Transcript Click to add title - PMF

Quantification of Credit Risk
(Croatian perspective)
Stjepan Anić, Dejan Donev
Erste & Steiermärkische Bank d.d.
ToC
1. Components of Credit risk
2. Quantification - You can manage what you can measure
3. First thing’s first - Scoring & Rating Models
4. Tasks of a modern risk manager
5. Required Competences
Risk Components
Risk
(two components)
Uncertainty
Exposure
Regulatory acknowledged types of risk
Credit Risk
Market Risk
Uncertainty
Exposure
Default risk
Recovery risk
Operational Risk
You can manage what you can measure
Credit Risk
Uncertainty
Default risk
PDi = f i (Rating grade)
i = 1, ... , n
( n number of exposure
classes )
Exposure
Recovery risk
LGD k,j= f ( k, j )
k = 1, ... , p; j=1, ... , q
EaD = f ( i , j )
( p  no. of collateral
types
l = 1, ... , r
q  no. of types of
facilities)
( r  no. of clients )
Scoring & Rating Models
• Credit quality of a client is analyzed, modeled and
ranked
• Credit Scoring  Transformation of input variables
describing banks’ client in numbers, sum of which (credit
score) gives numeric estimate of his credit quality
– Privates  socio-demographic data
– Corporates  financial ratios
• Credit Rating  grouping of score bins (plus some
other things)
• Predictive aspect of score/rating forecast default
tendency of a client within the one year horizon (PD
scoring/rating)
Problem with data
WARNING !
 Experience shows that many problems emerge from unsatisfactory quality and
availability of data
 Models are as good and accurate as are the data on which they are developed
 Time needed for preparation of raw data for the purposes of modeling is usually
dramatically underestimated (during the phase of project planning)
Scoring Model development
t
Loan applications
/
Annual financial
statements
t+12 m
Not
defaulted
Defaulted
Binomial
event
Methodology
Data mining
Techniques used to find patterns and relations within the data
 Proper usage of DM techniques for model building requires knowledge
about business problem we’re trying to solve
Statistics
Data-bases
Data Mining
Machine learning
Visualisation
IT technology
Tasks of a modern risk manager
 IDENTIFICATION OF RISK RELEVANT INFORMATION 
creating a list of necessary RM measures and procedures for
all types of products and clients
 PROPER RECORDING OF IDENTIFIED INFORMATION 
Centralized Risk DWH  Data–collection in hands of people
which understand the data and their usage
 CALCULATION OF RISK PARAMETERS  transformation
of recorded info into prediction of possible losses
(construction of a probability of loss distribution)
 INTERPRETATION AND USAGE OF RESULTS  RM must
insure that resulting risk parameters (PD, EL, CapReq, etc.)
are used throughout the bank in a consistent manner (loan
decisioning, portfolio mngmt, planning, provisioning, pricing,
etc.)
Required Competences
 TECHNICAL EXPERTISE  IT competences (knowing how to retrieve
data from data-bases, SQL, basic programming skills - VBA)
 METHODOLOGICAL EXPERTISE  skills in quantitative analytical
modeling (mathematical and statistical modeling, econometrics) and
skills in predictive data-mining (SAS, MATLAB, SPSS, etc.)
 BUSINESS EXPERTISE  knowing the business (banking & finance,
risk management, CRM, etc.)
 ANALYTICAL (AND ABSTRACT THINKING) MINDSET  can
transform business problems into abstract terms and solve them like
mathematical problems in algorithmic form
 MODERN RM ENVIRONMENT = CROSS-FUNCTIONAL TEAMS
Four major competences
Analytical
expertise
Business
expertise
Technical
expertise
Methodological
expertise
All four planets in this Risk
Orbit have to function
perfectly, otherwise we
could be facing...
consequences
of truly cosmic
proportions !
Thank you for your attention!