SAARC FINANCE,
REGIONAL SEMINAR, ISLAMABAD
30
PD Model
Development - Approaches
GOOD-BAD
SHADOW-BOND
CONSTRAINED EXPERT JUDGMENT
Data
Availability
Availability
Sufficient
data available for at least 100 defaulted and 200 non-defaulted clients
Insufficient
data for Good-Bad analysis, but external ratings available for a statistically significant sample of clients
Insufficient
data to develop a statistical model (few defaults or external ratings available)
Measure of Risk
(i.e. what the model will explain)
(i.e. what the model will explain)
Observed
1-year default rate of counterparties in the development sample
Implied
1-year default rate of the issuer rating of S&P rated counterparties in the development
sample
Ranking of
relative riskiness of counterparties in the development sample, supplemented by
"anchor points" of risk level such as long term default rate of portfolio or
S&P
ratings of limited number of counterparties
Development
Method
Method
Statistical
comparison of defaulted and non-defaulted clients to identify factors that are
predictive of
default;
this may be supplemented with expert input
this may be supplemented with expert input
Statistical
model built to mimic 1-year implied PD of external ratings, which can then be applied to unrated companies; this may be supplemented with expert input
Selection
and weighting of factors through experts, supplemented with statistical analysis in commensurate with quality and quantity of data, to mimic
relative riskiness
ranking
Application
>100 internal defaults
>150
external ratings
Internal
expert ranking
Middle Market
Emerging Middle Market
Emerging Middle Market
Large Corporate
Bank
Sovereign
Bank
Sovereign
Commodity
Traders & Buyers
Non Bank FI
Project Finance
Commercial Real Estate
Project Finance
Commercial Real Estate