Wednesday 3 October 2001 John P Ryan Financial Condition Reporting Practical Aspects GIRO / CAS Convention 2001

Q:\client\ryanjp\2001\jr0460sh.ppt 2 Financial Condition Reporting - Practical Aspects The FSA’s view Assessment of individual risk Modelling operational risk Importance of tail dependency Relevance of risk measures Overlaying hard to quantify risks with a DFA model Use of insurance to reduce capital requirements

Q:\client\ryanjp\2001\jr0460sh.ppt 3 Institute of Actuaries paper on FCA Provides a framework for evaluating a company’s financial position in relation to the risk it covers. Concentrates on non-life insurance but covers the principles for all companies. It covers both readily quantifiable risks and those not so readily quantifiable e.g. management succession risks. The Profession’s response to the FSA proposal. FSA will apply to all financial Institutions. Corley Report also calls for FCR reports for Life Co’s

Q:\client\ryanjp\2001\jr0460sh.ppt 4 Risk Management Circle Effective control requires quantification

Q:\client\ryanjp\2001\jr0460sh.ppt 5 Individual Risk Assessment.

Q:\client\ryanjp\2001\jr0460sh.ppt 6 Methods of Modelling Risk Financial Risk - investment models Financial Liabilities - actuarial models All Other - as operational risk

Q:\client\ryanjp\2001\jr0460sh.ppt 7 Insurance Company Risks

Q:\client\ryanjp\2001\jr0460sh.ppt 8 Financial Risks

Q:\client\ryanjp\2001\jr0460sh.ppt 9 Asset Risks Value, Bad debt Modelling volatility Market value  Reinsurance  Market ?Valuation  Concentration ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 10 Liability Risks  Discounting  Unexpired Risks  Unearned Premiums  Outstanding claims / IBNR ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 11 Liability Risks  Mismatch ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 12 Financial Risks

Q:\client\ryanjp\2001\jr0460sh.ppt 13 Underwriting Risks Impact on pricing assumptions But varies by class  Concentration ?Acceptance ?? Growth / new classes  Pricing ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 14 Exposure Risks ?Policyholders’ Reasonable Expectations  Claims frequency / severity  Reinsurance  PMLs ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 15 Business Risks  Investment Strategy ?Mergers & Acquisitions  Expenses ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 16 Financing Risks  Gearing  Debt interest / repayment  Dividend commitments ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 17 Insurance Company Risks

Q:\client\ryanjp\2001\jr0460sh.ppt 18 Operational Risks But might come across evidence But might help with system requirements Could help assess some procedures But might come across evidence XManagement XTechnology ?Administrative XFraud ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 19 Operational Risks DFA / Market Analysis XReputation ?Data quality / availability  Planning ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 20 External Risks Possible future. Some known changes XConfiscation / Nationalisation XPolitical Taxation ?Social ?? Legal / Legislative ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 21 External Risks XRegulatory  Group structure ?Dependency ACTUARIAL ASSESSMENTRISK

Q:\client\ryanjp\2001\jr0460sh.ppt 22 Operational Risk ASSESSMENT OF OPERATIONAL RISK

Q:\client\ryanjp\2001\jr0460sh.ppt 23 Management and Business Risk Some can be modelled using econometric or causal modelling techniques Some are really risks for shareholders rather than capital issues Stress testing can be a useful quantification technique Insurance often cannot be used for this type of risk

Q:\client\ryanjp\2001\jr0460sh.ppt 24 Quantification of Operational Risk Operational Risk Delphi Techniques Produce a Model Quantify Risk Corroborate Results Collect Data IndustrySpecific Model Quantify

Q:\client\ryanjp\2001\jr0460sh.ppt 25 Development of loss curves Budgeted Loss Amount of Loss Expected Level of Loss Probability of Loss Based on data

Q:\client\ryanjp\2001\jr0460sh.ppt 26 Quantification of Operational Risk It is more complex than pricing conventional insurance risk The risks are more under control of the institution than many insured perils Changes in practice can have a material impact Organisations do not like to admit to Operational Risk losses Some are not readily amenable to statistical analysis e.g. management succession risk

Q:\client\ryanjp\2001\jr0460sh.ppt 27 Scenarios Distributions may not be the best approach to evaluating certain types of operational risk Test the survival of the organisation to adverse scenarios Especially suitable for “people risks” e.g. succession planning

Q:\client\ryanjp\2001\jr0460sh.ppt 28 Data based approach Not many databases around Not all losses are disclosed Controls and mode of operation may render some data to be inappropriate Low frequency / high severity risk requires a different approach Some “operational risks” are budgeting items

Q:\client\ryanjp\2001\jr0460sh.ppt 29 Internal Interviews    Example Output from a Large Loss Study    50% 60% 70% 80% 90% Reliability of Loss Estimates Own Claims Claims Size (US$ billions)    External Interviews       External Research

Q:\client\ryanjp\2001\jr0460sh.ppt 30 Development of loss curves Budgeted Loss Amount of Loss Expected Level of Loss Probability of Loss Based on data Including interviews

Q:\client\ryanjp\2001\jr0460sh.ppt 31 Data for loss curve Loss Limit Amount of Loss in Data Amount of Losses in Data & Interviews

Q:\client\ryanjp\2001\jr0460sh.ppt 32 Questions The difficulty is the need to estimate the right tail in a skew distribution How good is the left of the curve at predicting the right tail Use of Bayesian statistics or credibility theory What distributions fit the data What techniques are best at supplementing the data for “missing large claims”

Q:\client\ryanjp\2001\jr0460sh.ppt 33 Conclusions - Data based methods Data based methods are the traditional actuarial technique for insurance claims, they are intuitively acceptable There are major deficiencies in using data based methods for operational risk not present in insurance data The major problem is non-reporting of large claims Useful check on the reality of other methods

Q:\client\ryanjp\2001\jr0460sh.ppt 34 What are the other methods? Delphi techniques Decision trees and casual modelling ? Fuzzy Logic ? Others ? Use data bases for left side and other techniques for right side

Q:\client\ryanjp\2001\jr0460sh.ppt 35 Delphi Technique Key Drivers of Business Unit What are the risks to each Business Unit What are the likely frequency If loss occurs what is the likely cost Fit curves following interview technique Model uncertainty Combine all results

Q:\client\ryanjp\2001\jr0460sh.ppt 36 Core Operational Risk Analysis Framework involves Business Process and Resource / Risk Classes Resource Classes Physical Assets TechnologyPeople Relationship (Liability) Other External Business Process Transactional Process Business Management Reputation

Q:\client\ryanjp\2001\jr0460sh.ppt 37 Risk profiles are linked to financial measures using a financial value tree approach Fixed AssetsDistribution Costs Price Taxes Free Cash Flow Op. Cash Flow Investment Gross Margin Revenues Volume Working Capital Event RiskFinancial RiskDistribution ChannelBusiness Risk

Q:\client\ryanjp\2001\jr0460sh.ppt 38 Importance of the risk measure Var implicitly assumes “elliptic risks” Operational risk does not satisfy this condition Market Risk needs to be frequently updated hence the importance of Var Operational Risk does not change rapidly Hence “equivalent Var” will not change rapidly

Q:\client\ryanjp\2001\jr0460sh.ppt 39 Adding the efficient frontiers will overstate the costs for a given risk as no adjustment is made for diversification credits This at a minimum changes the choices or risk loadings even if all strategies are ranked in the same way Risk Cost Adding Each Separately Evaluating risks altogether Consolidated largely independent risks Consolidated with many tail dependent risks

Q:\client\ryanjp\2001\jr0460sh.ppt 40 Risk Measures Var works well for symmetrical risks ECOR is better for skew risks such as most insurance risks A coherent measure needs to be used across the group as a whole Beware of tail dependency Other constraints are also needed such as a requirement to maintain a credit rating

Q:\client\ryanjp\2001\jr0460sh.ppt 41 ECOR reflects both the probability and the severity of ruin ECOR is the present value of expected deficits in excess of economic capital ECOR is derived as the probability of a loss times the severity of the loss In today’s dollars Sum of all loss events Reflects solvency risk tolerance measure and assigned capital The “ECOR ratio” is the ECOR divided by the present value of expected customer payments ECOR is a better solvency risk measure than probability of ruin because it reflects the cost of ruin, not simply the likelihood of event

Q:\client\ryanjp\2001\jr0460sh.ppt 42 Why Does This Matter? Combined Operational Risk Investment Risk Var ECOR The RBC’s are very different for different approaches

Q:\client\ryanjp\2001\jr0460sh.ppt 43 Coherent Risk Measures To be coherent a risk measure (p) must satisfy four conditions: (i)Translation Invariance p(x + .r) = p(x) -  (ii)Sub additivity p(x 1 + x 2 )  p(x 1 ) + p(x 2 ) (iii)Positive homogeneity for  o p( x)= p(x) (iv)Monotoniaty If x  y p(Y)  p(x) Var fails the sub additivity property

Q:\client\ryanjp\2001\jr0460sh.ppt 44 Insurance to cover Operational Risk This is a non-trivial subject. Basel has many doubts.

Q:\client\ryanjp\2001\jr0460sh.ppt 45 Coverage Gaps If complete cover is not available then capital will need to be held against remaining risk Insurance should mitigate operational risk cost and so should be allowable Operational Risk models would need to be run with and without insurance Contracts with material exclusions may not mitigate overall capital requirements much All Risks Cover is preferable Much operational risk violates an underwriting rule that the insured should not be able to manipulate his loss experience Some risks may not be insurable e.g. management succession risk

Q:\client\ryanjp\2001\jr0460sh.ppt 46 Claims Disputes Some financial impact as a dispute creates coverage gap Change insurance practice of conducting investigations at point of claim to investigating at point of sale Financial Enhancement Ratings (FER) Different in conditions (DIC) coverage