Scenario analysis 10 th OpRisk Asia July 30, 2015 Singapore Guntupalli Bharan Kumar Disclaimer Any views or opinions expressed are solely the presenter s and do not represent those of my current or past employers.
Agenda Scenario analysis Regulatory and/or business requirement? Best practices Developing scenarios Estimation of severities, frequencies Validation Capital estimate using scenario analysis Combining scenario analysis with stress testing Integrating scenarios in operational loss forecasts 2
What is scenario analysis? Scenario analysis typically involves the examination of rare, significant, yet plausible future events, taking into consideration the alternative possible outcomes for those events, and assigning probabilities to various scenarios APRA (Australian regulator) Scenario analysis is a process of obtaining expert opinion of business line and risk managers to identify potential operational risk events and assess their potential outcome. An effective tool to consider potential sources of significant operational risk and the need for additional risk management controls or mitigation solutions. Given the subjectivity of the scenario process, a robust governance framework is essential to ensure the integrity and consistency of the process. Scenario analysis is one of the four data elements that banks must incorporate into their AMA frameworks and modelling methodologies. BCBS 3 APRA: Australian Prudential Regulation Authority BCBS: Basel Committee on Banking Supervision
Scenario analysis is useful for.. Uses of scenario analysis Regulatory requirement Capital charge computation Operational risk management Coverage of tail risks Usage of scenarios Coverage of scenarios Process, assumptions, governance Role of scenario analysis in capital modelling: Direct model input One among other AMA elements Used only for validation of other inputs/elements Realisation of new risks (experienced externally) Manifestation of existing risks in new forms Predict/prepare for risks in changing environment 4
Scenario analysis: divergent practices Many banks use Scenario Analysis as a key operational risk tool Ad hoc basis Used only for risk measurement Conducting Scenario Analysis Second Line of Defence (LOD) leads and develops, or First LOD and topical experts lead SA process with guidance/challenge from 2 nd LOD Source: Review of principles for the Sound Management of Operational Risk BCBS 5
Scenario analysis: divergent practices Methods to gather scenario data Workshops involving multiple employees/units Individual meetings/interviews Questionnaires Voting Inputs in scenario analysis process Internal loss data External loss data Financial indicators Types of scenarios Group-wide scenarios affecting entire organisation Scenario specific to a business line, sub-group Source: Observed range of practice in key elements of AMA BCBS 6
Scenario analysis: best practices Granularity of implementation, governance Bank-wide, divisional and business unit levels Used only for risk measurement Governance committees to oversee Usage Assess existing controls, identify additional controls, monitoring mitigating plans Supplement RCSA process, focus on lowprobability/high-impact risks Enterprise-wide risk assessment process E.g. Catastrophes like earthquake, cyber attacks, impact on other risks like increase in defaults, market risks, lower revenue Few banks develop action plans from their scenario analysis 7 Source: Review of principles for the Sound Management of Operational Risk BCBS
Developing scenarios: approach Scenario analysis approach Methodology Process Capital model design Coverage, granularity Inputs collection Scenario parameterisation Frequency determination Loss quantification Desk research, preparation Workshops, interviews Computation Independent review/ challenge Sign-off Documentation Governance Roles and responsibilities of Committees, stake-holders Approval of model, change management Reporting, decision making 8
Scenario approach: dimensions Capital model design Coverage, granularity Validation of data inputs or outputs Incorporation in loss data Applying loss distribution to scenario data points Sole model driver Specific risks or organisation -wide Top-down or bottom-up approach Inputs collection Free form Structured through pre-defined templates Parameterisation Loss quantification Individual (one severity/one frequency); interval; percentile; modified percentile Experts driven or data driven Detailed High-level 9
Developing scenarios: key elements Severity/ Loss estimate Frequency estimate Elements to be covered Linkage to portfolio size Average and maximum ticket size: normal/ stress conditions Additional overheads Loss of revenue, compensation Regulatory strictures, if any Methods of estimate Internal Loss Data/ External Loss Data for body distribution; Scenarios for tail distribution Internal Loss Data and scenario distributions are combined by using simulation Elements to be covered Comparison with external loss data, internal loss data, related industry or government data Expert judgment Methods of estimate Individual or interval or percentile estimation method 10
Developing scenarios: validation Validation Inputs Completeness and integrity of data Cross verification with reported data Governance Model Output Roles, responsibilities, change management, assumptions/ parameters sign-off, documentation, audit trail Review clarity, reasonableness, consistency, granularity of methodology documentation Benchmarking with industry best practices Benchmarking with peers, industry data Back-testing of the model Sensitivity to change in input parameters 11
Stress testing framework (illustrative) Governance aspects of stress testing Sensitivity analysis Stress testing infrastructure Operational Risk Stress testing methodologies Severity Scenario selection Scenario analysis Reverse stress test Stress testing infrastructure Stress testing outputs Management actions Supervisory assessment Source: CEBS guidelines on stress testing 12
Stress testing: scenario analysis Different degrees of severity / frequency mild / moderate / severe Functional scenarios Increase in number of legal cases and consequent legal provision(s) Increase in employee turnover, delay in replacements/in-appropriate replacement Technical scenarios Illustration Element Normal scenario Stressed scenario Frequency distribution Poisson Negative Binomial Correlation Historical 0.99 Copulas Gaussian copula Students T with 1 degree of freedom 13
Scenarios: linkage to capital Element All Australia Europe Japan North America Internal Loss Data External Loss Data Scenario Analysis AMA elements: contribution to capital % contribution: 75 th percentile 50 36 46 16 83 45 45 48-40 84 93 75 85 38 BEICF 18-60 - 11 14 Source: Observed range of practice in key elements of AMA BCBS AMA: Advanced Measurement Approach BEICF: Business Environment and Internal Control Factor
Scenario analysis: capital estimate approaches Validation of data inputs or outputs Incorporation in loss data points Fitting/adjusting loss data distributions Fitting distribution, mixing with loss data distribution Sole model driver Validation of internal/external data points Scaling of extreme external loss data points Back-testing of capital output numbers Synthetic data points are added to loss data to address gaps/strengthen Shape of the distribution curve based on loss data is adjusted to cover scenario data Separate distributions for scenario and loss data derived Combining distributions by weights, Credibility theory or Bayesian model Scenario data is only input data into model Expert opinion/loss data used to adjust severities 15
Scenario analysis: other areas ICAAP and reverse stress testing Credit, market, liquidity and reputation risk scenarios; stress, reverse stress scenarios Business Continuity Management Dimensions: People, Process, Systems, Outsourcing Assessment: Critical/non-critical Insurance framework E.g. Crime policies, cyber risk, Directors & Officers Scenario analysis mechanism would determine risk appetite and the insurance limits New product launch E.g. launch of mobile banking application test scenarios: high concurrent logins, hacking 16 ICAAP: Internal Capital Adequacy Assessment Process
Scenario analysis: examples Ref # Scenario 1 Incorrect processing of transactions in core banking system and associated payment systems as a result of failed batch jobs or interfaces or change Linkage to External Losses Assumptions 2 Disruption in IT services due to non availability of primary and DR applications and systems Linkage to External Losses Basel Business Line/Loss Event category All business lines Execution, Delivery & Process Management All business lines Business Disruption & System Failures 17 DR: Disaster Recovery
Scenario analysis: examples Ref # Scenario 3 Information leakage of card holder data by employee/ vendors resulting in loss of data confidentiality and leading to major financial, legal & compliance issues, losses Linkage to External Losses Assumptions 4 Lawsuit against the company for harassment (mental, sexual), discrimination with regards to gender, age, nationalism Linkage to External Losses Basel Business Line/Loss Event category Retail Banking External Fraud All business lines Employment Practices and Workplace Safety 18
Scenario analysis: closing thoughts Scenario analysis is a powerful risk management tool Highlights major/catastrophic/solvency threatening risks Enhances awareness/understanding of risks and underlying drivers Facilitates to prioritise and evaluate mitigation/ control activities Enables migration to advanced measurement approach for capital computation and better capital management 19
20 Questions?