Contents Foreword Management report Corporate governance
Consolidated Financial Statements
Company Financial Statements
Pillar 3
developed credit risk models. These models are developed by
taking into account various risk factors including the sector,
country, size of the counterparty and type of counterparty.
When using the credit risk model, specific customer information
is entered, such as general customer behaviour, customer
financial data and market data. The credit risk models are used
as a credit decision supporting tool. The outcome of the credit
risk model is used as a starting point for determining the RRR.
Model results are combined with professional judgment and
risk management (e.g. credit committee) to take into account
relevant and material information, including those aspects
which are not (sufficiently) taken into account by the credit
risk model.
External agencies' credit ratings do not imply a specific PD,
although one can observe a default frequency for each
Standard Poor's (S&P) grade. The observed default frequency
is a backward-looking measure of PD. By matching the observed
default frequencies of the S&P grades with the average default
probabilities of associated internal RRR, a mapping has been
obtained from the external ratings by S&P to our internal ratings
for reference purposes.
The portfolio's average RRR is around R13 (PD between 0.92%
and 1.37%). For 2.4% of the portfolio, the commitments are not
fully met. If such a situation is expected an adequate allowance
will be formed for this part of the portfolio.
The IRB models calculate a client PD, which is subsequently
mapped to the RRR. For the IRB advanced portfolio, each
entity/type of credit facility has its own LGD models, which
are based upon the Rabobank LGD principles. Estimates for PD
and LGD, together with the exposure value (EAD), feed into the
calculation of EL and unexpected loss (UL). The latter is used to
determine regulatory and economic capital requirements.
Quality assurance credit risk models
Model governance
edtf 17 The Model Governance Committee (MGC) has the
responsibility to sign-off on credit risk models before
implementation (for De Lage Landen (DLL) a separate
arrangement on model validations is in place). Before MGC
sign-off is requested, all models are validated by
an independent Model Validation team. Implemented models
are reviewed on at least an annual basis including back testing
of predictions against realisations.
The Model Validation team assesses model performance
annually, based on statistical review complemented with
an in-depth analysis of model risks arising from changes in
model internal and external changes. For example, there
can be relevant changes in internal model usage, business
model, changes in external regulations and market conditions.
This periodic validation aims to assess the quality of the
model in terms of prudence, methodology, validity of key
assumptions, fit-for-purpose and compliance. The overall
conclusions on performance of the models are reported to the
MGC with a recommendation to either extend the usage of the
model, or to redevelop the model if necessary. If models are
tested as non-prudent, the MGC is informed and decides on
an appropriate capital add-on until the model is recalibrated to
a prudent level. Besides these internally reviewed risk models,
there are some risk models that are periodically reviewed by
external parties under supervision of the Model Validation team.
Assumptions used in our models are not disclosed as these are
considered proprietary.
Future Model Landscape
In 2016, it was decided to overhaul the credit modelling
landscape of Rabobank (excluding DLL) in light of the
strategical framework objectives, increased use of data-
analytics, new regulation and new modelling techniques.
In the coming years, new models will be built for the different
portfolios of Rabobank. In 2016, Rabobank started the
development of the following models:
The redevelopment of the PD, LGD and EAD models for
mortgages in the Netherlands is currently ongoing. The new
models improve the use of client behaviour, re-valuation
of collateral, downturn LGD, and using information from
defaulted exposures which are still in the workout process.
The new models will be submitted to the regulator for
a material change process at the end of June 2017.
The re-development of the PD, LGD and EAD models for SME
Retail in the Netherlands was recently initiated. It will lead to
a similar improvement of the models as the new mortgage
models.
Credit risk reporting
Credit risk reporting is based on the product administration
systems and the rating systems, which hold PD, LGD and EAD
information. Risk reporting is reconciled with financial reporting
data both at entity and Group level. Risk Management compiles
a Credit Quarterly Rabobank (CQR) report on the developments
in the credit portfolio, which is distributed among senior
management. Key risk indicators in this quarterly credit risk
report such as PD, EAD, LGD, EC and EL, are used to monitor
developments within the portfolio. Furthermore, trends in
loan impairment charges, loan impairment allowances, non-
performing loans, and number and amount of exposures are
analysed by Financial Restructuring Recovery (FR&R). Another
important periodic report is the semi-annual provisioning report.
323 6. Credit Risk