2 Credit Risk In A Model World BacktestingVolatility of capitalDatabaseRating requirementsConsistency of ratingParameters for portfolio risk
3 Impact of Loan Portfolio MTM Introduction:The movement towards marking to market loan portfolios will have a significant impact on P&L volatility.The principle drivers will be;volatility in credit spreads.the nature of the portfolio, in terms of credit rating and the tenor of the loans.
4 Impact of Loan Portfolio MTM A hypothetical 5 year $25bn loan portfolio has been modelled to show the impact of changes in credit spreads;
5 Use in Credit Assessment RATING MODELS HAVE NOW BEEN DEVELOPED SUFFICIENTLY TO BE USED FOR STAND-ALONE RATINGSBenefits of using robust, well-validated rating modelsConsistent: All factors inherent in ratings are imputed into final ratings via universally-accepted benchmarksUnbiased: Subjective judgement can be consistently applied within the rating process.Transparent: Models provide a complete description of methodology employedCoverage: Ability to assess corporates and banks beyond coverage of the major rating agencies.Efficiency: Ratings can be quickly generated using extensive on-line databases.
6 ImplementationCRS EMPLOYS ON-LINE FINANCIAL DATA TO GENERATE RATINGS BASED ON A PROVEN RATING METHODOLOGYCRS is an integrated system
8 Process of Development DEVELOPMENT OF CRS IS AN EVOLUTIONARY PROCESS, AND NOT PURELY QUANTITATIVEDevelopment involves significant analytical evaluation and feedbackRatings Agencies: Have developed quantitative models for several years employing all means of modelling from expert systems to neural networks. Now increasingly see their franchise as providing a rating methodology, rather than rating names (so have all begun to market quantitative models on this basis.). We are in discussion with Moody’s to benchmark our model to their unpublished neural network model.Third Party: Mostly publicly led by KMV. But also includes Alcar and Zeta Models (which are based off multi discriminant analysis). Moody’s also operate as a third-party supplier, but the methodology employed is very different.Chase/Chemical: Have publicised their methodology, which is now inherent in methodologies used at CIB. All factors are weighted. based on expert based logic to introduce consistency into the internal rating allocation.Others?: Have publicised their methodology, which is now inherent in methodologies used atWe really beleive the agencies and KMV represent our peer group in respect to the quantification of our credit risk at a counterparty level. But KMV is not yet a sufficiently accepted benchmark - which the proposals has done little to reinforce. An example of how they might impact our methodology is their approach to industry definitions. In terms of the agencies, we are most active in dialogue with them to understand how and where they are moving. A good recent example is the definition of ratings
9 MethodologyCHOICE OF APPROACH IS BASED ON SEVERAL CRITERIA, BUT MUST BE SIMPLE !!Quantitative models must be supportive to the analysis of credit riskCannot be a ‘black box’ – needs to be sufficiently transparent to allow interpretation of outputNeed for compatibility with benchmarks used within internal rating process
10 Approaches UsedCRS IS A HYBRID APPROACH WITH MODELS DEFINED BY SECTOR AND JURISDICTIONInput to the models
11 Fundamental DataHIGH CORRELATIONS BETWEEN VARIABLES ALLOW DEVELOPMENT OF SIMPLE, BUT EFFECTIVE MODELSExample - US food retailing
12 Fundamental DataNON-LINEAR METHODS ARE NECESSARY FOR OPTIMAL MODEL PERFORMANCEExample - Profits and Financial Strength Rating for European Banks
13 Model ValidationCRS HAS BEEN VALIDATED USING SEVERAL APPROACHES, RATHER THAN A SIMPLE “ONE ANSWER” APPROACHComprehensive validation should employ a multi-faceted approachWe start with and then dig down into the definition of credit ratings. This step is impt as we require a more precise definition to evaluate our success within both models we will talk about and
14 Default Prediction CRS DEFAULT EXPERIENCE Analysis of rated defaults shows similar ratings at ‘near default’Correspondence between CRS and public ratings by looking at ratings for a portfolio of names one year prior to default.Source: Moody’s Default Risk Service
15 Impact of SizeSTABILITY OF KEY DRIVERS, SUCH AS SIZE, IS CRITICAL TO USE ON DISPARATE PORTFOLIOSExample – Impact of asset size on model performance for chemicals
16 Credit Risk In A Model World Once adopted how do you integrate model use forCredit decisionsExposure methodologyProfit / risk maximisation minimisation or risk/rewardControl or business function
17 Credit Risk In A Model World Limit homogenisationWeighted approachBenchmarksNew deals into portfolioImmunisationCredit derivatives ?Next generation
18 Credit Risk In A Model World Out-performance through monitoringParameter adjustmentStaffing levelsInformation inputsFall back processPrayers