Seminar on Solvency II Using Internal Models to determine the SCR Paul Sharma Internal Models Expert Group Chair Bucharest, 13 June 2008 1
Outline Background Solvency Capital Requirement (SCR) principles Standard formula Internal models What is an internal model? What is the purpose of an internal model? What are the expected benefits? Approval process: the regulatory perspective Calibration using a third party vendor model Lessons from the CRD/Basel II Conclusions Questions Page 2
Adequacy of Financial Resources Free assets SCR Assets covering technical provisions, MCR and SCR assets MCR Risk margin Best estimate for non-hedgeable - risk components Technical provisions Market consistent valuation for hedgeable risk components Page 3
Solvency Capital Requirement (SCR) (Cfa 10.2) The SCR should deliver a level of capital that enables an insurance undertaking to absorb significant unforeseen losses over a specified time horizon and gives reasonable assurance to policyholders that payments will be made as they fall due Page 4
SCR: proposed principles Derived using either an approved internal model or a standard formula Partial internal model Full internal model Standard formula 99.5% confidence level over 1 year As a minimum to cover insurance, market, credit and operational risks Part of supervisory review process (SRP) Page 5
Pillar 1 SCR (std. formula) SCR BSCR SCRop SCRnl SCRmkt SCRdef SCRhealth SCRlife NLpr MKTeq LIFElapse LIFEmort NLcat MKTsp MKTconc LIFEexp LIFElong MKTint LIFEdis LIFEcat MKTprop MKTfx Health LT = Health ST Health WC = adjustment for the risk -mitigating effect of future profit sharing Page 6
What is an internal model? What is the purpose of an internal model? What is an internal model? A risk management system developed by an insurer to analyse the overall risk position, to quantify risks and to determine the economic capital required to meet those risks Solvency II Glossary (CEA/GC) What is the purpose of an internal model? To fully integrate processes of risk and capital management within the insurer Page 7
Internal Models Framework use test internal model (in the wider risk management sense) actions / steering internal risk control functions reporting / monitoring SCR (regulatory capital) Pillar-2 adjustment actuarial model (in the narrow sense) risk exposure data forecasts for P&L distributions SCR estimate adjusted SCR statistical quality test risk driver data myth1: main use is regulatory -> internal use more important myth2: main goal is computation of SCR -> SCR/EC = airbag : just one aspect of car safety myth3: one risk measure -> distributions & several risk measures/metrics for reporting calibration test Page 8
What are the expected benefits? Improved risk sensitivity of SCR related to the insurer s specific profile leading to a more adequate modelling of non-standard, especially non-linear, contracts, Better alignment of regulatory capital requirements with economic capital Encouragement of innovation in risk management methodology leading to higher competitiveness through better risk management and hence lower costs of capital, Page 9
What are the expected benefits? More effective pillar 2 discussion and familiarity of the supervisor with more detailed exposure data than is generally available in accounting records, Cost efficiencies through re-use of risk modelling infrastructure for discussion with supervisors, rating agencies, analysts and shareholders. Page 10
Approval process Formal application Documentary evidence that requirements are met (art. 118.3) The administrative or management bodies have overall responsibility for application role of internal committees (art.114.1) Systems must be in place to ensure internal model operates properly on a continuous basis (art 114.2) Supervisors have six month to grant approval (art. 118.4) reason for rejection must be provided (art. 118.6) Requirement to run standard approach in parallel for two years (art. 118.7) Page 11
Full Internal models Insurers must satisfy: Use test Statistical quality standards Calibration standards Validation standards Profit and Loss Attribution Documentation standards Use of external vendor models does not exempt insurers from any of the standards Page 12
Use test requirements (art. 118) Use test: process by which the internal model is assessed by the insurer in terms of its application within the undertaking s risk management processes Internal model must be central to: risk management and decision-making economic and solvency capital assessment system of governance Must be fully embedded in the operational and organisational structure of the insurer Demonstrate that the model remains useful and is applied consistently over time Board take full responsibility for model's use and validation Page 13
Statistical quality test requirements (art. 119) Probability distribution forecast: based on sound actuarial/ statistical techniques current, credible info/ realistic assumptions broadly consistent with approach to technical provisions assumptions justified by firms data appropriate and accurate No particular method, provided IM meets criteria of use test All material risks to be covered, including all those addressed in Standard Formula Page 14
Statistical quality test requirements Diversification approach to be sound and implemented with integrity Allowance for risk mitigation permitted provided counterparty credit risk and other risks arising reflected in internal model Financial guarantees or options to be modelled where material Management actions can be modelled provided firm reasonably expect to carry out in specific circumstances allowance for time to carry out management actions and obligations to policyholders (with regard to policyholder s reasonable expectations) Page 15
Calibration test requirements (art. 120) Where feasible, use VaR at 99.5% over 1 year confidence level Different risk measure/ time horizon permitted provided policyholders protection equivalent to Standard Formula i.e. 99.5% over 1 year, VaR Approximations permitted where firm demonstrates approach provides equivalent protection Supervisors may compare an insurer s internal model with industry benchmarks to verify specification meets market practice Page 16
Validation The undertaking has a primary responsibility for validation Validation is an iterative process There is no single validation method Validation should encompass both quantitative and qualitative elements Validation should be subject to independent review Page 17
Validation standards (art. 122) Regular cycle of model validation shall include: Monitoring the performance of the model Reviewing the ongoing appropriateness of its specification Testing its results against experience Analysis of stability of the model Testing the sensitivity of the results to changes in key underlying assumptions Assessment of the accuracy, completeness and appropriateness of data used by the internal model Page 18
Profit & Loss Attribution (art. 121) Analysis of profit and loss by cause/source for each major business unit, at least annually Demonstrate how categorisation of risk chosen explains the causes/sources of P&L Page 19
Documentation (art. 123) It must be thorough, detailed and complete enough to allow third parties to replicate the internal model It shall provide a detailed outline of: The theory Assumptions Mathematical and empirical basis underlying the model It is a crucial tool for the firm to demonstrate the supervisor that the undertaking really understands and has mastered the internal model it is using It shall indicate any circumstances under which the internal model does not work effectively (weaknesses) It must be revisited and, if necessary, updated in line with the complexity and stability of the risk profile of the insurer Page 20
Partial internal models Insurance and reinsurance undertakings may use them for the calculation of one or more: risk modules or sub-modules of the SCR operational risk adj. for the loss-absorbing capacity of TPs and deferred taxes In addition to: the whole business of insurance and reinsurance undertakings, or only one or major business unites Approval subject to the same requirements as full IM, plus additional conditions: Reason for the limited scope of application IM SCR reflects more appropriately the risk profile of the undertaking Page 21
Calibration using a third party vendor model (art. 124) Supervisors do not approve the model per se, but the use of the model!!!! Firms must be able to document and explain the role of vendor products and the extent to which they are used within their internal processes Firms must be able to demonstrate a thorough understanding of vendor products used in their internal processes Vendor products should be appropriate to the nature, scale and complexity of the risks incorporated within the firm s risk strategy and business objectives Firms must have clearly articulated strategies for regularly reviewing the performance of vendor model results and the integrity of external data used in their risk quantification process Page 22
Lessons from the CRD/Basel II Banks typically took longer than three years to implement the CRD/Basel II The length of Insurance liabilities can be much greater and data collection may take many years Data collection should be started as soon as possible so that embedded models are ready when Solvency II is implemented due to back testing, calibration and user test requirements Most of the CRD/Basel II expenditure was around the IT systems to capture the data Page 23
Lessons from the CRD/Basel II Banks were used to building models. However their models: were typically used for management purposes and not for capital purposes had not always been reviewed or subject to independent challenge Documentation issues Governance issues and commitment of Senior Management Competitive pressures Pricing Institutional investors and rating agencies Page 24
Summary Internal models are not only for large and sophisticated firms. However, the CRD/Basel II experience has shown that to develop and maintain an internal model four main conditions need to be satisfied: Good understanding of the business Commitment of senior management Knowledge, expertise to create and maintain model Good, deep and rich data sets Page 25
Questions?? Page 26