8/7/2019 Benos - SME Risk Rating
1/18
Group Risk Management Division 1
A Simple Credit Risk Modelfor SMEs
Dr. Alexandros Benos
National Bank of Greece
8/7/2019 Benos - SME Risk Rating
2/18
Group Risk Management Division 2
What is a credit risk model?
Its a mathematical function which,given a set of criteria as input(independent variables), returns ameasure for credit risk as an output(dependent variable).
8/7/2019 Benos - SME Risk Rating
3/18
Group Risk Management Division 3
Measures of Credit Risk Probability of Default measures
At OBLIGOR level PD Score Rating
Loss Given Default measure At FACILITY level
Depends mainly on the debt seniority and collateral
Portfolio measures Expected Loss Unexpected Loss VaR
Loss Distribution
Estimate for the obligors creditworthiness
8/7/2019 Benos - SME Risk Rating
4/18Group Risk Management Division 4
The use of PD models
Classification of Obligors depending ontheir probability to default
Used for
Credit Granting
Credit Pricing
8/7/2019 Benos - SME Risk Rating
5/18Group Risk Management Division 5
The Credit Granting problem
To lend or not to
lend?
To lendTo lend or not toor not to
lend?lend?
8/7/2019 Benos - SME Risk Rating
6/18Group Risk Management Division 6
An example (1)
Firm A (Clothes manufacture) requestsa loan of 5.000.000
To lend or not to lend ?
Can we answer the question?
No, we need more information!
8/7/2019 Benos - SME Risk Rating
7/18Group Risk Management Division 7
An example (2)
Firm A requests a loan of 5.000.000 Criteria:
Industry: Clothes manufacture HIGH RISK
Size: Annual sales 50.000.000 GOOD
Leverage: Equity / Liabilities = 1:12 BAD Liquidity: Current ratio = 1,1 MEDIUM
Earnings: NI / Sales = 15% GOOD
To lend or not to lend ? Can we answer the question NOW?
Maybe some statistics would help
8/7/2019 Benos - SME Risk Rating
8/18Group Risk Management Division 8
An example (3)
Statistical inferences on the historic data ofthe specific portfolio:
Mean default frequency = 3,8%
The default probability for a firm with the aboveproperties is 3,4%
To lend or not to lend ?
Can we answer the question NOW ?
Yes, the granting of the loan will improveimprove thequality of the portfolio.
8/7/2019 Benos - SME Risk Rating
9/18Group Risk Management Division 9
Pricing the Credit in example (4)
Take into account other parameters too: The LGD for the specific type of credit and
collateral.
Using it we can estimate EL(%) =PD x LGD The proposed spread must be greater than EL(%) for
the credit to be profitable
Example: In case LGD = 50%, EL(%) = 3,4% x 50% =
1,7%, hence
The minimum spread (due to credit risk) is170bps
8/7/2019 Benos - SME Risk Rating
10/18Group Risk Management Division 10
Why a model then?
The statistics make the difference, butthe initiating of a new statistical projectfor each credit request:
is unpractical,
charges the credit granting process with
additional time and costs and most interestingly it is not necessary!
Instead, we can use a PD model
8/7/2019 Benos - SME Risk Rating
11/18Group Risk Management Division 11
A PD model
Contains in it, all the correlations between theselected criteria and the historically observeddefaults.
Achieves the same results with the originalstatistical process.
Is costless in use and returns its results
instantly. Is created once but can be used to evaluate
any obligors PD for a long time period.
8/7/2019 Benos - SME Risk Rating
12/18
Group Risk Management Division 12
Different PD Measures The absolute value of PD is impractical
As a point-in-timemeasure, it is relatively unstable throughtime.
Credit scores Are indicators in a continuous numeric scale. Have lower and upper bounds, representing relative default
risk.
Credit ratings Are ranks which stand for similar credit risk profiles of the
obligors. They can be numerical (1,2,) or symbolic (AA+, A, BB- etc)
They are usually much more stable than instant PDs. They are often considered to be through-the-cycle
measures Compatible with Basle 2 IRB approaches
8/7/2019 Benos - SME Risk Rating
13/18
Group Risk Management Division 13
NBGs Model
Supports evaluation of firms located in different countries (mostly Balkan
ones),
with or without full financial statements, for which their shareholders liability extends to
their personal property
Forward looking features Through cash flow auto and manual projections
Uses both qualitative & financial criteria
8/7/2019 Benos - SME Risk Rating
14/18
Group Risk Management Division 14
Groups of Criteria Environment
Country rating Sector risk Competition
Management
Managers experience Stability
Enterprise Years of operation Customers concentration Operational risks
Credit history Court decisions Credit Bureau Black list
Defects in operations Signs of Illiquidity
Frequent past dues to Bankand other creditors
Leverage
Debt To Assets ratio Shareholders property
Expected cash flows Debt capacity and free capital Current past dues Financial projection
Cash flows volatility
8/7/2019 Benos - SME Risk Rating
15/18
Group Risk Management Division 15
NBGs Rating Scale
A2 to AaaA3
Baa1
Baa2
Baa3Ba1
Ba2
Ba3
B1
B2
B3
CaaMoodys
B+
0,08%A to AAAA10,12%A-
A2
0,20%BBB+A3
0,30%BBBB1
0,40%BBB-B20,60%BB+
B3
1,50%BBC1
2,30%BB-C2
3,50%C35,85%
D1
7,45%B
D2
12,20%B-D3
18,00%CCC+ES&PModel
Mean
PD
Rating
It directlycorresponds to S&Pand Moodys scales Naming convention
is indicative only.
It can be adaptedaccording to theconcentrations of thefirms in the ranks
8/7/2019 Benos - SME Risk Rating
16/18
Group Risk Management Division 16
The Application Supports the full evaluation process for corporate
obligors
Provides tools for financial and ratio analysis
cash flow analysis financial projections
Accommodates the Credit Risk Model
Can be structured either as a full-featured independent network application, or
as a module that can be incorporated in the banks coresystem
8/7/2019 Benos - SME Risk Rating
17/18
Group Risk Management Division 17
The ArchitectureClientsClients
User interface
Application ServerApplication Server
Model
Business logic
Data serverData server
Databases
Data maintenance
Internet
Core systemCore system
8/7/2019 Benos - SME Risk Rating
18/18
Group Risk Management Division 18
Adoption Steps Exploration of
the existing credit granting and risk measuring processes the risk managers and the credit officers point of view the available data the IT systems in place
Data analysis and initial model calibration Design for the incorporation of the model into the
banks processes and IT systems
Implementation and documentation of the newprocesses Users training Operation and maintenance Model validation and recalibration on an annual
basis