Home

User Manual: Product Name

Contents

1. Oracle Financial Services Software Confidential Restricted 45 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Index A Advanced Analytical Infrastructure Analysis across Models Anderson Darling Statistic B BEICF Scale Adjustment BEICF Shape Adjustment Bound Data Bucketing Business Metadata Documents Business Processor C Capital Calculation Chi Square test Conditional Value at Risk Copula failure errors Credibility Factor D Data Model Data Transformation Dataset Deductible Model Dependency Between Loss Frequencies E Error Messages Event Type analysis Expected and Unexpected Loss F Fct Operational Loss Filter Frequency Modeling G Goodness of Fit tests Oracle Financial Services Software Confidential Restricted 46 39 32 18 18 11 vi 39 15 38 34 21 38 22 38 12 37 41 32 38 23 37 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Graphical reports 32 H Hierarchy 39 I Insurance Eligibility 3 Internal Reporting Group 4 Internal RG 3 K Kolmogorov Smirnov Test 5 L Line of Business analysis 32 Log Transformation 10 Loss Data Frequency 18 Loss Data Severity 19 M Measure 39 Metadata 39 Missing Value treatment 22 Model Execution Bucketing Frequency 8 Model Execution Currency Conversion 8 Model Execution Frequency Generation 10 Model Execution Frequency Modeling 8 Mod
2. However in this case OREC application follows cumulative percentile calculation for each data point to find the correct value Due to this approach quartile points are always part of the loss data and thus they are approximately 25 5096 and 75 3 How are VaR numbers calculated VaR algorithm attempts to retain the respective weightage of various RGs with respect to each other when frequency for each RG is simulated Due to this when simulated losses are summed across the RGs to get bank level VaR it avoids forming simultaneous peaks for each RG This peculiar structure results in diversified VaR 4 Is model execution in Sandbox a compulsory step Sandbox is a trial area You can define various models and analyze the output Once the model is defined as per business requirement deploy the same for production set up If you are certain about the model execution results then deploy the same directly However it is advised to go through the sandbox as execution errors data non availability choice of distribution will be eliminated following this step 5 How to find which distribution fits the data properly To begin with for frequency analyses select the adaptive method In this method OREC application chooses the method which is more suitable for the data When it comes to severity analysis use the Modeling Framework Platform to get statistics regarding Goodness of Fit for a given data under various methods 6 Why do we
3. s o RORECSANDO v H Unified Metadata Manager EF Rules Framework H S Operations El System Configuration Administration amp Advanced Analytics Infrastructure Sandbox Definition Sandbox Data Maintenance Application Variable Stress Testing gA AMHM UMM Offline Population ff Limit Management Figure 2 OREC modeling Click Model Management to define a model Click 5 to create a new model Enter the Model Name Model Description Technique Model Objective in the Model Definition screen The seven tabs for data modeling are displayed and explanation of each tab is provided in the following sections 2 4 1 Tab 1 Capital Calculation The inputs in Capital Calculation Settings tab are the inputs given before modeling These are mandatory fields to be updated for Operational Risk Economic Capital calculation There are three sections under the Capital Calculation Settings tab namely Reporting Groups Parameters and Options Oracle Financial Services Software Confidential Restricted 15 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Model Definition Model Management gt Model Definit 2 Model Details Model Name OREC Insurance OREC Insurance Model Description Model Objective Portfolio Operational Risk EC Estimation Technique Loss Distribution Approach Capital Calculations Loss Data Frequency Loss Data Severity Scenario Data Credibility Factor
4. Operational Risk Economic Capital Modeling Model Execution After OREC application model is defined request for execution is made The Model Execution is available at the bottom of the screen Using this option a message Model Execution Triggered is displayed The execution of the model requires the MIS date and the number of executions to be greater than zero Model Output For more information on Model Output refer to Operational Risk Economic Capital Reporting on page 32 Model Deployment You can deploy the model in the production infodom and execute it However in the production infodom an end to end process to execute the Operational Risk Economic Capital model should be defined Screenshots relating to model deployment are as follows Oracle Financial Services Software Confidential Restricted 29 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Financial Services Analytical Applications Infrastructure User Imuser ORACLE Sandbox SECSAND Model Deployment e Model Deployment g Modeling E T Model A Search A e Model Management Model Name Model Execution Status L Model Outputs amp Model Deployment etoeoti0 E Model Deployment is Model Objective Model Name Version Log Model Deployment Status p A Rik ONEC surance 0 View N Available EC Estimation E Figure 21 Model Deployment Screen 1 Model Deploy
5. Scenario Definitions tret rr ERU IHRER E A ERER EE 26 Fipure 18 Stress Definition Scteen ausie ete testo te aie toii ib e di toe ee ipio e ricino 26 Figures t9 Sandbox Definition cas os ELE a RAM LUNA AME AM EET 28 Figure 20 Operational Risk Economic Capital Modeling Business Model Browser sess 29 Figure 21 Model Deployment Sereen 1 ssh bp irae Rea Epio dips 30 Figure 22 Model Deployment Screen 2 encres as 30 Figure 23 Expected and Unexpected LOS usina 38 Figure 24 Data Warehouse Schemas eene treten tai re lauri tu aeo aeta acercas 39 Oracle Financial Services Software Confidential Restricted iv User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 List of Tables Table Common conta viii Table 2 Document Convention Sesars iii viii Table 3 Standard Reporting Group ettet cr 4 Table 4 Reclassification Example tette ente testate oete sento orte beoe es tee ee kUg ke o lakes oo pe Ue deese aesa ia 4 Fable 5 Scalimo ii Rule Brad Oti ii dd et e e tte eni edens 7 Table 6 Deductible Model icasssasssssssssasssctssssassssssestandsacatindasoatintessssgnissesssedsagsdisdsgssssndvasatiedsiessatieauasseatseasindeanssdudn eaboeaaibena 12 Table 7 Proportionate Model with Single Offset ssssssssesseeeeeeeene eene coronan rara rar arrancar 12 Table 8 Credibility Factoria ia a eec eed AA detecte ec etude teet nete 22 Table Error Messages totetedededenecteieidte ias 43
6. EVT This is because scenario data is not realistic since it is generated by running model by the bank unlike internal loss However scenario data is the potential loss data as it is perceived as forward looking data Scenario data is arrived at on the basis of user judgment of the risk and control assessments within the bank and the scale of operation There are two different approaches for modeling scenario severity data However these changes should be done in the tool matrix in config schema Formula Based This would be done by using the formulas as described below Log Normal Shape Qsn p 2 4 In t1 4 1 2 Qsn p 2 Scale in m Shape 2 where p Percentile t1 percentile value m mode Weibull Shape In In 0 5 In 1 p In t2 t1 Scale In 0 5 t amp 9 Where p percentile t1 percentile value t2 median Exponential Shape In 1 p t1 Where p percentile T1 percentile value Bound Data In this method OREC application considers the scenario and severity buckets of each Reporting Group RG OREC application takes the mid value of each bucket and populates as many times frequencies simulated for a given RG To avoid zero variance minimum two buckets are required for each scenario The steps in severity modeling of internal data for estimating shape and scale are also applied to this data The difference between formula based and bound method is the way parameters are estimated In formula based f
7. Near miss values are optional and included only in frequency modeling as they do not hold any severity values Tolerance level as per industry standards given as a user input should normally be between 1 and 5 BEICF adjustments should be in percentage value and is given as a decimal value input External Data External data and Scenario data are not subject to any outlier adjustments Near Miss Near miss values are excluded from any missing value transformations Scenario Data Scenario data is independent of time bucket definition as the frequency and severity data is obtained directly as a download This data is directly used for shape and scale determination Scenario data does not undergo log transformation Dependency Between Loss Frequencies Dependency is measured only for frequencies of the internal loss frequency data across RGs Time Bucket Input The ratio of Time Window and Time Bucket Length should be greater than 3 failing which the model will not be saved Severity Modeling EVT is best applied for data with extreme deviations from the mean or technically the data with fat tail Insurance Insurance data should be mandatorily provided EEL and Agg Data if Consider Insurance is updated as Yes in Capital Calculation Settings Tab Ideally the EEL deductible amount or the EEL percentage in case of Proportional Model should be given as a download Else Net EEL claim Loss Amount All contract ID s in an RG should h
8. Select BEICF Shape Adjustment Factor Adaptive BEICF Scale Adjustment Factor Binomial Include Near Miss Events Neg Binomial User Specified Figure 7 Frequency Distribution Selection You also need to specify the BEICF Shape Adjustment and BEICF Scale Adjustment The Business Environment and Internal Control Factors BEICF are those measures that are usually Oracle Financial Services Software Confidential Restricted 18 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 associated with the day to day operations which are high frequency or low impact HFLI events In contrast the scenario data typically considers low frequency and high impact LFHI operational events In Frequency Modeling the fitted distribution s parameter like shape and scale are calculated by Method of Moments or Maximum Likelihood The parameters are rescaled to represent a one year period The parameters are multiplied by the number of time buckets in a year to rescale the parameters Rescaled parameters are adjusted for BEICF shape and BEICF scale adjustments by multiplying the factors given as an input This BEICF shape and scale adjustments should be a value greater than zero and is used as a percentage increment to the calculated parameters In order to scale down the parameters the BEICF shape and scale should be less than 1 For example if the Shape parameter is 0 5 and BEICF shape adjustment is 10 then the adjuste
9. Severity Shape and Scale Analysis for Event Type The following reports need to be calculated both for Standard RG combination of Standard LOB by Standard ET as well as Internal RG combination of Internal LOB by Internal ET For example if a bank has 10 Internal LOB and 3 Internal ET then the number of Internal RGs would be 10 3 30 However for reporting purposes the number of Standard RGs would be 8 2 16 where 2 of the Internal ETs map to 1 Standard ET Operational Risk Economic Capital generates the following 43 reports Tabular Reports OR Dependency Matrix Correlation Value Standard OR Dependency Matrix Covariance Value Standard OR EC Across Internal Event Type OR EC Across Internal Line of Business OR EC across Internal Reporting Groups OR EC Across Standard Event Type OR EC Across Standard Line of Business OR EC across Standard Reporting Groups OR Goodness of Fit Statistics for Internal Reporting Group OR Goodness of Fit Statistics for Standard Reporting Group OR Key Stats for Internal Data Internal Reporting Group OR Key Stats for Internal Data Standard Reporting Group OR RC Across Internal Event Type OR RC Across Internal Line of Business OR RC across Internal Reporting Groups OR RC Across Standard Event Type OR RC Across Standard Line of Business OR RC across Standard Reporting Groups OR Key Stats For External Loss INT RG OR Key Stats For External Loss STD RG OR Key Stats of Int and Scaled
10. Table 9 Error Messages Oracle Financial Services Software Confidential Restricted 43 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Some peculiar observations are noted as follows All simulated scenario severity values are the It is expected that even though due to value restrictions same frequency in simulations can be repeated severity values sampled should be different totaling to different sum of values for each sampling For example for a given RG whenever a Increase the threshold of the single largest severity value in the particular frequency say 194 repeats Gross FACT reporting group input table to get different gross values loss value is also repeating Table 10 Peculiar Observations Oracle Financial Services Software Confidential Restricted 44 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Acronyms and Glossary Terms BEICF BSP DT EDA EL ET EVT LDA LOB RG Standard Reporting Group UL VaR Business Environment and Internal Control Factors Business Solution Packs Data Transformation Exploratory Data Analysis Expected Loss Event Type Extreme Value Theory Loss Distribution Approach Line of Business Reporting Group The LOBs and ETs defined by the bank can be different from those prescribed by the Regulator which is referred as the Standard Reporting Groups Unexpected Loss Value at Risk
11. Unable to This mainly arises due to force fitting Check the distribution estimates the Parameters for the selected or modify the time bucket definition Fitted Distribution within permissible range Frequency Modeling Variance When all the data points are of the same value variance will is Zero so the Data cannot be be zero Check or modify the data as well as time buckets fitted for the Distribution Type defined Failed to calculate EDA Outputs Failure of statistics like Mean Variance Kurtosis and in Frequency or Severity Skewness for either frequency or severity modeling Modeling Review the data Goodness of Fit and how good the fit is Oracle Financial Services Software Confidential Restricted 41 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Goodness of Fit test failed Failure of goodness of fit test indicates that data does not follow the specified distribution Check or alter the distribution selected or time buckets 242 After BECIF adjustment Scale Check or modify BEICF parameters Parameter Probability is not within the range 0 to Report file Review the data points for further actions the given distribution 245 The mean cannot be less than Review the data points or distribution selected for further variance for the given action distribution type The Mean or Variance cannot Review the data points for further action be null or zero The mean cannot be gre
12. also possible to map a Scenario to multiple Baseline Runs This integrated stress testing enables you to assess the impact of a Scenario across multiple areas The Stress Definition New screen contains three panels namely Stress Definition Details Specification of the Stress Name and Description and selection of the scenario to be mapped to the particular Stress Definition Model Variable Shock Mapping Viewing the mapping of Variable Based Shocks that form part of the selected scenario to the models in the Baseline Run Mapped Rule Shocks Viewing rules and their respective shocks from scenario e The saved Stress definition can be either edited or deleted Stress Definition Details ROREC Bond MKT Impact Run Stress Stress Name ROREC Bond MKT Impact Run Stress Stress Definition Desc Base Line Run ROREC Bond MKT Impact Run Scenario Bond mkt Impact Scenario Model Variable Shock Mapping 1to 1 oft as Model Name Version i v Bond Market Impact 0 1to loft 2 Mapped Rule Shocks 3i Baseline Entity Rule Model Name Model Bond Market Impact 0 Version Shock Name No Shock Definition Figure 18 Stress Definition Screen Oracle Financial Services Software Confidential Restricted 26 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Oracle Financial Services Software Confidential Restricted 27 User Guide Oracle Financial Service
13. between different Reporting Groups RG to avoid any duplicity of biased event types Flexibility to adopt multiple scenarios multiple insurance policies across multiple Reporting Groups e Data cleansing methods like outlier and missing transformations Impact assessment of extreme risk scenarios on the risk factors and capital estimation can be done through the Stress Testing Framework Oracle Financial Services Operation Risk Economic Capital Release 2 1 has been integrated with Oracle Financial Services Operational Risk out of box application from a data mapping perspective Oracle Financial Services Software Confidential Restricted 2 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 2 Understanding the Application In Oracle Financial Services Operational Risk Economic Capital OREC Release 2 1 the objective is to calculate the expected and unexpected loss arising due to operational risk by executing loss distribution modeling on the historical data of the bank The historical loss data provided as a download by the bank is transformed and loaded to Fct Operational Loss after the Reclassification Rules are executed The internal data external data and scenario data are transformed according to the Reporting Group RG in this table OREC application business model fetches RG level data from Fct Operational Loss to calculate Economic Capital EC The estimation of loss distribution begin
14. business user is required to specify whether it is an alternative complimentary scenario or aggregated scenario Credibility Factor can be defined as the weight assigned to the scenario data while computing EC The simulation pertaining to scenario data would be selected based on the credibility factor Oracle Financial Services Software Confidential Restricted 12 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 For example In case of Complimentary method if the Credibility Factor is specified as 40 and the number of simulations generated is 1000 then 400 simulated losses are picked at random from simulated loss data and 600 values are taken at random from simulated scenario data for the calculation of VaR CVaR EL and UL In case of Alternative method the loss severity amounts are replaced according to the simulation For example 40 of worst scenario simulated losses are replaced by loss simulation and then the computation of VaR CVaR is applied to the aggregated data In case of Aggregate method the loss severity amounts are summed up with the scenario simulated losses For example If loss severity simulations are 10000 and scenario simulated losses are 10000 then the number of simulations as per the aggregate method will be 20000 2 2 4 Stress Testing Stress is applied on the shape and scale parameters of the distributions The shape and scale parameters cannot be negative or zero The stresse
15. cannot be same Parameter percentile value in Self Explanatory formula based approach is 0 or 1 Hence calculation for Weibull or Exponential Distribution has Oracle Financial Services Software Confidential Restricted 42 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 A CC NN EEL Threshold is less than EEL Self Explanatory Deductible EEL Threshold has to be greater than EEL Deductible Amount Aggregate Threshold is less than Self Explanatory EEL Deductible EEL Threshold has to be greater than Aggregate Deductible Amount The Correlation Matrix is a Zero Check correlation value in the dependency matrix table Correlation Matrix This type of Check for the condition if values are not uploaded matrix is not valid The given data does not suffice When RGI say ET 3 LOB 4 has correlation value of 0 5 then the condition to form a matrix approach requires that LOB 4 ET 3 also has value 0 5 Correlation Matrix If this is not the case then matrix preparation fails Also if data has 4 RGs and correlation data is available only for 3 RGs then the matrix generated will not suffice Check or modify the correlation values The Correlation Matrix is not Update the stage dependency matrix for the correlation value available even though the Loss Data is available for it Unable to convert the Semi For Gumbel copula non zero positive values of correlation Definite of the User Specified are require
16. co efficient or Kendall s Tau The correlation technique to be used is to be specified Dependency is measured only for frequencies of the loss data across RGs You have to specify the frequency distribution selection and Tolerance Level in the adaptive model There are 5 options available in Frequency Distribution selection and any one can be selected The available options are Adaptive Poisson Binomial Negative Binomial and User Specified When any one of the first four options are selected then all the cells follow the selected distribution that is adaptive poisson binomial or negative binomial When adaptive distribution is selected the distribution is selected based on the input data and the best fit distribution The decision for the best fit distribution for RG is done according to the mean and variance of the data The tolerance level for adaptive modeling should be a percentage value and should be greater than Zero Model Definition Model Management gt Model Definition 2 Model Details Model Name OREC Insurance OREC Insurance Model Description Model Objective Portfolio Operational Risk EC Estimation Technique Loss Distribution Approach Capital Calculations Loss Data Frequency Loss Data Severity Scenario Data CredibiltyFactor Data Transformation Fiter Copula Copula Not Required Correlation Method Kendal Parameters Frequency Distribution Selection Poisson EN Tolerance Level for Adaptive Model
17. dated June 2006 as the risk of loss resulting from inadequate or failed internal processes people and system or from external events Basel II Capital Accord has increased its focus on credit risk and market risk and faces a greater challenge in developing operational risk modeling Globalization and deregulation clubbed with emerging sophistication of financial technology are making the activities of the bank more complex and extremely sensitive to different risks Besides credit risk market risk and interest risk operational risk affects the stability and the functioning of the bank considerably Operational risk management is responsible for providing a framework for identifying measuring monitoring and managing all risks within the scope of the definition of operational risk Operational risk as generally seen is qualitative in nature Banks and supervisors need to manage operational risk due to a growing number of high profile operational losses As proposed in the New Basel Capital Accord the Basel committee realizes that risks other than Credit and Market Risk substantially affect the risk profiles of the bank The new accord emphasizes on financial institutions to make Operational Risk assessment as one of the integral components of their Risk Management System Oracle Financial Services Operational Risk Economic Capital OREC application enables you to model the distribution of potential losses due to operational risk In this applicati
18. eR P dE roads tusanhdasi bebe 18 2 4 3 Iab 3 L0ss Data Severitycasi sous NA 19 244 Tab 4 Scenario D t ia sais di dear ua jen da Pedal ed dedi aad dne E dod e d uoa ud da dad A 20 2445 Tab b5 Credibility EdctOEs ce ee e t a a Eee RU FERRE FER RUE EUR eR ERR HR YER 21 2 4 60 Tab 6 Data Transformation sessi esee einn annnn nnne th anas ases enata dada ss sess easi dadas sss sss a sa ada asas sn 22 224475 T bZFilterzsi eiat n dd a can nud unii 23 2 5 STRESS TESTING OVERVIEW air eee a EE AR Ead 23 3 PREPARING FOR EXECUTION cccccsccccscsccsccccscsccsccccscsccessccecsceecscsecscsecscsecsesecscsecseseces 28 3 1 SET UP DEFINITION esse ctae rr Gece eret er oro rore save nce rea Fewer eR Cre o revive Eve e rre RE aeree Ve Re ER eva way Fe EE EE RR ERE 28 3 2 STAGINGUAREA cect 3 nr Dover de ot rete eti cate ove deed A ato edd eb HO oov oae ve uet esr a Mns e rc Aeneas 30 4 EXECUTION edie ade tee RES 31 5 OPERATIONAL RISK ECONOMIC CAPITAL REPORTING ccccsssccssscccssceescecesceees 32 FREQUENTLY ASKED QUESTIONS cccssssssssecccccccessssssecccccceeaessssseccceeeeeaasssseececceeeeaasseseeseees 34 ANNEXURE A THINGS TO REMEMBER cccccsssccssssccecsccsssccessccscsccecsccecscsesscsacscscscsacseseces 37 ANNEXURE B UNDERSTANDING KEY TERMS AND CONCEPTS eene 38 ANNEXURE C ERROR MESSAGES AND OBSERVATIONS o occccoccncoccnonccnnnccnnnccnnaccnanccnnnccnanccnnnss 41 ACRONYMS AN
19. exceeding the deductible but with an upper threshold equal to the liability threshold 0 Xi X X lt d d d lt X lt l X l d I lt X 965 780 0 0 NA 200 000 00 2 140 000 00 1 374 220 00 765 780 00 0 Table 6 Deductible Model Proportional Model Proportional Model is characterized by a percentage instead of an absolute deductible amount It is mandatory to have percentage value in the proportional model Liability threshold is optional The process for proportional model is the same as deductible model A proportional retention model unbounded is characterized by factor 0 and liability threshold 1 1 0 The net loss incurred by the bank under each simulated scenario is given by H Q Xj 0X X l X 1 9 I lt X 965 780 00 51 492 547 80 1 718 619 58 1 245 387 38 473 232 20 0 Table 7 Proportionate Model with Single Offset Deductible model with multiple offset also works on the same basis however deductible amount 1s constant unlike in percentage model Model Execution Loss Calculation The aggregated yearly losses for each of the RGs are calculated using Monte Carlo simulation separately for the loss data internal loss data plus duly scaled external loss data if any and scenario data Loss scenarios are generated in scenario generation modeling These simulated losses are then added to the RG and to the entity as a whole While providing the scenarios for frequency and severity of a RG the
20. for Frequency Modeling Use this option extensively to avoid errors like data does not follow distribution data not available for frequency modeling along with the choice of distribution Specify the confidence level for Economic Capital EC and Regulatory Capital RC computation Confidence level is ideally between 95 and 100 However RC for operational risk is calculated at one rate for example 97 5 and EC for operational risk is calculated at another rate for example 99 Accordingly input any percentage values greater than zero Random number seed A seed is a number used to initialize a pseudorandom number generator A positive integer is expected as an input in the OREC application Parameters Reporting Currency US Dollar Number of Simulations 100 Time Window in days 1800 Bucket Length 30 Confidence Level for Regulatory Capital 96 97 5 Confidence Level for Economic Capital 99 Frequency Random Number Seed 1 Figure 5 Parameters Options in Capital Calculation Settings Tab 2 Options Distribution Fitting Methodology Method of Moments Calculate Allocation Factor No Consider External Data No y Consider Scenario Data No v Consider Insurance No v Maximum Insurance Benefit in Figure 6 Simulation Settings and Options Screen Oracle Financial Services Software Confidential Restricted 17 User Guide Oracle Financial Services Operational Risk Economic Capital Releas
21. logistic Log Gamma v COMPOUND LOSS DISTRIBUTION MONTE CARLO SIMULATION COMPOUND LOSS DISTRIBUTION MONTE CARLO SIMULATION Y Deductible Model INSURANCE Proportional Retention Model VaR Conditional VaR AGGREGATE LOSS DISTRIBUTION Expected Loss Unexpected Loss ALLOCATION Figure 1 Functional Flow of Operational Risk Economic Capital Application 2 4 Operational Risk Economic Capital Product Process Flow There is only one dataset on which modeling is performed as business models are defined on a single dataset Selection of the dataset for Operational Risk Economic Capital modeling is not required as after choosing the technique Loss Distribution Approach the dataset is automatically chosen Once the business model is selected the dataset browser becomes invisible There are seven tabs for data input namely Capital Calculation Loss Data Frequency Oracle Financial Services Software Confidential Restricted 14 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Loss Data Severity Scenario Data Credibility Factor Data Transformation Filter To access these tabs refer to the following steps 1 Click Modeling on the LHS pane of OREC application in sandbox infodom shown in the following figure Connected to 2 Home
22. observed sample distribution with the expected probability distribution Chi Square Goodness of Fit test determines how well theoretical distribution such as negative binomial binomial or Poisson fits the experimental distribution This is a test that is particularly adept at determining how well a model fits the observed data It evaluates how close the observed values are to expected values as given by the model in question thereby determining the best fit For a given data set and distribution the better the distribution fits the data the smaller this statistic will be Usually it is compared to the prescribed p value confidence value 2 2 Modeling Attributes e Rule Framework Loss Data Capture Reclassification Oracle Financial Services Software Confidential Restricted 5 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Eligibility Insurance and Modeling Currency Conversion Scaling of External Data EC Allocation as a last logical step after modeling e Model Execution Currency Conversion Frequency bucketing Frequency modeling EDA and Goodness of fit Severity modeling EDA and Goodness of fit Scenario Modeling Insurance Loss Calculation e Stress Testing 2 2 1 Rule Framework Loss Data and Loss Threshold Capture Loss data can be of two types based on their source namely internal loss data and external loss data In operational risk i
23. of the reporting functionality in OREC Application Common Icons The common icons incorporated into OREC application are as follows Use this icon to add a new entry Use this icon to view the details of an entry Use this icon to edit details of an existing entry Use this icon to delete an entry m8IRg EH E Enter the name of an entry and click this icon to search for an entry E Use this icon to refresh the screen E Use this icon to select an entry to delete edit or view the entry Oracle Financial Services Software Confidential Restricted vii User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Use this icon to view and select details in the particular browser Use this icon to navigate between pages Table 1 Common Icons Document Conventions Certain practices have been incorporated into this document to help you easily navigate through the document The table given below lists some of the document conventions incorporated into this User Guide Conventions Description Bold User Interface Terms Italics Cross References Emphasis Table 2 Document Conventions The other document conventions incorporated into this User Guide are as follows e Oracle Financial Services Operational Risk Economic Capital Release 2 1 has been referred to as OREC application in this User Guide In
24. scenario data is modeled with respect to the available formula The conversion of formula based approach to bound based can be done in a tool matrix the resulting effects can be viewed in the front end for scenario data severity Tab Level Validation UI can be modified for tab level validation or model level validation Oracle Financial Services Software Confidential Restricted 39 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 During tab level validations the value input is validated for its feasibility before moving to the next tab ALE EE VALIDATIONS gt TRUE TABLEVELVALIDATIONS gt lt This field is used t ontrol the tab level validations If required value should be TRUE else FA ontain value or i ot ve tab level va datin will be enforced gt lt CURRENCYHIERCODE gt l This tag is specifies ency hiera to be used t splay values for reporting rre gt lt SCENARIOPARAMESTIMATIONTYPE gt Insurance Allocation Factor If insurance allocation factor is selected as Yes then the allocation percentage is expected as a download Number of Buckets The number of buckets for the simulated values should be edited in the config xml Refer to the screen shot pasted below lt NUMBER_OF_BUCKETS gt 10 lt NUMBER_OF_BUCKETS gt lt POPULATE_SIMULATION_TABLE gt Y lt POPULATE_SIMULATION_TABLE gt lt UPD_SANDBOX_RSKEY gt SELECT MAX DIM_RUN N_RUN_SKEY FROM DIM_RUN WHERE F_HISTORICAL Y A
25. this document a Note is represented as follows Important or useful information has been represented as a Note Oracle Financial Services Software Confidential Restricted viii User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 1 Introduction Oracle Financial Services Analytical Applications Infrastructure OFSAAI provides the core foundation for delivering the Oracle Financial Services Analytical Applications an integrated suite of applications that is configured on a common account level relational data model and infrastructure components Oracle Financial Services Analytical Applications enable banks to measure and meet risk adjusted performance objectives cultivate a risk management culture through transparency manage their policy holders better improve the bank s profitability and lower the costs of compliance and regulation All Oracle Financial Services Analytical Applications processes including those related to business are metadata driven thereby providing a high degree of operational and usage flexibility and a single consistent view of information to all users Business Solution Packs BSP are pre packaged and ready to install analytical solutions and are available for specific analytical segments to aid the management in their strategic tactical and operational decision making 1 1 Overview of the Application Basel regulation defines operational risk in Para 644 BCBS 128 Basel II
26. to judge the fit 8 DolIneed to re execute the Runs for reports Additional executions for reporting purposes are not required Refer to the Run executed in the batch execution area along with execution date and access the required report Additional runs after batch execution are not required to be executed even if you are using the dashboard or not Dashboards can be linked to either sandbox or production set up 9 How does a Random Number Seed function Simulation programs make use of a pseudorandom number generator that requires a seed If none is provided a System Generated seed will be used This is used to separate the different simulations within a single run Any positive integers can be a random seed 10 Which are the hierarchies on which I can apply the filter In model definition OREC application provides a wide range of hierarchies on which filters can be applied You have a choice of selecting those hierarchies that form a part of OREC modeling data set 11 How will the outlier scaling factor be used and what would be the range for the same Outlier is used to identify those data values which should logically be a part of tail ends OREC application provides inter quartile method for this Q1 and Q3 are inter quartile ranges Upper Hinge Q1 K Q3 Q1 Lower Hinge Q3 K Q3 Q1 K is outlier scaling factor Higher values of K widens the range and vice versa There are no common values for this However outlie
27. April 2012 Version number 1 1 Oracle Corporation World Headquarters 500 Oracle Parkway Redwood Shores CA 94065 U S A Worldwide Inquiries Phone 1 650 506 7000 Fax 1 650 506 7200 www oracle com financial setvices Copyright O 2012 Oracle and or its affiliates All rights reserved No part of this work may be reproduced stored in a retrieval system adopted or transmitted in any form or by any means electronic mechanical photographic graphic optic recording or otherwise translated in any language or computer language without the prior written permission of Oracle Financial Services Software Limited Due cate has been taken to make this Oracle Financial Services Operational Risk Economic Capital User Guide and accompanying software package as accurate as possible However Oracle Financial Services Software Limited makes no representation or warranties with respect to the contents hereof and shall not be responsible for any loss or damage caused to the user by the direct or indirect use of this User Manual and the accompanying Software System Furthermore Oracle Financial Services Software Limited reserves the right to alter modify or otherwise change in any manner the content hereof without obligation of Oracle Financial Services Software Limited to notify any person of such revision or changes All company and product names are trademarks of the respective companies with which they are associated Oracle Financia
28. D GLOSSARY TERMS cccccccssscccssccccsccnsscccsceesceccscencscecescecescencsceecsceessceecsees 45 Ion NA D Y 46 Oracle Financial Services Software Confidential Restricted iii User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 List of Figures Figure 1 Functional Flow of Operational Risk Economic Capital Application see 14 Figure 2 COREG Model tete onere totes tec dilemma 15 Figure 3 Operational Risk Economic Capital Model Definition Screen macacos 16 Figure 4 Reporting Group Relevance Setting Screen sssssssssssseseeeeneeeene tenente tenete 16 Figut 5 Pata Eter S ena acd t rd rt n RR 17 Figure 6 Simulation Settings and Options Screen sese tenentes 17 Figure 7 Frequency Distribution Selection eene nennen ener 18 Figured Severity Distribution SelectlOtia saccos ettet e endete es ete meena eRe 20 Figure 9 scenario Analysis AE ia 20 Fip te 10 Credibility Factor etcetera tnnt rr gk a td 22 Figure 11 Data Transformation EE 23 Figures 12 Filtet i cau HTML eu dtu etum M dt dm 23 Figure 13 Variable Definition Select iii tete eti ce I Hd tei AS 24 Figure 14 Variable Management Screen eee nter nter retener cercar ceros 24 Figure 15 Variable Definition reinician dades 25 Figure 16 Variable Shock Definition coincidido 25 Figure 17
29. Data Transformation Filter Reporting Groups Reporting Group Selection Standard v Reporting Group Relevance Setting Yes El Parameters Reporting Currency US Dollar Number of Simulations 100 Time Window in days 1800 Bucket Length 30 Confidence Level for Regulatory Capital 97 5 Confidence Level for Economic Capital 99 Frequency Random Number Seed 1 Options Distribution Fitting Methodology Method of Moments y Calculate Allocation Factor No Consider External Data No v Consider Scenario Data No Consider Insurance No wv Maximum Insurance Benefit in Figure 3 Operational Risk Economic Capital Model Definition Screen Reporting Group Select the type of RG on which EC modeling is performed that is whether the modeling is to be done on regulator prescribed Standard Reporting Group or on Internal Reporting Group as specified by the bank In the Reporting Group section Reporting Group Relevance Setting should also be updated OREC application supports processing on 56 RG s or a selected group of RGs specified in the Relevance Setting Grid Reporting Group Relevance Setting Employment Practices and Workplace Safety Apply All Yes v No y Yes v Yes v Yes v Yes v Yes v Business Disruption Clients Products amp Damage to Physical amp System Failures Business Practices Assets Execution Delivery amp External Fraud Internal Fraud Process Management Annuities stemming from Non lif
30. Ext Loss INT RG OR Key Stats of Int and Scaled Ext Loss STD RG OR Dependency Matrix Correlation Value Internal OR Dependency Matrix Covariance Value Internal Oracle Financial Services Software Confidential Restricted 33 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Frequently Asked Questions 1 How is EVT calculated by OREC Application OREC application segregates the loss data as per given time buckets For this time bucket mean and variance is calculated OREC application begins with 50 value of loss data Only those loss values are chosen which falls above the threshold point These newly found values are again segregated in time buckets OREC application then checks if mean is greater than variance for this data The same process is repeated with a jump of 1 till mean is greater than variance The upper threshold of percentile is 99 9 While calculating monetary value of threshold point OREC application follows the function similar to percentile in Microsoft Excel This helps in finding the exact percentile value and results in a threshold value which may not be a part of loss data 2 How does OREC application calculate the outlier OREC application calculates outlier by using Inter Quartile method This is done to find out those data points which divide the data in a sorted manner in 4 equal parts Thus data points which are at 2596 50 and 75 percentile of the data are required
31. Management Before model definition LOB amp ET Reclassification needs to be performed The LOB and ET information is either reclassified according to the regulator rules or according to the bank s own reclassification rule Oracle Financial Services Operational Risk Economic Capital Release 2 1 Calculation is a business model You should select Operational Risk Economic Capital modeling from the Technique screen under the Business Models category as shown in the following screen Oracle Financial Services Software Confidential Restricted 28 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 2 Technique Hierarchical Members Selected Members E Business Models Loss Distribution Approach E Credit Risk Market Risk Operational Risk Loss Distribution Approach Statistical Techniques Figure 20 Operational Risk Economic Capital Modeling Business Model Browser Defining Operational Risk Economic Capital Model in Modeling Framework MF Provide data input for the calculation of Operational Risk Economic Capital Input the Model Name Model Objective and Technique for modeling in Business Models Loss Distribution Approach is selected Model Name Loss Distribution Approach Model Description Operational Risk Economic Capital Estimation for Standard RG with Financial Services and Scaled External Data Model Objective Portfolio Operational Risk EC Estimation Technique
32. ND DIM_RUN FIC_MIS_DATE MISDATE lt UPD_SANDBOX_RSKEY gt lt CURRENCY_CONVERSION gt lt DATA_TRANSFORMATION gt Populate Simulation Table The simulated values should be populated based on the settings in the config If this is remarked as Yes then the simulated values will be updated in fet_ Operational Risk Economic Capital _simulation table Hierarchy code dropdown The hierarchy filter code for internal and external classification has to be specified in config xml lt HIERARCHY_FILTERS gt lt HIERARCHY gt lt HIERARCHY_CODE gt HLOB003 lt HIERARCHY_CODE gt lt HIERARCHY_DESC gt INTERNAL OR EXTERNAL lt HIERARCHY_DESC gt lt HIERARCHY_NODES gt Oracle Financial Services Software Confidential Restricted 40 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Annexure C Error Messages and Observations Error Messages Functional Error Messages that may appear in OREC application are displayed in the following table Through this table any additional help or suggestive actions to be taken when an error is displayed is provided Time Bucket Generation Failed Ensure that time buckets are defined properly and data falls No Time Bucket found within these time buckets Frequency Modeling Failed This mainly occurs as data does not follow distribution This is usually seen as a very large or negative parameter value or inability to estimate them as well Check the distribution an
33. OW TO USE THIS USER GUIDE ss 62 7 32 3 52 ia ade Dueb eade id ao VII COMMON ICONS EEE EE A O RP EUR R IQ N USE RR EPIS PORE ERO E CERO ERE ORE E EU POE APR rere QE TA VII DOCUMENT CONVENTION CNRC UT TT VIII 1 INTRODUCTION UE 1 1 1 OVERVIEW OF THE APPLICATION 22 creare ao eae C nh ta DXRR EE CE DR DRE a Ca wa dna 1 2 UNDERSTANDING THE APPLIGA FION eet ree ERR ERR RE ERE 3 2 1 PRE MODEL nd ea 3 2 1 1 Standard Reporting Group isses eese eene nena haa ness essa ta asa assa essa sa dadas sss essa sa ada asas essa da 4 2 2 MODELING ATTRIBUTES cue eoe ee Re ret a ye vet evo ruego Qu Fus ave Fey Sev ev oae a 5 2 2 1 Rule Framework Loss Data and Loss Threshold Capture isses ener nnns enean 6 2 2 2 Rule Framework Reclassification sisse esses eene nnns nne th naa nsns sea ta aaa assa essa sa ra daas sees saa 7 2 2 3 MoOdel Execution ii e ie a ad eR E E OH I E d ca RE a X RO RR E e eu nn 8 2 2 4 SES Testing iei ee iie A e ds d RS tees 13 2 3 OPERATIONAL RISK ECONOMIC CAPITAL FUNCTIONAL PROCESS FLOW DIAGRAM eee ener nnne 14 2 4 OPERATIONAL RISK ECONOMIC CAPITAL PRODUCT PROCESS FLOW ococccononononononononononononononononononononononononononenoneness 14 24 1 Tab 1 Capital Calculation Ra estote Vae dias reda aa pi dead eu aa eod de a aaa Ve nen so ee hee Ula 15 2 42 FADQ 1058 Data Frequency uu est trees een tos edo er ce seed ovd peo rer NET PAYS Re PE eH
34. Oracle Financial Services Operational Risk Economic Capital User Guide Release 2 1 April 2012 ORACLE FINANCIAL SERVICES User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 What s New in this Release This section identifies updates in the Oracle Financial Services Operational Risk Economic Capital Release 2 1 Oracle Financial Services Operational Risk Economic Capital Release 2 1 now supports additional 10 truncated distributions for modeling the severity of internal loss data and duly scaled external loss data The additional distributions are as follows Truncated Burr Truncated Exponential e Truncated Gamma e Truncated Gumbel Truncated Log Gamma Truncated Log Logistic Truncated Log Normal Truncated Pareto e Truncated Uniform Truncated Weibull Oracle Financial Services Software Confidential Restricted ii User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Contents WHAT SNEWJIN THIS RELEASE ss rna RR RR RR RR REA R RENE RRRR CR RR RPRR ER RR RR RR RR RRRRRPRRS PME REPRE RR RARE II ABOUT THE GUIDE 5 eoeGebeecmiicennmiimimemipmmmie impri VI SCOPE OF THE GUIDE ss eot seri eee pte eret E o Y mede e v a Fever Veo test Ga eo over eodem yup ire creo tease as VI AUDIENCE diodos datan reet estesees dese ANE As eb de vena PURN BARON eaaet debat ene BANI RER Pe Eb aE evene NOES Eai VI WHERE TO FIND INFORMATION eee e aee voee e abate VI H
35. Table 10 Peculiar Observations tette trente rto re tauri cabecita scr 44 Oracle Financial Services Software Confidential Restricted V User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 About the Guide This section provides a brief description of the scope the audience the references the organization of the User Guide the common icons in the application and conventions incorporated into the User Guide The topics in this section are organized as follows Scope of the Guide Audience Where to Find Information e How to Use this User Guide e Common Icons e Document Conventions Scope of the Guide The objective of this User Guide is to provide a comprehensive working knowledge to the users on Oracle Financial Services Operational Risk Economic Capital Release 2 1 This User Guide is intended to help the user understand the key features and functions of Operational Risk Economic Capital application and use the application effectively However this User Guide is not meant to provide guidance on how to install and use Oracle Financial Services Analytical Application Infrastructure OFSAAI This User Guide is also not meant to provide details on installation of Oracle Financial Services Operational Risk Economic Capital Release 2 1 data model Audience This manual is intended for the following audience Technical Analyst This user ensures that the data is populated in the relevant tables
36. age Shift v Time Window Size Shock Curve Time Points bej 4 to1 of 1 Shock Values To 0 Day 15 Figure 16 Variable Shock Definition Scenario definition A Scenario is a set of multiple Variable shocks and Rule shocks These shocks may be Rule Based Shocks or Variable Based Shocks You can define a Scenario using a combination of Rule Based Shocks and Variable Based Shocks Navigate to New Scenario Definition which displays a new browser with Variable Name Oracle Financial Services Software Confidential Restricted 25 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Shock Name and Variable Type This would also display the list of shocks defined earlier The selection of multiple shocks together forms a new scenario This combination results in a New Scenario name cenario Details Scenario Name Economic Slow down 0 5 Economic Slow down 0 5 Scenario Description Variable Shocks gt Variable Name Y Frequency Scale Parameter Loss Data Rule Shocks si Baseline Entity Rule Model Name m 1to2of3 Shock Name Shock Description cy Scale Shock by 35 Bio 0160 of 0 Shock Name Shock Description Figure 17 Scenario Definition Stress Definition Stress Definition allows you to specify either a stand alone execution of a Scenario or maps a Scenario to a Baseline Run thereby creating a Stress Run It is
37. al EC figures the Undiversified EC represents EC on a stand alone basis for that particular RG and the Allocated EC represents EC as allocated at bank level Insurance Eligibility Model definition requires a certain eligibility criteria for insurance These eligibility criteria are handled by the Rules framework as follows Insurance provider has a minimum rating of A or an equivalent rating Policy should have a residual term of greater than 1 year Policy has a cancellation notice period of at least 90 days The insurance coverage for policies with a residual maturity of less than a year but more than 90 days is reduced by a specified hair cut The value of insurance is reduced by the hair cut The current release has incorporated multiple insurance policies with multiple RGs Insurance policies that are eligible or those which have passed the eligibility criteria are applied To change these setting from 90 1 year to other values make changes in the table at insurance contract level To change other settings modify the existing rules set up For more information on table structure refer to the technical metadata worksheets Oracle Financial Services Software Confidential Restricted 3 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 2 1 1 Standard Reporting Group The standard LOB types and ET prescribed by regulators are Payment and Settlement Damage to Physical As
38. ample If the shape parameter is 0 5 scale parameter is 0 9 and the Time bucket T is 4 3 then the scaled shape and scale up to 1 year would be 2 15 0 5 4 3 and 0 9 as the scale parameter does not undergo any adjustments The BEICF Adjustments given in the front end should be adjusted to the parameter by multiplying the factors to the respective parameters Shape 0 5 Time Bucket T 4 3 Scale 0 9 Rescaling Parameter T Shape 2 15 Oracle Financial Services Software Confidential Restricted 8 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 BEICF Shape 1 5 Scale 0 9 does not undergo any rescaling BEICF Scale 2 1 New Parameter Re scaled parameters BEICF Adjustments Shape 3 225 Scale 1 89 External Data is used for Severity modeling only and not for Frequency modeling Near miss values are those frequencies of losses which have not occurred but have the tendency of occurring Hence they do not have any severity values attached to it The use of near miss values is restricted to frequency modeling and is optional for the modeler Near miss values are not considered in severity modeling Additionally if a RG has all losses as near miss then the RG is excluded from the modeling process Model Execution Severity Modeling Severity data is assumed to be independent of frequency data while modeling The severity modeling process begins with the scaling of exte
39. apital Release 2 1 adequacy A what if analysis enables strategic planning in areas such as customer acquisition capital planning fund management and many more To access the Stress Testing framework click the option Stress Testing in the LHS pane shown in the following figure Connected to ORECPROD se a Home Unified Metadata Manager EY Rules Framework Operations System Configuration Administration amp Advanced Analytics Infrastructure Sandbox Definition Sandbox Data Maintenance Application Variable Modeling Stress Testing gf AMHM UMM Offline Population Figure 13 Variable Definition Selection To define a new variable which has to be stressed select Variable as shown in the preceding screenshot In the Variable Definition New E screen the name and description of the new variable is to be defined Enter a suitable name and description of the new variable in the Variable Name and Variable Description fields respectively From the Variable Type dropdown list select Idiosyncratic Variable Based on the drop down list select Measure to create a variable based on Measure The Variable Set Management pane is displayed on the RHS of the screen Variable Management ariable Management Search RA npn Variable Name Variable Management ES y 1 to 10 of 14 ES O Variable ID Varia
40. as per the specifications The user executes schedules and monitors the execution of Runs as batches Business User This user reviews the functional requirements and information sources like reports Data Analyst This user cleans validates and imports data into the OFSAAI Download Specification format Administrator The Administrator maintains user accounts and roles archives data loads data feeds and so on The administrator controls the access rights of users Where to Find Information For additional information on Oracle Financial Services Operational Risk Economic Capital Release 2 1 refer to the following documents Business Metadata Documents These documents are grouped into two sets as follows Oracle Financial Services Operational Risk Economic Capital Release 2 1 Business Metadata xls This document contains the definitions of the Business Metadata like Measures Business Processors Hierarchies Hierarchy Attributes Aliases Derived Entities and Datasets present in OREC Application Oracle Financial Services Operational Risk Economic Capital Release 2 1 Rule Metadata xls This document contains the definitions of Rules Pooling Optimizer and Processes Oracle Financial Services Software Confidential Restricted vi User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Technical Metadata This document contains the definitions of the Table to Table T2T used in va
41. ate method loss simulation values are summed up with scenario simulations values For example if there are 10000 loss simulations and 10000 scenario simulations then total number of simulations by aggregate method is 20000 simulations disabled Default correlation value is used only when loss data correlation is used Capital Calculations Loss Data Frequency Loss Data Severity Scenario Data Credibility Factor Data Transformation Parameters When scenario data is specified as No in the capital calculation tab the Credibility tab will be Filter Credibility Approach Alternative Credibility Factor Figure 10 Credibility Factor 2 4 6 Tab 6 Data Transformation The parameters in the data transformation include outlier determination outlier scaling factor and missing value treatment Outlier_ Determination The outliers can be identified across all the RGs by selecting Interquartile Method or can be made across each RG by selecting User Specified or this process can be excluded by selecting Not Required Missing Value treatment 3 options to replace the missing value in the loss data are available by the Mean of the data or you can Omit the data or customize the selection at the RG level by selecting the User Specified option Oracle Financial Services Software Confidential Restricted 22 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Capital Calculations Loss Da
42. ater than Review the data points or distribution selected for further variance for the given action distribution type The Data is not following the This is usually seen as very large or a negative parameter given Distribution Type value or inability to estimate them Check the distribution and time buckets Variance is zero so the data This happens if all the data points have the same numerical cannot be fitted for the given value Check the data points Distribution Type Relevant Event Type or LOB Check the LOB ET dimension table Data not found in the Dimension tables At any given percentile data This happens when you select Calculate for EVT threshold If doesn t follow specified at any percentile value Mean is not greater than or equal to the distribution The data is not Variance this message is displayed constant as per EVT modeling Either change the buckets or select Specify EVT threshold option While specifying the threshold value it should be above 50 No Relevant RGs have data Self Explanatory Cannot proceed further Input values for Scenario Update stage scenario data and scenario severity details as Severity Modeling cannot be required Null or Zero Parameter percentile value in Provide percentile value as greater than 4 formula based approach is less than 4 Hence calculation for Log Normal Distribution has failed Parameter percentile value and Self Explanatory median value in formula based approach
43. ave EEL Deductible Amount or Aggregate Deductible Amount if the insurance model is a Deductible Model For a Proportional Retention Model the percentage for EEL and Aggregate Loss should be given if deductible amount is not given Loss Calculation If the sum of EC values at RG level is not equal to that at bank level then check the Variance Covariance matrix to confirm Oracle Financial Services Software Confidential Restricted 37 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Annexure B Understanding Key Terms and Concepts Value at Risk Value at Risk VaR is the maximum loss not exceeding a given probability defined at the confidence level over a given period of time Conditional Value at Risk Conditional Value at Risk is defined as the mean of the loss a tail distribution CVaR is derived by taking a weighted average between the VaR and losses exceeding the VaR Expected and Unexpected Loss The expected loss is the mean annual aggregate loss and the unexpected loss is the annual aggregate loss in excess of this mean up to a particular confidence level say 95 or 99 per cent confidence The following diagram gives a detailed explanation of Expected and Unexpected loss Figure 23 Expected and Unexpected Loss Data Model is a logical map that represents the inherent properties of the data independent of software hardware or machine performance considerations The data model co
44. ble Name Variable Property Based On Created By Creation Date i vi27ss09417031 en elation Value Direct s ORECUSER 02 JUN 2010 03 17 56 PM dani ace V Direct ORECUSER 02 JUN 2010 03 19 42 PM Freq cy Scale Parameter Loss Data Direct Measure ORECUSER 25 MAY 2010 01 21 58 PM asc Scale Parameter Scenario Dat Direct Measure ORECUSER 25 MAY 2010 01 27 02 PM Frequency Shape Parameter Direct ORECUSER 25 MAY 2010 01 20 50 PM Direct 25 MAY 2010 01 25 22 PM Direct 20 MAY 2010 03 45 48 PM Direct 25 MAY 2010 01 24 19 PM Direct 25 MAY 2010 01 29 29 PM Direct 25 MAY 2010 01 22 52 PM Figure 14 Variable Management Screen Details such as Variable ID Variable Name and Creation details are displayed In the Variable Definition New screen the name and description of the new variable is to be defined Oracle Financial Services Software Confidential Restricted 24 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Variable Definition y ES 1 Frequency Scale Parameter Variable Name requency Scale Parameter Scenario D Variable Description Scenario Data Variable Type Idiosyncratic Variable bi Variable Structure Single Value v amp Based On Based On Measure de Variable Classification Numeric Variable M amp Measure Filter Do you want to apply filters MSR OR Scenario Frequency Scale Figure 15 Variable Definition S
45. credibility approach for scenario analysis method where Alternative Complimentary Aggregate or User Specified should be selected The second parameter is the credibility factor For example At the bank level if the number of simulations is 10000 then there would be 10000 loss simulations and 10000 scenario simulations These 20 000 simulations have to be converted to a final set of 10000 simulations In other words internal or external simulations have to be combined with scenario simulations for which three methods alternative complimentary and aggregate are available Under Alternative method given the credibility factor as 30 30 of worst severities of scenario simulations replace the corresponding simulated losses of internal or external data It should be noted that ordering simulated losses may be necessary as worst scenario simulations has to replace the worst internal or external simulations 096 credibility factor would mean only internal or external simulations will be considered for risk measure estimation 1 983 251 02 983 251 02 983 251 02 2 1 138 076 71 197 034 75 1 138 076 71 3 1 033 025 97 4 993 809 59 232 148 67 993 809 59 5 1 106 787 55 183 942 18 1 106 787 55 1 106 787 55 6 861 358 99 157 605 63 861 358 99 157 605 63 7 922 687 69 136 863 81 922 687 69 922 687 69 Oracle Financial Services Software Confidential Restricted 21 User Gui
46. creen Variable Shock To access the Variable Shock Library 1 Click Stress Testing 2 Click Shock Library 3 Click Variable Shock library The Variable Shock Library pane is displayed in the RHS of the screen Details such as Shock ID Variable Shock Name Variable and Creation Details are displayed Select the data set in this case Operational Risk Economic Capital Process Output and browse for the variable name The type of variable would be Idiosyncratic Variable Depending on the occurrence of the shock the shock type should be defined as either Instantaneous or Across Time The filter can also be applied by using New in the Shock Filter Specification This displays a new browser with the list of hierarchies For the shock parameter the modeling analyst should either shock the parameter on percentage base as Percentage shift or by absolute base as Absolute Shift The values of the shift should be defined at the Shock value Shock Details Shock Name Frequcny Shape Scen 1596 ROREC Process Outputs idiosyncratic Variable Shock Description Variable Name Variable Classification Frequency Shape Parameter S Numeric Variable Single Value Variable Property Direct Instantaneous M Shock in Reference to Time Point Standard Custom Time Points in Past Time Points in Future Shock current Value Tos as F ecification tA 0 to 0 of 0 d Dimen R Shock Parameters Shock Unit Percent
47. d Update the same in the dependency matrix table Loss Correlation Matrix Failed to convert Semi Definite For Gumbel copula non zero positive values of correlation of the Loss Correlation Matrix are required Update the same in the dependency matrix table Failed to convert Semi Definite For Gumbel copula non zero positive values of correlation of the Scenario Correlation are required Update the same in the dependency matrix table Matrix Negative association between For Gumbel copula non zero positive values of correlation the RG s Hence Gumbel Copula are required Update the same in the dependency matrix table failed The correlation value between For Gumbel copula non zero positive values of correlation RG s is greater than 1 Hence which are less than 1 are required Update the same in Gumbel Copula failed dependency matrix table Correlation value between the Check the Correlation values RG s cannot be 1 or more than 1 Theta computed is less than 1 For Gumbel copula non zero positive values of correlation Hence Gumbel Copula failed which are less than 1 are required Update the same in the dependency matrix table The correlation matrix is null Check the Correlation values in dependency matrix table So the theta cannot be computed Copula could not be generated Copula failure occurs either due to improper or insufficient for the given Copula type correlation values Check the values in the dependency matrix table
48. d shape is 5 You also have the flexibility to select the distribution for each cell by selecting the option User Specified The inclusion of near miss data would be in the interest of the modeling analyst as its inclusion or exclusion may change the capital calculation Near miss events would not have any severity value attached to it Hence they are considered only for frequency modeling The data needs to be passed to the NAG and shape and scale should be considered as output 2 4 3 Tab 3 Loss Data Severity Severity Distribution Selection The available dropdown options are Empirical Log Normal Extreme Value Theory EVT Weibull Gamma Exponential Pareto Burr Gumbel Uniform Log Gamma Log Logistic Truncated Burr Truncated Exponential Truncated Gamma Truncated Gumbel Truncated Log Gamma Truncated Log Logistic Truncated Log Normal Truncated Pareto Truncated Uniform and Truncated Weibull When any one of these options are selected then all the 56 cells or cells selected for processing follow the selected distribution but you also have the flexibility to select the distribution of each cell by selecting the option User Specified If EVT is used as the distribution then the modeling analyst has to specify whether to calculate the EVT Threshold or to specify the threshold value The specified EVT threshold value should always be greater than 50 EVT fitting is done after outlier treatment Due to this if outlier is opted fo
49. d time buckets However sometimes this error may be displayed due to data irregularities or if data exists but does not comply with the time buckets check the model definition and data as well Frequency Modeling Number Frequency modeling needs minimum five data points of Data Points is less than 5 buckets If not present the model fails Check the data and OFSAA Frequency Modeling time buckets defined Module requires minimum 5 data points 228 Frequency Modeling Failed to This mainly occurs as data does not follow distribution This Compute Mean and Variance in is usually seen when there is a very large or negative Adaptive Method Selection parameter value or inability to estimate them as well Check the distribution and time buckets Frequency Modeling Failed to This means in case of Negative Binomial or Binomial the Compute Number of Trials OREC application is unable to calculate N Check if distribution selected data and time buckets are defined Frequency Modeling Number This means in case of Negative Binomial or Binomial N is of Trials cannot be Negative negative Check if distribution selected data and time buckets are defined Frequency Modeling Failed to Check if distribution selected data and time buckets are Compute Number of Failures defined Frequency Modeling Number Check if distribution selected and data is proper and time of Failures cannot be Negative buckets are well defined Frequency Modeling
50. d value allocated to the shape and scale are either absolute or percentage shift Absolute value stresses the parameters directly For example if the shape is 0 5 and you input 0 25 as the absolute value then the stressed run would consider shape value as 0 75 If you input percentage shift of 25 then the shape would be multiplied by 1 25 and the new value is used for the Run execution Stress can also be applied on correlation matrix Further it should be noted that ensuring positive definiteness is mandatory Oracle Financial Services Software Confidential Restricted 13 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 2 3 Operational Risk Economic Capital Functional Process Flow Diagram INTERNAL LOSS DATA EXTERNAL LOSS DATA Scaling v FREQUENCY MODELING Poisson Binomial Neg Binomial y SEVERITY MODELING Log normal Emperical EVT Weibull Gamma Exponential Pareto Burr Gumbel Uniform Log logistic Log Gamma Truncated Burr Truncated Exponential Truncated Gamma Truncated Gumbel Truncated Log Gamma Truncated Log Logistic Truncated Log Normal Truncated Pareto Truncated Uniform Truncated Weibull SCENARIO DATA EREQUENCY MODELING Poisson Binomial Neg Binomial SEVERITY MODELING Log normal Weibull Gamma Exponential Pareto Burr Gumbel Uniform Log
51. dard currency The default currency would be USD and you can opt for any available currency If any currency exchange rates are missing then the previous exchange rate is considered and this is followed by rest of the risk computation The exchange rate data is obtained from Stg Exchange rate Hist Currency Conversion is handled by a DT For more information refer to the Business Metadata worksheets and Technical Metadata Worksheets Model Execution Bucketing Frequency The date of occurrence of each historical loss amount is expected as a download According to the length of the days called Time Window and Length of the Time Bucket given as an input in the UI Capital Calculation Settings the number of time buckets are calculated For each time bucket the frequency of loss amounts in the bucket is calculated Frequency modeling is done for these frequencies for each time bucket with respect to those particular RGs OREC application models the frequency of losses using either a Poisson binomial or negative binomial distribution If you are unsure of the method to be selected then select the adaptive method Model Execution Frequency Modeling Frequency modeling is performed after the time bucketing process The frequency data from each time bucket is used to determine the parameters Shape and Scale using distribution fitting and provides Goodness of Fit results After parameter estimation it is scaled to fit the annual horizon For Ex
52. de Oracle Financial Services Operational Risk Economic Capital Release 2 1 849 574 94 139 645 23 849 574 94 849 574 94 971 087 05 164 617 35 971 087 05 164 617 35 10 869 749 37 177 235 19 869 749 37 869 749 37 Table 8 Credibility Factor In the complimentary approach random replacement of internal or external simulations is done by randomly picking simulations from scenario data depending upon the credibility factor Random is the key here 100 credibility factor would mean only scenario simulations are considered for risk measure estimation Again 0 and 100 credibility factor can also be achieved While calculating bank level VaR it does not consider worst cases from both sides and also does not ignore the tail portion This is done by preserving the simulation order which ties the frequency generation done for all RGs It is preserved as OREC application uses copula to generate frequency so that real data is imitated as much as possible For example let s assume that there are 10 simulations from loss data and 10 from scenario all tied together by the simulation order If the first simulation number itself is the worst case for the scenario then OREC application ignores loss simulation It does not check if ignored loss value is worst case or not This results in Diversified VaR which applies portfolio effect to RG combinations by preserving the relationship generated by copula Under aggreg
53. ded with report filters to view data in different ways A model run using Run execution is all that the dashboard needs The dashboard queries the same database using filters like Run descriptions execution date and so on Graphical reports OR Risk Loss Distribution Risk Measures Across Time Internal Loss Statistics Economic Capital Allocation to Lines of Business Economic Capital Allocation to Event Types e Line of Business Analysis Key Risk Measures Economic Capital Trend Loss Event Statistics Economic Capital Allocated to Event Types Loss Frequency Analysis Loss Severity Analysis Event Type Analysis Key Risk Measures Economic Capital Trend Loss Event Statistics Economic Capital allocated to Event Types Loss Frequency Analysis Loss Severity Analysis Analysis across Models Risk Measures across Models Model General Details Simulation Reports Simulated Values of Frequency for Scenario Data Simulated Values of Severity for Scenario Data Simulated Values of Aggregated Loss for Scenario Data Distribution Parameters Stress Testing o OR Loss Distribution o Risk Measures across Time Oracle Financial Services Software Confidential Restricted 32 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 o Economic Capital Allocation to Line of Business o Economic Capital Allocation to Event Types o Frequency Shape and Scale Analysis for Event Type o
54. e 2 1 In this tab specify various conditions under which the simulations are to be done Here you can select the appropriate options depending upon the availability of external and scenario data Additionally you may or may not consider insurance You also have the option to Calculate Allocation Factor by selecting Yes in the available options after which OREC application generates the allocation factor On the other hand if Calculate Allocation Factor is No then the factor is expected as a download UL at the aggregate level is distributed back to the RGs in proportion to the variances and co variances of the individual RG losses You can specify the Distribution Fitting Methodology as Maximum Likelihood Estimate MLB Method of Moment MM or Maximum Likelihood Estimate BFGS 2 4 2 Tab 2 Loss Data Frequency The loss data frequency has two sections namely Correlation and Parameters You have to specify the copula method and correlation technique in OREC application Correlation The copula method has a drop down menu with four options namely Gaussian Gumbel Student s t and Not Required When Copula Method is selected as Not Required then the Correlation field remains disabled and the Copula is not used Parameters The OREC application estimates the dependence between the RG frequencies of internal loss data using any one of the correlation co efficient techniques Spearman s correlation co efficient or Pearson s correlation
55. e SLT Yes v Yes v No v Yes y Yes v Yes v Yes v Yes v Health Insurance Assistance Non Life Reinsurance Yes y Yes v No v Yes y Yes y Yes y Yes y Yes y Proportional Assistance Non Life Yes v Yes v ES 4 Yes v Yes v Yes y Yes y Yes y Casualty other than health Non Life REIns Yes v Yes v No v Yes v Yes v Yes y Yes y Yes y PROP Credit and suretyship Non Life y Reinsurance z No v No No No No No No Proportional Credit and oe l U No y No No No y No No No Death Accepted No m uc wr reinsurance Life Death Index linked and unit linked life No v No v No v No v No v No v No v No v insurance Life Death Life insurance with profit No v No v No v No v No wv No v No v No v participation Life Death Other life insurance Life Save Cancel Figure 4 Reporting Group Relevance Setting Screen The Reporting Group Relevance Setting remains disabled if the RG is not selected If the RG is changed before the model is saved the frequency and severity tab parameters changes to null if there are any new LOB and ETs then the dimension table should be resaved in the hierarchy Oracle Financial Services Software Confidential Restricted 16 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Parameters Reporting Currency An entity having operations in multiple geographic areas and those which have losses in multiple currencies EC should be calcu
56. e mandatory Data Quality DQ checks provided with OREC application to remove data inconsistency errors DQ checks can be defined through a simple GUI provided with OREC application However DQ checks need to be installed separately and is not part of the standard installation package Oracle Financial Services Software Confidential Restricted 6 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 2 2 2 Rule Framework Reclassification As Operational Risk Economic Capital calculation is based on the regulator prescribed Standard LOB and Standard ET the internal and external LOB and ET is mapped to the standard data This is done as a part of the Reclassification Rules Reclassification is handled in the Rules framework For more information refer to the Business Metadata document and Technical Metadata Worksheets Rule Framework Hligibility of Insurance and Model A Rule which checks insurance eligibility at this step has been configured in OREC application OREC application provides you with a choice to define upper or lower threshold of severity as well as a choice to select the external data source to be used in modeling Both checks are provided as a part of the Rule Model and Insurance Eligibility is handled in the Rules Framework For more information refer to Business Metadata document and Technical Metadata Worksheet Rule Framework Currency Conversion It is assumed that data is provided i
57. el Execution Insurance 11 Model Execution Loss Calculation 12 Model Execution Scenario Modeling 11 Model Execution Severity Generation 10 Model Execution Severity Modeling 9 Model Variable Shock Mapping 26 O Objective of Guide vi Operational Risk 1 Outlier calculation 34 Outlier Determination 22 Overview of the Application 1 P Oracle Financial Services Software Confidential Restricted 47 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Proportional Model 12 R Random Number Seed 35 Reclassification 4 Reporting Group 3 Rule Framework Currency Conversion 7 Rule Framework Eligibility of Insurance and Model 7 Rule Framework Loss Data Capture 6 Rule Framework Reclassification 7 Rule Framework Scaling of External Data 7 Rule Metadata vi S Sandbox 28 Sandbox Definition 28 Scenario Data 20 Scenario definition 25 Set up definition 28 Severity Distribution Selection 19 Simulation Reports 32 Stage tables 30 Standard LOB types and ET 4 Standard RG 3 Star Schema 39 Stress Testing 13 Stress Testing Framework 23 Stress Testing reports 32 T Technical Metadata vii V Value at Risk 38 VaR numbers calculation 34 Variable Definition 23 Variable Shock 25 Oracle Financial Services Software Confidential Restricted 48 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 ORACLE Oracle Financial Services Operational Risk Economic Capital Release 2 1 User Guide
58. erarchies may be based on either the FACT table or dimensional tables Measure A simple measure represents a quantum of data and is based on a specific attribute column of an entity table The measure by itself is an aggregation performed on the specific column such as summation count or a distinct count Business Processor This is a metric resulting from a computation performed on a simple measure The computation that is performed on the measure often involves the use of statistical mathematical or database functions Advanced Analytical Infrastructure The Oracle Operational Risk Economic Capital Modeling Environment performs estimations for a given input variable using historical data It relies on pre built statistical applications to build models The framework stores these applications so that models can be built easily by business users The metadata abstraction layer is actively used in the definition of models Underlying metadata objects such as Measures Hierarchies and Datasets are used along with statistical techniques in the definition of models Additional Things to Remember The current version of OREC application supports two different approaches for modeling Scenario Severity Bound Based During this approach the severity data is modeled similar to those of loss data The dropdown consists of all the distributions of severity distribution except EVT and Empirical Formula Based Usage of this approach the
59. g of these losses are considered This requirement is only for internal and external data Scenario data does not undergo any log transformations You are provided with an option to set log transformation at each RG level selectively Model Execution Frequency Generation Frequency generation using simulation can be done with or without the copula If used the copula retains the structure of loss data so that the simulated frequencies imitate the same pattern There is a good chance of having simultaneous peak amounts in RGs if the copula is not used However the desired results may be derived under specific business conditions The choice of copula affects the bank level loss figures more than individual loss figures For probability generation by using the copula correlation matrix is expected as an input OREC application supports Gaussian Gumbel and Student t copula However the copula selection is optional as you can opt for modeling without the copula This option is available in the model definition Correlation matrix generated is used reporting group wise for internal or standard data Once prepared it can also be used for scenario modeling There is no change in the way this correlation matrix is used for internal or standard data However the current requirements specify multiple scenarios for each RG and the simulation happens at the scenario level Hence correlation matrix also should be brought down to the scenario level There a
60. get copula failure errors Copula works mainly on correlation values between RGs There are a few obvious reasons for copula failure which are as follows Erroneous insertion of data under User Specified resulting in correlation value lt 1 or gt 1 Matrix approach is not followed For example RG ET 2 LOB 3 has correlation value of 0 5 but LOB 3 ET 2 has been inserted as 0 6 When Gumbel Copula is the correlation values are to be restricted between 0 and 1 as here non zero and positive values are required 7 Why we get error Data does not follow distribution error And how can we overcome this error Oracle Financial Services Software Confidential Restricted 34 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Every data has its own characteristic which is suitable for a particular distribution or class of distribution For example when mean is equal to variance Poisson is used If we try to fit the data which does not follow a particular distribution we end up doing a force fit which may lead to errors Another reason can be the choice of buckets Frequency modeling depends on bucketing Change in the bucket window and time period can substantially alter the data characteristics It is advised to choose a correct bucket structure for the output to match the expectation However it is always advisable to go though the Modeling Framework which provides a lot of tools like Goodness of Fit
61. ing The time window duration and the length of the bucket are used to define a bucket This is carried out to segregate frequency data in predefined time buckets Number of Buckets Time window in days Bucket length This will be rounded off to the nearest figure For example If the time window is 100 and bucket length is 10 then the number of buckets formed is 100 10 10 Oracle Financial Services Software Confidential Restricted 4 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 For frequency modeling the minimum number of buckets required is 3 Hence there is a front end validation to check this entry The model cannot be saved if the number of buckets is less than 3 You are required to have a bucket definition matching with the requirements This can be done using modeling framework where you can work around many bucket definitions to get the desired result Goodness of Fit tests The Goodness of Fit tests help in determining which distribution is to be used for frequency and severity modeling This is also separately made available in the modeling framework In OREC application model execution this is a compulsory step where all 3 tests are run and results stored While using modeling framework you are expected to choose a Goodness of Fit test suitable to the selected distribution Modeling framework provides the option of working with data for a variety of models so that the relevan
62. ing You can make changes in the tool matrix XML file which are valid across models and users which indicate that changes made here will have a global impact 16 How does log transformation help Log transformation helps in variance stabilization Increasing slopes in X in relation to another variable are linearized Positively skewed distributions of X are normalized 17 Where do we set threshold values for external loss data In the Rules framework you can specify upper and lower thresholds of external loss data Later outlier can be applied to ensure data selected for modeling resembles the expectations set or business requirements 18 For OBI reports Dashboard is execution compulsory in production set up OREC application provides the option to generate OBI reports from Sandbox or Production set up This is handled in OBI settings If sandbox is selected from OBI then Run in production set up for OBI need not be executed However by default the choice provided in OREC application setting is Production Oracle Financial Services Software Confidential Restricted 36 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Annexure A Things to Remember Frequency Modeling If you are unaware of the distribution type of the data for each RG select Adaptive modeling so that OREC application models on its own and selects the relevant distribution type for each RG to calculate the frequency parameters
63. ize 40000 All internal Insurance Company Asset size 10000 15000 25000 50000 EEI Scaling Factor 50000 40000 1 2 Scaling factor 50000 100000 0 5 Oracle Financial Services Software Confidential Restricted 7 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Scaling factor is calculated for each RG If the bank provides the scaling factor as a download for each RG then the scaling factor is not calculated OREC application calculates the scaling factor only when it is not given as a download If OREC application is unable to calculate the factor then the default value is assumed as 1 Rule Framework EC allocation Once risk measures are calculated they are allocated to RGs as per the allocation factor This factor is calculated as a co variance of simulated losses of a given RG divided by the variance of bank level losses 2 2 3 Model Execution Model Execution Currency Conversion The reporting currency is part of the model definition and can be defined in the Capital Calculation Settings tab Scaling of external data occurs outside the model definition Internal losses external losses and entity measures like asset size profit size insurance liability measures along with the loss severity values is converted to the reporting currency at the exchange rate available during the reporting date fic mis date or loss date for loss value Scaling factor is then determined using the stan
64. l Services Software Confidential Restricted 49
65. lated in the reporting currency The historical losses should be converted from standard currency to the reporting currency Scenario severity values should also be converted to the reporting currency prior to the model execution Number of simulations The number of simulations should be greater than Zero Severity Loss Simulation uses the temporary table space as a part of model execution The Database temporary table space requirement for running 56 Reporting Groups RG with 1 000 000 simulations is 12GB Time window in days The time window would specify the number of days the loss data would be used for risk computation Bucket length The length of the bucket as in the frequency data would be calculated based on this length These parameters are used in the simulation generation of frequencies and severities Enter the number of simulations of frequencies and severities to be generated for loss calculation The inputs for these parameters should follow the validation rule The input value for this field should be greater than 1 The parameters for Time window and Bucket length are used to form the time buckets number of time buckets Time window in days Bucket length Each time bucket contains the frequency of the loss event occurrence with respect to that particular RG The number of time buckets should always be greater than 3 else the model fails to save This is further used in calculating frequency in the relevant buckets
66. ment Authorization Model Deployment gt Model Deployment Authorization Model Deployment Request for Deployment Authorize amp Deploy Save Cancel Figure 22 Model Deployment Screen 2 3 2 Staging Area The uploading of data involves the loading of all the stage tables For more information on the stage tables to be populated refer to the Download Specifications DL Specs documents Oracle Financial Services Software Confidential Restricted 30 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 4 Execution A Run is executed in the production infodom to calculate Operational Risk Economic Capital A Run in the production infodom consists of the following T2T Stage to FACT Table data loading Reclassification Rules These are Rules to reclassify Internal to Standard External to Standard External to Standard LOB and Standard ET Model Execution Models defined in the sandbox are re executed through a Run in the production infodom For more information on execution of a Run refer to the Technical Metadata Worksheets Oracle Financial Services Software Confidential Restricted 31 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 5 Operational Risk Economic Capital Reporting A reporting dashboard is also provided with the OREC application This involves a combination of graphical as well as tabular reports You are provi
67. n value between scenarios within the same group Inter Group Default Correlation Value This value is replaced if there is no correlation value between RGs Frequency Distribution The OREC application supports Poisson Binomial and Negative Binomial distribution for modeling frequency of scenario data In this tab you need to specify the Frequency Distribution and Scenario Distribution The list of distributions for scenario data frequency modeling follows the same as in loss data frequency modeling For the adaptive model the tolerance level has to be specified and this can be done at RG level Capital Calculations Loss Data Frequency Loss Data Severity Scenario Data Credibility Factor Data Transformation Fitter ency Correlation cient Use Loss Data Correlation Inter Scenario Default Correlation Value 0 22 Inter Group Default Correlation Value 0 11 a rio Distribution Fre y Distribution Selection Adaptive v Tolerance Level for Adaptive Model 25 Severity Distribution Selection Exponential Figure 9 Scenario Analysis Screen Severity Modeling The OREC application supports all the distributions available for loss data severity modeling with Oracle Financial Services Software Confidential Restricted 20 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 the exception of Extreme Value Theory EVT Empirical and Truncated distribution Since EVT is be
68. n its natural currency However if a different reporting currency is provided then use this module to get the desired results OREC application provides a DT for this task When this DT performs data transformations the USD currency is hard coded However when the same is performed in model execution the currency defined in the model needs to be specified For more information refer to the Technical Metadata Worksheets Rule Framework Scaling of External Data You have the choice to include external data for modeling If the external data loss and threshold pertaining to a RG is not relevant exclude the external data for processing Further use the external data in an as is format that is without scaling the data or after scaling it using statistical technique such as least square method The external data is scaled to match the internal data OREC application scales data using the scaling factor A This can be either user specified or can be derived by a minimizing function A is multiplied with the external data to calculate the scaled data Also the computation of scaling factor using entity measures asset size profits happens as all the internal entity measures are summed and used as a denominator Scaling is performed after currency conversion in the overall process flow Scaling is handled in the Rules Framework For more information refer to the Business Metadata worksheets Table 5 Scaling in Rule Framework EE Asset S
69. nsists of entities tables and attributes columns and shows data elements grouped into records as well as the association around those records Dataset It is the simplest of data warehouse schemas This schema resembles a star diagram While the center contains one or more FACT tables the points rays contain the dimension tables as represented in the following diagram Oracle Financial Services Software Confidential Restricted 38 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Dimension Table Dimension Table Products Time Geography Sales Fact Table Customer Channel Figure 24 Data Warehouse Schemas Metadata A term used to denote data about data Business metadata objects are available in the form of Measures Business Processors Hierarchies Dimensions Datasets Cubes and so on The commonly used metadata definitions in this User Guide are Hierarchies Star Schema In a star schema only one join is required to establish the relationship between the FACT table and any one of the dimension tables which optimizes queries as all the information about each level is stored in a row The set of records resulting from this star join is known as a dataset Hierarchy A tree structure across which data is reported is known as a hierarchy The members that form the hierarchy are attributes of an entity Thus a hierarchy is necessarily based upon one or many columns of a table Hi
70. nternal loss data is defined as all losses experienced by the bank External loss data represents losses observed by the data consortium members and not necessarily by the bank itself However there is another set of loss data defined by the experts opinion based on their prior experiences These are called scenario data Scenario analysis in conjunction with external data evaluates exposure to high severity events This comprises of plausible severe losses in terms of likelihood of the loss corresponding to frequency and the likely amount of the same corresponding to severity Scenario analysis can be either alternative or complimentary in nature The capital is computed after taking into consideration the type of scenario data In certain situations you might want to censor the loss data below a specific threshold example while fitting truncated distribution Therefore threshold capture is also available as a part of OREC application release 2 1 For more information on loss data capture refer to the Download Specifications DL Specs and Technical Metadata Worksheets The major assumptions while integrating OR Operational Risk 4 6 out of box and OREC are as follows oLoss of value in OREC is net of insurance recovery oLoss date in OREC is the date created in operational risk oDimensional entities of both the applications should be in sync For more information on mapping details refer to the Technical Metadata Worksheets Follow th
71. on a Loss Distribution Based approach consistent with Basel II guidelines has been incorporated to estimate the Economic Capital EC of the operational risk at the firm level According to the Basel II guidelines financial institutions are required to develop their own internal measurement methods that estimate the expected and unexpected operational losses based on the combined use of internal relevant external and scenario data for Standard or Internal Reporting Groups RG Additionally the following functions have been incorporated in OREC application For an entity operating in multiple geographical locations the data from the sister company can be included for risk computation with the internal loss data by using currency conversion which is controlled through the User Interface UT Use of external entity data for better computation of EC through Scaling methodologies Use of Scenario data which are typically based on the expert s judgment to enrich internal or external historical data and estimate the capital requirement in a more informed manner Increased options to compute correlation and to fit distributions for frequency and severity on both loss data and scenario data analysis Oracle Financial Services Software Confidential Restricted 1 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Use of Copulas to simulate the frequency numbers thereby managing the correlation
72. ormulas as mentioned earlier are applied For bound data is prepared and then parameters are estimated similar to internal loss data modeling Here the inputs are scenario data like highest frequency single largest severity amount severity lower and upper bounds frequency and severity variance and mean Model Execution Insurance An insurance contract is characterized by two components namely Each and Every Loss EEL Insurance and Aggregate Agg Insurance EEL is applied at scenario level for each RG while aggregate insurance is applied on the aggregate sum of severity amounts generated for the RG You can choose level insurance in the model definition screen in percentage For example choose 20 with loss value without insurance being 10000 and then you can offset only 2000 part of the Oracle Financial Services Software Confidential Restricted 11 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 loss Even though the insurance contract allows for a better offset OREC application will cap the overall benefit OREC application supports two types of insurance models Deductible Model and Proportional Model Deductible Model In Deductible model the claim profile is characterized by a deductible d and a liability threshold 1 such that d lt 1 The insurance firm does not incur any liability till such time that the loss severity is less than the deductible It will cover the entire loss
73. r along with EVT the threshold point may be of a smaller value than expected USER CAPPING SELECTION YES with CAPPING SHAPE LOWER VAL 0 5 and CAPPING_SHAPE_HIGHER_VAL 1 This combination changes the fitted shape parameter of distribution if it falls outside the range defined Estimated values are replaced by the nearest threshold value in case the range is breached USER CAPPING SELECTION FAIL with CAPPING SHAPE LOWER VAL 0 5 and CAPPING_SHAPE_HIGHER_VAL 1 This combination changes the fitted or estimated shape parameter of distribution if it falls outside the range defined It also conveys that the Run would fail if estimated parameters are outside range You are required to change the data or redefine buckets for improvement in parameter values USER CAPPING SELECTION NO with CAPPING SHAPE LOWER VAL 0 5 and CAPPING_SHAPE_HIGHER_VAL 1 This combination changes the fitted shape parameter of the distribution if it falls outside the defined range In this option the OREC application performs simulation even though the parameters are outside range Simulation that results from such an execution should be read with caution In all 3 cases you can modify the lower and upper values depending on the experience However 0 5 and 0 75 or 1 are standard values As for shape parameter value of 1 and greater than or less than 0 5 the distribution has infinite variance and simulated values are erratic Apart from EVT there a
74. r scaling factor of 1 reduces the range to inter quartile range Values less than 1 are not advisable 12 When does model version change and when is the model asked to be saved with a different name Model changes can be saved in two ways First called version change involves minor changes like simulation numbers bucket parameters random number seed and insurance percentage Changing these parameters is not considered as the one which impacts the basic structure of the model and hence the same can be saved with the new version number However if any other changes are made ensure that you save it as a new model as all other changes impact the modeling structure 13 Where will I find outlier and missing value treatment results for view The relevant files are saved in the following location home RORECst ftpshare INFODOM scripts modeldatafiles Infodom values are generally RORECSAND or RORECPROD depending on which infodom the Run or model is executed on 14 Why is external data used only for severity modeling External loss data is not utilized for modeling of frequency as it is characterized by low frequency Oracle Financial Services Software Confidential Restricted 35 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 and high severity data and therefore is considered only at the time of modeling severity 15 How do I select between formula based and bound method for scenario severity model
75. re 11 other options available in severity distribution selection and any 1 can be selected Oracle Financial Services Software Confidential Restricted 19 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 Log Transformation is an option to apply log transformation on the loss data Capital Calculations Loss Data Frequency Loss Data Severity Scenario Data Credibility Factor Data Transformation Filter Parameters Severity Distribution Selection Log Transformation No v EVT Threshhold Specify v EVT Threshold Figure 8 Severity Distribution Selection If the severity distribution specified is user specified and EVT is selected at grid level then input different EVT Threshold at grid level for different RG s or specify one EVT Threshold for all the RG s M The RGs which are not selected for processing from the Reporting Group Relevance Settings field are disabled from selecting distribution 2 4 4 Tab 4 Scenario Data The scenario tab has two sections namely Frequency Correlation and Scenario Distribution Frequency Correlation In this tab specify the Correlation Coefficient Estimation between Use Loss Data Correlation and User Specified Correlation Based on Use Loss Data Correlation input the correlation default value There should be 2 inputs in the Scenario Analysis tab Inter Scenario Default Correlation Value This value is replaced if there is no correlatio
76. re two inputs in the Scenario Analysis tab e The correlation matrix can also be obtained as a download This matrix would replace any correlation calculated from the loss data and scenario data e Correlation matrix can be provided from the front end Care should be taken in relation to DQ check on the matrix for any values beyond 1 and 1 Model Execution Severity Generation Frequency and Severity scenarios are generated using Monte Carlo Simulation for each RG to predict the potential operational losses The frequency simulations are generated based on the fitted distribution parameters The frequency parameter for each RG is calculated during frequency modeling Severity simulations are generated based on the parameters as calculated in severity modeling Number of severity simulations for each scenario is equal to the value of frequency in that particular simulation For example if in RG 1 for Scenario ID 1 the frequency generated is 5 then 5 severity amounts Oracle Financial Services Software Confidential Restricted 10 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 S1 S2 S3 S4 S5 would be generated for that frequency 5 severity amounts for each scenario id In the computation of risk analytics all 5 severity values would be added to determine any RG level EC computation Model Execution Scenario Modeling OREC application supports all the distributions other than empirical and
77. rious portions of OREC application Download Specifications The format and structure of the RDBMS tables is specified in the Download Specifications DL Specs Download Specifications contain details of the attributes required for processing in OREC Application OFSAAI documents The set of OFSAAI documents packaged in the installer will help the user understand the functions of the various components of Oracle Financial Services Analytical Application Infrastructure OFSAAT used for OREC computation Application Installation Manual How to use this User Guide The information in this User Guide is divided into the following chapters Chapter 1 Introduction The objective of this chapter is to introduce the user to Oracle Financial Services Operational Risk Economic Capital Release 2 1 and provide an overview of OREC Application Chapter 2 Understanding the Application The objective of this chapter is to provide an understanding to the user on the various functions of OREC application Chapter 3 Preparing for Execution The objective of this chapter is to provide a detailed explanation of the activities involved before actual execution of Runs such as data maintenance and so on Chapter 4 Execution The objective of this chapter is to inform the user on the execution function of OREC application Chapter 5 Operational Risk Economic Capital Reporting The main objective of this chapter is to provide a brief description
78. rnal data to match internal data OREC application supports the following 22 distributions for modeling the severity of internal loss data and duly scaled external loss data e Burr e Empirical e Exponential e Extreme Value Theory GPD for tail e Gamma e Gumbel e Log gamma e Log Logistic e Log Normal e Pareto e Uniform e Weibull e Truncated Burr e Truncated Exponential e Truncated Gamma e Truncated Gumbel e Truncated Log Gamma e Truncated Log Logistic e Truncated Log Normal Oracle Financial Services Software Confidential Restricted 9 User Guide Oracle Financial Services Operational Risk Economic Capital Release 2 1 e Truncated Pareto e Truncated Uniform e Truncated Weibull The parameters of the truncated distributions are calculated using the Maximum Likelihood Estimate BFGS The parameters of the remaining distribution are calculated using the Method of Moments Maximum Likelihood Method Simplex or Maximum Likelihood Method BFGS Log Transformation Log transformation is used to stabilize the variance of sample data Log transformed values are always in uniform variance compared to the one before log transformation However this is an optional step and the selection is made in the Model Definition screen Before the data is sent for severity modeling the natural log is considered with respect to each value Before the computation of risk measures like Value added Risk VaR the exponential anti lo
79. s Operational Risk Economic Capital Release 2 1 3 Preparing for Execution 3 1 Set up definition Set up pertains to the ability of OFSAAI platform to calculate loss values OREC application provides two such infodoms namely Sandbox and Production Sandbox is a sort of trial area Various combinations of parameters can be tested till the desired results are achieved The model output can then be deployed after which it can be accessed in production In production the Model Management option is not available which restricts the changing of the model parameters However this can also be achieved through the sandbox Sandbox Definition Within modeling environment Sandbox Environment data would be extracted or imported from production infodom based on the dataset defined Also this can be the first data movement from the staging to the FACT table as well The objective of this step is to fetch data for all attributes for a particular time period The Sandbox Definition screen is reproduced below for easy reference Connected to ORECSAND v fi Home E Y Unified Metadata Manager E EY Rules Framework El Operations E El System Configuration E amp Administration E de Advanced Analytics Infrastructure Sandbox Definition Sandbox Data Maintenance Application Variable Modeling Stress Testing aA AMHM UMM Offline Population Figure 19 Sandbox Definition Sandbox Definition Screen includes the following Model
80. s with a separate modeling of the frequency and severity of losses Basic analysis of the data mean variance skewness and kurtosis frequency and severity modeling are performed on the reclassified data stored in the FACT table for each RG The compound loss distribution to calculate the risk measures is arrived at by merging the fitted frequency and severity distributions and simulating the resulting compound loss distribution A bank may mitigate the impact of operational risk losses by taking insurance against it Expected Loss EL and Unexpected Loss UL are calculated at RG level and at bank level The functional flow of OREC application can be classified in two broad categories namely Pre Model Activities and Model Execution 2 1 Pre Model OREC application calculates EC of operational risk using the Loss Distribution Approach The OREC application calculates the EC for each Reporting Group RG A Reporting Group can be a Standard RG or an Internal RG The Standard RG is a combination of regulator prescribed Lines of Business LOB and Event Types ET if any Similarly Internal RG is a combination of LOB and ETs as specified by the bank Internal RG classification is dependent on the bank s structure The number of RG s in case of Internal RG can be more or less than the standard classification OREC application also calculates EC at the bank level and then allocates the same to the RGs So at RG level there are two Economic Capit
81. sets Retail banking Employment Practices and Work place Safety Retail Brokerage Trading and Sales External Fraud Table 3 Standard Reporting Group Internal Reporting Group Typically a bank has its own definition of LOB and ETs The combination of such LOBs and ETs form the Internal RG For example 12 internal LOB types and 7 internal ET s can be present The bank has the flexibility to compute EC for Internal RG Standard RG or both Reclassification OREC reclassification process is the initial phase in model definition During this the bank s LOB and ETs which together forms a RG are mapped to a Standard RG as specified by the regulatory norms or the entity can define their own reporting group using Internal RG Reclassification is done either from external to standard classification or internal to standard classification Also a Data Transformation DT is provided which helps in mapping this classification in output tables as well For example if analysis is done for Internal RG and the same is required to be viewed as per Standard RG this DT can be used to get the desired result For example Industrial Finance Employment Practices and work place Other Retail safet Table 4 Reclassification Example To change or define the reclassification you can make changes to the DT For more information on DTs refer to the Technical Metadata Worksheet This step is included as a part of model execution Bucket
82. st applicable for the data with extreme deviation from the mean it is the outcome of user judgment based on the risk and control assessments within the bank and the scale of operation This peculiarity of the scenario data makes it difficult to fit Empirical distribution Since EVT approach comprises of Empirical distribution and Generalized Pareto distribution it is not advisable to fit EVT to scenario data Log Normal being a skewed tail amongst all thin tailed family of distributions proves to be the best distribution for scenario data The mode and percentile given as a download for specific reporting groups are used to fit a Log Normal distribution to the scenario data The process outputs calculated are stored in OREC application Process Output table OREC application supports two different approaches for modeling severity data Formula Based Usage of formula for severity modeling Bound Data Bound data approach follows modeling similar to severity modeling of loss data The convention between Formula based and Bound data can be handled in tool matrix Scenario data can also be defined at reporting group level by User Specified option available under the pane When scenario data is specified as No in the Capital Calculation tab the Scenario Data tab will be disabled Default correlation value is used only when loss data correlation is used 2 4 5 Tab 5 Credibility Factor You have to define two parameters the first one being the
83. t distribution method can be selected for frequency as well as severity Accordingly there are 3 approaches supported by OREC application namely KS test Anderson Darling Chi square test Kolmogorov Smirnov Test This test is used to decide if a sample comes from a hypothesized continuous distribution The Kolmogorov Smirnov KS test tries to determine if two datasets differ significantly The KS test has the advantage of making no assumptions about the distribution of data It is non parametric and distribution free The hypothesis regarding the distributional form is rejected at the chosen significance level alpha if the test statistic D is greater than the critical value obtained from a table and therefore determines the best fit Anderson Darling Statistic The Anderson Darling statistic measures how well the data follows a particular distribution thereby determining the best fit For a given data set and distribution the better the distribution fits the data the smaller this statistic will be Anderson Darling has almost replaced the usage of KS test as it is more sensitive to deviations in the tails of the distribution Use the corresponding p value when available to test whether the data comes from the chosen distribution If the p value is less than the selected alpha for example 0 05 reject the null hypothesis stipulating that data follows a particular distribution Chi Square test It is used to compare the
84. ta Frequency Loss Data Severity Scenario Data Credibility Factor Data Transformation Filter 2 Parameters Outlier interquartile Method v Missing Omit Y Outlier Scaling Factor Figure 11 Data Transformation 2 4 7 Tab 7 Filter Add multiple Hierarchy members as filters as shown in the screen Select filter tab in the Model Definition New screen This is mainly used for entity selection Technique Browser Webpage Dialog Search Mi 2 Technique Hierarchical Members Selected Members s Business Models Credit Risk Cash Flow Model Conditional Default Model D m Credit Metrics Structural Model Distribution Fitting based Future Value Model Historical Loss Distribution Fitting Model b Historical Pool Average Default Rate Model Merton Model 4 Time to Default Model VaR Reader Market Risk Operational Risk E ele Statistical Techniques 4 nm b Figure 12 Filters 2 5 Stress Testing Overview The Stress Testing Framework has four parts Variable definition Shock library Scenario definition Stress definition Variable Definition Stress Testing allows the bank to conduct an analysis that enables it to estimate the impact of movements in the variables on measures such as profitability capital Oracle Financial Services Software Confidential Restricted 23 User Guide Oracle Financial Services Operational Risk Economic C

Download Pdf Manuals

Related Search

Related Contents

Brochure commerciali - Bosch Security Systems  SK901 and 902 Owners Manual  3062064 Regelung Wärmepumpe Montage  Toshiba Satellite C855-S5134  コメン クロ  Pub 3189, Volunteer e-file Administrator Guide  MELSECNET/10 Remote I/O Module User`s Manual  Gateway FPD1975W Computer Monitor User Manual  インフルエンザ (H1N1) 2009の遺伝子診断キットの 感度試験  IP Door Phone IP39-4X Quick Installation Guide  

Copyright © All rights reserved. Failed to retrieve file