 8Sufficiency test (fit for purpose; addresses risks) 8Calibration test (confidence level; experience) 8Use test (sufficient discipline; extensive use; governance) 8Change test (reconcile and explain differences in results) 8
IAA Guidance Paper - Model Fundamentals Financial model Modeling process Proportionality Risk assessment framework Time horizon Risk Measure Confidence level Terminal provision Probability measures (real world & market consistency) Managing models Types of models 9
IAA Guidance Paper - Model Fundamentals Financial Model An internal model quantifies a range of possible future and current financial positions an insurer may find itself in as a result of the variety of risk factors ( risks ) to which it is exposed. An insurer s financial position is subject to many variables factors, such as: financial market performance (e.g., changes in interest rates, equity values, credit spreads, foreign exchange, etc.); insurance risk outcomes (e.g., claim frequencies, claim severities, change in policy interpretation, deterioration in pricing/underwriting, policyholder behaviour, trend uncertainty, level uncertainty, etc.); operational events (e.g., impact on financial states due to people, systems etc) or changes in the legal, regulatory and taxation environments, changes to the insurer s reputation and consequent effects on its ability to continue with its business plan; and management choices and strategies relating to investment options, marketing and new business, dividend and non-guaranteed element declarations and capital structure, etc. 10
IAA Guidance Paper - Model Fundamentals Modeling Process The financial position of an insurer at a particular time is dependent upon the values of a great many quantities. For example, the insurer s free surplus can only be obtained once other important components of the balance sheet are known. It is important to understand the dependence structure among all the assumptions and other variables in the model since this will determine the structure of the model and the optimal order in which calculations are carried out. Models typically begin with a projection of the raw elements (cash flows) that reflect the insurer s basic projected experience: volume of claims incurred, premiums received, expenses, taxes, investment income, business in-force, etc. 11
IAA Guidance Paper - Model Fundamentals Modeling Process It is essential that the resulting projected cash flows reflect the underlying future experience and are not simply those resulting from an intermediate pass-through facility or product (e.g. securitization of original financial product). Frequently it is important for the model to determine the present value (as at a current date or future dates) of a stream of future cash flows. This is determined through some valuation process which involves rates of return to be used for discounting. Once these valuation variables have been calculated, a model will often go through a process similar to that carried out in the preparation of an insurer s financial statements. It is only then that many other variables that the user is interested in, such as free surplus or capital ratios can be obtained and studied. 12
IAA Guidance Paper - Model Fundamentals Proportionality It would be easy to reach the conclusion that all internal models need to be very complex. This is not necessarily the case. The design of an internal model should be based upon the concept of proportionality: the structure of the model should depend upon the nature, size and complexity of an insurer s risks. Use of accepted or better practices Size of risk Breadth and diversity of risk Volatility or uncertainty of risk Partial models 13
IAA Guidance Paper - Model Fundamentals Risk Assessment Framework For risk and capital management, it is common for required capital to be set such that the total balance sheet assets can withstand extremely adverse experience over a defined period of time (say one year) and still have sufficient assets left over at the end of that time period to settle the remaining policyholder obligations Key elements of a risk assessment framework, Risks included Time horizon Risk measure Confidence level Terminal provision Real-world, risk neutral, market consistent 14
IAA Guidance Paper - Model Fundamentals Risk Assessment Framework Risks included: There should be explicit documentation and disclosure of the risks included in and excluded from the model. This is necessary for assessing the actual versus expected results from the internal model. Time horizon: The framework should specify the period over which the extremely adverse experience is assumed to occur. Frequently, this is called the time horizon. This should not be confused with the need to consider all risks over their full lifetime. 15
IAA Guidance Paper - Model Fundamentals Risk Assessment Framework Both One Year and Lifetime Perspectives Have Advocates One Year Consistent with other emerging regimes and economic frameworks (e.g. Basel II) Focuses risk analysis and management on actionable timeframe Appropriate terminal value methodology can reflect long term risks Long term models overweigh very subjective analysis of catastrophic long term risk modeling Lifetime Consistent with traditional actuarial approaches Some long term risk exposures cannot truly be captured in shorter term metric It is difficult to develop reliable terminal values for 1 year metric 16
IAA Guidance Paper - Model Fundamentals Risk Assessment Framework Risk measure: The risk measure is the statistical measure used to quantify the extreme experience. Measures often used are Value at Risk (VaR) or Conditional Tail Expectation (CTE, TailVaR, Expected Shortfall, or Average Downside Magnitude). Confidence level: The confidence level, in conjunction with a risk measure, specifies how extreme the adverse experience is over the time horizon 17
IAA Guidance Paper - Model Fundamentals Risk Assessment Framework 18
IAA Guidance Paper - Model Fundamentals Risk Assessment Framework Terminal Provision: The terminal provision defines what sufficient assets are at the end of the time horizon. It depends in part on how an insurer is assumed to act following an adverse extreme event. For example, whether the insurer would raise capital and continue operations, continue in a runoff mode, or liquidate the remaining risks. The terminal provision provides for the remaining risks beyond the time horizon according to the stressed assumptions as well as including some provision for future risk. 19
Risk Assessment Framework Example: One year risk horizon 1st year paths are stochastic Terminal value is path dependent and developed either stochastically or analytically Year 1 Path Take CTE(99) result with terminal values Path dependent Terminal values 1. Hedgeable risks: risk neutral 2. Non-hedgeable risks: real world CTE(60)-CTE (80) result is based on CTE(99) outcome of 1 year paths with calculated residual values 20
IAA Guidance Paper - Model Fundamentals Risk Assessment Framework Real world, risk neutral, market consistent: Real world probability measures project economic scenarios based on historical experience and actuarial techniques Risk neutral probability measures are designed to reproduce market prices. A capital model might use a real world approach over the risk horizon, but the terminal provision at the end of the risk horizon might follow either a real world or risk neutral approach, depending on the risk. Market-consistent valuation is used when it is necessary to price financial instruments that are not freely traded. Risk neutral or real world probabilities are first used to generate projected cash flows, then risk free interest rates (as generated by the model) are used to discount those cash flows. 21
IAA Guidance Paper - Model Fundamentals Risk Assessment Framework Terminal value based on available close out strategy for the risk availability of market prices directly use these prices (e.g. risk neutral prices for certain market risks) lack of market prices actuarial modeling using real world assumptions: Stochastic on stochastic modeling 22
IAA Guidance Paper Model Construction Time granularity Population of the model Product descriptions In-force data/assets Insurance experience assumptions for best estimates Insurance assumptions for projections Assumptions concerning the insurer Algorithms/random numbers/scenarios/extreme values Documentation 23
IAA Guidance Paper Model Construction Scenarios Random numbers Stochastic models depend upon random numbers. Random number generators do not in fact generate purely random numbers. These generators are algorithms. They require an initial number as a seed in order to produce the next number. A large collection of numbers produced sequentially in this way will appear to be distributed in proportion to the probability distribution upon which the generator is based. In order to de-bug models and for audit purposes, it will be necessary to be able to re-run models and verify their results. In stochastic models, this requires the ability to reproduce the random numbers that are used in a particular simulation 24
IAA Guidance Paper Model Construction Scenarios Insurance risks Assumptions about future experience for insurance risks are among the most difficult issues in constructing an internal model. Experience studies and credibility Future versus past experience New products with no experience Volatility, trend & level uncertainty 25
IAA Guidance Paper Model Construction Scenarios Extreme values Internal models are often used to study an insurer s capital needs. This requires knowledge of the extreme tail values of distributions of various financial results. A fundamental concern is whether the model captures sufficiently well all the extreme scenarios that can occur. A further concern is whether relationships between various components and assumptions that are embedded in the model remain valid in extreme scenarios. For example, there is ample evidence that in time of financial stress, certain correlations between economic factors or markets that normally hold may suddenly change significantly 26
IAA Guidance Paper Controls Sufficiency test Conceptual (model adequately provides for the risks) Implementation (model actually works as expected) Assumption (reasonableness of assumptions) Data (confirm accuracy, continuity and completeness of data used by model) Calibration test (calibration of parameters to experience) Use test (sufficient discipline; extensive use in risk and capital management) Change test (reconcile and explain differences in results from one run to next) 27
IAA Guidance Paper Governance Approvals Risk management policy Audit Review Documentation Compliance 28
IAA Guidance Paper Communication Fundamentals Identify Stakeholders Identify Communication Requirements Internal Management Communication Needs Examiner Communication Needs Public Communication Needs 29
IAA Guidance Paper Communication Fundamentals The development of internal models for an insurer s capital requirements is a complex task requiring a variety of technical skills. The key results from these internal models are very important to a variety of the insurer s internal and external stakeholders, helping them to assess the insurer s exposure to risk and its financial soundness (to name but two). These stakeholders will have a wide range of needs for information about the internal models, depending on their different perspectives and range of technical skills. 30
IAA Guidance Paper Communication Fundamentals Failure to communicate effectively throughout the development and on-going operation of internal models can pose serious risks to the insurer. For example, In the absence of good communication practices, internal models can easily be viewed by stakeholders as a black box or perhaps worse as being unreliable and not useful. Failure to communicate between the business unit whose risks are being modelled and the risk modellers will result in models which do not properly reflect the risks and risk management of the insurer. Failure to properly communicate the key assumptions, issues, and results to senior management and through public disclosure will result in lost credibility for the models and in the latter case perhaps even affect outside perception as to the value to be placed on the insurer s shares. 31
IAA Guidance Paper Communication Fundamentals CEO example Objective: To communicate with investor analysts publicly on a quarterly basis about the insurer s capital position. Requirements: The CEO requires a one page set of briefing notes and an attached dashboard of certain analytical information related to the capital requirements and consequent capital position of the insurer. These are to be available on a quarterly basis prior to any meetings with analysts. A prior briefing session with the CEO is also needed. The contents of the dashboard will be discussed and agreed upon well in advance. The CEO also needs to have additional validating or explanatory information to support the current quarter s results. The CEO needs to be able to succinctly and clearly describe the key purposes of the internal model, its key assumptions, methodologies, limitations, other issues and results. 32
IAA Guidance Paper Communication Fundamentals It is important for those involved in the development and use of internal models to use the fundamentals of good communication in their everyday work since: it is all too easy to communicate with others based solely on one s own perception of the other party s communication needs or in a manner with which we are comfortable; and the risks to the insurer of poor communication are significant and can be underestimated. 33
Internal Models Final thoughts It has been said that all models are wrong but some are useful. When properly embedded in an insurer s risk and capital management processes, internal models are a rich and vital source on information about the insurer s inherent and residual risks. Insurers must ensure through the sufficiency, calibration, use and change tests that their internal models are working properly as their results will be used throughout the risk and capital management processes (i.e. in strategic planning). The communication needs of all internal model stakeholders need to be carefully considered lest the internal model become a black box. Significant internal model implementation challenges will need to be addressed in the coming years by insurers and supervisors alike. 34
Guidance paper on the use of internal models for risk and capital management purposes by insurers October 1, 2008 Stuart Wason Chair, IAA Solvency Sub-Committee