Credit Risk Management in Banks: Hard Information, Soft Information and Manipulation
1. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Credit Risk Management in Banks
Hard Information, Soft Information and Manipulation
B. Godbillon-Camus and C.J. Godlewski
Institut d’Etudes Politiques
Universit´ Robert Schuman
e
Strasbourg 3
EFMA 2006 Annual Conference
28 June - 1 July
Universidade Complutense, Madrid, Spain
Godbillon-Camus and Godlewski Credit Risk Management and Information 1/ 24
2. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Outline
1 Background, Motivation & Aim
2 Literature Survey
3 Model
4 Results
5 Discussion
Godbillon-Camus and Godlewski Credit Risk Management and Information 2/ 24
3. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Background, Motivation & aim
Efficient information treatment is crucial for the banking
industry (Fama, 1985) ⇒ credit risk management process
Recent distinction of information’s type produced and treated by
banks
Hard versus Soft Information (Petersen, 2004)
Different lending technologies : Transaction Lending versus
Relationship Lending
Organizational structure adapted to information’s type (Stein,
2002; Takats, 2004)
AIM : Investigate the influence of information’s type
on bank’s risk taking behaviour
⇒ principal-agent model with moral hazard with hidden
information
Godbillon-Camus and Godlewski Credit Risk Management and Information 3/ 24
4. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Background, Motivation & aim
Efficient information treatment is crucial for the banking
industry (Fama, 1985) ⇒ credit risk management process
Recent distinction of information’s type produced and treated by
banks
Hard versus Soft Information (Petersen, 2004)
Different lending technologies : Transaction Lending versus
Relationship Lending
Organizational structure adapted to information’s type (Stein,
2002; Takats, 2004)
AIM : Investigate the influence of information’s type
on bank’s risk taking behaviour
⇒ principal-agent model with moral hazard with hidden
information
Godbillon-Camus and Godlewski Credit Risk Management and Information 3/ 24
5. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Background, Motivation & aim
Efficient information treatment is crucial for the banking
industry (Fama, 1985) ⇒ credit risk management process
Recent distinction of information’s type produced and treated by
banks
Hard versus Soft Information (Petersen, 2004)
Different lending technologies : Transaction Lending versus
Relationship Lending
Organizational structure adapted to information’s type (Stein,
2002; Takats, 2004)
AIM : Investigate the influence of information’s type
on bank’s risk taking behaviour
⇒ principal-agent model with moral hazard with hidden
information
Godbillon-Camus and Godlewski Credit Risk Management and Information 3/ 24
6. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Background, Motivation & aim
Efficient information treatment is crucial for the banking
industry (Fama, 1985) ⇒ credit risk management process
Recent distinction of information’s type produced and treated by
banks
Hard versus Soft Information (Petersen, 2004)
Different lending technologies : Transaction Lending versus
Relationship Lending
Organizational structure adapted to information’s type (Stein,
2002; Takats, 2004)
AIM : Investigate the influence of information’s type
on bank’s risk taking behaviour
⇒ principal-agent model with moral hazard with hidden
information
Godbillon-Camus and Godlewski Credit Risk Management and Information 3/ 24
7. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Background, Motivation & aim
Efficient information treatment is crucial for the banking
industry (Fama, 1985) ⇒ credit risk management process
Recent distinction of information’s type produced and treated by
banks
Hard versus Soft Information (Petersen, 2004)
Different lending technologies : Transaction Lending versus
Relationship Lending
Organizational structure adapted to information’s type (Stein,
2002; Takats, 2004)
AIM : Investigate the influence of information’s type
on bank’s risk taking behaviour
⇒ principal-agent model with moral hazard with hidden
information
Godbillon-Camus and Godlewski Credit Risk Management and Information 3/ 24
8. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Background, Motivation & aim
Efficient information treatment is crucial for the banking
industry (Fama, 1985) ⇒ credit risk management process
Recent distinction of information’s type produced and treated by
banks
Hard versus Soft Information (Petersen, 2004)
Different lending technologies : Transaction Lending versus
Relationship Lending
Organizational structure adapted to information’s type (Stein,
2002; Takats, 2004)
AIM : Investigate the influence of information’s type
on bank’s risk taking behaviour
⇒ principal-agent model with moral hazard with hidden
information
Godbillon-Camus and Godlewski Credit Risk Management and Information 3/ 24
9. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Hard versus Soft Information
“Hard information (. . . ) is when everyone agrees on its meaning.
(. . . ) Honest disagreements arise when two people perfectly observe
information yet interpret this information differently (i.e. soft
information)”
(Kirschenheiter, 2002)
Nature : quantitative vs qualitative (numbers vs words) / backward
versus forward looking
Collecting method : impersonal vs personal (production’s context,
role of the agent responsible for the production and treatment
process)
Cognitive factors : weakly present vs strongly present (subjective
judgment, interpretation and perception, opinions . . . )
Lending technology : transaction lending vs relationship lending
Organizational structure : centralized and hierarchical vs
decentralized and non-hierarchical
Godbillon-Camus and Godlewski Credit Risk Management and Information 4/ 24
10. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Hard versus Soft Information
“Hard information (. . . ) is when everyone agrees on its meaning.
(. . . ) Honest disagreements arise when two people perfectly observe
information yet interpret this information differently (i.e. soft
information)”
(Kirschenheiter, 2002)
Nature : quantitative vs qualitative (numbers vs words) / backward
versus forward looking
Collecting method : impersonal vs personal (production’s context,
role of the agent responsible for the production and treatment
process)
Cognitive factors : weakly present vs strongly present (subjective
judgment, interpretation and perception, opinions . . . )
Lending technology : transaction lending vs relationship lending
Organizational structure : centralized and hierarchical vs
decentralized and non-hierarchical
Godbillon-Camus and Godlewski Credit Risk Management and Information 4/ 24
11. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Hard versus Soft Information
“Hard information (. . . ) is when everyone agrees on its meaning.
(. . . ) Honest disagreements arise when two people perfectly observe
information yet interpret this information differently (i.e. soft
information)”
(Kirschenheiter, 2002)
Nature : quantitative vs qualitative (numbers vs words) / backward
versus forward looking
Collecting method : impersonal vs personal (production’s context,
role of the agent responsible for the production and treatment
process)
Cognitive factors : weakly present vs strongly present (subjective
judgment, interpretation and perception, opinions . . . )
Lending technology : transaction lending vs relationship lending
Organizational structure : centralized and hierarchical vs
decentralized and non-hierarchical
Godbillon-Camus and Godlewski Credit Risk Management and Information 4/ 24
12. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Hard versus Soft Information
“Hard information (. . . ) is when everyone agrees on its meaning.
(. . . ) Honest disagreements arise when two people perfectly observe
information yet interpret this information differently (i.e. soft
information)”
(Kirschenheiter, 2002)
Nature : quantitative vs qualitative (numbers vs words) / backward
versus forward looking
Collecting method : impersonal vs personal (production’s context,
role of the agent responsible for the production and treatment
process)
Cognitive factors : weakly present vs strongly present (subjective
judgment, interpretation and perception, opinions . . . )
Lending technology : transaction lending vs relationship lending
Organizational structure : centralized and hierarchical vs
decentralized and non-hierarchical
Godbillon-Camus and Godlewski Credit Risk Management and Information 4/ 24
13. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Hard versus Soft Information
“Hard information (. . . ) is when everyone agrees on its meaning.
(. . . ) Honest disagreements arise when two people perfectly observe
information yet interpret this information differently (i.e. soft
information)”
(Kirschenheiter, 2002)
Nature : quantitative vs qualitative (numbers vs words) / backward
versus forward looking
Collecting method : impersonal vs personal (production’s context,
role of the agent responsible for the production and treatment
process)
Cognitive factors : weakly present vs strongly present (subjective
judgment, interpretation and perception, opinions . . . )
Lending technology : transaction lending vs relationship lending
Organizational structure : centralized and hierarchical vs
decentralized and non-hierarchical
Godbillon-Camus and Godlewski Credit Risk Management and Information 4/ 24
14. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Pros and Cons
HARD information = low cost, durable, comparable, verifiable, not
manipulable
⇒ e.g.: scoring = increases credit’s availability and reduces credit’s
cost (risk adjusted pricing), BUT doesn’t increase risk
measurement’s precision as a complementary risk measurement tool
(Feldman, 1997; Berger et al. 2002; Frame et al., 2002)
SOFT information = multi-dimensional, richer, more precise, not
verifiable, manipulable
⇒ output of a bank-borrower relationship (private information,
multiple interactions) (Boot, 2000); can also increase credit’s
availability and reduce its cost
Godbillon-Camus and Godlewski Credit Risk Management and Information 5/ 24
15. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Pros and Cons
HARD information = low cost, durable, comparable, verifiable, not
manipulable
⇒ e.g.: scoring = increases credit’s availability and reduces credit’s
cost (risk adjusted pricing), BUT doesn’t increase risk
measurement’s precision as a complementary risk measurement tool
(Feldman, 1997; Berger et al. 2002; Frame et al., 2002)
SOFT information = multi-dimensional, richer, more precise, not
verifiable, manipulable
⇒ output of a bank-borrower relationship (private information,
multiple interactions) (Boot, 2000); can also increase credit’s
availability and reduce its cost
Godbillon-Camus and Godlewski Credit Risk Management and Information 5/ 24
16. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Impact of Soft Information on Default’s Risk Prediction
Empirical evidence by Grunert et al. (2002) and Lehmann (2003)
Soft factors are more stable and precise
Soft factors increase classification and discriminatory power of the
default’s prediction models
Godbillon-Camus and Godlewski Credit Risk Management and Information 6/ 24
17. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Impact of Soft Information on Default’s Risk Prediction
Empirical evidence by Grunert et al. (2002) and Lehmann (2003)
Soft factors are more stable and precise
Soft factors increase classification and discriminatory power of the
default’s prediction models
Godbillon-Camus and Godlewski Credit Risk Management and Information 6/ 24
18. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Organizational Structure and Information 1/2
Hyp.: Soft information is more precise but not verifiable and thus
manipulable
These characteristics imply an adapted organizational structure in
order to avoid consequences and costs of soft information
manipulation
Bank-borrower relationship, which gives access to soft
information, is a source of asymmetries between the agent in
charge of the information’s treatment and the principal who takes
his funds allocation and risk management decisions upon information
transmitted by the agent
The agent can extract private benefits and thus affect principal’s
decisions efficiency
Godbillon-Camus and Godlewski Credit Risk Management and Information 7/ 24
19. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Organizational Structure and Information 1/2
Hyp.: Soft information is more precise but not verifiable and thus
manipulable
These characteristics imply an adapted organizational structure in
order to avoid consequences and costs of soft information
manipulation
Bank-borrower relationship, which gives access to soft
information, is a source of asymmetries between the agent in
charge of the information’s treatment and the principal who takes
his funds allocation and risk management decisions upon information
transmitted by the agent
The agent can extract private benefits and thus affect principal’s
decisions efficiency
Godbillon-Camus and Godlewski Credit Risk Management and Information 7/ 24
20. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Organizational Structure and Information 1/2
Hyp.: Soft information is more precise but not verifiable and thus
manipulable
These characteristics imply an adapted organizational structure in
order to avoid consequences and costs of soft information
manipulation
Bank-borrower relationship, which gives access to soft
information, is a source of asymmetries between the agent in
charge of the information’s treatment and the principal who takes
his funds allocation and risk management decisions upon information
transmitted by the agent
The agent can extract private benefits and thus affect principal’s
decisions efficiency
Godbillon-Camus and Godlewski Credit Risk Management and Information 7/ 24
21. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Organizational Structure and Information 2/2
Stein (2002) : adequacy between organizational structure
(hierarchical & centralized vs non hierarchical & decentralized) and
information’s type (hard vs soft) (extensions by Takats, 2004)
Small banks seem to have an advantage in processing soft
information within a bank-borrower relationship framework (Berger
2004; De Young et al., 2004; Scott, 2004)
Empirical evidence : Berger and Udell (2002) and Berger et al.
(2001, 2002)
Bernardo et al. (2001), Ozerturk (2004), Ozbas (2005) : role of
the wages and budget allocation policy in implementing proper
incentives structure for the agent responsable for information’s
treatment
Godbillon-Camus and Godlewski Credit Risk Management and Information 8/ 24
22. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Organizational Structure and Information 2/2
Stein (2002) : adequacy between organizational structure
(hierarchical & centralized vs non hierarchical & decentralized) and
information’s type (hard vs soft) (extensions by Takats, 2004)
Small banks seem to have an advantage in processing soft
information within a bank-borrower relationship framework (Berger
2004; De Young et al., 2004; Scott, 2004)
Empirical evidence : Berger and Udell (2002) and Berger et al.
(2001, 2002)
Bernardo et al. (2001), Ozerturk (2004), Ozbas (2005) : role of
the wages and budget allocation policy in implementing proper
incentives structure for the agent responsable for information’s
treatment
Godbillon-Camus and Godlewski Credit Risk Management and Information 8/ 24
23. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Organizational Structure and Information 2/2
Stein (2002) : adequacy between organizational structure
(hierarchical & centralized vs non hierarchical & decentralized) and
information’s type (hard vs soft) (extensions by Takats, 2004)
Small banks seem to have an advantage in processing soft
information within a bank-borrower relationship framework (Berger
2004; De Young et al., 2004; Scott, 2004)
Empirical evidence : Berger and Udell (2002) and Berger et al.
(2001, 2002)
Bernardo et al. (2001), Ozerturk (2004), Ozbas (2005) : role of
the wages and budget allocation policy in implementing proper
incentives structure for the agent responsable for information’s
treatment
Godbillon-Camus and Godlewski Credit Risk Management and Information 8/ 24
24. Background, Motivation & Aim
Hard versus Soft Information
Literature Survey
Pros and Cons
Model
Impact of Soft Information on Default’s Risk Prediction
Results
Organizational Structure and Information
Discussion
Literature Survey
Organizational Structure and Information 2/2
Stein (2002) : adequacy between organizational structure
(hierarchical & centralized vs non hierarchical & decentralized) and
information’s type (hard vs soft) (extensions by Takats, 2004)
Small banks seem to have an advantage in processing soft
information within a bank-borrower relationship framework (Berger
2004; De Young et al., 2004; Scott, 2004)
Empirical evidence : Berger and Udell (2002) and Berger et al.
(2001, 2002)
Bernardo et al. (2001), Ozerturk (2004), Ozbas (2005) : role of
the wages and budget allocation policy in implementing proper
incentives structure for the agent responsable for information’s
treatment
Godbillon-Camus and Godlewski Credit Risk Management and Information 8/ 24
25. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 1/3
Bank’s Director (banker) = principal and Credit Officer = agent,
both risk averse
Banker’s decision = balance sheet’s structure, made upon the
information produced by the credit officer
Bank’s profit
˜ r
Π = ˜A A − rD D − w (˜A ) − c
r (1)
˜A : random assets’ (and credit officer’s budget) A return, rD : interest
r
rate on deposits D, w (˜A ) : credit officer salary (eventually function of the
r
random assets’ return ˜A ), c : credit officer’s unemployment insurance
r
cost (normalized to 0)
Godbillon-Camus and Godlewski Credit Risk Management and Information 9/ 24
26. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 1/3
Bank’s Director (banker) = principal and Credit Officer = agent,
both risk averse
Banker’s decision = balance sheet’s structure, made upon the
information produced by the credit officer
Bank’s profit
˜ r
Π = ˜A A − rD D − w (˜A ) − c
r (1)
˜A : random assets’ (and credit officer’s budget) A return, rD : interest
r
rate on deposits D, w (˜A ) : credit officer salary (eventually function of the
r
random assets’ return ˜A ), c : credit officer’s unemployment insurance
r
cost (normalized to 0)
Godbillon-Camus and Godlewski Credit Risk Management and Information 9/ 24
27. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 1/3
Bank’s Director (banker) = principal and Credit Officer = agent,
both risk averse
Banker’s decision = balance sheet’s structure, made upon the
information produced by the credit officer
Bank’s profit
˜ r
Π = ˜A A − rD D − w (˜A ) − c
r (1)
˜A : random assets’ (and credit officer’s budget) A return, rD : interest
r
rate on deposits D, w (˜A ) : credit officer salary (eventually function of the
r
random assets’ return ˜A ), c : credit officer’s unemployment insurance
r
cost (normalized to 0)
Godbillon-Camus and Godlewski Credit Risk Management and Information 9/ 24
28. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 2/3
Banker’s utility (β : constant risk aversion’s coefficient)
˜
UB = − exp−β(Π) (2)
Credit officer’s utility (γ : constant risk aversion’s coefficient)
UC = − exp−γ(˜A A+w (˜A ))
r r
(3)
Information concerns ˜A ⇒ modelled as a signal µ ∼ N(¯, υ 2 )
r ˜ µ
(following Bhattacharya and Pfleiderer, 1982) ⇒ linked to ˜A as
r
˜A = µ + ε,
r ˜ ˜ (4)
2
with ε ∼ N(0, σ ) ⇒ conditional distribution upon realization of µ is
˜
(˜A | µ) ∼ N(µ, σ 2 )
r
Godbillon-Camus and Godlewski Credit Risk Management and Information 10/ 24
29. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 2/3
Banker’s utility (β : constant risk aversion’s coefficient)
˜
UB = − exp−β(Π) (2)
Credit officer’s utility (γ : constant risk aversion’s coefficient)
UC = − exp−γ(˜A A+w (˜A ))
r r
(3)
Information concerns ˜A ⇒ modelled as a signal µ ∼ N(¯, υ 2 )
r ˜ µ
(following Bhattacharya and Pfleiderer, 1982) ⇒ linked to ˜A as
r
˜A = µ + ε,
r ˜ ˜ (4)
2
with ε ∼ N(0, σ ) ⇒ conditional distribution upon realization of µ is
˜
(˜A | µ) ∼ N(µ, σ 2 )
r
Godbillon-Camus and Godlewski Credit Risk Management and Information 10/ 24
30. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 2/3
Banker’s utility (β : constant risk aversion’s coefficient)
˜
UB = − exp−β(Π) (2)
Credit officer’s utility (γ : constant risk aversion’s coefficient)
UC = − exp−γ(˜A A+w (˜A ))
r r
(3)
Information concerns ˜A ⇒ modelled as a signal µ ∼ N(¯, υ 2 )
r ˜ µ
(following Bhattacharya and Pfleiderer, 1982) ⇒ linked to ˜A as
r
˜A = µ + ε,
r ˜ ˜ (4)
2
with ε ∼ N(0, σ ) ⇒ conditional distribution upon realization of µ is
˜
(˜A | µ) ∼ N(µ, σ 2 )
r
Godbillon-Camus and Godlewski Credit Risk Management and Information 10/ 24
31. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 3/3
Hyp.: Hard versus Soft Information ⇔ less / more precise ⇒
σ S < σH
Credit risk management ⇔ capital K allocation for Value at Risk
coverage
Banker states bank’s default probability α (exogenous) as
p (A(1 + ˜A ) − D(1 + rD ) < 0) = α
r (5)
following Broll and Wahl (2003) we infer VaR per risky assets unit
as follows
rD − µ − uα σ
rα = (6)
1 + rD
uα : fractile of a standardized distribution, rα = “VaR rate” ⇒ increases
with σ, as uα < 0,
VaRα = rα A (7)
Godbillon-Camus and Godlewski Credit Risk Management and Information 11/ 24
32. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 3/3
Hyp.: Hard versus Soft Information ⇔ less / more precise ⇒
σ S < σH
Credit risk management ⇔ capital K allocation for Value at Risk
coverage
Banker states bank’s default probability α (exogenous) as
p (A(1 + ˜A ) − D(1 + rD ) < 0) = α
r (5)
following Broll and Wahl (2003) we infer VaR per risky assets unit
as follows
rD − µ − uα σ
rα = (6)
1 + rD
uα : fractile of a standardized distribution, rα = “VaR rate” ⇒ increases
with σ, as uα < 0,
VaRα = rα A (7)
Godbillon-Camus and Godlewski Credit Risk Management and Information 11/ 24
33. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 3/3
Hyp.: Hard versus Soft Information ⇔ less / more precise ⇒
σ S < σH
Credit risk management ⇔ capital K allocation for Value at Risk
coverage
Banker states bank’s default probability α (exogenous) as
p (A(1 + ˜A ) − D(1 + rD ) < 0) = α
r (5)
following Broll and Wahl (2003) we infer VaR per risky assets unit
as follows
rD − µ − uα σ
rα = (6)
1 + rD
uα : fractile of a standardized distribution, rα = “VaR rate” ⇒ increases
with σ, as uα < 0,
VaRα = rα A (7)
Godbillon-Camus and Godlewski Credit Risk Management and Information 11/ 24
34. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 3/3
Hyp.: Hard versus Soft Information ⇔ less / more precise ⇒
σ S < σH
Credit risk management ⇔ capital K allocation for Value at Risk
coverage
Banker states bank’s default probability α (exogenous) as
p (A(1 + ˜A ) − D(1 + rD ) < 0) = α
r (5)
following Broll and Wahl (2003) we infer VaR per risky assets unit
as follows
rD − µ − uα σ
rα = (6)
1 + rD
uα : fractile of a standardized distribution, rα = “VaR rate” ⇒ increases
with σ, as uα < 0,
VaRα = rα A (7)
Godbillon-Camus and Godlewski Credit Risk Management and Information 11/ 24
35. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Main Framework 3/3
Hyp.: Hard versus Soft Information ⇔ less / more precise ⇒
σ S < σH
Credit risk management ⇔ capital K allocation for Value at Risk
coverage
Banker states bank’s default probability α (exogenous) as
p (A(1 + ˜A ) − D(1 + rD ) < 0) = α
r (5)
following Broll and Wahl (2003) we infer VaR per risky assets unit
as follows
rD − µ − uα σ
rα = (6)
1 + rD
uα : fractile of a standardized distribution, rα = “VaR rate” ⇒ increases
with σ, as uα < 0,
VaRα = rα A (7)
Godbillon-Camus and Godlewski Credit Risk Management and Information 11/ 24
36. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard Information 1/3
First step: Banker decides of the credit officer’s salary
Second step: Principal’s decisions concerning capital K , assets A
and deposits D are made upon the signal µ on the random assets’
˜
return ˜A distribution, transmitted by the agent
r
Hyp.: Hard Information (e.g. a score) provided by the credit officer
is verifiable and non manipulable
Credit officer’s salary
w (˜A ) = w0 .
r (8)
Godbillon-Camus and Godlewski Credit Risk Management and Information 12/ 24
37. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard Information 1/3
First step: Banker decides of the credit officer’s salary
Second step: Principal’s decisions concerning capital K , assets A
and deposits D are made upon the signal µ on the random assets’
˜
return ˜A distribution, transmitted by the agent
r
Hyp.: Hard Information (e.g. a score) provided by the credit officer
is verifiable and non manipulable
Credit officer’s salary
w (˜A ) = w0 .
r (8)
Godbillon-Camus and Godlewski Credit Risk Management and Information 12/ 24
38. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard Information 1/3
First step: Banker decides of the credit officer’s salary
Second step: Principal’s decisions concerning capital K , assets A
and deposits D are made upon the signal µ on the random assets’
˜
return ˜A distribution, transmitted by the agent
r
Hyp.: Hard Information (e.g. a score) provided by the credit officer
is verifiable and non manipulable
Credit officer’s salary
w (˜A ) = w0 .
r (8)
Godbillon-Camus and Godlewski Credit Risk Management and Information 12/ 24
39. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard Information 1/3
First step: Banker decides of the credit officer’s salary
Second step: Principal’s decisions concerning capital K , assets A
and deposits D are made upon the signal µ on the random assets’
˜
return ˜A distribution, transmitted by the agent
r
Hyp.: Hard Information (e.g. a score) provided by the credit officer
is verifiable and non manipulable
Credit officer’s salary
w (˜A ) = w0 .
r (8)
Godbillon-Camus and Godlewski Credit Risk Management and Information 12/ 24
40. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard Information 2/3
Optimization program with Hard information
max EUB ,
w0 ,K ,A,D
¯
EUC ≥ U,
K , A, D ∈ arg max EUB
ˆ ˆ ˆ
K ,A,D (9)
ˆ ˆ ˆ
K + D − A = 0,
ˆ
K − VaRα ≥ 0,
ˆ ˆ
VaRα = rα A = rD −µ−uα σ A.
1+rD
Godbillon-Camus and Godlewski Credit Risk Management and Information 13/ 24
41. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard Information 3/3
with
+∞
EUB = − exp[−β(˜A A−rD D−w0 )] η(˜A |µ)drA
r
r
−∞
+∞
EUC = − exp[−γ(˜A A+w0 )] η(˜A |µ)drA ,
r
r
−∞
¯
U = − exp−γv
Godbillon-Camus and Godlewski Credit Risk Management and Information 14/ 24
42. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Gains and Losses with Soft Information 1/2
GAINS : economy of capital K for VaR coverage (more precise
information)
LOSSES : manipulation problem if it gives the agent’s a higher
expected utility compared to its reservation value ⇒ the credit
officer transmits a signal µ while observing a signal µ + f with f > 0
or f < 0
f γ(µ−rD (1+uα σ))
M −
βσ 2 (1+rD )
EUC = − exp−γv exp (10)
¯
E (U)
Godbillon-Camus and Godlewski Credit Risk Management and Information 15/ 24
43. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Gains and Losses with Soft Information 1/2
GAINS : economy of capital K for VaR coverage (more precise
information)
LOSSES : manipulation problem if it gives the agent’s a higher
expected utility compared to its reservation value ⇒ the credit
officer transmits a signal µ while observing a signal µ + f with f > 0
or f < 0
f γ(µ−rD (1+uα σ))
M −
βσ 2 (1+rD )
EUC = − exp−γv exp (10)
¯
E (U)
Godbillon-Camus and Godlewski Credit Risk Management and Information 15/ 24
44. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Gains and Losses with Soft Information 2/2
Result : Downgrading manipulation ⇔ credit officer transmits a
signal µ while he observes µ + f
¯
⇒ in order to get more than U, the agent must induce the principal
in error so that the latter under-estimates what he actually
attributes to the credit officer
⇒ the agent’s utility depends upon his budget’s development and
his salary
⇒ the principal always guarantees the reservation utility level
⇒ under-estimating the signal allows the credit officer to limit the
expropriation by the banker and increases his expected utility
Godbillon-Camus and Godlewski Credit Risk Management and Information 16/ 24
45. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Gains and Losses with Soft Information 2/2
Result : Downgrading manipulation ⇔ credit officer transmits a
signal µ while he observes µ + f
¯
⇒ in order to get more than U, the agent must induce the principal
in error so that the latter under-estimates what he actually
attributes to the credit officer
⇒ the agent’s utility depends upon his budget’s development and
his salary
⇒ the principal always guarantees the reservation utility level
⇒ under-estimating the signal allows the credit officer to limit the
expropriation by the banker and increases his expected utility
Godbillon-Camus and Godlewski Credit Risk Management and Information 16/ 24
46. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Gains and Losses with Soft Information 2/2
Result : Downgrading manipulation ⇔ credit officer transmits a
signal µ while he observes µ + f
¯
⇒ in order to get more than U, the agent must induce the principal
in error so that the latter under-estimates what he actually
attributes to the credit officer
⇒ the agent’s utility depends upon his budget’s development and
his salary
⇒ the principal always guarantees the reservation utility level
⇒ under-estimating the signal allows the credit officer to limit the
expropriation by the banker and increases his expected utility
Godbillon-Camus and Godlewski Credit Risk Management and Information 16/ 24
47. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard and Soft Information 1/3
Hyp.: Soft Information produced by the credit officer (e.g.
relationship lending) is non verifiable and manipulable (moral hazard
problem with hidden information) but more precise ⇒ modification
of the credit decision process organization
Incentive salary package
w = w0 + w1 (˜A − bm)
r (11)
with m being the transmitted signal by the credit officer (“predicted mean
return”), which might be different from the observed signal µ, and b :
weighting factor for an objective to attain, with a bonus in case of
out-performance
Godbillon-Camus and Godlewski Credit Risk Management and Information 17/ 24
48. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard and Soft Information 1/3
Hyp.: Soft Information produced by the credit officer (e.g.
relationship lending) is non verifiable and manipulable (moral hazard
problem with hidden information) but more precise ⇒ modification
of the credit decision process organization
Incentive salary package
w = w0 + w1 (˜A − bm)
r (11)
with m being the transmitted signal by the credit officer (“predicted mean
return”), which might be different from the observed signal µ, and b :
weighting factor for an objective to attain, with a bonus in case of
out-performance
Godbillon-Camus and Godlewski Credit Risk Management and Information 17/ 24
49. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard and Soft Information 2/3
Optimization program with Soft information
max EUB (µ),
w0 ,w1 ,K ,A,D
EUC (µ) ≥ U, ¯
µ ∈ arg max EU (m),
C
m
K , A, D ∈ arg max EU (m),
B (12)
ˆ ˆ ˆ
K ,A,D
K + D − A = 0,
ˆ ˆ ˆ
ˆ − VaRα ≥ 0,
K
ˆ
VaRα = rα A = rD −µ−uα σ A.ˆ
1+rD
Godbillon-Camus and Godlewski Credit Risk Management and Information 18/ 24
50. Background, Motivation & Aim
Main Framework
Literature Survey
Banker & Credit Officer with Hard Information
Model
Gains and Losses with Soft Information
Results
Banker & Credit Officer with Hard and Soft Information
Discussion
Model
Banker & Credit Officer with Hard and Soft Information 3/3
with
+∞
EUB (m) = − exp[−β(˜A A−rD D−(w0 +w1 (˜A −bm)))] η(˜A |m)drA
r r
r
−∞
+∞
EUB (µ) = − exp[−β(˜A A−rD D−(w0 +w1 (˜A −bµ)))] η(˜A |µ)drA
r r
r
−∞
+∞
EUC (m) = − exp[−γ(˜A A+(w0 +w1 (˜A −bm)))] η(˜A |µ)drA
r r
r
−∞
+∞
EUC (µ) = − exp[−γ(˜A A+(w0 +w1 (˜A −bµ)))] η(˜A |µ)drA
r r
r
−∞
Godbillon-Camus and Godlewski Credit Risk Management and Information 19/ 24
51. Background, Motivation & Aim
Literature Survey
Analytical Results
Model
Comparison of Results with Hard versus Hard & Soft Information
Results
Discussion
Results
Analytical Results
Information Hard
(µ−rD (1+uα σ))(γ(µ−rD (1+uα σ))−2βµ(1+rD ))
Ew ∗ = v + 2β 2 σ 2 (1+rD )2
(µ−rD (1+uα σ))
A∗ = βσ 2 (1+r )
D
rD −µ−uα σ µ−rD (1+uα σ)
K∗ = 1+rD βσ 2 (1+rD )
(µ−rD (1+uα σ))(β(rD (1+uα σ)−µ(3+2rD ))+γ(µ−rD (1+uα σ)))
+βv
2βσ 2 (1+rD )2
EUB = − exp
∗
Information Hard et Soft
2µγ 2 (µ−rD (1+uα σ))
Ew ∗∗ = v − βσ 2 (1+rD )(bβ+2γ)2
+ b(µ − rD (1 + uα σ))Φ
(µ−rD (1+uα σ))(γ+bβ)+βµ(1+rD )
A∗∗ = βσ 2 (1+rD )(2γ+bβ)
(µ−rD (1+uα σ))(γ+bβ)+βµ(1+rD )
K ∗∗ rD −µ−uα σ
1+rD βσ 2 (bβ+2γ)
EUB∗∗ = . . .
Godbillon-Camus and Godlewski Credit Risk Management and Information 20/ 24
52. Background, Motivation & Aim
Literature Survey
Analytical Results
Model
Comparison of Results with Hard versus Hard & Soft Information
Results
Discussion
Results
Comparison of Results with Hard versus Hard & Soft Information
Comparison of the hard versus hard & soft solutions in terms of
differences of:
expected salary Ew , capital K , assets A and banker’s expected
utility E (UB )
Numerical simulations for fixed parameters except the signal µ
Parameters values are (with respect to the analytical constraints)
rD = 0.025, v = 0, uα=0.01 = −2.3263, β = γ = 1, σ = 0.2, λ = 0.1 (level
of hard’s signal imprecision as σH = σS + λ), b = 2 and µ : [0.03; 0.75].
Godbillon-Camus and Godlewski Credit Risk Management and Information 21/ 24
53. Background, Motivation & Aim
Literature Survey
Analytical Results
Model
Comparison of Results with Hard versus Hard & Soft Information
Results
Discussion
Results
Comparison of Results with Hard versus Hard & Soft Information
Comparison of the hard versus hard & soft solutions in terms of
differences of:
expected salary Ew , capital K , assets A and banker’s expected
utility E (UB )
Numerical simulations for fixed parameters except the signal µ
Parameters values are (with respect to the analytical constraints)
rD = 0.025, v = 0, uα=0.01 = −2.3263, β = γ = 1, σ = 0.2, λ = 0.1 (level
of hard’s signal imprecision as σH = σS + λ), b = 2 and µ : [0.03; 0.75].
Godbillon-Camus and Godlewski Credit Risk Management and Information 21/ 24
54. Background, Motivation & Aim
Literature Survey
Analytical Results
Model
Comparison of Results with Hard versus Hard & Soft Information
Results
Discussion
Results
Comparison of Results with Hard versus Hard & Soft Information
Comparison of the hard versus hard & soft solutions in terms of
differences of:
expected salary Ew , capital K , assets A and banker’s expected
utility E (UB )
Numerical simulations for fixed parameters except the signal µ
Parameters values are (with respect to the analytical constraints)
rD = 0.025, v = 0, uα=0.01 = −2.3263, β = γ = 1, σ = 0.2, λ = 0.1 (level
of hard’s signal imprecision as σH = σS + λ), b = 2 and µ : [0.03; 0.75].
Godbillon-Camus and Godlewski Credit Risk Management and Information 21/ 24
55. Background, Motivation & Aim
Literature Survey
Analytical Results
Model
Comparison of Results with Hard versus Hard & Soft Information
Results
Discussion
Results
3.5 4
3
3
2.5
2
2
1.5
1 1
0.5
0
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
dEw = Ew ∗ − Ew ∗∗ curve f (µ) dK = K ∗ − K ∗∗ curve f (µ)
0 0
-2 -0.05
-0.1
-4
-0.15
-6
-0.2
-8
-0.25
-10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
dA = A∗ − A∗∗ curve f (µ) dE (UB ) = E (UB ) − E (UB ) curve f (µ)
∗ ∗∗
Godbillon-Camus and Godlewski Credit Risk Management and Information 22/ 24
56. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Discussion
Importance of information for bank risk management (Hakenes, 2004)
Distinction of Hard versus Soft Information (Petersen, 2004) ⇔ Different
lending technologies : Transaction Lending versus Relationship lending
(Hakenes and Schnabel 2005; Berger, 2004; Danielsson et al., 2001)⇔
Different organization’s types : hierarchical and centralized vs non
hierarchical and decentralized (Stein, 2002; Takats, 2004)
Investigating the impact of information’s type on credit risk
management’ organization in a principal-agent framework with moral
hazard with hidden information
⇒ Soft information is more precise (pro) but manipulable (con) ⇒
decreases capital’s for VaR coverage and increases lending and
bankers’ utility although incentive salary package is implemented for
the credit officer
Godbillon-Camus and Godlewski Credit Risk Management and Information 23/ 24
57. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Discussion
Importance of information for bank risk management (Hakenes, 2004)
Distinction of Hard versus Soft Information (Petersen, 2004) ⇔ Different
lending technologies : Transaction Lending versus Relationship lending
(Hakenes and Schnabel 2005; Berger, 2004; Danielsson et al., 2001)⇔
Different organization’s types : hierarchical and centralized vs non
hierarchical and decentralized (Stein, 2002; Takats, 2004)
Investigating the impact of information’s type on credit risk
management’ organization in a principal-agent framework with moral
hazard with hidden information
⇒ Soft information is more precise (pro) but manipulable (con) ⇒
decreases capital’s for VaR coverage and increases lending and
bankers’ utility although incentive salary package is implemented for
the credit officer
Godbillon-Camus and Godlewski Credit Risk Management and Information 23/ 24
58. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Discussion
Importance of information for bank risk management (Hakenes, 2004)
Distinction of Hard versus Soft Information (Petersen, 2004) ⇔ Different
lending technologies : Transaction Lending versus Relationship lending
(Hakenes and Schnabel 2005; Berger, 2004; Danielsson et al., 2001)⇔
Different organization’s types : hierarchical and centralized vs non
hierarchical and decentralized (Stein, 2002; Takats, 2004)
Investigating the impact of information’s type on credit risk
management’ organization in a principal-agent framework with moral
hazard with hidden information
⇒ Soft information is more precise (pro) but manipulable (con) ⇒
decreases capital’s for VaR coverage and increases lending and
bankers’ utility although incentive salary package is implemented for
the credit officer
Godbillon-Camus and Godlewski Credit Risk Management and Information 23/ 24
59. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Current Research & Perspectives
Theoretical axe : take into account errors in soft information’s
processing (e.g. collusion between the credit officer and the debtor);
introduce soft information’s treatment cost (e.g. in terms of effort);
introduce trust; investigate the impact of different information’s type
on bank competition
Empirical strategy : focus on syndicated loans (decision based on
hard and soft information, following Dennis and Mullineux, 2000)
using Dealscan (LPC, Reuters) - build hard (scoring) and soft
(efficiency score) information’s proxies and investigate their impact
on bank debt’s contract characteristics, . . .
Godbillon-Camus and Godlewski Credit Risk Management and Information 24/ 24
60. Background, Motivation & Aim
Literature Survey
Model
Results
Discussion
Current Research & Perspectives
Theoretical axe : take into account errors in soft information’s
processing (e.g. collusion between the credit officer and the debtor);
introduce soft information’s treatment cost (e.g. in terms of effort);
introduce trust; investigate the impact of different information’s type
on bank competition
Empirical strategy : focus on syndicated loans (decision based on
hard and soft information, following Dennis and Mullineux, 2000)
using Dealscan (LPC, Reuters) - build hard (scoring) and soft
(efficiency score) information’s proxies and investigate their impact
on bank debt’s contract characteristics, . . .
Godbillon-Camus and Godlewski Credit Risk Management and Information 24/ 24