OpRisk. Loss Distribution Approach. Capital Allocation for Operational Risk Conference November 14 th - 16 th, Boston - G.

Transcription

1 Loss Distribution Approach Capital Allocation for Operational Risk Conference - G. Courage - OpRisk Global Risk Division/Risk Consulting

2 From Risk Information to Risk Capital Data Collection Loss Distributions for Single Risks Risk Map Integrated Annual Loss Density Function Target Risk Analyses Loss Data Exposure Data Interviews Experts Experience Press- / Expert opinion ERP-Systems CSA-Systems Near Misses-Syst. Knowledge- Systems Severity Frequency Volatility Conf. Level [%] Capital at Risk P&L/Balance Sheet Indicators WACC/Cash Flow/EVA VaR/CaR Risk Capital Balanced Scorecards Corporate Governance KonTraG Loss Distribution Approach/ G. Courage 2 Capital Allocation for Operational Risk Conference

3 Contents Prerequisites Treatment of Data Calibration of Parameters Loss Distribution Approach Integration of Risk Transfer Options Loss Distribution Approach/ G. Courage 3 Capital Allocation for Operational Risk Conference

4 A consistent Typology of Risks is the Prerequisite... War Act of Government Customer Intake and Documentation Strike/Riot/Civil Commotion Loss of Use Transaction Capture, Execution&Maintenance Customer/Client Account Management Unauthorized Activity Expropriation Diversity&Discrimination Monitoring and Reporting Suitability, Disclosure&Fiduciary Safe Environment - Third Party Product Flaws Advisory Activities Vendors&Suppliers Trade Counterparties Natural Disasters Political Risks Theft and Fraud System Security Employee Relations Improper Business or Market Practices Safe Environment - Workers Loss Distribution Approach/ G. Courage 4 Capital Allocation for Operational Risk Conference

5 ...to map Perils against the Definition of Operational Risk Market Strategy Operational Credit Financial Relationship People External Phys. Assets Technology Level 1 Unauthorized Activity Fraud, Embezzlement, Forgers, Infidelity Theft, Robbery,Computer Crime Other Criminal Activities Bombing Terrorism Strike/Riot/Civil Commotion Vandalism Natural Hazards Named Perils Defective Security System Electrical Overload Collision Fire/Explosion Lightning Sprinkler Leakage Overvoltage Mechanical Breakdown Collision of Aircraft/vehicle Pollution Property Damage Bodily Inury Employers Practice Liability, Discrimination or Harassment Financial Loss Storm Hurricane Hail Frost Flood Earthquake/Volcanic Eruption Landslide/Mudslide Avalanche Hardware Breakdown Software Breakdown Utility/Infrastructure Interruption Failure of Back-Up Failure of eexternal network Level 2 Level 3 e.g. Sprinkler, Fire Brigade Change in Tax Code Governmental Edict / Seizure of Assets / Blockage of Accounts Change in llegal Environment or Legislation Unsafe / Unhealthy Workplace Poor Water / Air Quality Wrong Ergonomic Design Lack of Control Lack of Reporting Exposure Measurement, Risk Quality, Early Warning Systems Level 4 Loss Distribution Approach/ G. Courage 5 Capital Allocation for Operational Risk Conference

6 ...and simultaneously to map Mitigation Options against Perils Market Strategy Operational Credit Financial Relationship People External Phys. Assets Technology Level 1 Unauthorized Activity Fraud, Embezzlement, Forgers, Infidelity Theft, Robbery,Computer Crime Other Criminal Activities Property Insurance 1.1 Fire Insurance 1.2 Extended Coverage 1.3 Increased Cost of Working 1.4 All Risks (Property) 1.5 CAR, EAR 1.6 Glass Insurance 1.7 Electronic Insurance 1.8 Transit Insurance 1.9 Insurance valuable items 1.10 Travel/Baggage Insurance 1.11 Fine Art Insurance 1.12 Business Interruption Insurance Natural Hazards Named Perils Defective Security System Electrical Overload Collision Bodily Inury Employers Practice Liability, Discrimination or Harassment Financial Loss Pollution Property Damage Special Lines 2.1 Bankers Blanket Bond (Fidelity/Crime/Computer Crime) 2.2 Burglary/Theft/Robbery 2.3 Kidnap and Ransom 2.4 Unauthorized Trading Hardware Breakdown Software Breakdown Utility/Infrastructure Interruption Failure of Back-Up Failure of external network Change in Tax Code Governmental Edict / Seizure of Assets / Blockage of Accounts Change in llegal Environment or Legislation Unsafe / Unhealthy Workplace Poor Water / Air Quality Wrong Ergonomic Design Lack of Control Lack of Reporting Casualty Insurance Level 2 Level General Liability 3.2 Professional Indemnity (Error&Omission) 3.3 D&O, Directors & Officers Liability 3.4 Property owner s liability Insurance 3.5 Environmental Impairment Liability 3.6 Workers Compensation 3.7 Employers Practice Liability 3.8 Personal Accident Insurance 3.9 Motor Insurance Level 4 Loss Distribution Approach/ G. Courage 6 Capital Allocation for Operational Risk Conference

7 How this works in practice? Operational Relationship People External Phys. Assets Technology Source: EFIRM (excerpt / last 5 years) 3 Direct Insurance Companies 2 Reinsurance Companies EFIRM Banks Insured Losses totaling to 3 billion US$ Unauthorized Activity Fraud, Embezzlement, Forgers, Infidelity Theft, Robbery,Computer Crime Other Criminal Activities Property Insurance 1.1 Fire Insurance 1.2 Extended Coverage 1.3 Increased Cost of Working 1.4 All Risks (Property) 1.5 CAR, EAR 1.6 Glass Insurance 1.7 Electronic Insurance 1.8 Transit Insurance 1.9 Insurance valuable items 1.10 Travel/Baggage Insurance 1.11 Fine Art Insurance 1.12 Business Interruption Insurance Natural Hazards Named Perils Defective Security System Electrical Overload Collision Bodily Inury Employers Practice Liability, Discrimination or Harassment Financial Loss Pollution Property Damage Special Lines 2.1 Bankers Blanket Bond (Fidelity/Crime/Computer Crime) 2.2 Burglary/Theft/Robbery 2.3 Kidnap and Ransom 2.4 Unauthorized Trading Hardware Breakdown Software Breakdown Utility/Infrastructure Interruption Failure of Back-Up Failure of eexternal network Change in Tax Code Governmental Edict / Seizure of Assets / Blockage of Accounts Change in llegal Environment or Legislation Unsafe / Unhealthy Workplace Poor Water / Air Quality Wrong Ergonomic Design Lack of Control Lack of Reporting Casualty Insurance 3.1 General Liability 3.2 Professional Indemnity (Error&Omission) 3.3 D&O, Directors & Officers Liability 3.4 Property owner s liability Insurance 3.5 Environmental Impairment Liability 3.6 Workers Compensation 3.7 Employers Practice Liability 3.8 Personal Accident Insurance 3.9 Motor Insurance Loss Distribution Approach/ G. Courage 7 Capital Allocation for Operational Risk Conference

8 Generating reliable Data from multi diversified Risk Information ( DataMining ) Source of Information Classification Filtering Norming Scaling Synchronization Risk Analyses ERP Systems CSA Systems Early Warning Systems Loss Data (internal) (external) Exposure Data Interviews Experts Experience Press- /Expert opinion Taxonomy from internal sources from external sources Loss Distribution Approach/ G. Courage 8 Capital Allocation for Operational Risk Conference

9 What (Loss describing) Data to collect (1) Month/Y ear Adusted Net Loss (Discounted Currency adusted incl. Risk Transfer) Gross Loss Severity Loss Event Type Date of Loss Country Currency Business Line Exposure Indicator(s) Type of Relief / Policy Loss Effect Type Risk Transfer / Relief Indicator(s) e.g. Premiums / Limits Frequency Time Trigger Loss Distribution Approach/ G. Courage 9 Capital Allocation for Operational Risk Conference

10 What (Loss describing) Data to collect (2) - Required Data (Proposal) - Above minimum threshold Gross loss Net loss Currency Country of occurrence Date of occurrence Event Type / Risk Category (RC) (at least level 2) Business Line (BL) Loss Effect type (LE) Above Expected Loss threshold Event Type / Risk Category (more detailed, level 3 or additional levels) Causative/ Contributory factor(s) (CC) Product/ Process/ Function type (PP) Type of insurance coverage / Relief Type (RT) applicable (e.g. bankers blanket bond, property, etc.) Date of discovery Date of insurance recovery Status of loss (open/ closed) Value of Exposure Indicators at time of loss (e.g. gross income, assets managed) (EI) Value of Relief Indicators at time of loss (e.g. insurance premium, limits, deductibles) (RI) Loss Distribution Approach/ G. Courage 10 Capital Allocation for Operational Risk Conference

11 Continuum of Approaches Approach Capital Charge Basic Standard K g K g, i = EI α = EI i β i K P P i RT, i = i (1 ) Li RT, p Capital Relief λ ( Lp Pp ) CBp K = i Method Premium Limit IMA K g, i, = EIi, PEi, LGEi, γ i, K RT RT, i, = Pi, pi, LRi, γ i, ( Li, Pi, ) CBi K RT, i, =, Premium Limit (Aggreg.) [ ] 1 N i. lee, i, ; Lagg, i, Pi, CBi K RT, i, = min, N i. Limit (eel) LDA K g = F 1 (0.99) EL agg Factors resembling actual bank specific risk situation Factors based on historical industry wide data Factors based on historical bank specific experience Notation: for single BL, RC, perfect RT mapping, without Credit Risk Adustment Loss Distribution Approach/ G. Courage 11 Capital Allocation for Operational Risk Conference

12 Calibration of Parameters - lower approaches - K g, i, = EIi, PEi, LGEi, γ i, EI Regression analysis MLE - Approach Gross-Loss Gross-Loss EI Loss Distribution Approach/ G. Courage 12 Capital Allocation for Operational Risk Conference

13 Calibration of Parameters - lower approaches - K RT RT, i, = Pi, pi, LRi, γ i, [ ] 1 N i. lee, i, ; Lagg, i, Pi, CBi K RT, i, = min, N i. EI, P, L Gross-Loss Net-Loss Loss Net-Loss EI P L Loss Distribution Approach/ G. Courage 13 Capital Allocation for Operational Risk Conference

14 There are specific Advantages and Disadvantages for each Approach Lower approaches Higher approaches facilitate the use: work with lower amounts of data simple formula approaches are conservative approaches as industry factors provide security margin require high implementation efforts require substantial data management (incomplete or insufficient data may lead to wrong assessment of actual risk situation) lead to lower risk sensitivity and to higher capital cost do not enable full use of Risk Management opportunities lead to more risk sensitivity and to lower capital costs enable feedback to improve Risk Management enable to manage risks enable to optimize risk transfer options Loss Distribution Approach/ G. Courage 14 Capital Allocation for Operational Risk Conference

15 Frequency and Severity Distributions are generated Severity classes equation: border i = 10 1 c i+ i 0 Frequency / Severity estimation fitting 1,0 0,9 0,8 0,7 Lognorm2(11,375; 0,63474) Accuracy of approximation by ChiSquare Values x 10^-5 0,6 0,5 0,4 0,3 0,2 0,1 0, Values in Thousands 90,0% 5,0% 30, ,4658 > Loss Distribution Approach/ G. Courage 15 Capital Allocation for Operational Risk Conference

16 But still there are Gaps within the Landscape - Changing the risk map - Shaping the risk landscape - Inclusion of external data Inclusion of worst case scenarios Inclusion of expert opinion Inclusion of risk transfer Frequency and severity distributions will change P Mathematical Model E Loss Distribution Approach/ G. Courage 16 Capital Allocation for Operational Risk Conference

17 External Data have to be included using Scaling Techniques Norming/Scaling equation: b EI Bank. X X Bank. X = X Bank. A 1+ a 1 EI Bank. A) Two parameters (a, b) allow various relations linear with different factors positive and negative exponential Possibility to stack multiple Norming/Scaling Parameters Possibility to generate anonymous losses and to merge data Parameter in relation to Lines of Banking Business Classes of Sizes of Banks Losses Bank A 100 / 1000 Bank B 100 / / 800 Bank C Norming Parameters e.g. EI, L, P, RQ Norming / Scaling Procedure Direct Scaling Anonymous Bank Bank X 1500 Loss Distribution Approach/ G. Courage 17 Capital Allocation for Operational Risk Conference

18 X Calibration of Parameters - Norming / Scaling - Bank. X b EI Bank. X = X Bank. A 1+ a 1 EI Bank. A) Regression analysis MLE - Approach EI Gross-Loss Gross-Loss EI Loss Distribution Approach/ G. Courage 18 Capital Allocation for Operational Risk Conference

19 How to account for Worst Case Scenarios? - High Impact / Low Frequency Events - Natural Disasters Bank of Credit&Commerce, 1991 Barings, Severity Scaled! P Confidence Level [%] 1 0,99 0,98 0,97 E 0,96 Frequency 0, Value at Risk [$ Mio.] Loss Distribution Approach/ G. Courage 19 Capital Allocation for Operational Risk Conference

20 Risk Quality does have a Substantial Influence - How to account for? - Increase in Loss Frequency and Severity Industry Standard Based upon requirements from Regulators Authorities Insurers (recommendations) Risk Control Department Risk Management Department Eventually Other Parties involved Decrease in Loss Frequency and Severity Sample: Fire Protection Annual Probability of Occurrence Source: Wuppertaler Berichte zum Brand- und Explosionsschutz Band 3 Loss Distribution Approach/ G. Courage 20 Capital Allocation for Operational Risk Conference

21 Adustment and Aggregation of Distributions - Overall Picture - Convolution/Aggregation Annual Loss Density Function (per Peril/ Loss Cause, Business Line) Quality Dimension Management Organization Technical Controlling OR Man. Stand. Other Business Adustment Factors Business Business Line A Line B Business Line C Distribution Scenario based assumptions using: Data Analysis approaches (empirically, parameterization) Expert Testimony (Delphi technique, bets ) Modeling (stochastic simulation, analytical model, Bayesian model, segmentation) Type of Distribution Probability Scaling Severity Scaling Scenario A Weibull 1,0 LogNormal 1,0 C,R Scenario B Pareto 1,5 Beta 1,0 C,I,R Scenario C Poisson 1,0 LogNormal 1,0 C,I,R Approved bylast Update Loss Distribution Approach/ G. Courage 21 Capital Allocation for Operational Risk Conference

22 Aggregated Annual Loss Density Function (Single Risk / Gross) Calculated Capital at Risk (Gross) Confidence Level = 0,99 % Unexpected Loss Expected Loss set equal to Mean No Mitigation Per Single Risk Per Business Line Confidence Level [%] 1,000 Mean=56194,94 0,800 0,600 0,400 Subect to Capital Charge 0,200 0, EL Values in Thousands 90% Value at Risk [$ Mio.] 0 322,32 UL 5% > Loss Distribution Approach/ G. Courage 22 Capital Allocation for Operational Risk Conference

23 Aggregated Annual Loss Density Function (Single Risk / Net) Calculated Capital at Risk (Net) Confidence Level = 0,99 % Unexpected Loss being reduced Expected Loss set equal to Mean Mitigation -->Prop. Insurance -->Deductibles -->Limits Per Single Risk Per Business Line 1,000 Mean=39448,57 Mean=56194,94 Mean=24329,62 Mean=35424,19 0,800 Confidence Level [%] 0,600 0,400 0,200 0,000 Subect to Capital Charge % 0 322,32 Risk Transfer Options: 1 Uninsured 2 Percentage Deductible per Annual Loss 3 Deductible per Annual Loss 4 Deductible per Single Loss Value at Risk [$ Mio.] 5% > Loss Distribution Approach/ G. Courage 23 Capital Allocation for Operational Risk Conference

24 From single Risks to aggregated Risks and VaR (Gross and Net) Risk Landscape Aggregated Annual Loss Density Function Integrated Annual Loss Density Function Confidence Level [%] Capital at Risk Bottom up figures/function taking into account: - loss prevention (avoidance, reduction) - effective near misses systems - self retention - quality factors = gross but excluding: - mitigation by risk transfer Operational Risk Capital after mitigation/reduction through risk transfer = net Loss Distribution Approach/ G. Courage 24 Capital Allocation for Operational Risk Conference

25 Overall Distributions Calculated aggregated Capital at Risk (Gross/Net) Confidence Level = 0,99 % Unexpected Loss Expected Loss set equal to Mean Mitigation on Single Risk Basis: --> Proportional Insurance --> Deductibles --> Limits Mitigation on aggregated Risk Basis: --> Stop Loss System --> Deductible --> Limit Per Business Line 1,000 0,800 0,600 0,400 0,200 0,000 Mean= Distribution for Gros (1) / Sum:/B2 Mean= Finding the precise Capital Charge Finding the optimal Risk Mitigation strategy Values in Millions Based on Simulated Data Loss Distribution Approach/ G. Courage 25 Capital Allocation for Operational Risk Conference

26 Summary - Loss Distribution Approach - The Loss Distribution Approach is/will be a feasible solution to: calculate operational risk capital including capital charge with a high degree of risk sensitivity integrate risk management quality and best practice standards into the overall process and therefore offer the opportunity to grant benefit for good risk management design risk mitigation by insurance as effective and comprehensive as possible Loss Distribution Approach/ G. Courage 26 Capital Allocation for Operational Risk Conference

27 Summary Discussion on Taxonomy and Data Collection (Standards) is coming close to an end: Final summit of involved parties (regulators, banks & insurers) and final approval through regulators suggested It is important and possible to reflect the effects of risk management and risk transfer in the risk landscape, henceforth in the capital charge Discussion on approaches to calculate capital charge and relief: Further analytic, scientific and practical work as well as open exchange of ideas needed. LDA still regarded as most risk sensitive approach: Prototypes show feasibility - Pilot implementations will prove practicability Loss Distribution Approach/ G. Courage 27 Capital Allocation for Operational Risk Conference