ING Direct 2015 Annual Report Download - page 260

Contents
Report of the
Executive Board
Corporate
Governance
Consolidated
annual accounts
Parent company
annual accounts
Other
information
Additional
information
Additional Pillar III information - continued
Financial
institutions
83,191 10.5% 17,091 8.5%
PD
3
The main PD model applied is Bank Commercial based
upon financial,
qualitative and country information. Other
PD models for different types of financial institutions are
built using a similar framework, but are more specialised
for the specific characteristics of the financial institution.
>7 years
LGD
1
This LGD model was developed based on expert
judgement, supported by limited internal and external
data. The developed LGD model is based on ultimate
recovery rates.
>7
years
EAD
1
The Low Default EAD model is a hybrid model used that
pools default
information from multiple low default
portfolios, including financial institutions.
>7
years
Corporates
-
Specialized
lending
72,911 9.2% 26,053 12.9%
PD
3
Expert based scorecards Models predict a rating for
Commercial Property Finance, Project Finance,
Trade and
Commodity Finance.
>7
years
LGD
3
Hybrid LGD Models predict loss given default for
Commercial Property Finance, Project Finance, Trade and
Commodity Fi
nance.
>7
years
EAD
1
There is a dedicated EAD model for commercial property
finance due to the specificities of this portfolio.
>7
years
Large
Corporates
135,286 17.0% 63,923 31.7%
PD
1
The Corporate Large model is a global hybrid model build
on 13
years of data, including balance sheet and
qualitative information as well as
country risk and parent
influence
.
>7
years
LGD
1
Loss Given Default for Large corporates are predicted by a
dedicated hybrid LGD model using both no loss rates as
well as secured/unsecured recovery rates.
>7
years
EAD
1
The Low Default EAD model is a hybrid model used that
pools default information from multiple low default
portfolios, including large corporates.
>7
years
SME
64,151 8.1% 32,958 16.3%
PD
6
The SME PD models are estimated statistically and
directly predict a PD. Most of
these models are developed
locally to reflect regional/jurisdiction circumstances
.
>7
years
LGD
6
Local statistical models or hybrid models use various data
inputs on cure behaviour as well as cost and recovery.
>7
years
EAD
6
Local statistical models that use various data inputs,
including product type and geography.
>7
years
Secured by
Res.
Mortgage
278,565 35.1% 42,297 20.9%
PD
71
The PD mortgages models are all developed statistically
and include borrower specific information,
payment
behaviour and product related information. These are
statistical models that directly predict a PD
.
>7
years
LGD
7
Local statistical models or hybrid models use various data
inputs on cure behaviour as well as cost and recovery
>7
years
EAD
7
Local statistical models that use various data inputs,
including product type and geography.
>7
years
Private
Individuals
22,236 2.8% 9,859 4.9%
PD
41
The PD models for private individuals are all developed
statistically and include borrower
specific information,
payment behaviour and product related information.
These are statistical models that directly predict a PD
.
>7
years
LGD
4
Local statistical models use various data inputs on cure
behaviour as well as cost and recovery.
>7
years
EAD
4
Local statistical models that use various data inputs,
including product type and geography.
>7
years
Other1
16,199
2.0%
1,975
1.0%
Other
(Covered Bonds, Structured assets)
Total
794,148
100%
201,967
100%
1 Belgian PD models provide a rating at a customer level, covering both secured and unsecured loans.
Includes the AIRB portfolio only; excludes securitisations, CVA RWA, equities and ONCOA.
AIRB models and methodology - continued
ING Bank Annual Report 2015 258