Treatment of technical provisions under Solvency II Quantitative methods, qualitative requirements and disclosure obligations Authors Martin Brosemer Dr. Susanne Lepschi Dr. Katja Lord Contact solvency-solutions@munichre.com You can download the Knowledge Series at www.munichre.com July 2011 Introduction The European insurance supervisory authority (EIOPA) carried out the fifth quantitative impact study, QIS5, in the second half of 2010. The experience gained in the study has demonstrated that the calculation of technical provisions is still causing problems for the insurance industry. For example: The segmentation by line of business stipulated by the framework directive posed a problem for many insurance companies because their data was often not sufficiently granular to divide the provisions into the required risk groups. Figure 1 illustrates the treatment of technical provisions in the three pillars of Solvency II. The framework directive (level 1) and the implementing measures (level 2) 2 form the overriding legal basis. The individual quantitative, qualitative and information requirements are derived from regulations, which have been allocated to the appropriate pillars. The majority considered the calculation of the risk margin to be too complex most insurance companies used simplified methods. Many life insurers were unable to cope with the valuation methodology for options, guarantees and future profit commissions. Many supervisors also noted an incoherent use of the illiquidity premium. 1 1 Since illiquidity can be taken into account in the discounting of technical provisions (through the use of a higher yield curve), the risk of economic capital decreasing in the event of a fall in the yield curve is to be captured in the market module. The fall will be taken into account in a scenario in which a 65% reduction in the stress factors for the relevant curve is assumed. 2 In future, the implementing measures will be referred to as delegated acts.
Page 2/8 The aim of this publication is to illustrate the overall treatment of technical provisions under Solvency II and to discuss various methods that could be used to determine bestestimate provisions in life and nonlife insurance for the first pillar. They complement the methods already presented in the Solvency Consulting Knowledge Series Best estimates for Solvency II technical provisions. 3 The first pillar How should technical PROvisions be calculated? In addition to the risk model, the economic balance sheet plays a key role in determining the regulatory capital requirements under Solvency II. For the economic balance sheet, all items are shown at market value. It is particularly difficult to value technical provisions because there is generally no market from which prices can be derived. We therefore base our valuation of technical provisions on the exit value, which, according to the framework directive, is the amount an insurance or reinsurance undertaking would have to pay if it transferred its contractual rights and obligations immediately to another undertaking and hence equates to a sale price. The figure must be determined in accordance with market principles and is calculated by adding the best estimate to the risk margin (cf. Figure 2). Fig. 1: Regulatory requirements for technical provisions under Solvency II Treatment of technical provisions under Solvency II Market value of assets Pillar 1 Level 1: Art. 76 (2), (3) Art. 77 81 Level 2: IM-TP TP: Technical provisions; PD: Public disclosure. Fig. 2: Solvency II balance sheet Economic capital Risk margin Best estimate The best estimate is equal to the probability-weighted average of future cash flows and takes into account the time value of money. Article 76 of the framework directive stipulates that the calculation of the best estimate shall be based upon up-to-date and credible information and realistic assumptions and be performed using adequate, applicable and relevant actuarial and statistical methods. Level 1 Level 2 Pillar 2 Level 1: Art. 76 (4), (5) Art. 77 (2) Art. 79 Art. 82 Level 2: IM-TP Pillar 3 Level 1: Art. 51 (1) (d) Level 2: IM-TP Market value of technical provisions Economic capital = Market value of assets Market value of technical provisions The risk margin reflects the present value of the future capital costs. However, where future cash flows associated with insurance obligations can be replicated using a financial instrument for which a reliable market value is observable, the value of the technical provisions can be determined on the basis of that financial instrument. In such cases, separate calculations of the best estimate and the risk margin will not be required. 3 Published in April 2008
Page 3/8 We consider below technical and methodological issues of importance for the calculation of the best-estimate provisions. Best estimate in life Insurance Best estimates for the main actuarial assumptions for life insurance products are needed to determine the best-estimate provisions for life insurance business, particularly for mortality, longevity, disability, lapses, costs and interest rates. Data requirements and segmentation Ideally, best-estimate actuarial assumptions should be based on company-specific insurance portfolios at policy level over a period of years (e.g. five years). Of fundamental importance will be whether it will be possible for the actuarial assumptions to be classified by sub-portfolio. In addition to the traditional distinction for age and gender, differentiations could also be made, for example, for smokers/non-smokers, extent of risk assessment, duration of policies, occupation class and sales channel. The aim is to break the portfolio down into groups that are as homogeneous as possible on the basis of the criteria specified. Since the validation and preparation of the portfolio data is an annually recurring process, it makes sense to use standardised tools. Methods With the traditional actuarial approach, which has been used, for example, to derive the DAV 4 tables, the portfolio is divided into risk groups that are as homogeneous as possible. Approximated best-estimate rates are calculated for these sub-portfolios, i.e. the ratio of losses to the exposures 5 at risk is calculated. However, the sub-portfolios decrease in size with each differentiation by risk characteristic and soon become too small for valid conclusions to be drawn, so that the rates derived from them vary considerably and are subject to a high degree of uncertainty. Generalised linear models (GLMs) offer one solution to this problem. GLMs are statistical models in which all risk factors are analysed simultaneously. Thus, robust estimates of general effects such as dependence on incidence rates of age and gender can be made on the basis of the entire portfolio. The rates serve as a starting point and are adjusted for individual sub-groups such as smokers or certain occupation groups to reflect the risk, enabling valid conclusions to be drawn even for smaller sub-groups. Due to the simultaneous consideration of all risk factors, the effects of individual factors can be clearly defined. Effect overlaps can be detected and duplication avoided. GLMs permit a better understanding of portfolios, revealing correlations and interactions within the data and hence delivering more than traditional methods: they clarify the correlation between different factors, thus enabl ing us to take prompt and focused action in response to changes. Example: In addition to a normal product, a BI portfolio includes a comfort product, which offers policyholders better terms and conditions, so that a claim can be made at an earlier stage. The comfort product has only been sold in the last few years. An analysis of disablement rates by product using the traditional method demonstrates that the comfort product has lower incidence rates even though higher rates would be expected as the conditions are better than those for the normal product. If the portfolio is analysed with a GLM, factors other than the product itself can be taken into account, such as the time the insured has been in the portfolio or the occupation class. The GLM analysis shows that the product does not affect the disablement risk, though a clear selection effect is apparent: insureds who have been in the portfolio for only a few years have lower disablement rates because of the only recently performed risk assessment, regardless of the product. Thus, the lower rates observed for the comfort product when using the traditional method were wrongly attributed to the product, when they were in fact due to the insureds for that product still being in the selection phase. With GLM on the other hand,consideration of all factors enabled the sub-portfolio to which an insured belonged to be identified as the decisive risk factor. 4 Deutsche Aktuarvereinigung (German Association of Actuaries). 5 Exposures can be weighted sums insured, numbers of policies or person-years at risk.
Page 4/8 Challenges/difficulties Large insurance companies have sufficiently large portfolios to be able to derive their own company-specific tables, whereas this is often difficult for companies with smaller portfolios. Deriving accurate best-estimate actuarial assumptions can be a problem even for large companies if insufficient loss experience is available (for example for new products or where there are major differences in product features). In such cases, tables produced by actuarial associations or pool analyses can be used as a guide. Insurers can use the progressions in the tables and adjust them as appropriate by a factor specific to the port folio and independent of age. Trends can be distinguished from random fluctuations at an early stage, enabling companies to react to them more rapidly. Best estimate in non-life insurance Data requirements and segmentation Historical loss triangles will generally be used in the calculation of bestestimate reserves in non-life, except for claims which involve subsequent payment of a pension, which should be valued using the same methodology as for life. The loss triangles should be divided into homogeneous portfolios, broken down by line of business as a minimum. Where the data situation permits, further subdivisions can be performed, for example: personal and industrial business, short-tail and long-tail run-off, claims payments and claims handling expenses, loss drivers (e.g. personal injury and property damage), currency (discounting with differing yield curves). Methods The most common methods of determining best estimates are: Method Advantages Disadvantages Chain ladder Independent of exposure measure (e.g. premium, sum insured, number of vehicles), and hence of market cycles and rate adjustments. Heavily dependent on current calendar-year balances, especially latest and oldest accident/underwriting year (extreme current calendar-year balances result in extreme ultimate figures). The following items must be calculated to determine the best-estimate provisions for non-life insurance business: loss reserves for losses already incurred, Additive Independent of current calendar-year balances. Can be used even if no calendar-year information available. Market information usable. Cannot be used if no calendar-year information available. Dependent on exposure measure, market cycles and rate adjustments for premiums to be taken into account. Dependence on last accident year. Lower dependence on current calendar-year balances. premium reserves for liabilities that have not yet arisen. 6 Bornhuetter- Ferguson All cash inflows and outflows will have to be included, for example premiums, claims payments, loss adjustment expenses, bonus payments and guaranteed premium refunds. Subjectivity in choice of market information possible, dependent on exposure measure. 6 Under Solvency II, contracts written in a financial year that have not commenced by the year-end will in future have to be recognised in that financial year.
Page 5/8 Challenges/difficulties In the fifth quantitative impact study, most property insurers used either the chain-ladder or the Bornhuetter- Ferguson method, which raised the following practical issues: 1. Data situation Is the history of the run-off triangles adequate? Is the history of the run-off triangles still representative (e.g. regarding court decisions, technological change, economic developments, etc.)? Is the company s internal data situation adequate, or does it need to be complemented/replaced by external data sources? Can the future course of inflation be extrapolated from the historical data observed? 2. Methodology Which method for taking account of inflation best reflects the exposure (e.g. eliminating past inflation from data and projecting future inflation)? What effect does monitoring of reserves have on methodology and underlying assumptions (e.g. subsequent reserving behaviour, especially for business with long run-off periods)? What effect do large losses have on the projection of the run-off triangles (elimination of large and catastrophe losses from data)? How can large and catastrophe losses be assessed (e.g. analyses of individual losses or scenarios)? Risk margin As mentioned at the outset, the risk margin reflects the present value of the future capital costs. Calculating the risk margin is complicated, because the risk-capital components used (underwriting risk, default risk, operational risk and unavoidable market risk 7 ) have to projected over the residual term of the portfolio. The regulators therefore allowed other calculation methods with varying degrees of simplification for QIS5. For example, the simplest approach entailed calculation of the risk margin in proportion to the best-estimate provisions. Many companies took advantage of the simplifications, which leads to other problems: The different calculation methods produce highly divergent results and hence arbitrage possibilities. The comparability of economic balance sheets is restricted. Dubious results are possible, for example proportional calculation with negative best-estimate provisions also produces a negative risk margin. For these reasons it is important for stronger guidelines to be defined for the calculation of the risk margin or for the methodology to be refined so that usable results can be achieved. 7 According to the current state of the debate, the unavoidable market risk will no longer be taken into account in the future.
Page 6/8 The second pillar What key factors need to be borne in mind for the process? Fig. 3: The governance system Principle-based Solvency II requirements Professional suitability and personal reliability of board Information, assumptions and methodology play an important part in determining technical provisions. The information base must be credible and consistent, the assumptions realistic and the actuarial/statistical methods appropriate as required by Article 77 of the framework directive. The actuarial function will have a significant role to play in meeting the future requirements. This separate function is a component of the governance system illustrated in Figure 3. Principle of proportionality Board is responsible for business and risk strategy Governance requirements Transparent structure Clear separation of responsibilities written guidelines Regular reviews Internal control system and compliance function Risk management system Quantitative requirements Risk management function Contingency plans Internal model ORSA (internal assessment of risk capital requirement) Actuarial function Outsourcing Information, documentation, reports Internal audit The actuarial function defined in Article 48 of the framework directive must have an understanding of the stochastic nature of insurance business and of the risks that can arise on both sides of the balance sheet. The range of responsibilities is very wide and includes the following: Coordinating the calculation of technical provisions Ensuring the appropriateness of methods, models and assumptions Assessing the quality of the data used in the calculations Best-estimate analyses Appraisal of underwriting policy Assessing the adequacy of reinsurance arrangements Reporting to management and supervisory authorities The coordination function will be described in greater detail in the draft of the Level 2 implementing measures. For example, methods and procedures must be developed and used to assess the accuracy with which reserves reflect reality. Moreover, the calculation must be consistent with the requirements of the framework directive and the degree of uncertainty inherent in the technical provisions must be assessed and explained. Article 82 of the directive also requires companies to have internal processes and procedures in place to ensure the appropriateness, completeness and accuracy of the data used in the calculation of their technical provisions. Monitoring, data standards, methodology and documentation requirements will play a major role. Monitoring Determining best-estimate actuarial assumptions is a recurring activity. Regular comparison of actual claims experience with the best-estimate expectation will permit companies to detect systematic deviations, enabling them to adjust assumptions quickly. They will recognise trends and changes in portfolio structure and the product landscape early and be able to target their reactions to them.
Page 7/8 Data standards/methodology Guidelines must be put in place that cover data quality, the underlying assumptions, expert estimates, data updates and the methodologies used. Quality standards for errors, completeness and appropriateness should be defined for the purpose of assessing the data, including their input, processing and use. Documentation requirements Methods, assumptions, data sources and their use, and particular features of data must be documented. A written record must be made of changes between periods, including the reasons for the changes. To summarise, insurance companies must be in a position to provide information on their procedures for supervisory audits and to explain and justify methods and assumptions used for the calculation of technical provisions. The third pillar What information will have to be published in the future? The framework directive sets out the Solvency II disclosure requirements. In future, insurance companies will have to produce two reports. Article 35 (1) stipulates that supervisory authorities must be provided with the information necessary for the purposes of supervision. The Regular Supervisory Report is a report prepared solely for the supervisory authorities. The second report, the Report on solvency and financial condition defined in Article 51, provides the general public with annual information on a company s solvency and financial position. Both reports have a similar structure and contain quantitative components as well as qualitative information. Quantitative Reporting Templates (QRTs) will be used to request quantitative information for reports. They are in a format prescribed by the supervisory authority, but so far we have only seen drafts. A characteristic feature of the templates is the considerable scope of the information and details requested, but also the high level of granularity. On the basis of the draft versions available, we expect an increase in requirements with regard to processes, data storage and reporting systems. The main areas covered by the qualitative reporting are business and performance, system of governance, risk management, the regulatory balance sheet, capital management. The link between reporting and technical provisions is created in Article 51 (d), which stipulates that separate descriptions must be provided of the bases and methods used for the valuations of assets, technical provisions, and other liabilities. The amount of the technical provisions for all significant lines of business must be disclosed, together with the bases, methodologies and assumptions used. To summarise, The information requirements for Pillar 3 can only be met if the necessary processes, datasets and methodology for determining reserves have been put in place.
Page 8/8 Outlook Munich Re can provide support in the determination and regular review of best estimates. Life Munich Re performs an annual analysis of biometric risks for a data pool representative of the market. Munich Re has accumulated a high level of expertise over many years of experience and makes use of professional and standardised tools and processes. A pool analysis provides a comparison of a company s own loss experience with the pool. Furthermore, because of the size of the pool, analyses based on a wide range of characteristics, such as occupation or higher risks, can be performed, enabling trends that it would not be possible to detect with small portfolios to be highlighted at an early stage. The results for different risk groups can be applied to smaller companies. Insurance companies can use the results of the biometric pool analyses to carry out regular monitoring of their best-estimate assumptions. Non-life Munich Re s worldwide expertise and methodological competence can be particularly useful for smaller insurers who may not have the required resources. Furthermore, underwriting at insurers can benefit from Munich Re s assistance with compiling relevant statistics and rating structures. Our expertise can help with the assessment of typical claims experience, thus substantially improving claims handling and reserving. 2011 Münchener Rückversicherungs-Gesellschaft Königinstrasse 107, 80802 München, Germany Order number 302-07063