Aufrufe
vor 4 Jahren

RISIKO MANAGER 10.2019

  • Text
  • Banken
  • Informationen
  • Beispielsweise
  • Parameter
  • Historical
  • Blockchain
  • Schadenanzahl
  • Unternehmen
  • Risiken
  • Risiko
RISIKO MANAGER ist das führende Medium für alle Experten des Financial Risk Managements in Banken, Sparkassen und Versicherungen. Mit Themen aus den Bereichen Kreditrisiko, Marktrisiko, OpRisk, ERM und Regulierung vermittelt RISIKO MANAGER seinen Lesern hochkarätige Einschätzungen und umfassendes Wissen für fortschrittliches Risikomanagement.

12

12 RISIKO MANAGER 10|2019 should be similar. Validation of the pool They should also verify regularly that the results of the credit bureau score continue to have appropriate input variables in their credit rating process. Even when the credit bureau score is the main driver of the rating result, the bank should ensure that all relevant internal information regarding the creditworthiness of the obligor is considered in the rating. Pooled data The use of pooled data is treated similarly to the process of combining internal data with data derived from a different set of obligors or facilities. Estimates of PD are generated from the data used for risk quantification. If several banks use a common rating, each one should ensure that its rating process is aligned with all identically defined input risk drivers across all segments, and that all assessments of qualitative components of the model will be performed in a comparable manner. In the case of a pooled estimation of LGD parameters, the workout processes model, including testing of discriminatory power and predictive power, should be applied by each bank on its portfolio. Risk drivers Relevant risk drivers can be taken into consideration in several ways: i. when assigning exposures to different rating systems; ii. at rating system level when assigning exposures to different models; iii. as explanatory variables in models, and as drivers in the process for the assignment of PDs to grades or pools. Banks are expected to demonstrate that their models also perform adequately in terms of discriminatory and predictive power on economically significant applications of the rating systems. The subranges are identified by partitioning the full range of applications based on potential drivers for risk differentiation. For rating systems co- vering retail exposures, the following information has to be provided: » client type (e. g. private/high net worth banking, other individuals, SMEs, self-employed); » product type (e. g. consumer credit, credit card, other); » region (e. g. nomenclature of territorial units for statistics (NUTS) as defined by Eurostat); » past delinquency (e. g. obligors with delinquency events, i.e. days past due, in the last 12 months); » maturity (e. g. original or remaining maturity and for rating systems covering retail exposures secured by real estate: » type of real estate (e. g. residential, commercial, other); » past delinquency (e. g. obligors with delinquency events, i. e. days past due, in the last 12 months); » maturity (e. g. original or remaining maturity). Rating systems covering retail exposure must provide information including the default rate denominator, credit obligations, and the numerator. The default rate denominator should consist of the number of non-defaulted obligors observed at the one-year observation period, along with any further credit obligations. A credit obligation refers to any amount of principle, interest, and fees, as well as any off-balance sheet items, including guarantees. The numerator should include all obligors considered in the denominator with at least of default event during the one-year observation period where the one-year default rate is calculated by the rating grade or pool. The denominator should be assigned to a rating grade or pool at the beginning of the observation account overrides, excluding any substitution effects due to credit risk. Banks should calculate the one-year default rate, also for the subset of obligors without a rating at the start of the relevant observation period, who were in the range of the model under consideration. This remains true, even if these obligors were conservatively assigned to a rating grade of pool for the calculation of capital requirements and are “missing ratings.”

Kreditrisiko 13 To avoid doubt, the obligor should be included in all relevant calculations, including in case of a migration to a different rating grad, pool, or rating model, rating system, or approach to calculation of capital requirements within the observation period. If the long-run average default rate does not equal the average of all observed oneyear default rates, banks should compare the adjusted long-run average default rates to the maximum of the observed average from the one-year default rates of the past 5 years, and the observed average of all available one-year default rates. Testing of the probability of default (PD): PD Completeness check First of all, missing value analysis of the raw data needs to be performed. Missing information can be depicted by computing the proportion of obligors/facilities with non-populated fields (blanks or similarly default values), or with erroneous fields (not present in the internal nomenclature). All fields used when computing the default flag within all databases across the IRB workflow, from source databases through to historical PD must be shown. For reference, dates disclose the proportion of missing values for the dataset identified in the scope in terms of number and exposure. Information (name, identifier, description, particularities, etc.) relating to the databases/datasets used must also be shown. Missing value analysis is performed by computing the proportion of obligors / facilities without obligor/facility identifiers (blank values or default values, or with erroneous identifiers. All databases used in the IRB workflow, from source databases through to historical PD data, must be identified. The proportion of missing values for each database/dataset identified in the scope should be presented in terms of number and exposure. Information (name, identifier, description, particularities, etc.) relating to the databases/datasets used must be noted (refer to [EBA (2016a)]). PD Accuracy check Profile statistics (distribution analysis of obligors/facilities) of the data used when computing the default flag indicator are calculated and the distribution for the different databases is analysed. For the distribution comparison of the categorical variables, the diversity index (e.g. Shannon index) can be computed and compared in addition to the graphical comparison. All fields used when computing the default flag within all databases across the IRB workflow, from source databases through to historical PD data must be judged. For the reference, dates summarize the metrics (e.g. Shannon index) and the graphics (e.g. histograms, box plots, etc.) computed for each database/dataset are identified. Information (name, identifier, description, etc.) relating to the databases used must be documented. PD (Cross) consistency check Data transformations and treatments must be applied before checks and tests are performed (e.g. queries/computer programs used). The outcome/results should be tested, and the controls on the process or outcomes of the test must be performed. The relevant criteria and calculations used when computing the test (e.g. clarification of the criteria used to identify such missing values – blanks, default values, etc.) has to be clearly stated and documented for missing value analysis. Information on whether this test (or a similar one) is being performed by the bank, outlining current measures applied by the bank to mitigate the issues arising from the tests, if any. The assessment has to be supported with other documentation/information (technical/ functional documentation, database layouts, etc.). Historical PD data and historical LGD data should be reconciled. It must be verified that the same obligor and facility identifiers are used for the same obligors/ facilities through historical PD data and LGD data. For reference dates, the difference of number and exposure between the reconciled databases should be explained in addition to differences that must be covered. PD Uniqueness check Duplicates count for obligor/facility identifiers across the IRB workflow are checked by the consultant. Obligor and facility primary keys must be identified in all databases used in the IRB workflow. Duplicates should be accounted in each database identified in the scope in terms of number and exposure. Database related information (name, identifier, description, particularities, etc.) relating to the databases/datasets used must be addressed. LGD Analysis The modeling of LGD should be based on loss or recovery experience. Banks are not allowed to use any realized LGD inferred from market prices. Realized LGD Banks calculate the realized LGD at facility level default. The definition of a default should be identical to the one used for purposes of PD estimation, and any deviation should be justified. Banks are expected to calculate realized LGD as the ratio of losses to the observed exposure at default (EAD). The EAD definition should be the one used for CCF Credit Conversion Factor estimation. The treatment of post-default drawings should be identical for the EAD used in both the LGD and loss estimations. At the date of default, the amount forgiven or written off is not an outstanding obligation at the moment of default. The amount is added to the outstanding obligation at the moment of default included denominated of the realized LGD. The amount of exposure that is also expected to be added to realized losses. How are multiple defaults treated? An exposure that is classified as defaulted after the return to non-defaulted status should be considered as constantly defaulted starting with the first default that occurred, provided the time between the defaults is shorter than one year. As a discount rate, the primary interbank offered rate (e.g. 1 year Euribor – at the date of default-) + 5% is applied. LGD structure Potential risk drivers relevant to the specific characteristics of the type of exposures covered by the rating system must be collected and identified. Potential risk drivers should include the following:

RISIKO MANAGER

 

Copyright Risiko Manager © 2004-2017. All Rights Reserved.