Economic Capital: Building Upon Embedded Value Calculations Hosted by AIROC

Transcription

1 Economic Capital: Building Upon Embedded Value Calculations Hosted by AIROC Deloitte Actuarial and Insurance Solutions (Hong Kong) Limited 14 January 2014 Agenda 1. Why economic capital: business benefits and imperatives 2. Economic capital calculations general methodology 3. Break 4. General modelling requirements 5. Economic capital practices around the world 6. Roadmap for implementing the calculation methodology for Taiwan 7. Q&A 1

2 Why Economic Capital: Business Benefits and Imperatives 2 Definition of Economic Capital The excess of market value of the assets over the fair value of the liabilities required to ensure that obligations can be satisfied at a given confidence level, over a specified time horizon. Market Value of Assets Base Scenario Available Economic Capital Market Value of Liabilities Required Economic Capital Free Surplus Include NB contribution and available EC released Market Value of Assets + Net Worth created in a year at X% confidence level Available Economic Capital at an X% confidence level Market Value of Liabilities at an X% confidence level X% confidence level over a given time horizon 3

3 The Economic Capital Process The calculation of Economic Capital is part of a wider cycle Risk Appetite Risk Identification Financial Model Risk Quantification Model Use Set Risk Appetite and Monitor Risk Policies Identify Significant Risk Independent assessment Risk Model Business model Capital Models Aggregation Report Capital requirements Diversification Analysis of change Capital Management Business applications Risk management Control Cycle Monitor exposure against appetite Assess risk and return Calculate longterm value Develop the business As well as solvency reporting, business applications could include: Product design Product pricing and risk assessment Investment strategy 4 What can EC tell you? 1. The balance sheet on a realistic basis, given the risk profile of the company. Different capital requirements include Benchmark Capital Align with peers and Maintain profitability and avoid ruin Economic Capital remain competitive. Maintain credit rating Rating Agency Capital Avoid regulatory action Regulatory Capital 2. Impact to balance sheet for a given activity in the company. e.g. product launch, investment, etc Projected balance sheet After Activity Current Balance Before Activity Sheet Market Value of Assets Market Value of Liabilities Market Value of Assets Available Capital Market Value of Liabilities Market Value of Assets Available Capital Market Value of Liabilities 3. Description of the risk profile of the company, using EC as a metric for risks Required Economic Capital Risk Market e.g. equity Non-market e.g. lapse Diversification effects Economic Capital (TWD) X Y (Z) 5

4 Why use EC? 1. Regulators are increasingly requiring it. 2. Creates long-term shareholder value 3. Obtain market advantage Understand risk and value China Developing C-ROSS risk-based reporting. Europe Solvency II being developed, implementation 2016 Globally International Association of Insurance Supervisors (ICP 16) Supervisors establish ERM for solvency to address all relevant risks Consider the effect of a diverse portfolio 4. Single measure to reflect risk and reward 5. Industry best practice 6. Links to other areas e.g. Own Risk Solvency Assessment (ORSA) and IFRS (realistic) reporting. 6 Key questions around EC (1) 1. What is the fair value of the liabilities? For assets, this is generally market value. For liabilities, market consistent valuation techniques are applied. Simple, except when there are embedded options and guarantees in the products. For products with options and guarantees, option pricing techniques are used. Easier in markets where there is a deep and liquid derivative market. 2. What time horizon? The time horizon is variable, it could be: o One Year o More than One Year o Life time 7

5 Key questions around EC (2) 3. What is the given level of confidence? The confidence level is variable, but possibilities may include: o 95%, to survive a 1-in-20 year event o 99%, to survive a 1-in-100 year event o 99.5%, to survive a 1-in-200 year event What is an acceptable probability of ruin? Tail events small probability of ruin Credit rating AAA AA A BBB BB B CCC CC C Probability of default (ruin) Short Term 0.04% 0.06% 0.13% 0.47% 2.40% 7.49% 16.90% Long-Term 1.08% 1.74% 2.33% 2.55% 2.85% 3.9% 5.28% 6.36% 10.46% 4. What risk are considered in an EC calculation? Market Risk Credit Risk Insurance Risk Operational Risk What else? 8 Issues Arising from Economic Capital Insurer s considerations some examples Products with onerous options and guarantees Marketable, but increase risk and hence capital requirement. Investment strategy Assets which offer higher average returns but are more volatile will increase risk capital. Guaranteed issuance of term insurance policies Potential increase to volatility of mortality rates, increase to risk capital. Running significant asset-liability mismatch to get high yields Questions Why do we think that they are risky? How is risk defined? Is there a way to measure risk quantitatively? How can we measure risk-return tradeoff? How do we know we have taken too much risks in aggregate? Should there be a limit of total risk quantum for the company? Should the total risk limits be different for different companies? How should companies decide? 9

6 Pricing Example (1) Let s consider two non-par single premium endowment products. Total guaranteed return for 10 years of 5% p.a. and 0% p.a. respectively Calculate the profit margins (e.g. present value of future statutory profits over present value of future premiums) Both have the same profit margin (EV not EC basis), which exceeds the company s threshold. Questions to consider: Which product should be launched, you can launch only one? 0% guarantee 5% guarantee Projected sales Would it be different if 10-year government bond yield is 5% or 0%? Would it be different if the companies invest in 50% equities or 10% equities? 10 Pricing Example (2) Clearly, there are some business benefits for offering the endowment with the higher guarantee. BUT, what about the risks? Interest Rate Risk The guarantee increases the interest rate risk 0% guarantee 5% guarantee Projected sales Unless there is a secure bond that can match the cash-flows. Equity risk Holding more equity increases the equity risk capital requirement. Profitability reflecting Economic Capital The profitability on Profit margin allowing for economic capital 0% guarantee 5% guarantee 11

7 Pricing Example (3) Observations Overall Risks and returns should be looked at together. Both should be considered for any activity in the company. External and internal factors need to be considered. Different companies would have different risk appetite. i.e. different companies could reach different conclusions. Risk and Return Tradeoff Profits per capital invested tells us about returns on investments. We need either the profits or capital to be risk-adjusted to give us a better picture. Economic capital is a risk-adjusted capital and a way to measure risks. Profits per economic capital is a measure for risk-return tradeoff. Economic capital is central to all of these 12 Risk Appetite Risk appetite: The level of risk that an organisation is prepared to accept, before action is deemed necessary to reduce it. It represents a balance between the potential benefits of innovation and the threats that change inevitably brings. Initial steps include: Risks to accept Risks to reject Performance measures, for example: Earning not to fall by x%, with y% confidence Risk Appetite Statement Risk Policy Risk Measures Why use risk appetite Risk management Focus on value 13

8 Setting a Risk Appetite Risk appetite is led by the Board and Senior Management. Board and Senior Management Propensity to Exercise Risk Risk Assuran ce Vision Strategy Risk Appetite Controls Data Operatin g Model Propensity to Take Control Process Policy 14 Using Risk Appetite & Risk Limits Risk Capacity Accept the right amount of risk: Too little inefficient use of capital Too much capital shortfall Know your limits, manage the business in line with the triggers defined. Risk Appetite Upper Risk Limit Upper Risk Trigger Lower Risk Trigger Lower Risk Limit 15 Too little risk Just right Warning almost to upper limit Warning above risk appetite Warning above risk capacity

9 EC Calculations General Methodology 16 Economic Capital and Risk Economic Capital focuses on overall tail events. Difference compared with the 50 th percentile is the value at risk ( VaR ) which is also the economic capital requirement, how is this determined? Best Estimate Assumption 50 th percentile Risk Appetite 1 in 20 year event (95 th percentile) EC 1 in 2,000 year event (99.95 th percentile) Weaker Realistic balance sheet strength Stronger 17

10 Choice of aggregation methodology To determine the realistic balance sheet in the tail events, risks under considerations need to be modelled and their potential impact to the realistic balance sheet aggregated. Variance-covariance approach Sophistication Monte Carlo approach 18 Method 1 Variance-covariance approach Steps for finding VaR Step 1 Calculate loss values under individual risk stresses at the required confidence level Step 2 Determine a correlation matrix for the relationship between risk pairs Step 3 Calculate VaR using matrix multiplication Risk Loss (L) 1 ρ 12 ρ 13 R 1 L 1 R 2 L 2 C = ρ 21 1 ρ 23 ρ 31 ρ 32 1 VaR = (L T CL) R 3 L 3 Example with just two risks: VaR = ( L L ρ 12 L 1 L 2 ) 19

11 Method 1 Variance-covariance approach Just need one stressed calculation for each risk Simple, intuitive calculation Ignores non-linearity of losses for risks Ignores the interaction between different risks Without further assessment, it does not show the VaR of each individual risk after allowing for diversification benefits 20 Method 2 Monte Carlo approach What s the idea? Generate lots of random scenarios for your risks (individual distributions + risk correlations) Evaluate realistic balance sheets under each scenario Market risk distributions Correlation matrix Risk scenarios, typically 100,000 or more Net asset distribution 21 Non-market risk distributions

12 Method 2 Monte Carlo approach Very flexible can model whatever distributions Allows for diversification and interaction effects explicitly Captures non-linearity impact of risks Produces a full distribution of net assets Possible to identify scenarios that matters most in contributing to VaR Mathematically intensive (scenario generation) Parametric (requires distribution and correlation assumptions) Slow, so need proxy business models source of error 22 Choice of overall aggregation framework Comparing the two approaches Variance Covariance Monte Carlo Ease of implementation Non-normal distributions Modelling non-linearity and interaction effects Diversification effects Undiversified capital by risk Insights on identifying scenarios that matters Full distribution of net assets Explainability More complex methods will free us from the wrong assumptions of simple models, instead they let us come up with our own. 23

13 Risk modelling Comparison between aggregation approaches Variance-covariance Approach Only the shock at the desired level of confidence is required The correlation assumptions used theoretically should describe the relationship between risk capital, which is not easily defined/observed Monte Carlo Approach Requires full distribution of risks to be modelled The correlation assumptions used input describes relationship between risks drivers only Risk modelling parameter setting considerations Internal data analysis A/E ratio, change in assumptions etc External data analysis for limited internal data such as life catastrophe risk Expert Judgement qualitative override such as truncating irrelevant data, proxy data, allowing for future outlook versus past experience 24 Dependencies Setting the correlation matrix Similar to modelling of other risks, correlation assumptions may be set considering internal data analysis, external data analysis and expert judgement. Scenario analysis can be used to assess the risks and to a lesser extent, factor in their interactions. Expert judgement is an important part in determining the dependencies/correlations between risks. The final step is to ensure that the matrix is internally consistent (Positive Semi-Definite PSD ). If the matrix isn t PSD then there are mathematical algorithms which can be used to rectify this problem. 25

14 Lapse risk modelling example (1) Risk Modelling Data Analysis and Model Fitting Available data includes actual/expected experience Model Validation Max. and min. A/E observations are within the body of the distribution suggested by the fitted model Expert Judgement It may be appropriate to make manual adjustments to the models to reflect items, such as the level of uncertainty, based on expert judgement: The high volatility in 2009 was a result of the effects of the financial crisis Expert sources suggest a shock twice as large is industry practice, so shock is increased by 50% Available Data Data Analysis Model Fitting Lapse rate (%) Lapse A/E Actual and expected lapse experience % 150% 100% 50% 0% A/E for lapse data Model of future lapse A/E ratios Expected Actual 26 Lapse risk modelling example (2) Correlation Modelling Data Analysis Correlation between any two data sets can be calculated to provide a data-driven correlation suggestion Data Analysis Risk A/E Lapse 105% 110% 97% 86% Equity 93% 178% 62% 115% Interest 113% 134% 75% 84% Mortality 55% 72% 28% 90% Lapse-equity correlation Lapse-interest correlation Lapse-mortality correlation Scenario Analysis Scenario analysis can create data sets to use for correlation modelling, particularly where data is limited Primary and Secondary Correlations Focus on getting the primary correlations right Back-solve the secondary correlations based on the values determined for the primary correlations. Positive Semi-Definite The final check is that the correlation matrix is positive semi-definite (PSD) does it all hang together? Scenario Analysis Primary & Secondary Correlations Scenario Likelihood Risk driver Lapse Equity A. High B Medium C Low Rate the Risk Drivers for each scenario Equity Lapse Interest Primary correlations Secondary correlations 27

15 General Modelling Requirements 28 Market Value Surplus for Base Scenario Required Outputs from Projection Model Market Value Surplus (MVS) Market Value Asset (MVA) Market Value Liability (MVL) MVS = MVA - MVL Based on EEV Market values of the assets held by the company at the valuation date MVL = Statutory Reserve PVFP + TVOG where PVFP = Discounted profit using risk discount rate TVOG = cost of options and guarantees based on stochastic projections of real world scenarios MVS = MVA - MVL Based on MCEV Market values of the assets held by the company at the valuation date MVL = Statutory Reserve PVFP + TVOG + RM where PVFP = Discounted profit using risk neutral interest rates TVOG = cost of options and guarantees based on stochastic projections of risk neutral scenarios RM = Risk margin which is required to capture the uncertainty of the projection of cash flows Stochastic ALM model is required to properly reflect the linkage between the investment performance and the liability cash flows when calculating TVOG 29

16 Modelling Requirement for Base Scenario High level calculation flow of an ALM model DATA CALCULATIONS RESULTS COMPANY Equity Data Bonds Data Asset Calculations Asset Results Bal Sheet / Rev Account FUND LEVEL DECISIONS Asset Allocation Economic Scenarios Liability Results Bal Sheet / Rev Account Asset Trading Bonus Policy Prophet Liability Calculations 30 Modelling Requirement for Base Scenario Overall ALM Model Structure Asset & Fund Model Asset Categories Cash Equity Bond Option Treatment of Funds Investment Asset mix strategy + Accounting Asset-liab = process matching Model Output Market values Book values Investment returns URG/URL Stochastic Scenarios Equity Index Liability Model Bond Return Index Discount rates MPs Benefit outgo Par/NP A&F Dynamic Calculations Assumption Bonus outgo Liab. Dynamic Calculations Two-way data transfer between Asset & Fund Model and Liability Model takes place at userdefined time point, Updates calculation results. The more frequent the data transfer, the longer the run time Universal Life MPs Assumptions + Benefit Crediting = outgo rates From A&F Inv. returns Bonus policy Model Output CFs Profits From Liab. Reserves Cash flows Reserves Solvency 31

17 Modelling Requirement for Base Scenario Dynamic Actions to be considered Some questions need to be addressed when modelling dynamic calculations Management Decisions Policyholder Behaviour Dynamic Bonus / Dividend Dynamic Crediting Rate Dynamic Lapse Method on dynamic actions What method is used to determine dividend revisions? What method is used to determine crediting rates? What method is used to determine lapse rate increases / cuts? Level What grouping of policies will be used to determine dividend revisions? Product / Product Group / Entity? What grouping of policies will be used to determine changes to crediting rate? Product / Product Group? What grouping of policies will be used to determine changes to lapse rates? Policy / Product? Limits on revisions How much can dividends be revised? Absolute limit / % change in rates compared to previous year? How much can crediting rates be revised? Absolute limit / % change in rates compared to previous year? How much can lapse rates change in each year and is there a total accumulative limit? Cannot be negative or exceed 100% Timing of revisions When and how often will dividends be altered? Annual revision at year-end / Annual revision at a particular date? When and how often will crediting rates be altered? Annual revision at year-end / Annual revision at a particular date? When and how often will lapse rates be allowed to be altered? Annual revision at year-end / Annual revision at a particular date? 32 Modelling Requirement for Base Scenario Additional considerations for market consistent valuation methodology Market consistent scenarios are required Cannot use the prescribed scenarios used in Appointed Actuary s report Yield curves used has to be consistent with the deflator used Need to have extra care if ESG cannot cover the longer term bonds and extrapolation of yield curves in the model is required Have to perform Martingale test (i.e. 1 = 1 test) on both scenarios and the output of the ALM model to ensure market consistency Exchange rate must be set using interest rate parity, otherwise arbitrage opportunities will exist Check if BEL plus present value future profits (PVFP) equals Statutory reserves (i.e. leakage test ) Revisit dynamic calculations in the ALM model to ensure it works well under market consistent valuation For example, dynamic bonus / dividend strategy 33

18 Modelling Requirement for Base Scenario Improving runtime performance Hardware Use more modern hardware Use a stable standalone network for processing Move to enterprise level processing Software Move to the latest system version Use an algorithmic grouping tool to reduce the number of model points Model setup Limit the variables to be calculated, based on a particular target output variable Limit variables to be output Limit projection term Limit period for which dynamic decisions are made Ensure all debug messages and diagnostic tools are switched off for production 34 Market Value Surplus at One-year Horizon Requirements and Possible Options Requirements To calculate EC, distribution of MVS one-year horizon under realworld scenarios (outer simulations) are needed For MVS under each outer scenarios, another simulated scenarios (inner simulations) are needed to calculate the TVOG for products with options and guarantees EC therefore requires stochastic on stochastic (SOS) calculations Full stochastic on stochastic models Outer Simulations Inner simulations used to calculate realistic liabilities Economic Capital MVS curve Possible Options Stress Test Replicating portfolio Curve fitting Percentile Best Estimate 35

19 Options for SOS Calculations Full Stochastic on Stochastic Perform full stochastic on stochastic run Algorithm grouping could help to reduce the runtimes to a level where this approach would become feasible Total simulations = No. Outer Simulations x no. Inner Simulations 1000 Simulations Inner simulations used to calculate TVOG Advantages The most theoretically correct method of projecting realistic liabilities Relatively easy to justify the choice of model These two distributions may be the same (simple implementation) or simulation dependent. If simulation dependent, need to generate a lot of stochastic values Disadvantages Extremely computationally intensive. Additional development work may be needed to ensure that the tailored code can be used with the SoS feature 36 T=0 T=1 Options for SOS Calculations Stress Testing Economic capital is determined by taking the difference of MVS calculated using best estimate scenario and MVS calculated using stress test scenario Best estimate scenario Inner simulations used to calculate realistic liabilities Advantages Commonly used approach Simple model Short development period Disadvantages Stress test scenario (e.g percentile) T=0 T=1 The stress test scenario needs to be carefully calibrated. Difficult to calibrate the correlation matrix used for risk aggregation 37

20 Options for SOS Calculations Replicating Portfolios Use candidate assets to replicated the cash flows, as well as the present values or sensitivities of the asset and liability Calculate the market value of the candidate assets under different market risk factors and determine EC based on the distribution of MVS 1000 Simulations Realistic liabilities calculated by referring to the replicating assets Advantages Widely used in the market for fitting time 0 liability values Simplifies communication by expressing the liabilities in terms of a portfolio of assets Fit the replicating portfolio to the liabilities at time 0. The fit may be based on individual cashflows or PVs Disadvantages Candidate assets need to be carefully chosen taking into account the liability characteristics this can be difficult to do The effect of management actions is difficult to replicate 38 T=0 T=1 Options for SOS Calculations Curve Fitting The Curve Fit model finds a polynomial to fit the change in asset and liability corresponding to the change in risk factors Use regression method to determine the coefficients of the polynomial based on results of calibration points 1000 Simulations Fit a curve instead of inner simulations Advantages Widely used in the market for fitting time 0 liability values Runtimes will be much quicker than full SoS Disadvantages Calibration of a time 0 model is a multi-stage process involving expert judgment and often some iteration 39 T=0 T=1

21 EC Developments in Rest of the World 40 Economic Capital Calculations Developments in various countries UK (1) Current statutory solvency capital regime: Solvency I, which takes the general form of x% reserves + y% NAR Being tested: Solvency II regime Companies can choose a standard formula, a full internal model or a partial internal model Implies 99.5th percentile over a 1-year time horizon 41

22 Economic Capital Calculations Developments in various countries UK (2) Illustration of various stress scenarios Market Value after Stress Market Value after Stress 100% Base case 80%Property stress test 68% global equity 55% hedge fund 42 Economic Capital Calculations Developments in various countries UK (3) For the intra-risk diversification benefit, take equity risk as example, Step 1: Calculate the capital requirement under both Global and Other equity shock Global Other Equity shock i 32% 45% Mkt eq,i = max (ΔNAV equity shock i ;0) Step 2: Apply the correlation matrix of intra equity risks CorrIndex Global Other Global Other Mkt eq = SQRT ( RxC CorrIndex RxC Mkt R Mkt C ) = SQRT (Mkt 2 Global + Mkt 2 Other Mkt Global Mkt Other ) 43

23 Economic Capital Calculations Developments in various countries UK (4) The standard formula allows for inter-risk diversification benefit as follows: Market Default Life Health Non-life Market 100% 25% 25% 25% 25% Default 25% 100% 25% 25% 50% Life 25% 25% 100% 25% 0% Health 25% 25% 25% 100% 0% Non-life 25% 50% 0% 0% 100% 44 Economic Capital Calculations Developments in various countries Australia (1) Regulatory solvency formula for banks and insurers based on parametric estimates of economic capital, calibrated to something like the 99.5th percentile based on industry data where available. Stress testing is mandated. Regulator (APRA) encourages use of economic capital models. 4 risks involved: 1. Asset risk Calculated by determining the fall in the capital base of the fund in 7 prescribed stress tests Aggregated through prescribed correlation matrix 2. Insurance risk Calculated as the Reduction in the capital base that would occur if the adjusted policy liabilities were changed to an amount equal to the stressed policy liabilities 9 prescribed stress tests, determined at a 99.5 per cent probability of sufficiency over a 12 month period Aggregated through prescribed correlation matrix 3. Operational risk Calculated by prescribed formula 4. Concentration risk Calculated by prescribed exposure limit 45

24 Economic Capital Calculations Developments in various countries Australia (2) Major banks and insurers all have economic capital models at various levels of sophistication for internal use. General approach is to model risk categories (asset risk, catastrophe risk, liquidity risk etc.) separately and determine overall level of capital by aggregating using correlation matrices or copulas. One general insurer has had its internal capital model approved by APRA for determining capital. There is a floor of 90% of the standard formula for internally derived economic capital. A lot of regulatory and industry focus on modelling operational risk. 46 Economic Capital Calculations Developments in various countries South Africa Current Statutory solvency capital regime: Capital Adequacy Requirement ( CAR ) Being tested: Solvency Assessment and Management ( SAM ) Regime similar to Solvency II 3 impact studies performed Full implementation expected to be effective from 1 January 2016 Feedback from latest impact study: 33% indicated they intend using a standard formula capital model for regulatory capital 11% intend applying for a partial internal model 11% indicated that they would use a full internal model 34% expect their model to vary by licence 10% are still undecided Approach to economic capital aggregation No clear preference evidenced by the respondents to this question with components of Monte Carlo and Copulas as well as Correlation matrices being mentioned. 47

25 Economic Capital Calculations Practices in various countries China Current statutory solvency capital regime: Solvency I (), which takes the general form of x% of statutory reserves and y% of NAR Under development: C-ROSS (), which is similar to the Solvency II framework Larger companies tend to do EC calculations for internal use: Market value balance sheet approach Covers similar types of risks but have seen variations in inclusion/exclusion of widening of credit risk spread, property etc. Tend to use stochastic calculations for market risks (commonly 1,000 scenarios) Level of confidence varies by company Aggregation commonly done using variance-covariance approach Some smaller companies tend to do the following: Calculations performed by stressing GAAP profit for different risks Shock factors developed by companies, basing on 90% to 95% of level of confidence Do not necessarily use stochastic calculations Aggregation also commonly done by using variance-covariance approach (some companies directly use or refer to the correlation matrix under Solvency II) 48 Roadmap for EC Implementation for Taiwan 49

26 Current Position in Taiwan Solvency capital and risk management regime (1) Statutory solvency capital Enterprise risk management Regime/Guidance Risk-Based Capital regime Description First introduced in 2003 Rule-based regime with prescribed risk factors for most risks Covers assets, insurance, interest rate and general risks Risk factors imply a 95% VaR over a 1-year time horizon Statutory minimum capital requirement 200% RBC Certain risk types not covered by the regime, e.g. catastrophe, credit risk spread Required capital calculation for interest rate risk follows deterministic calculations Follows statutory valuation basis Certain assets are valued at book or amortized book value and some at market value Liabilities are statutory liabilities calculated with locked-in assumptions, but current development towards market valuation of liabilities Half-yearly RBC reporting reports RBC position as at valuation date, but AA reporting requires projected RBC position in next 3 years with AA s recommendations on remedies if RBC ratio falls below statutory minimum First introduced in 2009 Proposed framework for enterprise risk management Companies are advised to start developing EC techniques and ORSA mechanism Explicitly sets out types of risks to be considered, e.g. market risk, credit risk, liquidity risk, operational risk, insurance risk, asset liability mismatch risk 50 Current Position in Taiwan ERM framework in risk management guidance 51

27 Current Position in Taiwan Solvency capital and risk management regime (2) Annual stress testing on financial position Regime/Guidance Description Prescribed scenarios and stress factors Results submitted to Taiwan Insurance Institute Scenario Details (Best Estimate) (Spread )*1.5 (Spread )+100bp (Spread ) *0.5 (Spread )-100bp 0.0% 40%( )(beta ) 40%( )(beta ) 10%() 10% 10%( )10% 52 Current Position in Taiwan Solvency capital and risk management regime (3) Scenario Details A () (5) ()10% 10%150% * * * SARS % S&P % 7.6% 10 4bp bp(9) % S&P % 14.0% 10 85bp bp(9) 53

28 Proposed EC Approach Short-term solution Develop a Taiwan Standard Formula with the following features: Features Types of risks included Risk types included in the Solvency II standard formula Time horizon 1-year time horizon Level of confidence Start with 95% level of confidence, in line with level of confidence underlying current RBC regime Scenarios Sets of real-world economic scenarios for market risks under each shock Assume zero correlation between non-market risks and market risks in the short-term, e.g. assume no correlation between lapses and market risks Relax this assumption in later phases of EC development see roadmap Calculation methodology Stress EEV for each risk Use variance-covariance method for aggregation Not too difficult to implement Choice of level of confidence not critical for short-term solution Centrally developed standard formula, economic scenarios and correlation matrix 54 Gap Analysis What is needed for short-term EC calculation approach (1) Calculation Requirements Types of risks included 55 Current Position Explicit guidance in on types of risks to consider Covers almost all risks in Solvency II Standard Formula What else is needed? Reflect these risks plus a few others in Taiwan Standard Formula Time horizon 1-year time horizon in annual stress testing Continue with 1-year time horizon in Taiwan Standard Formula Level of confidence RBC risk factors imply 95% level of confidence over a year Annual stress testing factors developed with reference to historic experience collected Scenarios Real-world economic scenarios (interest rates for a number of economies, a number of equity indices, property indices, FX rates) already exist from AA reporting Prescribed stress testing factors for nonmarket risks from annual stress testing All of these are centrally set and updated each year by the Taiwan Insurance Institute Calculation methodology Industry now familiar with TEV calculations (which will likely become an annual filing item) No aggregation of stressed scenarios under annual stress testing Start with 95% to be in line with current level implied by RBC Generate sets of real-world economic scenarios following chosen level of confidence (i.e. 95%) for market risks under each shock Assume zero correlation between non-market risks and market risks in the short-term, e.g. assume lapse shocks have no impact on any market risks To be centrally developed for the industry Move one step forward to calculate EEV and different stressed scenarios of EEV Use variance-covariance approach to start with

29 Gap Analysis Comparison of risk types in ERM guidance and S2 standard formula Explicitly mentioned in Indirectly mentioned in 56 Gap Analysis What is needed for short-term EC calculation approach (2) Calculation Requirements Valuation assumptions Current Position Readily available best estimate assumptions Dynamic assumptions in place from AA reporting Projection model Set up with asset and liability projection capability Set up to perform stochastic and dynamic calculations What else is needed? Well prepared for EC calculations Not expecting too much extra work Well prepared for EC calculations Not expecting too much extra work for EEV approach 57

30 Proposed EC Approach Roadmap for later developments Short Term Medium Term Long Term Types of risks included Time horizon Level of Confidence Scenarios Risk types included in the Solvency II standard formula Start with 95% level of confidence, in line with level of confidence underlying current RBC regime Assumes zero correlation between market and non-market risks Scenarios generated centrally for the industry following Taiwan SF Companies may consider doing their own risk identification and include company-specific risks in calculations 1-year time horizon Level of confidence should be aligned with purpose of calculation, e.g. 1) Achieving a particular level of credit rating; 2) level set by regulator suitable for the industry Relax the short-term assumption, e.g. consider correlation between lapses and market risks As companies move to using company-specific risks in calculations, start to derive company-specific shocks As the industry moves to a market value balance sheet, generate riskneutral scenarios for calculations Calculation Methodology Use variance-covariance method for aggregation Bring in solutions such as killer scenario to address non-linearity issue in variance-covariance app. May consider moving to Monte Carlos + proxy modelling approach Actuarial Projection Models Enhance models to allow for different shock factors by products / product groups Use algorithm grouping to reduce number of model point s for products which require stochastic runs for TVOG As calculations change to a marketconsistent basis, some modeling work may be required 58 Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee, and its network of member firms, each of which is a legally separate and independent entity. Please see for a detailed description of the legal structure of Deloitte Touche Tohmatsu Limited and its member firms. About Deloitte Global Deloitte provides audit, tax, consulting, and financial advisory services to public and private clients spanning multiple industries. With a globally connected network of member firms in more than 150 countries, Deloitte brings world-class capabilities and high-quality service to clients, delivering the insights they need to address their most complex business challenges. More than 200,000 Deloitte professionals are committed to becoming the standard of excellence. About Deloitte in Greater China We are one of the leading professional services providers with 21 offices in Beijing, Hong Kong, Shanghai, Taipei, Chongqing, Dalian, Guangzhou, Hangzhou, Harbin, Hsinchu, Jinan, Kaohsiung, Macau, Nanjing, Shenzhen, Suzhou, Taichung, Tainan, Tianjin, Wuhan and Xiamen in Greater China. We have nearly 13,500 people working on a collaborative basis to serve clients, subject to local applicable laws. About Deloitte China In the Chinese Mainland, Hong Kong and Macau, services are provided by Deloitte Touche Tohmatsu, its affiliates, including Deloitte Touche Tohmatsu Certified Public Accountants LLP, and their respective subsidiaries and affiliates. Deloitte Touche Tohmatsu is a member firm of Deloitte Touche Tohmatsu Limited (DTTL). As early as 1917, we opened an office in Shanghai. Backed by our global network, we deliver a full range of audit, tax, consulting and financial advisory services to national, multinational and growth enterprise clients in China. We have considerable experience in China and have been a significant contributor to the development of China's accounting standards, taxation system and local professional accountants. We provide services to around one-third of all companies listed on the Stock Exchange of Hong Kong. Disclaimer This publication contains general information only, and none of Deloitte Touche Tohmatsu Limited, its member firms, or their related entities (collectively the Deloitte Network ) is by means of this publication, rendering professional advice or services. Before making any decision or taking any action that may affect your finances or your business, you should consult a qualified professional adviser. No entity in the Deloitte Network shall be responsible for any loss whatsoever sustained by any person who relies on this publication. Hong Kong 2013 Deloitte Touche Tohmatsu 2012 Deloitte Touche Tohmatsu Limited